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
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 108 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 109 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 110
 111 #ifdef CONFIG_SYSCTL
 112 static void addrconf_sysctl_register(struct inet6_dev *idev);
 113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 114 #else
 115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 116 {
 117 }
 118
 119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 120 {
 121 }
 122 #endif
 123
 124 #ifdef CONFIG_IPV6_PRIVACY
 125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 127 static void ipv6_regen_rndid(unsigned long data);
 128 #endif
 129
 130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 131 static int ipv6_count_addresses(struct inet6_dev *idev);
 132
 133 /*
 134  *      Configured unicast address hash table
 135  */
 136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 137 static DEFINE_SPINLOCK(addrconf_hash_lock);
 138
 139 static void addrconf_verify(unsigned long);
 140
 141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 142 static DEFINE_SPINLOCK(addrconf_verify_lock);
 143
 144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 146
 147 static void addrconf_type_change(struct net_device *dev,
 148                                  unsigned long event);
 149 static int addrconf_ifdown(struct net_device *dev, int how);
 150
 151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 152 static void addrconf_dad_timer(unsigned long data);
 153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_run(struct inet6_dev *idev);
 155 static void addrconf_rs_timer(unsigned long data);
 156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158
 159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 160                                 struct prefix_info *pinfo);
 161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 162                                struct net_device *dev);
 163
 164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .dad_transmits          = 1,
 175         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 176         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 177         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 178 #ifdef CONFIG_IPV6_PRIVACY
 179         .use_tempaddr           = 0,
 180         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 181         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 182         .regen_max_retry        = REGEN_MAX_RETRY,
 183         .max_desync_factor      = MAX_DESYNC_FACTOR,
 184 #endif
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199 };
 200
 201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 202         .forwarding             = 0,
 203         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 204         .mtu6                   = IPV6_MIN_MTU,
 205         .accept_ra              = 1,
 206         .accept_redirects       = 1,
 207         .autoconf               = 1,
 208         .dad_transmits          = 1,
 209         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 210         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 211         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 212 #ifdef CONFIG_IPV6_PRIVACY
 213         .use_tempaddr           = 0,
 214         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 215         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 216         .regen_max_retry        = REGEN_MAX_RETRY,
 217         .max_desync_factor      = MAX_DESYNC_FACTOR,
 218 #endif
 219         .max_addresses          = IPV6_MAX_ADDRESSES,
 220         .accept_ra_defrtr       = 1,
 221         .accept_ra_pinfo        = 1,
 222 #ifdef CONFIG_IPV6_ROUTER_PREF
 223         .accept_ra_rtr_pref     = 1,
 224         .rtr_probe_interval     = 60 * HZ,
 225 #ifdef CONFIG_IPV6_ROUTE_INFO
 226         .accept_ra_rt_info_max_plen = 0,
 227 #endif
 228 #endif
 229         .proxy_ndp              = 0,
 230         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 231         .disable_ipv6           = 0,
 232         .accept_dad             = 1,
 233 };
 234
 235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 240
 241 /* Check if a valid qdisc is available */
 242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 243 {
 244         return !qdisc_tx_is_noop(dev);
 245 }
 246
 247 /* Check if a route is valid prefix route */
 248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 249 {
 250         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 251 }
 252
 253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 254 {
 255         if (del_timer(&ifp->timer))
 256                 __in6_ifa_put(ifp);
 257 }
