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->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         rcu_read_lock();
 497         for_each_netdev_rcu(net, dev) {
 498                 idev = __in6_dev_get(dev);
 499                 if (idev) {
 500                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 501                         idev->cnf.forwarding = newf;
 502                         if (changed)
 503                                 dev_forward_change(idev);
 504                 }
 505         }
 506         rcu_read_unlock();
 507 }
 508
 509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 510 {
 511         struct net *net;
 512         int old;
 513
 514         if (!rtnl_trylock())
 515                 return restart_syscall();
 516
 517         net = (struct net *)table->extra2;
 518         old = *p;
 519         *p = newf;
 520
 521         if (p == &net->ipv6.devconf_dflt->forwarding) {
 522                 rtnl_unlock();
 523                 return 0;
 524         }
 525
 526         if (p == &net->ipv6.devconf_all->forwarding) {
 527                 net->ipv6.devconf_dflt->forwarding = newf;
 528                 addrconf_forward_change(net, newf);
 529         } else if ((!newf) ^ (!old))
 530                 dev_forward_change((struct inet6_dev *)table->extra1);
 531         rtnl_unlock();
 532
 533         if (newf)
 534                 rt6_purge_dflt_routers(net);
 535         return 1;
 536 }
 537 #endif
 538
 539 /* Nobody refers to this ifaddr, destroy it */
 540 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 541 {
 542         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 543
 544 #ifdef NET_REFCNT_DEBUG
 545         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 546 #endif
 547
 548         in6_dev_put(ifp->idev);
 549
 550         if (del_timer(&ifp->timer))
 551                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 552
 553         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 554                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 555                 return;
 556         }
 557         dst_release(&ifp->rt->dst);
 558
 559         kfree_rcu(ifp, rcu);
 560 }
 561
 562 static void
 563 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 564 {
 565         struct list_head *p;
 566         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 567
 568         /*
 569          * Each device address list is sorted in order of scope -
 570          * global before linklocal.
 571          */
 572         list_for_each(p, &idev->addr_list) {
 573                 struct inet6_ifaddr *ifa
 574                         = list_entry(p, struct inet6_ifaddr, if_list);
 575                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 576                         break;
 577         }
 578
 579         list_add_tail(&ifp->if_list, p);
 580 }
 581
 582 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 583 {
 584         /*
 585          * We perform the hash function over the last 64 bits of the address
 586          * This will include the IEEE address token on links that support it.
 587          */
 588         return jhash_2words((__force u32)addr->s6_addr32[2],
 589                             (__force u32)addr->s6_addr32[3], 0)
 590                 & (IN6_ADDR_HSIZE - 1);
 591 }
 592
 593 /* On success it returns ifp with increased reference count */
 594
 595 static struct inet6_ifaddr *
 596 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 597               int scope, u32 flags)
 598 {
 599         struct inet6_ifaddr *ifa = NULL;
 600         struct rt6_info *rt;
 601         unsigned int hash;
 602         int err = 0;
 603         int addr_type = ipv6_addr_type(addr);
 604
 605         if (addr_type == IPV6_ADDR_ANY ||
 606             addr_type & IPV6_ADDR_MULTICAST ||
 607             (!(idev->dev->flags & IFF_LOOPBACK) &&
 608              addr_type & IPV6_ADDR_LOOPBACK))
 609                 return ERR_PTR(-EADDRNOTAVAIL);
 610
 611         rcu_read_lock_bh();
 612         if (idev->dead) {
 613                 err = -ENODEV;                  /*XXX*/
 614                 goto out2;
 615         }
 616
 617         if (idev->cnf.disable_ipv6) {
 618                 err = -EACCES;
 619                 goto out2;
 620         }
 621
 622         spin_lock(&addrconf_hash_lock);
 623
 624         /* Ignore adding duplicate addresses on an interface */
 625         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 626                 ADBG(("ipv6_add_addr: already assigned\n"));
 627                 err = -EEXIST;
 628                 goto out;
 629         }
 630
 631         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 632
 633         if (ifa == NULL) {
 634                 ADBG(("ipv6_add_addr: malloc failed\n"));
 635                 err = -ENOBUFS;
 636                 goto out;
 637         }
 638
 639         rt = addrconf_dst_alloc(idev, addr, false);
 640         if (IS_ERR(rt)) {
 641                 err = PTR_ERR(rt);
 642                 goto out;
 643         }
 644
 645         ifa->addr = *addr;
 646
 647         spin_lock_init(&ifa->lock);
 648         spin_lock_init(&ifa->state_lock);
 649         init_timer(&ifa->timer);
 650         INIT_HLIST_NODE(&ifa->addr_lst);
 651         ifa->timer.data = (unsigned long) ifa;
 652         ifa->scope = scope;
 653         ifa->prefix_len = pfxlen;
 654         ifa->flags = flags | IFA_F_TENTATIVE;
 655         ifa->cstamp = ifa->tstamp = jiffies;
 656
 657         ifa->rt = rt;
 658
 659         ifa->idev = idev;
 660         in6_dev_hold(idev);
 661         /* For caller */
 662         in6_ifa_hold(ifa);
 663
 664         /* Add to big hash table */
 665         hash = ipv6_addr_hash(addr);
 666
 667         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 668         spin_unlock(&addrconf_hash_lock);
 669
 670         write_lock(&idev->lock);
 671         /* Add to inet6_dev unicast addr list. */
 672         ipv6_link_dev_addr(idev, ifa);
 673
 674 #ifdef CONFIG_IPV6_PRIVACY
 675         if (ifa->flags&IFA_F_TEMPORARY) {
 676                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 677                 in6_ifa_hold(ifa);
 678         }
 679 #endif
 680
 681         in6_ifa_hold(ifa);
 682         write_unlock(&idev->lock);
 683 out2:
 684         rcu_read_unlock_bh();
 685
 686         if (likely(err == 0))
 687                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 688         else {
 689                 kfree(ifa);
 690                 ifa = ERR_PTR(err);
 691         }
 692
 693         return ifa;
 694 out:
 695         spin_unlock(&addrconf_hash_lock);
 696         goto out2;
 697 }
 698
 699 /* This function wants to get referenced ifp and releases it before return */
 700
 701 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 702 {
 703         struct inet6_ifaddr *ifa, *ifn;
 704         struct inet6_dev *idev = ifp->idev;
 705         int state;
 706         int deleted = 0, onlink = 0;
 707         unsigned long expires = jiffies;
 708
 709         spin_lock_bh(&ifp->state_lock);
 710         state = ifp->state;
 711         ifp->state = INET6_IFADDR_STATE_DEAD;
 712         spin_unlock_bh(&ifp->state_lock);
 713
 714         if (state == INET6_IFADDR_STATE_DEAD)
 715                 goto out;
 716
 717         spin_lock_bh(&addrconf_hash_lock);
 718         hlist_del_init_rcu(&ifp->addr_lst);
 719         spin_unlock_bh(&addrconf_hash_lock);
 720
 721         write_lock_bh(&idev->lock);
 722 #ifdef CONFIG_IPV6_PRIVACY
 723         if (ifp->flags&IFA_F_TEMPORARY) {
 724                 list_del(&ifp->tmp_list);
 725                 if (ifp->ifpub) {
 726                         in6_ifa_put(ifp->ifpub);
 727                         ifp->ifpub = NULL;
 728                 }
 729                 __in6_ifa_put(ifp);
 730         }
 731 #endif
 732
 733         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 734                 if (ifa == ifp) {
 735                         list_del_init(&ifp->if_list);
 736                         __in6_ifa_put(ifp);
 737
 738                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 739                                 break;
 740                         deleted = 1;
 741                         continue;
 742                 } else if (ifp->flags & IFA_F_PERMANENT) {
 743                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 744                                               ifp->prefix_len)) {
 745                                 if (ifa->flags & IFA_F_PERMANENT) {
 746                                         onlink = 1;
 747                                         if (deleted)
 748                                                 break;
 749                                 } else {
 750                                         unsigned long lifetime;
 751
 752                                         if (!onlink)
 753                                                 onlink = -1;
 754
 755                                         spin_lock(&ifa->lock);
 756
 757                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 758                                         /*
 759                                          * Note: Because this address is
 760                                          * not permanent, lifetime <
 761                                          * LONG_MAX / HZ here.
 762                                          */
 763                                         if (time_before(expires,
 764                                                         ifa->tstamp + lifetime * HZ))
 765                                                 expires = ifa->tstamp + lifetime * HZ;
 766                                         spin_unlock(&ifa->lock);
 767                                 }
 768                         }
 769                 }
 770         }
 771         write_unlock_bh(&idev->lock);
 772
 773         addrconf_del_timer(ifp);
 774
 775         ipv6_ifa_notify(RTM_DELADDR, ifp);
 776
 777         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 778
 779         /*
 780          * Purge or update corresponding prefix
 781          *
 782          * 1) we don't purge prefix here if address was not permanent.
 783          *    prefix is managed by its own lifetime.
 784          * 2) if there're no addresses, delete prefix.
 785          * 3) if there're still other permanent address(es),
 786          *    corresponding prefix is still permanent.
 787          * 4) otherwise, update prefix lifetime to the
 788          *    longest valid lifetime among the corresponding
 789          *    addresses on the device.
 790          *    Note: subsequent RA will update lifetime.
 791          *
 792          * --yoshfuji
 793          */
 794         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 795                 struct in6_addr prefix;
 796                 struct rt6_info *rt;
 797                 struct net *net = dev_net(ifp->idev->dev);
 798                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 799                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 800
 801                 if (rt && addrconf_is_prefix_route(rt)) {
 802                         if (onlink == 0) {
 803                                 ip6_del_rt(rt);
 804                                 rt = NULL;
 805                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 806                                 rt6_set_expires(rt, expires);
 807                         }
 808                 }
 809                 dst_release(&rt->dst);
 810         }
 811
 812         /* clean up prefsrc entries */
 813         rt6_remove_prefsrc(ifp);
 814 out:
 815         in6_ifa_put(ifp);
 816 }
 817
 818 #ifdef CONFIG_IPV6_PRIVACY
 819 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 820 {
 821         struct inet6_dev *idev = ifp->idev;
 822         struct in6_addr addr, *tmpaddr;
 823         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 824         unsigned long regen_advance;
 825         int tmp_plen;
 826         int ret = 0;
 827         int max_addresses;
 828         u32 addr_flags;
 829         unsigned long now = jiffies;
 830
 831         write_lock(&idev->lock);
 832         if (ift) {
 833                 spin_lock_bh(&ift->lock);
 834                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 835                 spin_unlock_bh(&ift->lock);
 836                 tmpaddr = &addr;
 837         } else {
 838                 tmpaddr = NULL;
 839         }
 840 retry:
 841         in6_dev_hold(idev);
 842         if (idev->cnf.use_tempaddr <= 0) {
 843                 write_unlock(&idev->lock);
 844                 printk(KERN_INFO
 845                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 846                 in6_dev_put(idev);
 847                 ret = -1;
 848                 goto out;
 849         }
 850         spin_lock_bh(&ifp->lock);
 851         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 852                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 853                 spin_unlock_bh(&ifp->lock);
 854                 write_unlock(&idev->lock);
 855                 printk(KERN_WARNING
 856                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 857                 in6_dev_put(idev);
 858                 ret = -1;
 859                 goto out;
 860         }
 861         in6_ifa_hold(ifp);
 862         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 863         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 864                 spin_unlock_bh(&ifp->lock);
 865                 write_unlock(&idev->lock);
 866                 printk(KERN_WARNING
 867                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 868                 in6_ifa_put(ifp);
 869                 in6_dev_put(idev);
 870                 ret = -1;
 871                 goto out;
 872         }
 873         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 874         age = (now - ifp->tstamp) / HZ;
 875         tmp_valid_lft = min_t(__u32,
 876                               ifp->valid_lft,
 877                               idev->cnf.temp_valid_lft + age);
 878         tmp_prefered_lft = min_t(__u32,
 879                                  ifp->prefered_lft,
 880                                  idev->cnf.temp_prefered_lft + age -
 881                                  idev->cnf.max_desync_factor);
 882         tmp_plen = ifp->prefix_len;
 883         max_addresses = idev->cnf.max_addresses;
 884         tmp_tstamp = ifp->tstamp;
 885         spin_unlock_bh(&ifp->lock);
 886
 887         regen_advance = idev->cnf.regen_max_retry *
 888                         idev->cnf.dad_transmits *
 889                         idev->nd_parms->retrans_time / HZ;
 890         write_unlock(&idev->lock);
 891
 892         /* A temporary address is created only if this calculated Preferred
 893          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 894          * an implementation must not create a temporary address with a zero
 895          * Preferred Lifetime.
 896          */
 897         if (tmp_prefered_lft <= regen_advance) {
 898                 in6_ifa_put(ifp);
 899                 in6_dev_put(idev);
 900                 ret = -1;
 901                 goto out;
 902         }
 903
 904         addr_flags = IFA_F_TEMPORARY;
 905         /* set in addrconf_prefix_rcv() */
 906         if (ifp->flags & IFA_F_OPTIMISTIC)
 907                 addr_flags |= IFA_F_OPTIMISTIC;
 908
 909         ift = !max_addresses ||
 910               ipv6_count_addresses(idev) < max_addresses ?
 911                 ipv6_add_addr(idev, &addr, tmp_plen,
 912                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 913                               addr_flags) : NULL;
 914         if (!ift || IS_ERR(ift)) {
 915                 in6_ifa_put(ifp);
 916                 in6_dev_put(idev);
 917                 printk(KERN_INFO
 918                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 919                 tmpaddr = &addr;
 920                 write_lock(&idev->lock);
 921                 goto retry;
 922         }
 923
 924         spin_lock_bh(&ift->lock);
 925         ift->ifpub = ifp;
 926         ift->valid_lft = tmp_valid_lft;
 927         ift->prefered_lft = tmp_prefered_lft;
 928         ift->cstamp = now;
 929         ift->tstamp = tmp_tstamp;
 930         spin_unlock_bh(&ift->lock);
 931
 932         addrconf_dad_start(ift, 0);
 933         in6_ifa_put(ift);
 934         in6_dev_put(idev);
 935 out:
 936         return ret;
 937 }
 938 #endif
 939
 940 /*
 941  *      Choose an appropriate source address (RFC3484)
 942  */
 943 enum {
 944         IPV6_SADDR_RULE_INIT = 0,
 945         IPV6_SADDR_RULE_LOCAL,
 946         IPV6_SADDR_RULE_SCOPE,
 947         IPV6_SADDR_RULE_PREFERRED,
 948 #ifdef CONFIG_IPV6_MIP6
 949         IPV6_SADDR_RULE_HOA,
 950 #endif
 951         IPV6_SADDR_RULE_OIF,
 952         IPV6_SADDR_RULE_LABEL,
 953 #ifdef CONFIG_IPV6_PRIVACY
 954         IPV6_SADDR_RULE_PRIVACY,
 955 #endif
 956         IPV6_SADDR_RULE_ORCHID,
 957         IPV6_SADDR_RULE_PREFIX,
 958         IPV6_SADDR_RULE_MAX
 959 };
 960
 961 struct ipv6_saddr_score {
 962         int                     rule;
 963         int                     addr_type;
 964         struct inet6_ifaddr     *ifa;
 965         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 966         int                     scopedist;
 967         int                     matchlen;
 968 };
 969
 970 struct ipv6_saddr_dst {
 971         const struct in6_addr *addr;
 972         int ifindex;
 973         int scope;
 974         int label;
 975         unsigned int prefs;
 976 };
 977
 978 static inline int ipv6_saddr_preferred(int type)
 979 {
 980         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 981                 return 1;
 982         return 0;
 983 }
 984
 985 static int ipv6_get_saddr_eval(struct net *net,
 986                                struct ipv6_saddr_score *score,
 987                                struct ipv6_saddr_dst *dst,
 988                                int i)
 989 {
 990         int ret;
 991
 992         if (i <= score->rule) {
 993                 switch (i) {
 994                 case IPV6_SADDR_RULE_SCOPE:
 995                         ret = score->scopedist;
 996                         break;
 997                 case IPV6_SADDR_RULE_PREFIX:
 998                         ret = score->matchlen;
 999                         break;
1000                 default:
1001                         ret = !!test_bit(i, score->scorebits);
1002                 }
1003                 goto out;
1004         }
1005
1006         switch (i) {
1007         case IPV6_SADDR_RULE_INIT:
1008                 /* Rule 0: remember if hiscore is not ready yet */
1009                 ret = !!score->ifa;
1010                 break;
1011         case IPV6_SADDR_RULE_LOCAL:
1012                 /* Rule 1: Prefer same address */
1013                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1014                 break;
1015         case IPV6_SADDR_RULE_SCOPE:
1016                 /* Rule 2: Prefer appropriate scope
1017                  *
1018                  *      ret
1019                  *       ^
1020                  *    -1 |  d 15
1021                  *    ---+--+-+---> scope
1022                  *       |
1023                  *       |             d is scope of the destination.