 258
 259 enum addrconf_timer_t {
 260         AC_NONE,
 261         AC_DAD,
 262         AC_RS,
 263 };
 264
 265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 266                                enum addrconf_timer_t what,
 267                                unsigned long when)
 268 {
 269         if (!del_timer(&ifp->timer))
 270                 in6_ifa_hold(ifp);
 271
 272         switch (what) {
 273         case AC_DAD:
 274                 ifp->timer.function = addrconf_dad_timer;
 275                 break;
 276         case AC_RS:
 277                 ifp->timer.function = addrconf_rs_timer;
 278                 break;
 279         default:
 280                 break;
 281         }
 282         ifp->timer.expires = jiffies + when;
 283         add_timer(&ifp->timer);
 284 }
 285
 286 static int snmp6_alloc_dev(struct inet6_dev *idev)
 287 {
 288         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 289                           sizeof(struct ipstats_mib),
 290                           __alignof__(struct ipstats_mib)) < 0)
 291                 goto err_ip;
 292         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 293                           sizeof(struct icmpv6_mib),
 294                           __alignof__(struct icmpv6_mib)) < 0)
 295                 goto err_icmp;
 296         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 297                           sizeof(struct icmpv6msg_mib),
 298                           __alignof__(struct icmpv6msg_mib)) < 0)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305 err_icmp:
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static void snmp6_free_dev(struct inet6_dev *idev)
 312 {
 313         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 314         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 315         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 316 }
 317
 318 /* Nobody refers to this device, we may destroy it. */
 319
 320 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 321 {
 322         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 323         kfree(idev);
 324 }
 325
 326 void in6_dev_finish_destroy(struct inet6_dev *idev)
 327 {
 328         struct net_device *dev = idev->dev;
 329
 330         WARN_ON(!list_empty(&idev->addr_list));
 331         WARN_ON(idev->mc_list != NULL);
 332
 333 #ifdef NET_REFCNT_DEBUG
 334         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 335 #endif
 336         dev_put(dev);
 337         if (!idev->dead) {
 338                 pr_warning("Freeing alive inet6 device %p\n", idev);
 339                 return;
 340         }
 341         snmp6_free_dev(idev);
 342         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 343 }
 344
 345 EXPORT_SYMBOL(in6_dev_finish_destroy);
 346
 347 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 348 {
 349         struct inet6_dev *ndev;
 350
 351         ASSERT_RTNL();
 352
 353         if (dev->mtu < IPV6_MIN_MTU)
 354                 return NULL;
 355
 356         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 357
 358         if (ndev == NULL)
 359                 return NULL;
 360
 361         rwlock_init(&ndev->lock);
 362         ndev->dev = dev;
 363         INIT_LIST_HEAD(&ndev->addr_list);
 364
 365         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 366         ndev->cnf.mtu6 = dev->mtu;
 367         ndev->cnf.sysctl = NULL;
 368         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 369         if (ndev->nd_parms == NULL) {
 370                 kfree(ndev);
 371                 return NULL;
 372         }
 373         if (ndev->cnf.forwarding)
 374                 dev_disable_lro(dev);
 375         /* We refer to the device */
 376         dev_hold(dev);
 377
 378         if (snmp6_alloc_dev(ndev) < 0) {
 379                 ADBG((KERN_WARNING
 380                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 381                         __func__, dev->name));
 382                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 383                 ndev->dead = 1;
 384                 in6_dev_finish_destroy(ndev);
 385                 return NULL;
 386         }
 387
 388         if (snmp6_register_dev(ndev) < 0) {
 389                 ADBG((KERN_WARNING
 390                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 391                         __func__, dev->name));
 392                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 393                 ndev->dead = 1;
 394                 in6_dev_finish_destroy(ndev);
 395                 return NULL;
 396         }
 397
 398         /* One reference from device.  We must do this before
 399          * we invoke __ipv6_regen_rndid().
 400          */
 401         in6_dev_hold(ndev);
 402
 403         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 404                 ndev->cnf.accept_dad = -1;
 405
 406 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 407         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 408                 printk(KERN_INFO
 409                        "%s: Disabled Multicast RS\n",
 410                        dev->name);
 411                 ndev->cnf.rtr_solicits = 0;
 412         }
 413 #endif
 414
 415 #ifdef CONFIG_IPV6_PRIVACY
 416         INIT_LIST_HEAD(&ndev->tempaddr_list);
 417         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 418         if ((dev->flags&IFF_LOOPBACK) ||
 419             dev->type == ARPHRD_TUNNEL ||
 420             dev->type == ARPHRD_TUNNEL6 ||
 421             dev->type == ARPHRD_SIT ||
 422             dev->type == ARPHRD_NONE) {
 423                 ndev->cnf.use_tempaddr = -1;
 424         } else {
 425                 in6_dev_hold(ndev);
 426                 ipv6_regen_rndid((unsigned long) ndev);
 427         }
 428 #endif
 429
 430         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 431                 ndev->if_flags |= IF_READY;
 432
 433         ipv6_mc_init_dev(ndev);
 434         ndev->tstamp = jiffies;
 435         addrconf_sysctl_register(ndev);
 436         /* protected by rtnl_lock */
 437         rcu_assign_pointer(dev->ip6_ptr, ndev);
 438
 439         /* Join all-node multicast group */
 440         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 441
 442         return ndev;
 443 }
 444
 445 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 446 {
 447         struct inet6_dev *idev;
 448
 449         ASSERT_RTNL();
 450
 451         idev = __in6_dev_get(dev);
 452         if (!idev) {
 453                 idev = ipv6_add_dev(dev);
 454                 if (!idev)
 455                         return NULL;
 456         }
 457
 458         if (dev->flags&IFF_UP)
 459                 ipv6_mc_up(idev);
 460         return idev;
 461 }
 462
 463 #ifdef CONFIG_SYSCTL
 464 static void dev_forward_change(struct inet6_dev *idev)
 465 {
 466         struct net_device *dev;
 467         struct inet6_ifaddr *ifa;
 468
 469         if (!idev)
 470                 return;
 471         dev = idev->dev;
 472         if (idev->cnf.forwarding)
 473                 dev_disable_lro(dev);
 474         if (dev && (dev->flags & IFF_MULTICAST)) {
 475                 if (idev->cnf.forwarding)
 476                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 477                 else
 478                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 479         }
 480
 481         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 482                 if (ifa->flags&IFA_F_TENTATIVE)
 483                         continue;
 484                 if (idev->cnf.forwarding)
 485                         addrconf_join_anycast(ifa);
 486                 else
 487                         addrconf_leave_anycast(ifa);
 488         }
 489 }
 490
 491
 492 static void addrconf_forward_change(struct net *net, __s32 newf)
 493 {
 494         struct net_device *dev;
 495         struct inet6_dev *idev;
 496
 497         rcu_read_lock();
 498         for_each_netdev_rcu(net, dev) {
 499                 idev = __in6_dev_get(dev);
 500                 if (idev) {
 501                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 502                         idev->cnf.forwarding = newf;
 503                         if (changed)
 504                                 dev_forward_change(idev);
 505                 }
 506         }
 507         rcu_read_unlock();
 508 }
 509
 510 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 511 {
 512         struct net *net;
 513
 514         net = (struct net *)table->extra2;
 515         if (p == &net->ipv6.devconf_dflt->forwarding)
 516                 return 0;
 517
 518         if (!rtnl_trylock()) {
 519                 /* Restore the original values before restarting */
 520                 *p = old;
 521                 return restart_syscall();
 522         }
 523
 524         if (p == &net->ipv6.devconf_all->forwarding) {
 525                 __s32 newf = net->ipv6.devconf_all->forwarding;
 526                 net->ipv6.devconf_dflt->forwarding = newf;
 527                 addrconf_forward_change(net, newf);
 528         } else if ((!*p) ^ (!old))
 529                 dev_forward_change((struct inet6_dev *)table->extra1);
 530         rtnl_unlock();
 531
 532         if (*p)
 533                 rt6_purge_dflt_routers(net);
 534         return 1;
 535 }
 536 #endif
 537
 538 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 539 {
 540         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 541         kfree(ifp);
 542 }
 543
 544 /* Nobody refers to this ifaddr, destroy it */
 545 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 546 {
 547         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 548
 549 #ifdef NET_REFCNT_DEBUG
 550         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 551 #endif
 552
 553         in6_dev_put(ifp->idev);
 554
 555         if (del_timer(&ifp->timer))
 556                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 557
 558         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 559                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 560                 return;
 561         }
 562         dst_release(&ifp->rt->dst);
 563
 564         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 565 }
 566
 567 static void
 568 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 569 {
 570         struct list_head *p;
 571         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 572
 573         /*
 574          * Each device address list is sorted in order of scope -
 575          * global before linklocal.
 576          */
 577         list_for_each(p, &idev->addr_list) {
 578                 struct inet6_ifaddr *ifa
 579                         = list_entry(p, struct inet6_ifaddr, if_list);
 580                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 581                         break;
 582         }
 583
 584         list_add_tail(&ifp->if_list, p);
 585 }
 586
 587 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 588 {
 589         /*
 590          * We perform the hash function over the last 64 bits of the address
 591          * This will include the IEEE address token on links that support it.