1024                  *  B-d  |  \
1025                  *       |   \      <- smaller scope is better if
1026                  *  B-15 |    \        if scope is enough for destinaion.
1027                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1028                  * d-C-1 | /
1029                  *       |/         <- greater is better
1030                  *   -C  /             if scope is not enough for destination.
1031                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1032                  *
1033                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1034                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1035                  * Assume B = 0 and we get C > 29.
1036                  */
1037                 ret = __ipv6_addr_src_scope(score->addr_type);
1038                 if (ret >= dst->scope)
1039                         ret = -ret;
1040                 else
1041                         ret -= 128;     /* 30 is enough */
1042                 score->scopedist = ret;
1043                 break;
1044         case IPV6_SADDR_RULE_PREFERRED:
1045                 /* Rule 3: Avoid deprecated and optimistic addresses */
1046                 ret = ipv6_saddr_preferred(score->addr_type) ||
1047                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1048                 break;
1049 #ifdef CONFIG_IPV6_MIP6
1050         case IPV6_SADDR_RULE_HOA:
1051             {
1052                 /* Rule 4: Prefer home address */
1053                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1054                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1055                 break;
1056             }
1057 #endif
1058         case IPV6_SADDR_RULE_OIF:
1059                 /* Rule 5: Prefer outgoing interface */
1060                 ret = (!dst->ifindex ||
1061                        dst->ifindex == score->ifa->idev->dev->ifindex);
1062                 break;
1063         case IPV6_SADDR_RULE_LABEL:
1064                 /* Rule 6: Prefer matching label */
1065                 ret = ipv6_addr_label(net,
1066                                       &score->ifa->addr, score->addr_type,
1067                                       score->ifa->idev->dev->ifindex) == dst->label;
1068                 break;
1069 #ifdef CONFIG_IPV6_PRIVACY
1070         case IPV6_SADDR_RULE_PRIVACY:
1071             {
1072                 /* Rule 7: Prefer public address
1073                  * Note: prefer temporary address if use_tempaddr >= 2
1074                  */
1075                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1076                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1077                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1078                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1079                 break;
1080             }
1081 #endif
1082         case IPV6_SADDR_RULE_ORCHID:
1083                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1084                  *          non-ORCHID vs non-ORCHID
1085                  */
1086                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1087                         ipv6_addr_orchid(dst->addr));
1088                 break;
1089         case IPV6_SADDR_RULE_PREFIX:
1090                 /* Rule 8: Use longest matching prefix */
1091                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1092                                                        dst->addr);
1093                 break;
1094         default:
1095                 ret = 0;
1096         }
1097
1098         if (ret)
1099                 __set_bit(i, score->scorebits);
1100         score->rule = i;
1101 out:
1102         return ret;
1103 }
1104
1105 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1106                        const struct in6_addr *daddr, unsigned int prefs,
1107                        struct in6_addr *saddr)
1108 {
1109         struct ipv6_saddr_score scores[2],
1110                                 *score = &scores[0], *hiscore = &scores[1];
1111         struct ipv6_saddr_dst dst;
1112         struct net_device *dev;
1113         int dst_type;
1114
1115         dst_type = __ipv6_addr_type(daddr);
1116         dst.addr = daddr;
1117         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1118         dst.scope = __ipv6_addr_src_scope(dst_type);
1119         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1120         dst.prefs = prefs;
1121
1122         hiscore->rule = -1;
1123         hiscore->ifa = NULL;
1124
1125         rcu_read_lock();
1126
1127         for_each_netdev_rcu(net, dev) {
1128                 struct inet6_dev *idev;
1129
1130                 /* Candidate Source Address (section 4)
1131                  *  - multicast and link-local destination address,
1132                  *    the set of candidate source address MUST only
1133                  *    include addresses assigned to interfaces
1134                  *    belonging to the same link as the outgoing
1135                  *    interface.
1136                  * (- For site-local destination addresses, the
1137                  *    set of candidate source addresses MUST only
1138                  *    include addresses assigned to interfaces
1139                  *    belonging to the same site as the outgoing
1140                  *    interface.)
1141                  */
1142                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1143                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1144                     dst.ifindex && dev->ifindex != dst.ifindex)
1145                         continue;
1146
1147                 idev = __in6_dev_get(dev);
1148                 if (!idev)
1149                         continue;
1150
1151                 read_lock_bh(&idev->lock);
1152                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1153                         int i;
1154
1155                         /*
1156                          * - Tentative Address (RFC2462 section 5.4)
1157                          *  - A tentative address is not considered
1158                          *    "assigned to an interface" in the traditional
1159                          *    sense, unless it is also flagged as optimistic.
1160                          * - Candidate Source Address (section 4)
1161                          *  - In any case, anycast addresses, multicast
1162                          *    addresses, and the unspecified address MUST
1163                          *    NOT be included in a candidate set.
1164                          */
1165                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1166                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1167                                 continue;
1168
1169                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1170
1171                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1172                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1173                                 LIMIT_NETDEBUG(KERN_DEBUG
1174                                                "ADDRCONF: unspecified / multicast address "
1175                                                "assigned as unicast address on %s",
1176                                                dev->name);
1177                                 continue;
1178                         }
1179
1180                         score->rule = -1;
1181                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1182
1183                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1184                                 int minihiscore, miniscore;
1185
1186                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1187                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1188
1189                                 if (minihiscore > miniscore) {
1190                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1191                                             score->scopedist > 0) {
1192                                                 /*
1193                                                  * special case:
1194                                                  * each remaining entry
1195                                                  * has too small (not enough)
1196                                                  * scope, because ifa entries
1197                                                  * are sorted by their scope
1198                                                  * values.
1199                                                  */
1200                                                 goto try_nextdev;
1201                                         }
1202                                         break;
1203                                 } else if (minihiscore < miniscore) {
1204                                         if (hiscore->ifa)
1205                                                 in6_ifa_put(hiscore->ifa);
1206
1207                                         in6_ifa_hold(score->ifa);
1208
1209                                         swap(hiscore, score);
1210
1211                                         /* restore our iterator */
1212                                         score->ifa = hiscore->ifa;
1213
1214                                         break;
1215                                 }
1216                         }
1217                 }
1218 try_nextdev:
1219                 read_unlock_bh(&idev->lock);
1220         }
1221         rcu_read_unlock();
1222
1223         if (!hiscore->ifa)
1224                 return -EADDRNOTAVAIL;
1225
1226         *saddr = hiscore->ifa->addr;
1227         in6_ifa_put(hiscore->ifa);
1228         return 0;
1229 }
1230 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1231
1232 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1233                     unsigned char banned_flags)
1234 {
1235         struct inet6_dev *idev;
1236         int err = -EADDRNOTAVAIL;
1237
1238         rcu_read_lock();
1239         idev = __in6_dev_get(dev);
1240         if (idev) {
1241                 struct inet6_ifaddr *ifp;
1242
1243                 read_lock_bh(&idev->lock);
1244                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1245                         if (ifp->scope == IFA_LINK &&
1246                             !(ifp->flags & banned_flags)) {
1247                                 *addr = ifp->addr;
1248                                 err = 0;
1249                                 break;
1250                         }
1251                 }
1252                 read_unlock_bh(&idev->lock);
1253         }
1254         rcu_read_unlock();
1255         return err;
1256 }
1257
1258 static int ipv6_count_addresses(struct inet6_dev *idev)
1259 {
1260         int cnt = 0;
1261         struct inet6_ifaddr *ifp;
1262
1263         read_lock_bh(&idev->lock);
1264         list_for_each_entry(ifp, &idev->addr_list, if_list)
1265                 cnt++;
1266         read_unlock_bh(&idev->lock);
1267         return cnt;
1268 }
1269
1270 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1271                   struct net_device *dev, int strict)
1272 {
1273         struct inet6_ifaddr *ifp;
1274         struct hlist_node *node;
1275         unsigned int hash = ipv6_addr_hash(addr);
1276
1277         rcu_read_lock_bh();
1278         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1279                 if (!net_eq(dev_net(ifp->idev->dev), net))
1280                         continue;
1281                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1282                     !(ifp->flags&IFA_F_TENTATIVE) &&
1283                     (dev == NULL || ifp->idev->dev == dev ||
1284                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1285                         rcu_read_unlock_bh();
1286                         return 1;
1287                 }
1288         }
1289
1290         rcu_read_unlock_bh();
1291         return 0;
1292 }
1293 EXPORT_SYMBOL(ipv6_chk_addr);
1294
1295 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1296                                struct net_device *dev)
1297 {
1298         unsigned int hash = ipv6_addr_hash(addr);
1299         struct inet6_ifaddr *ifp;
1300         struct hlist_node *node;
1301
1302         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1303                 if (!net_eq(dev_net(ifp->idev->dev), net))
1304                         continue;
1305                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1306                         if (dev == NULL || ifp->idev->dev == dev)
1307                                 return true;
1308                 }
1309         }
1310         return false;
1311 }
1312
1313 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1314 {
1315         struct inet6_dev *idev;
1316         struct inet6_ifaddr *ifa;
1317         int     onlink;
1318
1319         onlink = 0;
1320         rcu_read_lock();
1321         idev = __in6_dev_get(dev);
1322         if (idev) {
1323                 read_lock_bh(&idev->lock);
1324                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1325                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1326                                                    ifa->prefix_len);
1327                         if (onlink)
1328                                 break;
1329                 }
1330                 read_unlock_bh(&idev->lock);
1331         }
1332         rcu_read_unlock();
1333         return onlink;
1334 }
1335
1336 EXPORT_SYMBOL(ipv6_chk_prefix);
1337
1338 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1339                                      struct net_device *dev, int strict)
1340 {
1341         struct inet6_ifaddr *ifp, *result = NULL;
1342         unsigned int hash = ipv6_addr_hash(addr);
1343         struct hlist_node *node;
1344
1345         rcu_read_lock_bh();
1346         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1347                 if (!net_eq(dev_net(ifp->idev->dev), net))
1348                         continue;
1349                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1350                         if (dev == NULL || ifp->idev->dev == dev ||
1351                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1352                                 result = ifp;
1353                                 in6_ifa_hold(ifp);
1354                                 break;
1355                         }
1356                 }
1357         }
1358         rcu_read_unlock_bh();
1359
1360         return result;
1361 }
1362
1363 /* Gets referenced address, destroys ifaddr */
1364
1365 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1366 {
1367         if (ifp->flags&IFA_F_PERMANENT) {
1368                 spin_lock_bh(&ifp->lock);
1369                 addrconf_del_timer(ifp);
1370                 ifp->flags |= IFA_F_TENTATIVE;
1371                 if (dad_failed)
1372                         ifp->flags |= IFA_F_DADFAILED;
1373                 spin_unlock_bh(&ifp->lock);
1374                 if (dad_failed)
1375                         ipv6_ifa_notify(0, ifp);
1376                 in6_ifa_put(ifp);
1377 #ifdef CONFIG_IPV6_PRIVACY
1378         } else if (ifp->flags&IFA_F_TEMPORARY) {
1379                 struct inet6_ifaddr *ifpub;
1380                 spin_lock_bh(&ifp->lock);
1381                 ifpub = ifp->ifpub;
1382                 if (ifpub) {
1383                         in6_ifa_hold(ifpub);
1384                         spin_unlock_bh(&ifp->lock);
1385                         ipv6_create_tempaddr(ifpub, ifp);
1386                         in6_ifa_put(ifpub);
1387                 } else {
1388                         spin_unlock_bh(&ifp->lock);
1389                 }
1390                 ipv6_del_addr(ifp);
1391 #endif
1392         } else
1393                 ipv6_del_addr(ifp);
1394 }
1395
1396 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1397 {
1398         int err = -ENOENT;
1399
1400         spin_lock(&ifp->state_lock);
1401         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1402                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1403                 err = 0;
1404         }
1405         spin_unlock(&ifp->state_lock);
1406
1407         return err;
1408 }
1409
1410 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1411 {
1412         struct inet6_dev *idev = ifp->idev;
1413
1414         if (addrconf_dad_end(ifp)) {
1415                 in6_ifa_put(ifp);
1416                 return;
1417         }
1418
1419         if (net_ratelimit())
1420                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1421                         ifp->idev->dev->name, &ifp->addr);
1422
1423         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1424                 struct in6_addr addr;
1425
1426                 addr.s6_addr32[0] = htonl(0xfe800000);
1427                 addr.s6_addr32[1] = 0;
1428
1429                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1430                     ipv6_addr_equal(&ifp->addr, &addr)) {
1431                         /* DAD failed for link-local based on MAC address */
1432                         idev->cnf.disable_ipv6 = 1;
1433
1434                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1435                                 ifp->idev->dev->name);
1436                 }
1437         }
1438
1439         addrconf_dad_stop(ifp, 1);
1440 }
1441
1442 /* Join to solicited addr multicast group. */
1443
1444 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1445 {
1446         struct in6_addr maddr;
1447
1448         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1449                 return;
1450
1451         addrconf_addr_solict_mult(addr, &maddr);
1452         ipv6_dev_mc_inc(dev, &maddr);
1453 }
1454
1455 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1456 {
1457         struct in6_addr maddr;
1458
1459         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1460                 return;
1461
1462         addrconf_addr_solict_mult(addr, &maddr);
1463         __ipv6_dev_mc_dec(idev, &maddr);
1464 }
1465
1466 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1467 {
1468         struct in6_addr addr;
1469         if (ifp->prefix_len == 127) /* RFC 6164 */
1470                 return;
1471         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1472         if (ipv6_addr_any(&addr))
1473                 return;
1474         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1475 }
1476
1477 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1478 {
1479         struct in6_addr addr;
1480         if (ifp->prefix_len == 127) /* RFC 6164 */
1481                 return;
1482         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1483         if (ipv6_addr_any(&addr))
1484                 return;
1485         __ipv6_dev_ac_dec(ifp->idev, &addr);
1486 }
1487
1488 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1489 {
1490         if (dev->addr_len != ETH_ALEN)
1491                 return -1;
1492         memcpy(eui, dev->dev_addr, 3);
1493         memcpy(eui + 5, dev->dev_addr + 3, 3);
1494
1495         /*
1496          * The zSeries OSA network cards can be shared among various
1497          * OS instances, but the OSA cards have only one MAC address.
1498          * This leads to duplicate address conflicts in conjunction
1499          * with IPv6 if more than one instance uses the same card.
1500          *
1501          * The driver for these cards can deliver a unique 16-bit
1502          * identifier for each instance sharing the same card.  It is
1503          * placed instead of 0xFFFE in the interface identifier.  The
1504          * "u" bit of the interface identifier is not inverted in this
1505          * case.  Hence the resulting interface identifier has local
1506          * scope according to RFC2373.