 592          */
 593         return jhash_2words((__force u32)addr->s6_addr32[2],
 594                             (__force u32)addr->s6_addr32[3], 0)
 595                 & (IN6_ADDR_HSIZE - 1);
 596 }
 597
 598 /* On success it returns ifp with increased reference count */
 599
 600 static struct inet6_ifaddr *
 601 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 602               int scope, u32 flags)
 603 {
 604         struct inet6_ifaddr *ifa = NULL;
 605         struct rt6_info *rt;
 606         unsigned int hash;
 607         int err = 0;
 608         int addr_type = ipv6_addr_type(addr);
 609
 610         if (addr_type == IPV6_ADDR_ANY ||
 611             addr_type & IPV6_ADDR_MULTICAST ||
 612             (!(idev->dev->flags & IFF_LOOPBACK) &&
 613              addr_type & IPV6_ADDR_LOOPBACK))
 614                 return ERR_PTR(-EADDRNOTAVAIL);
 615
 616         rcu_read_lock_bh();
 617         if (idev->dead) {
 618                 err = -ENODEV;                  /*XXX*/
 619                 goto out2;
 620         }
 621
 622         if (idev->cnf.disable_ipv6) {
 623                 err = -EACCES;
 624                 goto out2;
 625         }
 626
 627         spin_lock(&addrconf_hash_lock);
 628
 629         /* Ignore adding duplicate addresses on an interface */
 630         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 631                 ADBG(("ipv6_add_addr: already assigned\n"));
 632                 err = -EEXIST;
 633                 goto out;
 634         }
 635
 636         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 637
 638         if (ifa == NULL) {
 639                 ADBG(("ipv6_add_addr: malloc failed\n"));
 640                 err = -ENOBUFS;
 641                 goto out;
 642         }
 643
 644         rt = addrconf_dst_alloc(idev, addr, 0);
 645         if (IS_ERR(rt)) {
 646                 err = PTR_ERR(rt);
 647                 goto out;
 648         }
 649
 650         ipv6_addr_copy(&ifa->addr, addr);
 651
 652         spin_lock_init(&ifa->lock);
 653         spin_lock_init(&ifa->state_lock);
 654         init_timer(&ifa->timer);
 655         INIT_HLIST_NODE(&ifa->addr_lst);
 656         ifa->timer.data = (unsigned long) ifa;
 657         ifa->scope = scope;
 658         ifa->prefix_len = pfxlen;
 659         ifa->flags = flags | IFA_F_TENTATIVE;
 660         ifa->cstamp = ifa->tstamp = jiffies;
 661
 662         ifa->rt = rt;
 663
 664         /*
 665          * part one of RFC 4429, section 3.3
 666          * We should not configure an address as
 667          * optimistic if we do not yet know the link
 668          * layer address of our nexhop router
 669          */
 670
 671         if (rt->rt6i_nexthop == NULL)
 672                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 673
 674         ifa->idev = idev;
 675         in6_dev_hold(idev);
 676         /* For caller */
 677         in6_ifa_hold(ifa);
 678
 679         /* Add to big hash table */
 680         hash = ipv6_addr_hash(addr);
 681
 682         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 683         spin_unlock(&addrconf_hash_lock);
 684
 685         write_lock(&idev->lock);
 686         /* Add to inet6_dev unicast addr list. */
 687         ipv6_link_dev_addr(idev, ifa);
 688
 689 #ifdef CONFIG_IPV6_PRIVACY
 690         if (ifa->flags&IFA_F_TEMPORARY) {
 691                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 692                 in6_ifa_hold(ifa);
 693         }
 694 #endif
 695
 696         in6_ifa_hold(ifa);
 697         write_unlock(&idev->lock);
 698 out2:
 699         rcu_read_unlock_bh();
 700
 701         if (likely(err == 0))
 702                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 703         else {
 704                 kfree(ifa);
 705                 ifa = ERR_PTR(err);
 706         }
 707
 708         return ifa;
 709 out:
 710         spin_unlock(&addrconf_hash_lock);
 711         goto out2;
 712 }
 713
 714 /* This function wants to get referenced ifp and releases it before return */