1507          */
1508         if (dev->dev_id) {
1509                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1510                 eui[4] = dev->dev_id & 0xFF;
1511         } else {
1512                 eui[3] = 0xFF;
1513                 eui[4] = 0xFE;
1514                 eui[0] ^= 2;
1515         }
1516         return 0;
1517 }
1518
1519 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1520 {
1521         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1522         if (dev->addr_len != ARCNET_ALEN)
1523                 return -1;
1524         memset(eui, 0, 7);
1525         eui[7] = *(u8*)dev->dev_addr;
1526         return 0;
1527 }
1528
1529 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1530 {
1531         if (dev->addr_len != INFINIBAND_ALEN)
1532                 return -1;
1533         memcpy(eui, dev->dev_addr + 12, 8);
1534         eui[0] |= 2;
1535         return 0;
1536 }
1537
1538 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1539 {
1540         if (addr == 0)
1541                 return -1;
1542         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1543                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1544                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1545                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1546                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1547                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1548         eui[1] = 0;
1549         eui[2] = 0x5E;
1550         eui[3] = 0xFE;
1551         memcpy(eui + 4, &addr, 4);
1552         return 0;
1553 }
1554
1555 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1556 {
1557         if (dev->priv_flags & IFF_ISATAP)
1558                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1559         return -1;
1560 }
1561
1562 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1563 {
1564         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1565 }
1566
1567 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1568 {
1569         switch (dev->type) {
1570         case ARPHRD_ETHER:
1571         case ARPHRD_FDDI:
1572         case ARPHRD_IEEE802_TR:
1573                 return addrconf_ifid_eui48(eui, dev);
1574         case ARPHRD_ARCNET:
1575                 return addrconf_ifid_arcnet(eui, dev);
1576         case ARPHRD_INFINIBAND:
1577                 return addrconf_ifid_infiniband(eui, dev);
1578         case ARPHRD_SIT:
1579                 return addrconf_ifid_sit(eui, dev);
1580         case ARPHRD_IPGRE:
1581                 return addrconf_ifid_gre(eui, dev);
1582         }
1583         return -1;
1584 }
1585
1586 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1587 {
1588         int err = -1;
1589         struct inet6_ifaddr *ifp;
1590
1591         read_lock_bh(&idev->lock);
1592         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1593                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1594                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1595                         err = 0;
1596                         break;
1597                 }
1598         }
1599         read_unlock_bh(&idev->lock);
1600         return err;
1601 }
1602
1603 #ifdef CONFIG_IPV6_PRIVACY
1604 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1605 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1606 {
1607 regen:
1608         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1609         idev->rndid[0] &= ~0x02;
1610
1611         /*
1612          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1613          * check if generated address is not inappropriate
1614          *
1615          *  - Reserved subnet anycast (RFC 2526)
1616          *      11111101 11....11 1xxxxxxx
1617          *  - ISATAP (RFC4214) 6.1
1618          *      00-00-5E-FE-xx-xx-xx-xx
1619          *  - value 0
1620          *  - XXX: already assigned to an address on the device
1621          */
1622         if (idev->rndid[0] == 0xfd &&
1623             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1624             (idev->rndid[7]&0x80))
1625                 goto regen;
1626         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1627                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1628                         goto regen;
1629                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1630                         goto regen;
1631         }
1632
1633         return 0;
1634 }
1635
1636 static void ipv6_regen_rndid(unsigned long data)
1637 {
1638         struct inet6_dev *idev = (struct inet6_dev *) data;
1639         unsigned long expires;
1640
1641         rcu_read_lock_bh();
1642         write_lock_bh(&idev->lock);
1643
1644         if (idev->dead)
1645                 goto out;
1646
1647         if (__ipv6_regen_rndid(idev) < 0)
1648                 goto out;
1649
1650         expires = jiffies +
1651                 idev->cnf.temp_prefered_lft * HZ -
1652                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1653                 idev->cnf.max_desync_factor * HZ;
1654         if (time_before(expires, jiffies)) {
1655                 printk(KERN_WARNING
1656                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1657                         idev->dev->name);
1658                 goto out;
1659         }
1660
1661         if (!mod_timer(&idev->regen_timer, expires))
1662                 in6_dev_hold(idev);
1663
1664 out:
1665         write_unlock_bh(&idev->lock);
1666         rcu_read_unlock_bh();
1667         in6_dev_put(idev);
1668 }
1669
1670 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1671         int ret = 0;
1672
1673         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1674                 ret = __ipv6_regen_rndid(idev);
1675         return ret;
1676 }
1677 #endif
1678
1679 /*
1680  *      Add prefix route.
1681  */
1682
1683 static void
1684 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1685                       unsigned long expires, u32 flags)
1686 {
1687         struct fib6_config cfg = {
1688                 .fc_table = RT6_TABLE_PREFIX,
1689                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1690                 .fc_ifindex = dev->ifindex,
1691                 .fc_expires = expires,
1692                 .fc_dst_len = plen,
1693                 .fc_flags = RTF_UP | flags,
1694                 .fc_nlinfo.nl_net = dev_net(dev),
1695                 .fc_protocol = RTPROT_KERNEL,
1696         };
1697
1698         cfg.fc_dst = *pfx;
1699
1700         /* Prevent useless cloning on PtP SIT.
1701            This thing is done here expecting that the whole
1702            class of non-broadcast devices need not cloning.
1703          */
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1706                 cfg.fc_flags |= RTF_NONEXTHOP;
1707 #endif
1708
1709         ip6_route_add(&cfg);
1710 }
1711
1712
1713 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1714                                                   int plen,
1715                                                   const struct net_device *dev,
1716                                                   u32 flags, u32 noflags)
1717 {
1718         struct fib6_node *fn;
1719         struct rt6_info *rt = NULL;
1720         struct fib6_table *table;
1721
1722         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1723         if (table == NULL)
1724                 return NULL;
1725
1726         write_lock_bh(&table->tb6_lock);
1727         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1728         if (!fn)
1729                 goto out;
1730         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1731                 if (rt->dst.dev->ifindex != dev->ifindex)
1732                         continue;
1733                 if ((rt->rt6i_flags & flags) != flags)
1734                         continue;
1735                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1736                         continue;
1737                 dst_hold(&rt->dst);
1738                 break;
1739         }
1740 out:
1741         write_unlock_bh(&table->tb6_lock);
1742         return rt;
1743 }
1744
1745
1746 /* Create "default" multicast route to the interface */
1747
1748 static void addrconf_add_mroute(struct net_device *dev)
1749 {
1750         struct fib6_config cfg = {
1751                 .fc_table = RT6_TABLE_LOCAL,
1752                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1753                 .fc_ifindex = dev->ifindex,
1754                 .fc_dst_len = 8,
1755                 .fc_flags = RTF_UP,
1756                 .fc_nlinfo.nl_net = dev_net(dev),
1757         };
1758
1759         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1760
1761         ip6_route_add(&cfg);
1762 }
1763
1764 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1765 static void sit_route_add(struct net_device *dev)
1766 {
1767         struct fib6_config cfg = {
1768                 .fc_table = RT6_TABLE_MAIN,
1769                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1770                 .fc_ifindex = dev->ifindex,
1771                 .fc_dst_len = 96,
1772                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1773                 .fc_nlinfo.nl_net = dev_net(dev),
1774         };
1775
1776         /* prefix length - 96 bits "::d.d.d.d" */
1777         ip6_route_add(&cfg);
1778 }
1779 #endif
1780
1781 static void addrconf_add_lroute(struct net_device *dev)
1782 {
1783         struct in6_addr addr;
1784
1785         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1786         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1787 }
1788
1789 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1790 {
1791         struct inet6_dev *idev;
1792
1793         ASSERT_RTNL();
1794
1795         idev = ipv6_find_idev(dev);
1796         if (!idev)
1797                 return ERR_PTR(-ENOBUFS);
1798
1799         if (idev->cnf.disable_ipv6)
1800                 return ERR_PTR(-EACCES);
1801
1802         /* Add default multicast route */
1803         if (!(dev->flags & IFF_LOOPBACK))
1804                 addrconf_add_mroute(dev);
1805
1806         /* Add link local route */
1807         addrconf_add_lroute(dev);
1808         return idev;
1809 }
1810
1811 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1812 {
1813         struct prefix_info *pinfo;
1814         __u32 valid_lft;
1815         __u32 prefered_lft;
1816         int addr_type;
1817         struct inet6_dev *in6_dev;
1818         struct net *net = dev_net(dev);
1819
1820         pinfo = (struct prefix_info *) opt;
1821
1822         if (len < sizeof(struct prefix_info)) {
1823                 ADBG(("addrconf: prefix option too short\n"));
1824                 return;
1825         }
1826
1827         /*
1828          *      Validation checks ([ADDRCONF], page 19)
1829          */
1830
1831         addr_type = ipv6_addr_type(&pinfo->prefix);
1832
1833         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1834                 return;
1835
1836         valid_lft = ntohl(pinfo->valid);
1837         prefered_lft = ntohl(pinfo->prefered);
1838
1839         if (prefered_lft > valid_lft) {
1840                 if (net_ratelimit())
1841                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1842                 return;
1843         }
1844
1845         in6_dev = in6_dev_get(dev);
1846
1847         if (in6_dev == NULL) {
1848                 if (net_ratelimit())
1849                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1850                 return;
1851         }
1852
1853         /*
1854          *      Two things going on here:
1855          *      1) Add routes for on-link prefixes
1856          *      2) Configure prefixes with the auto flag set
1857          */
1858
1859         if (pinfo->onlink) {
1860                 struct rt6_info *rt;
1861                 unsigned long rt_expires;
1862
1863                 /* Avoid arithmetic overflow. Really, we could
1864                  * save rt_expires in seconds, likely valid_lft,
1865                  * but it would require division in fib gc, that it
1866                  * not good.
1867                  */
1868                 if (HZ > USER_HZ)
1869                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1870                 else
1871                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1872
1873                 if (addrconf_finite_timeout(rt_expires))
1874                         rt_expires *= HZ;
1875
1876                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1877                                                pinfo->prefix_len,
1878                                                dev,
1879                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1880                                                RTF_GATEWAY | RTF_DEFAULT);
1881
1882                 if (rt) {
1883                         /* Autoconf prefix route */
1884                         if (valid_lft == 0) {
1885                                 ip6_del_rt(rt);
1886                                 rt = NULL;
1887                         } else if (addrconf_finite_timeout(rt_expires)) {
1888                                 /* not infinity */
1889                                 rt6_set_expires(rt, jiffies + rt_expires);
1890                         } else {
1891                                 rt6_clean_expires(rt);
1892                         }
1893                 } else if (valid_lft) {
1894                         clock_t expires = 0;
1895                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1896                         if (addrconf_finite_timeout(rt_expires)) {
1897                                 /* not infinity */
1898                                 flags |= RTF_EXPIRES;
1899                                 expires = jiffies_to_clock_t(rt_expires);
1900                         }
1901                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1902                                               dev, expires, flags);
1903                 }
1904                 if (rt)
1905                         dst_release(&rt->dst);
1906         }
1907
1908         /* Try to figure out our local address for this prefix */
1909
1910         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1911                 struct inet6_ifaddr * ifp;
1912                 struct in6_addr addr;
1913                 int create = 0, update_lft = 0;
1914
1915                 if (pinfo->prefix_len == 64) {
1916                         memcpy(&addr, &pinfo->prefix, 8);
1917                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1918                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1919                                 in6_dev_put(in6_dev);
1920                                 return;
1921                         }
1922                         goto ok;
1923                 }
1924                 if (net_ratelimit())
1925                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1926                                pinfo->prefix_len);
1927                 in6_dev_put(in6_dev);
1928                 return;
1929
1930 ok:
1931
1932                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1933
1934                 if (ifp == NULL && valid_lft) {
1935                         int max_addresses = in6_dev->cnf.max_addresses;
1936                         u32 addr_flags = 0;
1937
1938 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1939                         if (in6_dev->cnf.optimistic_dad &&
1940                             !net->ipv6.devconf_all->forwarding && sllao)
1941                                 addr_flags = IFA_F_OPTIMISTIC;
1942 #endif
1943
1944                         /* Do not allow to create too much of autoconfigured
1945                          * addresses; this would be too easy way to crash kernel.
1946                          */
1947                         if (!max_addresses ||
1948                             ipv6_count_addresses(in6_dev) < max_addresses)
1949                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1950                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1951                                                     addr_flags);
1952
1953                         if (!ifp || IS_ERR(ifp)) {
1954                                 in6_dev_put(in6_dev);
1955                                 return;
1956                         }
1957
1958                         update_lft = create = 1;
1959                         ifp->cstamp = jiffies;
1960                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1961                 }
1962
1963                 if (ifp) {
1964                         int flags;
1965                         unsigned long now;
1966 #ifdef CONFIG_IPV6_PRIVACY
1967                         struct inet6_ifaddr *ift;
1968 #endif
1969                         u32 stored_lft;
1970
1971                         /* update lifetime (RFC2462 5.5.3 e) */
1972                         spin_lock(&ifp->lock);
1973                         now = jiffies;
1974                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1975                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1976                         else
1977                                 stored_lft = 0;
1978                         if (!update_lft && stored_lft) {
1979                                 if (valid_lft > MIN_VALID_LIFETIME ||
1980                                     valid_lft > stored_lft)
1981                                         update_lft = 1;
1982                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1983                                         /* valid_lft <= stored_lft is always true */
1984                                         /*
1985                                          * RFC 4862 Section 5.5.3e:
1986                                          * "Note that the preferred lifetime of
1987                                          *  the corresponding address is always
1988                                          *  reset to the Preferred Lifetime in
1989                                          *  the received Prefix Information
1990                                          *  option, regardless of whether the
1991                                          *  valid lifetime is also reset or
1992                                          *  ignored."
1993                                          *
1994                                          *  So if the preferred lifetime in
1995                                          *  this advertisement is different
1996                                          *  than what we have stored, but the
1997                                          *  valid lifetime is invalid, just
1998                                          *  reset prefered_lft.
1999                                          *
2000                                          *  We must set the valid lifetime
2001                                          *  to the stored lifetime since we'll
2002                                          *  be updating the timestamp below,
2003                                          *  else we'll set it back to the
2004                                          *  minimum.
2005                                          */
2006                                         if (prefered_lft != ifp->prefered_lft) {
2007                                                 valid_lft = stored_lft;
2008                                                 update_lft = 1;
2009                                         }
2010                                 } else {
2011                                         valid_lft = MIN_VALID_LIFETIME;
2012                                         if (valid_lft < prefered_lft)
2013                                                 prefered_lft = valid_lft;
2014                                         update_lft = 1;
2015                                 }
2016                         }
2017
2018                         if (update_lft) {
2019                                 ifp->valid_lft = valid_lft;
2020                                 ifp->prefered_lft = prefered_lft;
2021                                 ifp->tstamp = now;
2022                                 flags = ifp->flags;
2023                                 ifp->flags &= ~IFA_F_DEPRECATED;
2024                                 spin_unlock(&ifp->lock);
2025
2026                                 if (!(flags&IFA_F_TENTATIVE))
2027                                         ipv6_ifa_notify(0, ifp);
2028                         } else
2029                                 spin_unlock(&ifp->lock);
2030
2031 #ifdef CONFIG_IPV6_PRIVACY
2032                         read_lock_bh(&in6_dev->lock);
2033                         /* update all temporary addresses in the list */
2034                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2035                                             tmp_list) {
2036                                 int age, max_valid, max_prefered;
2037
2038                                 if (ifp != ift->ifpub)
2039                                         continue;
2040
2041                                 /*
2042                                  * RFC 4941 section 3.3:
2043                                  * If a received option will extend the lifetime
2044                                  * of a public address, the lifetimes of
2045                                  * temporary addresses should be extended,
2046                                  * subject to the overall constraint that no
2047                                  * temporary addresses should ever remain
2048                                  * "valid" or "preferred" for a time longer than
2049                                  * (TEMP_VALID_LIFETIME) or
2050                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2051                                  * respectively.
2052                                  */
2053                                 age = (now - ift->cstamp) / HZ;
2054                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2055                                 if (max_valid < 0)
2056                                         max_valid = 0;
2057
2058                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2059                                                in6_dev->cnf.max_desync_factor -
2060                                                age;
2061                                 if (max_prefered < 0)
2062                                         max_prefered = 0;
2063
2064                                 if (valid_lft > max_valid)
2065                                         valid_lft = max_valid;
2066
2067                                 if (prefered_lft > max_prefered)
2068                                         prefered_lft = max_prefered;
2069
2070                                 spin_lock(&ift->lock);
2071                                 flags = ift->flags;
2072                                 ift->valid_lft = valid_lft;
2073                                 ift->prefered_lft = prefered_lft;
2074                                 ift->tstamp = now;
2075                                 if (prefered_lft > 0)
2076                                         ift->flags &= ~IFA_F_DEPRECATED;
2077
2078                                 spin_unlock(&ift->lock);
2079                                 if (!(flags&IFA_F_TENTATIVE))
2080                                         ipv6_ifa_notify(0, ift);
2081                         }
2082
2083                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2084                                 /*
2085                                  * When a new public address is created as
2086                                  * described in [ADDRCONF], also create a new
2087                                  * temporary address. Also create a temporary
2088                                  * address if it's enabled but no temporary
2089                                  * address currently exists.
2090                                  */
2091                                 read_unlock_bh(&in6_dev->lock);
2092                                 ipv6_create_tempaddr(ifp, NULL);
2093                         } else {
2094                                 read_unlock_bh(&in6_dev->lock);
2095                         }
2096 #endif
2097                         in6_ifa_put(ifp);
2098                         addrconf_verify(0);
2099                 }
2100         }
2101         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2102         in6_dev_put(in6_dev);
2103 }
2104
2105 /*
2106  *      Set destination address.
2107  *      Special case for SIT interfaces where we create a new "virtual"
2108  *      device.
2109  */
2110 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2111 {
2112         struct in6_ifreq ireq;
2113         struct net_device *dev;
2114         int err = -EINVAL;
2115
2116         rtnl_lock();
2117
2118         err = -EFAULT;
2119         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2120                 goto err_exit;
2121
2122         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2123
2124         err = -ENODEV;
2125         if (dev == NULL)
2126                 goto err_exit;
2127
2128 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2129         if (dev->type == ARPHRD_SIT) {
2130                 const struct net_device_ops *ops = dev->netdev_ops;
2131                 struct ifreq ifr;
2132                 struct ip_tunnel_parm p;
2133
2134                 err = -EADDRNOTAVAIL;
2135                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2136                         goto err_exit;
2137
2138                 memset(&p, 0, sizeof(p));
2139                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2140                 p.iph.saddr = 0;
2141                 p.iph.version = 4;
2142                 p.iph.ihl = 5;
2143                 p.iph.protocol = IPPROTO_IPV6;
2144                 p.iph.ttl = 64;
2145                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2146
2147                 if (ops->ndo_do_ioctl) {
2148                         mm_segment_t oldfs = get_fs();
2149
2150                         set_fs(KERNEL_DS);
2151                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2152                         set_fs(oldfs);
2153                 } else
2154                         err = -EOPNOTSUPP;
2155
2156                 if (err == 0) {
2157                         err = -ENOBUFS;
2158                         dev = __dev_get_by_name(net, p.name);
2159                         if (!dev)
2160                                 goto err_exit;
2161                         err = dev_open(dev);
2162                 }
2163         }
2164 #endif
2165
2166 err_exit:
2167         rtnl_unlock();
2168         return err;
2169 }
2170
2171 /*
2172  *      Manual configuration of address on an interface
2173  */
2174 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2175                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2176                           __u32 valid_lft)
2177 {
2178         struct inet6_ifaddr *ifp;
2179         struct inet6_dev *idev;
2180         struct net_device *dev;
2181         int scope;
2182         u32 flags;
2183         clock_t expires;
2184         unsigned long timeout;
2185
2186         ASSERT_RTNL();
2187
2188         if (plen > 128)
2189                 return -EINVAL;
2190
2191         /* check the lifetime */
2192         if (!valid_lft || prefered_lft > valid_lft)
2193                 return -EINVAL;
2194
2195         dev = __dev_get_by_index(net, ifindex);
2196         if (!dev)
2197                 return -ENODEV;
2198
2199         idev = addrconf_add_dev(dev);
2200         if (IS_ERR(idev))
2201                 return PTR_ERR(idev);
2202
2203         scope = ipv6_addr_scope(pfx);
2204
2205         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2206         if (addrconf_finite_timeout(timeout)) {
2207                 expires = jiffies_to_clock_t(timeout * HZ);
2208                 valid_lft = timeout;
2209                 flags = RTF_EXPIRES;
2210         } else {
2211                 expires = 0;
2212                 flags = 0;
2213                 ifa_flags |= IFA_F_PERMANENT;
2214         }
2215
2216         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2217         if (addrconf_finite_timeout(timeout)) {
2218                 if (timeout == 0)
2219                         ifa_flags |= IFA_F_DEPRECATED;
2220                 prefered_lft = timeout;
2221         }
2222
2223         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2224
2225         if (!IS_ERR(ifp)) {
2226                 spin_lock_bh(&ifp->lock);
2227                 ifp->valid_lft = valid_lft;
2228                 ifp->prefered_lft = prefered_lft;
2229                 ifp->tstamp = jiffies;
2230                 spin_unlock_bh(&ifp->lock);
2231
2232                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2233                                       expires, flags);
2234                 /*
2235                  * Note that section 3.1 of RFC 4429 indicates
2236                  * that the Optimistic flag should not be set for
2237                  * manually configured addresses
2238                  */
2239                 addrconf_dad_start(ifp, 0);
2240                 in6_ifa_put(ifp);
2241                 addrconf_verify(0);
2242                 return 0;
2243         }
2244
2245         return PTR_ERR(ifp);
2246 }
2247
2248 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2249                           unsigned int plen)
2250 {
2251         struct inet6_ifaddr *ifp;
2252         struct inet6_dev *idev;
2253         struct net_device *dev;
2254
2255         if (plen > 128)
2256                 return -EINVAL;
2257
2258         dev = __dev_get_by_index(net, ifindex);
2259         if (!dev)
2260                 return -ENODEV;
2261
2262         if ((idev = __in6_dev_get(dev)) == NULL)
2263                 return -ENXIO;
2264
2265         read_lock_bh(&idev->lock);
2266         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2267                 if (ifp->prefix_len == plen &&
2268                     ipv6_addr_equal(pfx, &ifp->addr)) {
2269                         in6_ifa_hold(ifp);
2270                         read_unlock_bh(&idev->lock);
2271
2272                         ipv6_del_addr(ifp);
2273
2274                         /* If the last address is deleted administratively,
2275                            disable IPv6 on this interface.
2276                          */
2277                         if (list_empty(&idev->addr_list))
2278                                 addrconf_ifdown(idev->dev, 1);
2279                         return 0;
2280                 }
2281         }
2282         read_unlock_bh(&idev->lock);
2283         return -EADDRNOTAVAIL;
2284 }
2285
2286
2287 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2288 {
2289         struct in6_ifreq ireq;
2290         int err;
2291
2292         if (!capable(CAP_NET_ADMIN))
2293                 return -EPERM;
2294
2295         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2296                 return -EFAULT;
2297
2298         rtnl_lock();
2299         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2300                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2301                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2302         rtnl_unlock();
2303         return err;
2304 }
2305
2306 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2307 {
2308         struct in6_ifreq ireq;
2309         int err;
2310
2311         if (!capable(CAP_NET_ADMIN))
2312                 return -EPERM;
2313
2314         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2315                 return -EFAULT;
2316
2317         rtnl_lock();
2318         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2319                              ireq.ifr6_prefixlen);
2320         rtnl_unlock();
2321         return err;
2322 }
2323
2324 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2325                      int plen, int scope)
2326 {
2327         struct inet6_ifaddr *ifp;
2328
2329         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2330         if (!IS_ERR(ifp)) {
2331                 spin_lock_bh(&ifp->lock);
2332                 ifp->flags &= ~IFA_F_TENTATIVE;
2333                 spin_unlock_bh(&ifp->lock);
2334                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2335                 in6_ifa_put(ifp);
2336         }
2337 }
2338
2339 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2340 static void sit_add_v4_addrs(struct inet6_dev *idev)
2341 {
2342         struct in6_addr addr;
2343         struct net_device *dev;
2344         struct net *net = dev_net(idev->dev);
2345         int scope;
2346
2347         ASSERT_RTNL();
2348
2349         memset(&addr, 0, sizeof(struct in6_addr));
2350         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2351
2352         if (idev->dev->flags&IFF_POINTOPOINT) {
2353                 addr.s6_addr32[0] = htonl(0xfe800000);
2354                 scope = IFA_LINK;
2355         } else {
2356                 scope = IPV6_ADDR_COMPATv4;
2357         }
2358
2359         if (addr.s6_addr32[3]) {
2360                 add_addr(idev, &addr, 128, scope);
2361                 return;
2362         }
2363
2364         for_each_netdev(net, dev) {
2365                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2366                 if (in_dev && (dev->flags & IFF_UP)) {
2367                         struct in_ifaddr * ifa;
2368
2369                         int flag = scope;
2370
2371                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2372                                 int plen;
2373
2374                                 addr.s6_addr32[3] = ifa->ifa_local;
2375
2376                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2377                                         continue;
2378                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2379                                         if (idev->dev->flags&IFF_POINTOPOINT)
2380                                                 continue;
2381                                         flag |= IFA_HOST;
2382                                 }
2383                                 if (idev->dev->flags&IFF_POINTOPOINT)
2384                                         plen = 64;
2385                                 else
2386                                         plen = 96;
2387
2388                                 add_addr(idev, &addr, plen, flag);
2389                         }
2390                 }
2391         }
2392 }
2393 #endif
2394
2395 static void init_loopback(struct net_device *dev)
2396 {
2397         struct inet6_dev  *idev;
2398
2399         /* ::1 */
2400
2401         ASSERT_RTNL();
2402
2403         if ((idev = ipv6_find_idev(dev)) == NULL) {
2404                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2405                 return;
2406         }
2407
2408         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2409 }
2410
2411 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2412 {
2413         struct inet6_ifaddr * ifp;
2414         u32 addr_flags = IFA_F_PERMANENT;
2415
2416 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2417         if (idev->cnf.optimistic_dad &&
2418             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2419                 addr_flags |= IFA_F_OPTIMISTIC;
2420 #endif
2421
2422
2423         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2424         if (!IS_ERR(ifp)) {
2425                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2426                 addrconf_dad_start(ifp, 0);
2427                 in6_ifa_put(ifp);
2428         }
2429 }
2430
2431 static void addrconf_dev_config(struct net_device *dev)
2432 {
2433         struct in6_addr addr;
2434         struct inet6_dev    * idev;
2435
2436         ASSERT_RTNL();
2437
2438         if ((dev->type != ARPHRD_ETHER) &&
2439             (dev->type != ARPHRD_FDDI) &&
2440             (dev->type != ARPHRD_IEEE802_TR) &&
2441             (dev->type != ARPHRD_ARCNET) &&
2442             (dev->type != ARPHRD_INFINIBAND)) {
2443                 /* Alas, we support only Ethernet autoconfiguration. */
2444                 return;
2445         }
2446
2447         idev = addrconf_add_dev(dev);
2448         if (IS_ERR(idev))
2449                 return;
2450
2451         memset(&addr, 0, sizeof(struct in6_addr));
2452         addr.s6_addr32[0] = htonl(0xFE800000);
2453
2454         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2455                 addrconf_add_linklocal(idev, &addr);
2456 }
2457
2458 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2459 static void addrconf_sit_config(struct net_device *dev)
2460 {
2461         struct inet6_dev *idev;
2462
2463         ASSERT_RTNL();
2464
2465         /*
2466          * Configure the tunnel with one of our IPv4
2467          * addresses... we should configure all of
2468          * our v4 addrs in the tunnel
2469          */
2470
2471         if ((idev = ipv6_find_idev(dev)) == NULL) {
2472                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2473                 return;
2474         }
2475
2476         if (dev->priv_flags & IFF_ISATAP) {
2477                 struct in6_addr addr;
2478
2479                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2480                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2481                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2482                         addrconf_add_linklocal(idev, &addr);
2483                 return;
2484         }
2485
2486         sit_add_v4_addrs(idev);
2487
2488         if (dev->flags&IFF_POINTOPOINT) {
2489                 addrconf_add_mroute(dev);
2490                 addrconf_add_lroute(dev);
2491         } else
2492                 sit_route_add(dev);
2493 }
2494 #endif
2495
2496 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2497 static void addrconf_gre_config(struct net_device *dev)
2498 {
2499         struct inet6_dev *idev;
2500         struct in6_addr addr;
2501
2502         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2503
2504         ASSERT_RTNL();
2505
2506         if ((idev = ipv6_find_idev(dev)) == NULL) {
2507                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2508                 return;
2509         }
2510
2511         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2512         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2513
2514         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2515                 addrconf_add_linklocal(idev, &addr);
2516 }
2517 #endif
2518
2519 static inline int
2520 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2521 {
2522         struct in6_addr lladdr;
2523
2524         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2525                 addrconf_add_linklocal(idev, &lladdr);
2526                 return 0;
2527         }
2528         return -1;
2529 }
2530
2531 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2532 {
2533         struct net_device *link_dev;
2534         struct net *net = dev_net(idev->dev);
2535
2536         /* first try to inherit the link-local address from the link device */
2537         if (idev->dev->iflink &&
2538             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2539                 if (!ipv6_inherit_linklocal(idev, link_dev))
2540                         return;
2541         }
2542         /* then try to inherit it from any device */
2543         for_each_netdev(net, link_dev) {
2544                 if (!ipv6_inherit_linklocal(idev, link_dev))
2545                         return;
2546         }
2547         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2548 }
2549
2550 /*
2551  * Autoconfigure tunnel with a link-local address so routing protocols,
2552  * DHCPv6, MLD etc. can be run over the virtual link
2553  */
2554
2555 static void addrconf_ip6_tnl_config(struct net_device *dev)
2556 {
2557         struct inet6_dev *idev;
2558
2559         ASSERT_RTNL();
2560
2561         idev = addrconf_add_dev(dev);
2562         if (IS_ERR(idev)) {
2563                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2564                 return;
2565         }
2566         ip6_tnl_add_linklocal(idev);
2567 }
2568
2569 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2570                            void * data)
2571 {
2572         struct net_device *dev = (struct net_device *) data;
2573         struct inet6_dev *idev = __in6_dev_get(dev);
2574         int run_pending = 0;
2575         int err;
2576
2577         switch (event) {
2578         case NETDEV_REGISTER:
2579                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2580                         idev = ipv6_add_dev(dev);
2581                         if (!idev)
2582                                 return notifier_from_errno(-ENOMEM);
2583                 }
2584                 break;
2585
2586         case NETDEV_UP:
2587         case NETDEV_CHANGE:
2588                 if (dev->flags & IFF_SLAVE)
2589                         break;
2590
2591                 if (event == NETDEV_UP) {
2592                         if (!addrconf_qdisc_ok(dev)) {
2593                                 /* device is not ready yet. */
2594                                 printk(KERN_INFO
2595                                         "ADDRCONF(NETDEV_UP): %s: "
2596                                         "link is not ready\n",
2597                                         dev->name);
2598                                 break;
2599                         }
2600
2601                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2602                                 idev = ipv6_add_dev(dev);
2603
2604                         if (idev) {
2605                                 idev->if_flags |= IF_READY;
2606                                 run_pending = 1;
2607                         }
2608                 } else {
2609                         if (!addrconf_qdisc_ok(dev)) {
2610                                 /* device is still not ready. */
2611                                 break;
2612                         }
2613
2614                         if (idev) {
2615                                 if (idev->if_flags & IF_READY)
2616                                         /* device is already configured. */
2617                                         break;
2618                                 idev->if_flags |= IF_READY;
2619                         }
2620
2621                         printk(KERN_INFO
2622                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2623                                         "link becomes ready\n",
2624                                         dev->name);
2625
2626                         run_pending = 1;
2627                 }
2628
2629                 switch (dev->type) {
2630 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2631                 case ARPHRD_SIT:
2632                         addrconf_sit_config(dev);
2633                         break;
2634 #endif
2635 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2636                 case ARPHRD_IPGRE:
2637                         addrconf_gre_config(dev);
2638                         break;
2639 #endif
2640                 case ARPHRD_TUNNEL6:
2641                         addrconf_ip6_tnl_config(dev);
2642                         break;
2643                 case ARPHRD_LOOPBACK:
2644                         init_loopback(dev);
2645                         break;
2646
2647                 default:
2648                         addrconf_dev_config(dev);
2649                         break;
2650                 }
2651
2652                 if (idev) {
2653                         if (run_pending)
2654                                 addrconf_dad_run(idev);
2655
2656                         /*
2657                          * If the MTU changed during the interface down,
2658                          * when the interface up, the changed MTU must be
2659                          * reflected in the idev as well as routers.
2660                          */
2661                         if (idev->cnf.mtu6 != dev->mtu &&
2662                             dev->mtu >= IPV6_MIN_MTU) {
2663                                 rt6_mtu_change(dev, dev->mtu);
2664                                 idev->cnf.mtu6 = dev->mtu;
2665                         }
2666                         idev->tstamp = jiffies;
2667                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2668
2669                         /*
2670                          * If the changed mtu during down is lower than
2671                          * IPV6_MIN_MTU stop IPv6 on this interface.