 715
 716 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 717 {
 718         struct inet6_ifaddr *ifa, *ifn;
 719         struct inet6_dev *idev = ifp->idev;
 720         int state;
 721         int hash;
 722         int deleted = 0, onlink = 0;
 723         unsigned long expires = jiffies;
 724
 725         hash = ipv6_addr_hash(&ifp->addr);
 726
 727         spin_lock_bh(&ifp->state_lock);
 728         state = ifp->state;
 729         ifp->state = INET6_IFADDR_STATE_DEAD;
 730         spin_unlock_bh(&ifp->state_lock);
 731
 732         if (state == INET6_IFADDR_STATE_DEAD)
 733                 goto out;
 734
 735         spin_lock_bh(&addrconf_hash_lock);
 736         hlist_del_init_rcu(&ifp->addr_lst);
 737         spin_unlock_bh(&addrconf_hash_lock);
 738
 739         write_lock_bh(&idev->lock);
 740 #ifdef CONFIG_IPV6_PRIVACY
 741         if (ifp->flags&IFA_F_TEMPORARY) {
 742                 list_del(&ifp->tmp_list);
 743                 if (ifp->ifpub) {
 744                         in6_ifa_put(ifp->ifpub);
 745                         ifp->ifpub = NULL;
 746                 }
 747                 __in6_ifa_put(ifp);
 748         }
 749 #endif
 750
 751         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 752                 if (ifa == ifp) {
 753                         list_del_init(&ifp->if_list);
 754                         __in6_ifa_put(ifp);
 755
 756                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 757                                 break;
 758                         deleted = 1;
 759                         continue;
 760                 } else if (ifp->flags & IFA_F_PERMANENT) {
 761                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 762                                               ifp->prefix_len)) {
 763                                 if (ifa->flags & IFA_F_PERMANENT) {
 764                                         onlink = 1;
 765                                         if (deleted)
 766                                                 break;
 767                                 } else {
 768                                         unsigned long lifetime;
 769
 770                                         if (!onlink)
 771                                                 onlink = -1;
 772
 773                                         spin_lock(&ifa->lock);
 774
 775                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 776                                         /*
 777                                          * Note: Because this address is
 778                                          * not permanent, lifetime <
 779                                          * LONG_MAX / HZ here.
 780                                          */
 781                                         if (time_before(expires,
 782                                                         ifa->tstamp + lifetime * HZ))
 783                                                 expires = ifa->tstamp + lifetime * HZ;
 784                                         spin_unlock(&ifa->lock);
 785                                 }
 786                         }
 787                 }
 788         }
 789         write_unlock_bh(&idev->lock);
 790
 791         addrconf_del_timer(ifp);
 792
 793         ipv6_ifa_notify(RTM_DELADDR, ifp);
 794
 795         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 796
 797         /*
 798          * Purge or update corresponding prefix
 799          *
 800          * 1) we don't purge prefix here if address was not permanent.
 801          *    prefix is managed by its own lifetime.
 802          * 2) if there're no addresses, delete prefix.
 803          * 3) if there're still other permanent address(es),
 804          *    corresponding prefix is still permanent.
 805          * 4) otherwise, update prefix lifetime to the
 806          *    longest valid lifetime among the corresponding
 807          *    addresses on the device.
 808          *    Note: subsequent RA will update lifetime.