2672                          */
2673                         if (dev->mtu < IPV6_MIN_MTU)
2674                                 addrconf_ifdown(dev, 1);
2675                 }
2676                 break;
2677
2678         case NETDEV_CHANGEMTU:
2679                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2680                         rt6_mtu_change(dev, dev->mtu);
2681                         idev->cnf.mtu6 = dev->mtu;
2682                         break;
2683                 }
2684
2685                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2686                         idev = ipv6_add_dev(dev);
2687                         if (idev)
2688                                 break;
2689                 }
2690
2691                 /*
2692                  * MTU falled under IPV6_MIN_MTU.
2693                  * Stop IPv6 on this interface.
2694                  */
2695
2696         case NETDEV_DOWN:
2697         case NETDEV_UNREGISTER:
2698                 /*
2699                  *      Remove all addresses from this interface.
2700                  */
2701                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2702                 break;
2703
2704         case NETDEV_CHANGENAME:
2705                 if (idev) {
2706                         snmp6_unregister_dev(idev);
2707                         addrconf_sysctl_unregister(idev);
2708                         addrconf_sysctl_register(idev);
2709                         err = snmp6_register_dev(idev);
2710                         if (err)
2711                                 return notifier_from_errno(err);
2712                 }
2713                 break;
2714
2715         case NETDEV_PRE_TYPE_CHANGE:
2716         case NETDEV_POST_TYPE_CHANGE:
2717                 addrconf_type_change(dev, event);
2718                 break;
2719         }
2720
2721         return NOTIFY_OK;
2722 }
2723
2724 /*
2725  *      addrconf module should be notified of a device going up
2726  */
2727 static struct notifier_block ipv6_dev_notf = {
2728         .notifier_call = addrconf_notify,
2729 };
2730
2731 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2732 {
2733         struct inet6_dev *idev;
2734         ASSERT_RTNL();
2735
2736         idev = __in6_dev_get(dev);
2737
2738         if (event == NETDEV_POST_TYPE_CHANGE)
2739                 ipv6_mc_remap(idev);
2740         else if (event == NETDEV_PRE_TYPE_CHANGE)
2741                 ipv6_mc_unmap(idev);
2742 }
2743
2744 static int addrconf_ifdown(struct net_device *dev, int how)
2745 {
2746         struct net *net = dev_net(dev);
2747         struct inet6_dev *idev;
2748         struct inet6_ifaddr *ifa;
2749         int state, i;
2750
2751         ASSERT_RTNL();
2752
2753         rt6_ifdown(net, dev);
2754         neigh_ifdown(&nd_tbl, dev);
2755
2756         idev = __in6_dev_get(dev);
2757         if (idev == NULL)
2758                 return -ENODEV;
2759
2760         /*
2761          * Step 1: remove reference to ipv6 device from parent device.
2762          *         Do not dev_put!
2763          */
2764         if (how) {
2765                 idev->dead = 1;
2766
2767                 /* protected by rtnl_lock */
2768                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2769
2770                 /* Step 1.5: remove snmp6 entry */
2771                 snmp6_unregister_dev(idev);
2772
2773         }
2774
2775         /* Step 2: clear hash table */
2776         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2777                 struct hlist_head *h = &inet6_addr_lst[i];
2778                 struct hlist_node *n;
2779
2780                 spin_lock_bh(&addrconf_hash_lock);
2781         restart:
2782                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2783                         if (ifa->idev == idev) {
2784                                 hlist_del_init_rcu(&ifa->addr_lst);
2785                                 addrconf_del_timer(ifa);
2786                                 goto restart;
2787                         }
2788                 }
2789                 spin_unlock_bh(&addrconf_hash_lock);
2790         }
2791
2792         write_lock_bh(&idev->lock);
2793
2794         /* Step 2: clear flags for stateless addrconf */
2795         if (!how)
2796                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2797
2798 #ifdef CONFIG_IPV6_PRIVACY
2799         if (how && del_timer(&idev->regen_timer))
2800                 in6_dev_put(idev);
2801
2802         /* Step 3: clear tempaddr list */
2803         while (!list_empty(&idev->tempaddr_list)) {
2804                 ifa = list_first_entry(&idev->tempaddr_list,
2805                                        struct inet6_ifaddr, tmp_list);
2806                 list_del(&ifa->tmp_list);
2807                 write_unlock_bh(&idev->lock);
2808                 spin_lock_bh(&ifa->lock);
2809
2810                 if (ifa->ifpub) {
2811                         in6_ifa_put(ifa->ifpub);
2812                         ifa->ifpub = NULL;
2813                 }
2814                 spin_unlock_bh(&ifa->lock);
2815                 in6_ifa_put(ifa);
2816                 write_lock_bh(&idev->lock);
2817         }
2818 #endif
2819
2820         while (!list_empty(&idev->addr_list)) {
2821                 ifa = list_first_entry(&idev->addr_list,
2822                                        struct inet6_ifaddr, if_list);
2823                 addrconf_del_timer(ifa);
2824
2825                 list_del(&ifa->if_list);
2826
2827                 write_unlock_bh(&idev->lock);
2828
2829                 spin_lock_bh(&ifa->state_lock);
2830                 state = ifa->state;
2831                 ifa->state = INET6_IFADDR_STATE_DEAD;
2832                 spin_unlock_bh(&ifa->state_lock);
2833
2834                 if (state != INET6_IFADDR_STATE_DEAD) {
2835                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2836                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2837                 }
2838                 in6_ifa_put(ifa);
2839
2840                 write_lock_bh(&idev->lock);
2841         }
2842
2843         write_unlock_bh(&idev->lock);
2844
2845         /* Step 5: Discard multicast list */
2846         if (how)
2847                 ipv6_mc_destroy_dev(idev);
2848         else
2849                 ipv6_mc_down(idev);
2850
2851         idev->tstamp = jiffies;
2852
2853         /* Last: Shot the device (if unregistered) */
2854         if (how) {
2855                 addrconf_sysctl_unregister(idev);
2856                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2857                 neigh_ifdown(&nd_tbl, dev);
2858                 in6_dev_put(idev);
2859         }
2860         return 0;
2861 }
2862
2863 static void addrconf_rs_timer(unsigned long data)
2864 {
2865         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2866         struct inet6_dev *idev = ifp->idev;
2867
2868         read_lock(&idev->lock);
2869         if (idev->dead || !(idev->if_flags & IF_READY))
2870                 goto out;
2871
2872         if (idev->cnf.forwarding)
2873                 goto out;
2874
2875         /* Announcement received after solicitation was sent */
2876         if (idev->if_flags & IF_RA_RCVD)
2877                 goto out;
2878
2879         spin_lock(&ifp->lock);
2880         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2881                 /* The wait after the last probe can be shorter */
2882                 addrconf_mod_timer(ifp, AC_RS,
2883                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2884                                    idev->cnf.rtr_solicit_delay :
2885                                    idev->cnf.rtr_solicit_interval);
2886                 spin_unlock(&ifp->lock);
2887
2888                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2889         } else {
2890                 spin_unlock(&ifp->lock);
2891                 /*
2892                  * Note: we do not support deprecated "all on-link"
2893                  * assumption any longer.
2894                  */
2895                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2896                        idev->dev->name);
2897         }
2898
2899 out:
2900         read_unlock(&idev->lock);
2901         in6_ifa_put(ifp);
2902 }
2903
2904 /*
2905  *      Duplicate Address Detection
2906  */
2907 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2908 {
2909         unsigned long rand_num;
2910         struct inet6_dev *idev = ifp->idev;
2911
2912         if (ifp->flags & IFA_F_OPTIMISTIC)
2913                 rand_num = 0;
2914         else
2915                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2916
2917         ifp->probes = idev->cnf.dad_transmits;
2918         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2919 }
2920
2921 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2922 {
2923         struct inet6_dev *idev = ifp->idev;
2924         struct net_device *dev = idev->dev;
2925
2926         addrconf_join_solict(dev, &ifp->addr);
2927
2928         net_srandom(ifp->addr.s6_addr32[3]);
2929
2930         read_lock_bh(&idev->lock);
2931         spin_lock(&ifp->lock);
2932         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2933                 goto out;
2934
2935         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2936             !(dev->flags&IFF_MULTICAST) ||
2937             idev->cnf.accept_dad < 1 ||
2938             !(ifp->flags&IFA_F_TENTATIVE) ||
2939             ifp->flags & IFA_F_NODAD) {
2940                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2941                 spin_unlock(&ifp->lock);
2942                 read_unlock_bh(&idev->lock);
2943
2944                 addrconf_dad_completed(ifp);
2945                 return;
2946         }
2947
2948         if (!(idev->if_flags & IF_READY)) {
2949                 spin_unlock(&ifp->lock);
2950                 read_unlock_bh(&idev->lock);
2951                 /*
2952                  * If the device is not ready:
2953                  * - keep it tentative if it is a permanent address.
2954                  * - otherwise, kill it.
2955                  */
2956                 in6_ifa_hold(ifp);
2957                 addrconf_dad_stop(ifp, 0);
2958                 return;
2959         }
2960
2961         /*
2962          * Optimistic nodes can start receiving
2963          * Frames right away
2964          */
2965         if (ifp->flags & IFA_F_OPTIMISTIC)
2966                 ip6_ins_rt(ifp->rt);
2967
2968         addrconf_dad_kick(ifp);
2969 out:
2970         spin_unlock(&ifp->lock);
2971         read_unlock_bh(&idev->lock);
2972 }
2973
2974 static void addrconf_dad_timer(unsigned long data)
2975 {
2976         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2977         struct inet6_dev *idev = ifp->idev;
2978         struct in6_addr mcaddr;
2979
2980         if (!ifp->probes && addrconf_dad_end(ifp))
2981                 goto out;
2982
2983         read_lock(&idev->lock);
2984         if (idev->dead || !(idev->if_flags & IF_READY)) {
2985                 read_unlock(&idev->lock);
2986                 goto out;
2987         }
2988
2989         spin_lock(&ifp->lock);
2990         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2991                 spin_unlock(&ifp->lock);
2992                 read_unlock(&idev->lock);
2993                 goto out;
2994         }
2995
2996         if (ifp->probes == 0) {
2997                 /*
2998                  * DAD was successful
2999                  */
3000
3001                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3002                 spin_unlock(&ifp->lock);
3003                 read_unlock(&idev->lock);
3004
3005                 addrconf_dad_completed(ifp);
3006
3007                 goto out;
3008         }
3009
3010         ifp->probes--;
3011         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3012         spin_unlock(&ifp->lock);
3013         read_unlock(&idev->lock);
3014
3015         /* send a neighbour solicitation for our addr */
3016         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3017         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3018 out:
3019         in6_ifa_put(ifp);
3020 }
3021
3022 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3023 {
3024         struct net_device *dev = ifp->idev->dev;
3025
3026         /*
3027          *      Configure the address for reception. Now it is valid.
3028          */
3029
3030         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3031
3032         /* If added prefix is link local and we are prepared to process
3033            router advertisements, start sending router solicitations.
3034          */
3035
3036         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3037              ifp->idev->cnf.accept_ra == 2) &&
3038             ifp->idev->cnf.rtr_solicits > 0 &&
3039             (dev->flags&IFF_LOOPBACK) == 0 &&
3040             (dev->flags&IFF_MULTICAST) &&
3041             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3042                 /*
3043                  *      If a host as already performed a random delay
3044                  *      [...] as part of DAD [...] there is no need
3045                  *      to delay again before sending the first RS
3046                  */
3047                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3048
3049                 spin_lock_bh(&ifp->lock);
3050                 ifp->probes = 1;
3051                 ifp->idev->if_flags |= IF_RS_SENT;
3052                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3053                 spin_unlock_bh(&ifp->lock);
3054         }
3055 }
3056
3057 static void addrconf_dad_run(struct inet6_dev *idev)
3058 {
3059         struct inet6_ifaddr *ifp;
3060
3061         read_lock_bh(&idev->lock);
3062         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3063                 spin_lock(&ifp->lock);
3064                 if (ifp->flags & IFA_F_TENTATIVE &&
3065                     ifp->state == INET6_IFADDR_STATE_DAD)
3066                         addrconf_dad_kick(ifp);
3067                 spin_unlock(&ifp->lock);
3068         }
3069         read_unlock_bh(&idev->lock);
3070 }
3071
3072 #ifdef CONFIG_PROC_FS
3073 struct if6_iter_state {
3074         struct seq_net_private p;
3075         int bucket;
3076         int offset;
3077 };
3078
3079 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3080 {
3081         struct inet6_ifaddr *ifa = NULL;
3082         struct if6_iter_state *state = seq->private;
3083         struct net *net = seq_file_net(seq);
3084         int p = 0;
3085
3086         /* initial bucket if pos is 0 */
3087         if (pos == 0) {
3088                 state->bucket = 0;
3089                 state->offset = 0;
3090         }
3091
3092         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3093                 struct hlist_node *n;
3094                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3095                                          addr_lst) {
3096                         /* sync with offset */
3097                         if (p < state->offset) {
3098                                 p++;
3099                                 continue;
3100                         }
3101                         state->offset++;
3102                         if (net_eq(dev_net(ifa->idev->dev), net))
3103                                 return ifa;
3104                 }
3105
3106                 /* prepare for next bucket */
3107                 state->offset = 0;
3108                 p = 0;
3109         }
3110         return NULL;
3111 }
3112
3113 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3114                                          struct inet6_ifaddr *ifa)
3115 {
3116         struct if6_iter_state *state = seq->private;
3117         struct net *net = seq_file_net(seq);
3118         struct hlist_node *n = &ifa->addr_lst;
3119
3120         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3121                 state->offset++;
3122                 if (net_eq(dev_net(ifa->idev->dev), net))
3123                         return ifa;
3124         }
3125
3126         while (++state->bucket < IN6_ADDR_HSIZE) {
3127                 state->offset = 0;
3128                 hlist_for_each_entry_rcu_bh(ifa, n,
3129                                      &inet6_addr_lst[state->bucket], addr_lst) {
3130                         state->offset++;
3131                         if (net_eq(dev_net(ifa->idev->dev), net))
3132                                 return ifa;
3133                 }
3134         }
3135
3136         return NULL;
3137 }
3138
3139 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3140         __acquires(rcu_bh)
3141 {
3142         rcu_read_lock_bh();
3143         return if6_get_first(seq, *pos);
3144 }
3145
3146 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3147 {
3148         struct inet6_ifaddr *ifa;
3149
3150         ifa = if6_get_next(seq, v);
3151         ++*pos;
3152         return ifa;
3153 }
3154
3155 static void if6_seq_stop(struct seq_file *seq, void *v)
3156         __releases(rcu_bh)
3157 {
3158         rcu_read_unlock_bh();
3159 }
3160
3161 static int if6_seq_show(struct seq_file *seq, void *v)
3162 {
3163         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3164         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3165                    &ifp->addr,
3166                    ifp->idev->dev->ifindex,
3167                    ifp->prefix_len,
3168                    ifp->scope,
3169                    ifp->flags,
3170                    ifp->idev->dev->name);
3171         return 0;
3172 }
3173
3174 static const struct seq_operations if6_seq_ops = {
3175         .start  = if6_seq_start,
3176         .next   = if6_seq_next,
3177         .show   = if6_seq_show,
3178         .stop   = if6_seq_stop,
3179 };
3180
3181 static int if6_seq_open(struct inode *inode, struct file *file)
3182 {
3183         return seq_open_net(inode, file, &if6_seq_ops,
3184                             sizeof(struct if6_iter_state));
3185 }
3186
3187 static const struct file_operations if6_fops = {
3188         .owner          = THIS_MODULE,
3189         .open           = if6_seq_open,
3190         .read           = seq_read,
3191         .llseek         = seq_lseek,
3192         .release        = seq_release_net,
3193 };
3194
3195 static int __net_init if6_proc_net_init(struct net *net)
3196 {
3197         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3198                 return -ENOMEM;
3199         return 0;
3200 }
3201
3202 static void __net_exit if6_proc_net_exit(struct net *net)
3203 {
3204        proc_net_remove(net, "if_inet6");
3205 }
3206
3207 static struct pernet_operations if6_proc_net_ops = {
3208        .init = if6_proc_net_init,
3209        .exit = if6_proc_net_exit,
3210 };
3211
3212 int __init if6_proc_init(void)
3213 {
3214         return register_pernet_subsys(&if6_proc_net_ops);
3215 }
3216
3217 void if6_proc_exit(void)
3218 {
3219         unregister_pernet_subsys(&if6_proc_net_ops);
3220 }
3221 #endif  /* CONFIG_PROC_FS */
3222
3223 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3224 /* Check if address is a home address configured on any interface. */
3225 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3226 {
3227         int ret = 0;
3228         struct inet6_ifaddr *ifp = NULL;
3229         struct hlist_node *n;
3230         unsigned int hash = ipv6_addr_hash(addr);
3231
3232         rcu_read_lock_bh();
3233         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3234                 if (!net_eq(dev_net(ifp->idev->dev), net))
3235                         continue;
3236                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3237                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3238                         ret = 1;
3239                         break;
3240                 }
3241         }
3242         rcu_read_unlock_bh();
3243         return ret;
3244 }
3245 #endif
3246
3247 /*
3248  *      Periodic address status verification
3249  */
3250
3251 static void addrconf_verify(unsigned long foo)
3252 {
3253         unsigned long now, next, next_sec, next_sched;
3254         struct inet6_ifaddr *ifp;
3255         struct hlist_node *node;
3256         int i;
3257
3258         rcu_read_lock_bh();
3259         spin_lock(&addrconf_verify_lock);
3260         now = jiffies;
3261         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3262
3263         del_timer(&addr_chk_timer);
3264
3265         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3266 restart:
3267                 hlist_for_each_entry_rcu_bh(ifp, node,
3268                                          &inet6_addr_lst[i], addr_lst) {
3269                         unsigned long age;
3270
3271                         if (ifp->flags & IFA_F_PERMANENT)
3272                                 continue;
3273
3274                         spin_lock(&ifp->lock);
3275                         /* We try to batch several events at once. */
3276                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3277
3278                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3279                             age >= ifp->valid_lft) {
3280                                 spin_unlock(&ifp->lock);
3281                                 in6_ifa_hold(ifp);
3282                                 ipv6_del_addr(ifp);
3283                                 goto restart;
3284                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3285                                 spin_unlock(&ifp->lock);
3286                                 continue;
3287                         } else if (age >= ifp->prefered_lft) {
3288                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3289                                 int deprecate = 0;
3290
3291                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3292                                         deprecate = 1;
3293                                         ifp->flags |= IFA_F_DEPRECATED;
3294                                 }
3295
3296                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3297                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3298
3299                                 spin_unlock(&ifp->lock);
3300
3301                                 if (deprecate) {
3302                                         in6_ifa_hold(ifp);
3303
3304                                         ipv6_ifa_notify(0, ifp);
3305                                         in6_ifa_put(ifp);
3306                                         goto restart;
3307                                 }
3308 #ifdef CONFIG_IPV6_PRIVACY
3309                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3310                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3311                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3312                                         ifp->idev->cnf.dad_transmits *
3313                                         ifp->idev->nd_parms->retrans_time / HZ;
3314
3315                                 if (age >= ifp->prefered_lft - regen_advance) {
3316                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3317                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3318                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3319                                         if (!ifp->regen_count && ifpub) {
3320                                                 ifp->regen_count++;
3321                                                 in6_ifa_hold(ifp);
3322                                                 in6_ifa_hold(ifpub);
3323                                                 spin_unlock(&ifp->lock);
3324
3325                                                 spin_lock(&ifpub->lock);
3326                                                 ifpub->regen_count = 0;
3327                                                 spin_unlock(&ifpub->lock);
3328                                                 ipv6_create_tempaddr(ifpub, ifp);
3329                                                 in6_ifa_put(ifpub);
3330                                                 in6_ifa_put(ifp);
3331                                                 goto restart;
3332                                         }
3333                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3334                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3335                                 spin_unlock(&ifp->lock);
3336 #endif
3337                         } else {
3338                                 /* ifp->prefered_lft <= ifp->valid_lft */
3339                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3340                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3341                                 spin_unlock(&ifp->lock);
3342                         }
3343                 }
3344         }
3345
3346         next_sec = round_jiffies_up(next);
3347         next_sched = next;
3348
3349         /* If rounded timeout is accurate enough, accept it. */
3350         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3351                 next_sched = next_sec;
3352
3353         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3354         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3355                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3356
3357         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3358               now, next, next_sec, next_sched));
3359
3360         addr_chk_timer.expires = next_sched;
3361         add_timer(&addr_chk_timer);
3362         spin_unlock(&addrconf_verify_lock);
3363         rcu_read_unlock_bh();
3364 }
3365
3366 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3367 {
3368         struct in6_addr *pfx = NULL;
3369
3370         if (addr)
3371                 pfx = nla_data(addr);
3372
3373         if (local) {
3374                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3375                         pfx = NULL;
3376                 else
3377                         pfx = nla_data(local);
3378         }
3379
3380         return pfx;
3381 }
3382
3383 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3384         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3385         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3386         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3387 };
3388
3389 static int
3390 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3391 {
3392         struct net *net = sock_net(skb->sk);
3393         struct ifaddrmsg *ifm;
3394         struct nlattr *tb[IFA_MAX+1];
3395         struct in6_addr *pfx;
3396         int err;
3397
3398         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3399         if (err < 0)
3400                 return err;
3401
3402         ifm = nlmsg_data(nlh);
3403         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3404         if (pfx == NULL)
3405                 return -EINVAL;
3406
3407         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3408 }
3409
3410 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3411                              u32 prefered_lft, u32 valid_lft)
3412 {
3413         u32 flags;
3414         clock_t expires;
3415         unsigned long timeout;
3416
3417         if (!valid_lft || (prefered_lft > valid_lft))
3418                 return -EINVAL;
3419
3420         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3421         if (addrconf_finite_timeout(timeout)) {
3422                 expires = jiffies_to_clock_t(timeout * HZ);
3423                 valid_lft = timeout;
3424                 flags = RTF_EXPIRES;
3425         } else {
3426                 expires = 0;
3427                 flags = 0;
3428                 ifa_flags |= IFA_F_PERMANENT;
3429         }
3430
3431         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3432         if (addrconf_finite_timeout(timeout)) {
3433                 if (timeout == 0)
3434                         ifa_flags |= IFA_F_DEPRECATED;
3435                 prefered_lft = timeout;
3436         }
3437
3438         spin_lock_bh(&ifp->lock);
3439         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3440         ifp->tstamp = jiffies;
3441         ifp->valid_lft = valid_lft;
3442         ifp->prefered_lft = prefered_lft;
3443
3444         spin_unlock_bh(&ifp->lock);
3445         if (!(ifp->flags&IFA_F_TENTATIVE))
3446                 ipv6_ifa_notify(0, ifp);
3447
3448         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3449                               expires, flags);
3450         addrconf_verify(0);
3451
3452         return 0;
3453 }
3454
3455 static int
3456 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3457 {
3458         struct net *net = sock_net(skb->sk);
3459         struct ifaddrmsg *ifm;
3460         struct nlattr *tb[IFA_MAX+1];
3461         struct in6_addr *pfx;
3462         struct inet6_ifaddr *ifa;
3463         struct net_device *dev;
3464         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3465         u8 ifa_flags;
3466         int err;
3467
3468         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3469         if (err < 0)
3470                 return err;
3471
3472         ifm = nlmsg_data(nlh);
3473         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3474         if (pfx == NULL)
3475                 return -EINVAL;
3476
3477         if (tb[IFA_CACHEINFO]) {
3478                 struct ifa_cacheinfo *ci;
3479
3480                 ci = nla_data(tb[IFA_CACHEINFO]);
3481                 valid_lft = ci->ifa_valid;
3482                 preferred_lft = ci->ifa_prefered;
3483         } else {
3484                 preferred_lft = INFINITY_LIFE_TIME;
3485                 valid_lft = INFINITY_LIFE_TIME;
3486         }
3487
3488         dev =  __dev_get_by_index(net, ifm->ifa_index);
3489         if (dev == NULL)
3490                 return -ENODEV;
3491
3492         /* We ignore other flags so far. */
3493         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3494
3495         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3496         if (ifa == NULL) {
3497                 /*
3498                  * It would be best to check for !NLM_F_CREATE here but
3499                  * userspace alreay relies on not having to provide this.
3500                  */
3501                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3502                                       ifm->ifa_prefixlen, ifa_flags,
3503                                       preferred_lft, valid_lft);
3504         }
3505
3506         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3507             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3508                 err = -EEXIST;
3509         else
3510                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3511
3512         in6_ifa_put(ifa);
3513
3514         return err;
3515 }
3516
3517 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3518                           u8 scope, int ifindex)
3519 {
3520         struct ifaddrmsg *ifm;
3521
3522         ifm = nlmsg_data(nlh);
3523         ifm->ifa_family = AF_INET6;
3524         ifm->ifa_prefixlen = prefixlen;
3525         ifm->ifa_flags = flags;
3526         ifm->ifa_scope = scope;
3527         ifm->ifa_index = ifindex;
3528 }
3529
3530 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3531                          unsigned long tstamp, u32 preferred, u32 valid)
3532 {
3533         struct ifa_cacheinfo ci;
3534
3535         ci.cstamp = cstamp_delta(cstamp);
3536         ci.tstamp = cstamp_delta(tstamp);
3537         ci.ifa_prefered = preferred;
3538         ci.ifa_valid = valid;
3539
3540         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3541 }
3542
3543 static inline int rt_scope(int ifa_scope)
3544 {
3545         if (ifa_scope & IFA_HOST)
3546                 return RT_SCOPE_HOST;
3547         else if (ifa_scope & IFA_LINK)
3548                 return RT_SCOPE_LINK;
3549         else if (ifa_scope & IFA_SITE)
3550                 return RT_SCOPE_SITE;
3551         else
3552                 return RT_SCOPE_UNIVERSE;
3553 }
3554
3555 static inline int inet6_ifaddr_msgsize(void)
3556 {
3557         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3558                + nla_total_size(16) /* IFA_ADDRESS */
3559                + nla_total_size(sizeof(struct ifa_cacheinfo));
3560 }
3561
3562 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3563                              u32 pid, u32 seq, int event, unsigned int flags)
3564 {
3565         struct nlmsghdr  *nlh;
3566         u32 preferred, valid;
3567
3568         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3569         if (nlh == NULL)
3570                 return -EMSGSIZE;
3571
3572         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3573                       ifa->idev->dev->ifindex);
3574
3575         if (!(ifa->flags&IFA_F_PERMANENT)) {
3576                 preferred = ifa->prefered_lft;
3577                 valid = ifa->valid_lft;
3578                 if (preferred != INFINITY_LIFE_TIME) {
3579                         long tval = (jiffies - ifa->tstamp)/HZ;
3580                         if (preferred > tval)
3581                                 preferred -= tval;
3582                         else
3583                                 preferred = 0;
3584                         if (valid != INFINITY_LIFE_TIME) {
3585                                 if (valid > tval)
3586                                         valid -= tval;
3587                                 else
3588                                         valid = 0;
3589                         }
3590                 }
3591         } else {
3592                 preferred = INFINITY_LIFE_TIME;
3593                 valid = INFINITY_LIFE_TIME;
3594         }
3595
3596         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3597             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3598                 nlmsg_cancel(skb, nlh);
3599                 return -EMSGSIZE;
3600         }
3601
3602         return nlmsg_end(skb, nlh);
3603 }
3604
3605 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3606                                 u32 pid, u32 seq, int event, u16 flags)
3607 {
3608         struct nlmsghdr  *nlh;
3609         u8 scope = RT_SCOPE_UNIVERSE;
3610         int ifindex = ifmca->idev->dev->ifindex;
3611
3612         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3613                 scope = RT_SCOPE_SITE;
3614
3615         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3616         if (nlh == NULL)
3617                 return -EMSGSIZE;
3618
3619         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3620         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3621             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3622                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3623                 nlmsg_cancel(skb, nlh);
3624                 return -EMSGSIZE;
3625         }
3626
3627         return nlmsg_end(skb, nlh);
3628 }
3629
3630 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3631                                 u32 pid, u32 seq, int event, unsigned int flags)
3632 {
3633         struct nlmsghdr  *nlh;
3634         u8 scope = RT_SCOPE_UNIVERSE;
3635         int ifindex = ifaca->aca_idev->dev->ifindex;
3636
3637         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3638                 scope = RT_SCOPE_SITE;
3639
3640         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3641         if (nlh == NULL)
3642                 return -EMSGSIZE;
3643
3644         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3645         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3646             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3647                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3648                 nlmsg_cancel(skb, nlh);
3649                 return -EMSGSIZE;
3650         }
3651
3652         return nlmsg_end(skb, nlh);
3653 }
3654
3655 enum addr_type_t {
3656         UNICAST_ADDR,
3657         MULTICAST_ADDR,
3658         ANYCAST_ADDR,
3659 };
3660
3661 /* called with rcu_read_lock() */
3662 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3663                           struct netlink_callback *cb, enum addr_type_t type,
3664                           int s_ip_idx, int *p_ip_idx)
3665 {
3666         struct ifmcaddr6 *ifmca;
3667         struct ifacaddr6 *ifaca;
3668         int err = 1;
3669         int ip_idx = *p_ip_idx;
3670
3671         read_lock_bh(&idev->lock);
3672         switch (type) {
3673         case UNICAST_ADDR: {
3674                 struct inet6_ifaddr *ifa;
3675
3676                 /* unicast address incl. temp addr */
3677                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3678                         if (++ip_idx < s_ip_idx)
3679                                 continue;
3680                         err = inet6_fill_ifaddr(skb, ifa,
3681                                                 NETLINK_CB(cb->skb).pid,
3682                                                 cb->nlh->nlmsg_seq,
3683                                                 RTM_NEWADDR,
3684                                                 NLM_F_MULTI);
3685                         if (err <= 0)
3686                                 break;
3687                 }
3688                 break;
3689         }
3690         case MULTICAST_ADDR:
3691                 /* multicast address */
3692                 for (ifmca = idev->mc_list; ifmca;
3693                      ifmca = ifmca->next, ip_idx++) {
3694                         if (ip_idx < s_ip_idx)
3695                                 continue;
3696                         err = inet6_fill_ifmcaddr(skb, ifmca,
3697                                                   NETLINK_CB(cb->skb).pid,
3698                                                   cb->nlh->nlmsg_seq,
3699                                                   RTM_GETMULTICAST,
3700                                                   NLM_F_MULTI);
3701                         if (err <= 0)
3702                                 break;
3703                 }
3704                 break;
3705         case ANYCAST_ADDR:
3706                 /* anycast address */
3707                 for (ifaca = idev->ac_list; ifaca;
3708                      ifaca = ifaca->aca_next, ip_idx++) {
3709                         if (ip_idx < s_ip_idx)
3710                                 continue;
3711                         err = inet6_fill_ifacaddr(skb, ifaca,
3712                                                   NETLINK_CB(cb->skb).pid,
3713                                                   cb->nlh->nlmsg_seq,
3714                                                   RTM_GETANYCAST,
3715                                                   NLM_F_MULTI);
3716                         if (err <= 0)
3717                                 break;
3718                 }
3719                 break;
3720         default:
3721                 break;
3722         }
3723         read_unlock_bh(&idev->lock);
3724         *p_ip_idx = ip_idx;
3725         return err;
3726 }
3727
3728 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3729                            enum addr_type_t type)
3730 {
3731         struct net *net = sock_net(skb->sk);
3732         int h, s_h;
3733         int idx, ip_idx;
3734         int s_idx, s_ip_idx;
3735         struct net_device *dev;
3736         struct inet6_dev *idev;
3737         struct hlist_head *head;
3738         struct hlist_node *node;
3739
3740         s_h = cb->args[0];
3741         s_idx = idx = cb->args[1];
3742         s_ip_idx = ip_idx = cb->args[2];
3743
3744         rcu_read_lock();
3745         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3746                 idx = 0;
3747                 head = &net->dev_index_head[h];
3748                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3749                         if (idx < s_idx)
3750                                 goto cont;
3751                         if (h > s_h || idx > s_idx)
3752                                 s_ip_idx = 0;
3753                         ip_idx = 0;
3754                         idev = __in6_dev_get(dev);
3755                         if (!idev)
3756                                 goto cont;
3757
3758                         if (in6_dump_addrs(idev, skb, cb, type,
3759                                            s_ip_idx, &ip_idx) <= 0)
3760                                 goto done;
3761 cont:
3762                         idx++;
3763                 }
3764         }
3765 done:
3766         rcu_read_unlock();
3767         cb->args[0] = h;
3768         cb->args[1] = idx;
3769         cb->args[2] = ip_idx;
3770
3771         return skb->len;
3772 }
3773
3774 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3775 {
3776         enum addr_type_t type = UNICAST_ADDR;
3777
3778         return inet6_dump_addr(skb, cb, type);
3779 }
3780
3781 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3782 {
3783         enum addr_type_t type = MULTICAST_ADDR;
3784
3785         return inet6_dump_addr(skb, cb, type);
3786 }
3787
3788
3789 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3790 {
3791         enum addr_type_t type = ANYCAST_ADDR;
3792
3793         return inet6_dump_addr(skb, cb, type);
3794 }
3795
3796 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3797                              void *arg)
3798 {
3799         struct net *net = sock_net(in_skb->sk);
3800         struct ifaddrmsg *ifm;
3801         struct nlattr *tb[IFA_MAX+1];
3802         struct in6_addr *addr = NULL;
3803         struct net_device *dev = NULL;
3804         struct inet6_ifaddr *ifa;
3805         struct sk_buff *skb;
3806         int err;
3807
3808         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3809         if (err < 0)
3810                 goto errout;
3811
3812         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3813         if (addr == NULL) {
3814                 err = -EINVAL;
3815                 goto errout;
3816         }
3817
3818         ifm = nlmsg_data(nlh);
3819         if (ifm->ifa_index)
3820                 dev = __dev_get_by_index(net, ifm->ifa_index);
3821
3822         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3823         if (!ifa) {
3824                 err = -EADDRNOTAVAIL;
3825                 goto errout;
3826         }
3827
3828         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3829         if (!skb) {
3830                 err = -ENOBUFS;
3831                 goto errout_ifa;
3832         }
3833
3834         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3835                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3836         if (err < 0) {
3837                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3838                 WARN_ON(err == -EMSGSIZE);
3839                 kfree_skb(skb);
3840                 goto errout_ifa;
3841         }
3842         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3843 errout_ifa:
3844         in6_ifa_put(ifa);
3845 errout:
3846         return err;
3847 }
3848
3849 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3850 {
3851         struct sk_buff *skb;
3852         struct net *net = dev_net(ifa->idev->dev);
3853         int err = -ENOBUFS;
3854
3855         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3856         if (skb == NULL)
3857                 goto errout;
3858
3859         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3860         if (err < 0) {
3861                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3862                 WARN_ON(err == -EMSGSIZE);
3863                 kfree_skb(skb);
3864                 goto errout;
3865         }
3866         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3867         return;
3868 errout:
3869         if (err < 0)
3870                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3871 }
3872
3873 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3874                                 __s32 *array, int bytes)
3875 {
3876         BUG_ON(bytes < (DEVCONF_MAX * 4));
3877
3878         memset(array, 0, bytes);
3879         array[DEVCONF_FORWARDING] = cnf->forwarding;
3880         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3881         array[DEVCONF_MTU6] = cnf->mtu6;
3882         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3883         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3884         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3885         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3886         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3887         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3888                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3889         array[DEVCONF_RTR_SOLICIT_DELAY] =
3890                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3891         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3892 #ifdef CONFIG_IPV6_PRIVACY
3893         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3894         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3895         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3896         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3897         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3898 #endif
3899         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3900         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3901         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3902 #ifdef CONFIG_IPV6_ROUTER_PREF
3903         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3904         array[DEVCONF_RTR_PROBE_INTERVAL] =
3905                 jiffies_to_msecs(cnf->rtr_probe_interval);
3906 #ifdef CONFIG_IPV6_ROUTE_INFO
3907         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3908 #endif
3909 #endif
3910         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3911         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3912 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3913         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3914 #endif
3915 #ifdef CONFIG_IPV6_MROUTE
3916         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3917 #endif
3918         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3919         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3920         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3921 }
3922
3923 static inline size_t inet6_ifla6_size(void)
3924 {
3925         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3926              + nla_total_size(sizeof(struct ifla_cacheinfo))
3927              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3928              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3929              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3930 }
3931
3932 static inline size_t inet6_if_nlmsg_size(void)
3933 {
3934         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3935                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3936                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3937                + nla_total_size(4) /* IFLA_MTU */
3938                + nla_total_size(4) /* IFLA_LINK */
3939                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3940 }
3941
3942 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3943                                       int items, int bytes)
3944 {
3945         int i;
3946         int pad = bytes - sizeof(u64) * items;
3947         BUG_ON(pad < 0);
3948
3949         /* Use put_unaligned() because stats may not be aligned for u64. */
3950         put_unaligned(items, &stats[0]);
3951         for (i = 1; i < items; i++)
3952                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3953
3954         memset(&stats[items], 0, pad);
3955 }
3956
3957 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3958                                       int items, int bytes, size_t syncpoff)
3959 {
3960         int i;
3961         int pad = bytes - sizeof(u64) * items;
3962         BUG_ON(pad < 0);
3963
3964         /* Use put_unaligned() because stats may not be aligned for u64. */
3965         put_unaligned(items, &stats[0]);
3966         for (i = 1; i < items; i++)
3967                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3968
3969         memset(&stats[items], 0, pad);
3970 }
3971
3972 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3973                              int bytes)
3974 {
3975         switch (attrtype) {
3976         case IFLA_INET6_STATS:
3977                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3978                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3979                 break;
3980         case IFLA_INET6_ICMP6STATS:
3981                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3982                 break;
3983         }
3984 }
3985
3986 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3987 {
3988         struct nlattr *nla;
3989         struct ifla_cacheinfo ci;
3990
3991         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3992
3993         ci.max_reasm_len = IPV6_MAXPLEN;
3994         ci.tstamp = cstamp_delta(idev->tstamp);
3995         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3996         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3997         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3998
3999         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4000         if (nla == NULL)
4001                 goto nla_put_failure;
4002         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4003
4004         /* XXX - MC not implemented */
4005
4006         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4007         if (nla == NULL)
4008                 goto nla_put_failure;
4009         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4010
4011         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4012         if (nla == NULL)
4013                 goto nla_put_failure;
4014         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4015
4016         return 0;
4017
4018 nla_put_failure:
4019         return -EMSGSIZE;
4020 }
4021
4022 static size_t inet6_get_link_af_size(const struct net_device *dev)
4023 {
4024         if (!__in6_dev_get(dev))
4025                 return 0;
4026
4027         return inet6_ifla6_size();
4028 }
4029
4030 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4031 {
4032         struct inet6_dev *idev = __in6_dev_get(dev);
4033
4034         if (!idev)
4035                 return -ENODATA;
4036
4037         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4038                 return -EMSGSIZE;
4039
4040         return 0;
4041 }
4042
4043 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4044                              u32 pid, u32 seq, int event, unsigned int flags)
4045 {
4046         struct net_device *dev = idev->dev;
4047         struct ifinfomsg *hdr;
4048         struct nlmsghdr *nlh;
4049         void *protoinfo;
4050
4051         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4052         if (nlh == NULL)
4053                 return -EMSGSIZE;
4054
4055         hdr = nlmsg_data(nlh);
4056         hdr->ifi_family = AF_INET6;
4057         hdr->__ifi_pad = 0;
4058         hdr->ifi_type = dev->type;
4059         hdr->ifi_index = dev->ifindex;
4060         hdr->ifi_flags = dev_get_flags(dev);
4061         hdr->ifi_change = 0;
4062
4063         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4064
4065         if (dev->addr_len)
4066                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4067
4068         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4069         if (dev->ifindex != dev->iflink)
4070                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4071
4072         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4073         if (protoinfo == NULL)
4074                 goto nla_put_failure;
4075
4076         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4077                 goto nla_put_failure;
4078
4079         nla_nest_end(skb, protoinfo);
4080         return nlmsg_end(skb, nlh);
4081
4082 nla_put_failure:
4083         nlmsg_cancel(skb, nlh);
4084         return -EMSGSIZE;
4085 }
4086
4087 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4088 {
4089         struct net *net = sock_net(skb->sk);
4090         int h, s_h;
4091         int idx = 0, s_idx;
4092         struct net_device *dev;
4093         struct inet6_dev *idev;
4094         struct hlist_head *head;
4095         struct hlist_node *node;
4096
4097         s_h = cb->args[0];
4098         s_idx = cb->args[1];
4099
4100         rcu_read_lock();
4101         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4102                 idx = 0;
4103                 head = &net->dev_index_head[h];
4104                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4105                         if (idx < s_idx)
4106                                 goto cont;
4107                         idev = __in6_dev_get(dev);
4108                         if (!idev)
4109                                 goto cont;
4110                         if (inet6_fill_ifinfo(skb, idev,
4111                                               NETLINK_CB(cb->skb).pid,
4112                                               cb->nlh->nlmsg_seq,
4113                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4114                                 goto out;
4115 cont:
4116                         idx++;
4117                 }
4118         }
4119 out:
4120         rcu_read_unlock();
4121         cb->args[1] = idx;
4122         cb->args[0] = h;
4123
4124         return skb->len;
4125 }
4126
4127 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4128 {
4129         struct sk_buff *skb;
4130         struct net *net = dev_net(idev->dev);
4131         int err = -ENOBUFS;
4132
4133         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4134         if (skb == NULL)
4135                 goto errout;
4136
4137         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4138         if (err < 0) {
4139                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4140                 WARN_ON(err == -EMSGSIZE);
4141                 kfree_skb(skb);
4142                 goto errout;
4143         }
4144         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4145         return;
4146 errout:
4147         if (err < 0)
4148                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4149 }
4150
4151 static inline size_t inet6_prefix_nlmsg_size(void)
4152 {
4153         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4154                + nla_total_size(sizeof(struct in6_addr))
4155                + nla_total_size(sizeof(struct prefix_cacheinfo));
4156 }
4157
4158 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4159                              struct prefix_info *pinfo, u32 pid, u32 seq,
4160                              int event, unsigned int flags)
4161 {
4162         struct prefixmsg *pmsg;
4163         struct nlmsghdr *nlh;
4164         struct prefix_cacheinfo ci;
4165
4166         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4167         if (nlh == NULL)
4168                 return -EMSGSIZE;
4169
4170         pmsg = nlmsg_data(nlh);
4171         pmsg->prefix_family = AF_INET6;
4172         pmsg->prefix_pad1 = 0;
4173         pmsg->prefix_pad2 = 0;
4174         pmsg->prefix_ifindex = idev->dev->ifindex;
4175         pmsg->prefix_len = pinfo->prefix_len;
4176         pmsg->prefix_type = pinfo->type;
4177         pmsg->prefix_pad3 = 0;
4178         pmsg->prefix_flags = 0;
4179         if (pinfo->onlink)
4180                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4181         if (pinfo->autoconf)
4182                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4183
4184         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4185
4186         ci.preferred_time = ntohl(pinfo->prefered);
4187         ci.valid_time = ntohl(pinfo->valid);
4188         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4189
4190         return nlmsg_end(skb, nlh);
4191
4192 nla_put_failure:
4193         nlmsg_cancel(skb, nlh);
4194         return -EMSGSIZE;
4195 }
4196
4197 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4198                          struct prefix_info *pinfo)
4199 {
4200         struct sk_buff *skb;
4201         struct net *net = dev_net(idev->dev);
4202         int err = -ENOBUFS;
4203
4204         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4205         if (skb == NULL)
4206                 goto errout;
4207
4208         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4209         if (err < 0) {
4210                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4211                 WARN_ON(err == -EMSGSIZE);
4212                 kfree_skb(skb);
4213                 goto errout;
4214         }
4215         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4216         return;
4217 errout:
4218         if (err < 0)
4219                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4220 }
4221
4222 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4223 {
4224         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4225
4226         switch (event) {
4227         case RTM_NEWADDR:
4228                 /*
4229                  * If the address was optimistic
4230                  * we inserted the route at the start of
4231                  * our DAD process, so we don't need
4232                  * to do it again
4233                  */
4234                 if (!(ifp->rt->rt6i_node))
4235                         ip6_ins_rt(ifp->rt);
4236                 if (ifp->idev->cnf.forwarding)
4237                         addrconf_join_anycast(ifp);
4238                 break;
4239         case RTM_DELADDR:
4240                 if (ifp->idev->cnf.forwarding)
4241                         addrconf_leave_anycast(ifp);
4242                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4243                 dst_hold(&ifp->rt->dst);
4244
4245                 if (ip6_del_rt(ifp->rt))
4246                         dst_free(&ifp->rt->dst);
4247                 break;
4248         }
4249 }
4250
4251 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4252 {
4253         rcu_read_lock_bh();
4254         if (likely(ifp->idev->dead == 0))
4255                 __ipv6_ifa_notify(event, ifp);
4256         rcu_read_unlock_bh();
4257 }
4258
4259 #ifdef CONFIG_SYSCTL
4260
4261 static
4262 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4263                            void __user *buffer, size_t *lenp, loff_t *ppos)
4264 {
4265         int *valp = ctl->data;
4266         int val = *valp;
4267         loff_t pos = *ppos;
4268         ctl_table lctl;
4269         int ret;
4270
4271         /*
4272          * ctl->data points to idev->cnf.forwarding, we should
4273          * not modify it until we get the rtnl lock.
4274          */
4275         lctl = *ctl;
4276         lctl.data = &val;
4277
4278         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4279
4280         if (write)
4281                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4282         if (ret)
4283                 *ppos = pos;
4284         return ret;
4285 }
4286
4287 static void dev_disable_change(struct inet6_dev *idev)
4288 {
4289         if (!idev || !idev->dev)
4290                 return;
4291
4292         if (idev->cnf.disable_ipv6)
4293                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4294         else
4295                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4296 }
4297
4298 static void addrconf_disable_change(struct net *net, __s32 newf)
4299 {
4300         struct net_device *dev;
4301         struct inet6_dev *idev;
4302
4303         rcu_read_lock();
4304         for_each_netdev_rcu(net, dev) {
4305                 idev = __in6_dev_get(dev);
4306                 if (idev) {
4307                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4308                         idev->cnf.disable_ipv6 = newf;
4309                         if (changed)
4310                                 dev_disable_change(idev);
4311                 }
4312         }
4313         rcu_read_unlock();
4314 }
4315
4316 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4317 {
4318         struct net *net;
4319         int old;
4320
4321         if (!rtnl_trylock())
4322                 return restart_syscall();
4323
4324         net = (struct net *)table->extra2;
4325         old = *p;
4326         *p = newf;
4327
4328         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4329                 rtnl_unlock();
4330                 return 0;
4331         }
4332
4333         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4334                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4335                 addrconf_disable_change(net, newf);
4336         } else if ((!newf) ^ (!old))
4337                 dev_disable_change((struct inet6_dev *)table->extra1);
4338
4339         rtnl_unlock();
4340         return 0;
4341 }
4342
4343 static
4344 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4345                             void __user *buffer, size_t *lenp, loff_t *ppos)
4346 {
4347         int *valp = ctl->data;
4348         int val = *valp;
4349         loff_t pos = *ppos;
4350         ctl_table lctl;
4351         int ret;
4352
4353         /*
4354          * ctl->data points to idev->cnf.disable_ipv6, we should
4355          * not modify it until we get the rtnl lock.
4356          */
4357         lctl = *ctl;
4358         lctl.data = &val;
4359
4360         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4361
4362         if (write)
4363                 ret = addrconf_disable_ipv6(ctl, valp, val);
4364         if (ret)
4365                 *ppos = pos;
4366         return ret;
4367 }
4368
4369 static struct addrconf_sysctl_table
4370 {
4371         struct ctl_table_header *sysctl_header;
4372         ctl_table addrconf_vars[DEVCONF_MAX+1];
4373         char *dev_name;
4374 } addrconf_sysctl __read_mostly = {
4375         .sysctl_header = NULL,
4376         .addrconf_vars = {
4377                 {
4378                         .procname       = "forwarding",
4379                         .data           = &ipv6_devconf.forwarding,
4380                         .maxlen         = sizeof(int),
4381                         .mode           = 0644,
4382                         .proc_handler   = addrconf_sysctl_forward,
4383                 },
4384                 {
4385                         .procname       = "hop_limit",
4386                         .data           = &ipv6_devconf.hop_limit,
4387                         .maxlen         = sizeof(int),
4388                         .mode           = 0644,
4389                         .proc_handler   = proc_dointvec,
4390                 },
4391                 {
4392                         .procname       = "mtu",
4393                         .data           = &ipv6_devconf.mtu6,
4394                         .maxlen         = sizeof(int),
4395                         .mode           = 0644,
4396                         .proc_handler   = proc_dointvec,
4397                 },
4398                 {
4399                         .procname       = "accept_ra",
4400                         .data           = &ipv6_devconf.accept_ra,
4401                         .maxlen         = sizeof(int),
4402                         .mode           = 0644,
4403                         .proc_handler   = proc_dointvec,
4404                 },
4405                 {
4406                         .procname       = "accept_redirects",
4407                         .data           = &ipv6_devconf.accept_redirects,
4408                         .maxlen         = sizeof(int),
4409                         .mode           = 0644,
4410                         .proc_handler   = proc_dointvec,
4411                 },
4412                 {
4413                         .procname       = "autoconf",
4414                         .data           = &ipv6_devconf.autoconf,
4415                         .maxlen         = sizeof(int),
4416                         .mode           = 0644,
4417                         .proc_handler   = proc_dointvec,
4418                 },
4419                 {
4420                         .procname       = "dad_transmits",
4421                         .data           = &ipv6_devconf.dad_transmits,
4422                         .maxlen         = sizeof(int),
4423                         .mode           = 0644,
4424                         .proc_handler   = proc_dointvec,
4425                 },
4426                 {
4427                         .procname       = "router_solicitations",
4428                         .data           = &ipv6_devconf.rtr_solicits,
4429                         .maxlen         = sizeof(int),
4430                         .mode           = 0644,
4431                         .proc_handler   = proc_dointvec,
4432                 },
4433                 {
4434                         .procname       = "router_solicitation_interval",
4435                         .data           = &ipv6_devconf.rtr_solicit_interval,
4436                         .maxlen         = sizeof(int),
4437                         .mode           = 0644,
4438                         .proc_handler   = proc_dointvec_jiffies,
4439                 },
4440                 {
4441                         .procname       = "router_solicitation_delay",
4442                         .data           = &ipv6_devconf.rtr_solicit_delay,
4443                         .maxlen         = sizeof(int),
4444                         .mode           = 0644,
4445                         .proc_handler   = proc_dointvec_jiffies,
4446                 },
4447                 {
4448                         .procname       = "force_mld_version",
4449                         .data           = &ipv6_devconf.force_mld_version,
4450                         .maxlen         = sizeof(int),
4451                         .mode           = 0644,
4452                         .proc_handler   = proc_dointvec,
4453                 },
4454 #ifdef CONFIG_IPV6_PRIVACY
4455                 {
4456                         .procname       = "use_tempaddr",
4457                         .data           = &ipv6_devconf.use_tempaddr,
4458                         .maxlen         = sizeof(int),
4459                         .mode           = 0644,
4460                         .proc_handler   = proc_dointvec,
4461                 },
4462                 {
4463                         .procname       = "temp_valid_lft",
4464                         .data           = &ipv6_devconf.temp_valid_lft,
4465                         .maxlen         = sizeof(int),
4466                         .mode           = 0644,
4467                         .proc_handler   = proc_dointvec,
4468                 },
4469                 {
4470                         .procname       = "temp_prefered_lft",
4471                         .data           = &ipv6_devconf.temp_prefered_lft,
4472                         .maxlen         = sizeof(int),
4473                         .mode           = 0644,
4474                         .proc_handler   = proc_dointvec,
4475                 },
4476                 {
4477                         .procname       = "regen_max_retry",
4478                         .data           = &ipv6_devconf.regen_max_retry,
4479                         .maxlen         = sizeof(int),
4480                         .mode           = 0644,
4481                         .proc_handler   = proc_dointvec,
4482                 },
4483                 {
4484                         .procname       = "max_desync_factor",
4485                         .data           = &ipv6_devconf.max_desync_factor,
4486                         .maxlen         = sizeof(int),
4487                         .mode           = 0644,
4488                         .proc_handler   = proc_dointvec,
4489                 },
4490 #endif
4491                 {
4492                         .procname       = "max_addresses",
4493                         .data           = &ipv6_devconf.max_addresses,
4494                         .maxlen         = sizeof(int),
4495                         .mode           = 0644,
4496                         .proc_handler   = proc_dointvec,
4497                 },
4498                 {
4499                         .procname       = "accept_ra_defrtr",
4500                         .data           = &ipv6_devconf.accept_ra_defrtr,
4501                         .maxlen         = sizeof(int),
4502                         .mode           = 0644,
4503                         .proc_handler   = proc_dointvec,
4504                 },
4505                 {
4506                         .procname       = "accept_ra_pinfo",
4507                         .data           = &ipv6_devconf.accept_ra_pinfo,
4508                         .maxlen         = sizeof(int),
4509                         .mode           = 0644,
4510                         .proc_handler   = proc_dointvec,
4511                 },
4512 #ifdef CONFIG_IPV6_ROUTER_PREF
4513                 {
4514                         .procname       = "accept_ra_rtr_pref",
4515                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4516                         .maxlen         = sizeof(int),
4517                         .mode           = 0644,
4518                         .proc_handler   = proc_dointvec,
4519                 },
4520                 {
4521                         .procname       = "router_probe_interval",
4522                         .data           = &ipv6_devconf.rtr_probe_interval,
4523                         .maxlen         = sizeof(int),
4524                         .mode           = 0644,
4525                         .proc_handler   = proc_dointvec_jiffies,
4526                 },
4527 #ifdef CONFIG_IPV6_ROUTE_INFO
4528                 {
4529                         .procname       = "accept_ra_rt_info_max_plen",
4530                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4531                         .maxlen         = sizeof(int),
4532                         .mode           = 0644,
4533                         .proc_handler   = proc_dointvec,
4534                 },
4535 #endif
4536 #endif
4537                 {
4538                         .procname       = "proxy_ndp",
4539                         .data           = &ipv6_devconf.proxy_ndp,
4540                         .maxlen         = sizeof(int),
4541                         .mode           = 0644,
4542                         .proc_handler   = proc_dointvec,
4543                 },
4544                 {
4545                         .procname       = "accept_source_route",
4546                         .data           = &ipv6_devconf.accept_source_route,
4547                         .maxlen         = sizeof(int),
4548                         .mode           = 0644,
4549                         .proc_handler   = proc_dointvec,
4550                 },
4551 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4552                 {
4553                         .procname       = "optimistic_dad",
4554                         .data           = &ipv6_devconf.optimistic_dad,
4555                         .maxlen         = sizeof(int),
4556                         .mode           = 0644,
4557                         .proc_handler   = proc_dointvec,
4558
4559                 },
4560 #endif
4561 #ifdef CONFIG_IPV6_MROUTE
4562                 {
4563                         .procname       = "mc_forwarding",
4564                         .data           = &ipv6_devconf.mc_forwarding,
4565                         .maxlen         = sizeof(int),
4566                         .mode           = 0444,
4567                         .proc_handler   = proc_dointvec,
4568                 },
4569 #endif
4570                 {
4571                         .procname       = "disable_ipv6",
4572                         .data           = &ipv6_devconf.disable_ipv6,
4573                         .maxlen         = sizeof(int),
4574                         .mode           = 0644,
4575                         .proc_handler   = addrconf_sysctl_disable,
4576                 },
4577                 {
4578                         .procname       = "accept_dad",
4579                         .data           = &ipv6_devconf.accept_dad,
4580                         .maxlen         = sizeof(int),
4581                         .mode           = 0644,
4582                         .proc_handler   = proc_dointvec,
4583                 },
4584                 {
4585                         .procname       = "force_tllao",
4586                         .data           = &ipv6_devconf.force_tllao,
4587                         .maxlen         = sizeof(int),
4588                         .mode           = 0644,
4589                         .proc_handler   = proc_dointvec
4590                 },
4591                 {
4592                         /* sentinel */
4593                 }
4594         },
4595 };
4596
4597 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4598                 struct inet6_dev *idev, struct ipv6_devconf *p)
4599 {
4600         int i;
4601         struct addrconf_sysctl_table *t;
4602
4603 #define ADDRCONF_CTL_PATH_DEV   3
4604
4605         struct ctl_path addrconf_ctl_path[] = {
4606                 { .procname = "net", },
4607                 { .procname = "ipv6", },
4608                 { .procname = "conf", },
4609                 { /* to be set */ },
4610                 { },
4611         };
4612
4613
4614         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4615         if (t == NULL)
4616                 goto out;
4617
4618         for (i = 0; t->addrconf_vars[i].data; i++) {
4619                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4620                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4621                 t->addrconf_vars[i].extra2 = net;
4622         }
4623
4624         /*
4625          * Make a copy of dev_name, because '.procname' is regarded as const
4626          * by sysctl and we wouldn't want anyone to change it under our feet
4627          * (see SIOCSIFNAME).
4628          */
4629         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4630         if (!t->dev_name)
4631                 goto free;
4632
4633         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4634
4635         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4636                         t->addrconf_vars);
4637         if (t->sysctl_header == NULL)
4638                 goto free_procname;
4639
4640         p->sysctl = t;
4641         return 0;
4642
4643 free_procname:
4644         kfree(t->dev_name);
4645 free:
4646         kfree(t);
4647 out:
4648         return -ENOBUFS;
4649 }
4650
4651 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4652 {
4653         struct addrconf_sysctl_table *t;
4654
4655         if (p->sysctl == NULL)
4656                 return;
4657
4658         t = p->sysctl;
4659         p->sysctl = NULL;
4660         unregister_net_sysctl_table(t->sysctl_header);
4661         kfree(t->dev_name);
4662         kfree(t);
4663 }
4664
4665 static void addrconf_sysctl_register(struct inet6_dev *idev)
4666 {
4667         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4668                               &ndisc_ifinfo_sysctl_change);
4669         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4670                                         idev, &idev->cnf);
4671 }
4672
4673 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4674 {
4675         __addrconf_sysctl_unregister(&idev->cnf);
4676         neigh_sysctl_unregister(idev->nd_parms);
4677 }
4678
4679
4680 #endif
4681
4682 static int __net_init addrconf_init_net(struct net *net)
4683 {
4684         int err;
4685         struct ipv6_devconf *all, *dflt;
4686
4687         err = -ENOMEM;
4688         all = &ipv6_devconf;
4689         dflt = &ipv6_devconf_dflt;
4690
4691         if (!net_eq(net, &init_net)) {
4692                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4693                 if (all == NULL)
4694                         goto err_alloc_all;
4695
4696                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4697                 if (dflt == NULL)
4698                         goto err_alloc_dflt;
4699         } else {
4700                 /* these will be inherited by all namespaces */
4701                 dflt->autoconf = ipv6_defaults.autoconf;
4702                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4703         }
4704
4705         net->ipv6.devconf_all = all;
4706         net->ipv6.devconf_dflt = dflt;
4707
4708 #ifdef CONFIG_SYSCTL
4709         err = __addrconf_sysctl_register(net, "all", NULL, all);
4710         if (err < 0)
4711                 goto err_reg_all;
4712
4713         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4714         if (err < 0)
4715                 goto err_reg_dflt;
4716 #endif
4717         return 0;
4718
4719 #ifdef CONFIG_SYSCTL
4720 err_reg_dflt:
4721         __addrconf_sysctl_unregister(all);
4722 err_reg_all:
4723         kfree(dflt);
4724 #endif
4725 err_alloc_dflt:
4726         kfree(all);
4727 err_alloc_all:
4728         return err;
4729 }
4730
4731 static void __net_exit addrconf_exit_net(struct net *net)
4732 {
4733 #ifdef CONFIG_SYSCTL
4734         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4735         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4736 #endif
4737         if (!net_eq(net, &init_net)) {
4738                 kfree(net->ipv6.devconf_dflt);
4739                 kfree(net->ipv6.devconf_all);
4740         }
4741 }
4742
4743 static struct pernet_operations addrconf_ops = {
4744         .init = addrconf_init_net,
4745         .exit = addrconf_exit_net,
4746 };
4747
4748 /*
4749  *      Device notifier
4750  */
4751
4752 int register_inet6addr_notifier(struct notifier_block *nb)
4753 {
4754         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4755 }
4756 EXPORT_SYMBOL(register_inet6addr_notifier);
4757
4758 int unregister_inet6addr_notifier(struct notifier_block *nb)
4759 {
4760         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4761 }
4762 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4763
4764 static struct rtnl_af_ops inet6_ops = {
4765         .family           = AF_INET6,
4766         .fill_link_af     = inet6_fill_link_af,
4767         .get_link_af_size = inet6_get_link_af_size,
4768 };
4769
4770 /*
4771  *      Init / cleanup code
4772  */
4773
4774 int __init addrconf_init(void)
4775 {
4776         int i, err;
4777
4778         err = ipv6_addr_label_init();
4779         if (err < 0) {
4780                 printk(KERN_CRIT "IPv6 Addrconf:"
4781                        " cannot initialize default policy table: %d.\n", err);
4782                 goto out;
4783         }
4784
4785         err = register_pernet_subsys(&addrconf_ops);
4786         if (err < 0)
4787                 goto out_addrlabel;
4788
4789         /* The addrconf netdev notifier requires that loopback_dev
4790          * has it's ipv6 private information allocated and setup
4791          * before it can bring up and give link-local addresses
4792          * to other devices which are up.
4793          *
4794          * Unfortunately, loopback_dev is not necessarily the first
4795          * entry in the global dev_base list of net devices.  In fact,
4796          * it is likely to be the very last entry on that list.
4797          * So this causes the notifier registry below to try and
4798          * give link-local addresses to all devices besides loopback_dev
4799          * first, then loopback_dev, which cases all the non-loopback_dev
4800          * devices to fail to get a link-local address.
4801          *
4802          * So, as a temporary fix, allocate the ipv6 structure for
4803          * loopback_dev first by hand.
4804          * Longer term, all of the dependencies ipv6 has upon the loopback
4805          * device and it being up should be removed.
4806          */
4807         rtnl_lock();
4808         if (!ipv6_add_dev(init_net.loopback_dev))
4809                 err = -ENOMEM;
4810         rtnl_unlock();
4811         if (err)
4812                 goto errlo;
4813
4814         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4815                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4816
4817         register_netdevice_notifier(&ipv6_dev_notf);
4818
4819         addrconf_verify(0);
4820
4821         err = rtnl_af_register(&inet6_ops);
4822         if (err < 0)
4823                 goto errout_af;
4824
4825         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4826                               NULL);
4827         if (err < 0)
4828                 goto errout;
4829
4830         /* Only the first call to __rtnl_register can fail */
4831         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4832         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4833         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4834                         inet6_dump_ifaddr, NULL);
4835         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4836                         inet6_dump_ifmcaddr, NULL);
4837         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4838                         inet6_dump_ifacaddr, NULL);
4839
4840         ipv6_addr_label_rtnl_register();
4841
4842         return 0;
4843 errout:
4844         rtnl_af_unregister(&inet6_ops);
4845 errout_af:
4846         unregister_netdevice_notifier(&ipv6_dev_notf);
4847 errlo:
4848         unregister_pernet_subsys(&addrconf_ops);
4849 out_addrlabel:
4850         ipv6_addr_label_cleanup();
4851 out:
4852         return err;
4853 }
4854
4855 void addrconf_cleanup(void)
4856 {
4857         struct net_device *dev;
4858         int i;
4859
4860         unregister_netdevice_notifier(&ipv6_dev_notf);
4861         unregister_pernet_subsys(&addrconf_ops);
4862         ipv6_addr_label_cleanup();
4863
4864         rtnl_lock();
4865
4866         __rtnl_af_unregister(&inet6_ops);
4867
4868         /* clean dev list */
4869         for_each_netdev(&init_net, dev) {
4870                 if (__in6_dev_get(dev) == NULL)
4871                         continue;
4872                 addrconf_ifdown(dev, 1);
4873         }
4874         addrconf_ifdown(init_net.loopback_dev, 2);
4875
4876         /*
4877          *      Check hash table.
4878          */
4879         spin_lock_bh(&addrconf_hash_lock);
4880         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4881                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4882         spin_unlock_bh(&addrconf_hash_lock);
4883
4884         del_timer(&addr_chk_timer);
4885         rtnl_unlock();
4886 }