 809          *
 810          * --yoshfuji
 811          */
 812         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 813                 struct in6_addr prefix;
 814                 struct rt6_info *rt;
 815                 struct net *net = dev_net(ifp->idev->dev);
 816                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 817                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 818
 819                 if (rt && addrconf_is_prefix_route(rt)) {
 820                         if (onlink == 0) {
 821                                 ip6_del_rt(rt);
 822                                 rt = NULL;
 823                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 824                                 rt->rt6i_expires = expires;
 825                                 rt->rt6i_flags |= RTF_EXPIRES;
 826                         }
 827                 }
 828                 dst_release(&rt->dst);
 829         }
 830
 831 out:
 832         in6_ifa_put(ifp);
 833 }
 834
 835 #ifdef CONFIG_IPV6_PRIVACY
 836 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 837 {
 838         struct inet6_dev *idev = ifp->idev;
 839         struct in6_addr addr, *tmpaddr;
 840         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 841         unsigned long regen_advance;
 842         int tmp_plen;
 843         int ret = 0;
 844         int max_addresses;
 845         u32 addr_flags;
 846
 847         write_lock(&idev->lock);
 848         if (ift) {
 849                 spin_lock_bh(&ift->lock);
 850                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 851                 spin_unlock_bh(&ift->lock);
 852                 tmpaddr = &addr;
 853         } else {
 854                 tmpaddr = NULL;
 855         }
 856 retry:
 857         in6_dev_hold(idev);
 858         if (idev->cnf.use_tempaddr <= 0) {
 859                 write_unlock(&idev->lock);
 860                 printk(KERN_INFO
 861                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 862                 in6_dev_put(idev);
 863                 ret = -1;
 864                 goto out;
 865         }
 866         spin_lock_bh(&ifp->lock);
 867         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 868                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 869                 spin_unlock_bh(&ifp->lock);
 870                 write_unlock(&idev->lock);
 871                 printk(KERN_WARNING
 872                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 873                 in6_dev_put(idev);
 874                 ret = -1;
 875                 goto out;
 876         }
 877         in6_ifa_hold(ifp);
 878         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 879         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 880                 spin_unlock_bh(&ifp->lock);
 881                 write_unlock(&idev->lock);
 882                 printk(KERN_WARNING
 883                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 884                 in6_ifa_put(ifp);
 885                 in6_dev_put(idev);
 886                 ret = -1;
 887                 goto out;
 888         }
 889         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 890         age = (jiffies - ifp->tstamp) / HZ;
 891         tmp_valid_lft = min_t(__u32,
 892                               ifp->valid_lft,
 893                               idev->cnf.temp_valid_lft + age);
 894         tmp_prefered_lft = min_t(__u32,
 895                                  ifp->prefered_lft,
 896                                  idev->cnf.temp_prefered_lft + age -
 897                                  idev->cnf.max_desync_factor);
 898         tmp_plen = ifp->prefix_len;
 899         max_addresses = idev->cnf.max_addresses;
 900         tmp_cstamp = ifp->cstamp;
 901         tmp_tstamp = ifp->tstamp;
 902         spin_unlock_bh(&ifp->lock);
 903
 904         regen_advance = idev->cnf.regen_max_retry *
 905                         idev->cnf.dad_transmits *
 906                         idev->nd_parms->retrans_time / HZ;
 907         write_unlock(&idev->lock);
 908
 909         /* A temporary address is created only if this calculated Preferred
 910          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 911          * an implementation must not create a temporary address with a zero
 912          * Preferred Lifetime.
 913          */
 914         if (tmp_prefered_lft <= regen_advance) {
 915                 in6_ifa_put(ifp);
 916                 in6_dev_put(idev);
 917                 ret = -1;
 918                 goto out;
 919         }
 920
 921         addr_flags = IFA_F_TEMPORARY;
 922         /* set in addrconf_prefix_rcv() */
 923         if (ifp->flags & IFA_F_OPTIMISTIC)
 924                 addr_flags |= IFA_F_OPTIMISTIC;
 925
 926         ift = !max_addresses ||
 927               ipv6_count_addresses(idev) < max_addresses ?
 928                 ipv6_add_addr(idev, &addr, tmp_plen,
 929                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 930                               addr_flags) : NULL;
 931         if (!ift || IS_ERR(ift)) {
 932                 in6_ifa_put(ifp);
 933                 in6_dev_put(idev);
 934                 printk(KERN_INFO
 935                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 936                 tmpaddr = &addr;
 937                 write_lock(&idev->lock);
 938                 goto retry;
 939         }
 940
 941         spin_lock_bh(&ift->lock);
 942         ift->ifpub = ifp;
 943         ift->valid_lft = tmp_valid_lft;
 944         ift->prefered_lft = tmp_prefered_lft;
 945         ift->cstamp = tmp_cstamp;
 946         ift->tstamp = tmp_tstamp;
 947         spin_unlock_bh(&ift->lock);
 948
 949         addrconf_dad_start(ift, 0);
 950         in6_ifa_put(ift);
 951         in6_dev_put(idev);
 952 out:
 953         return ret;
 954 }
 955 #endif
 956
 957 /*
 958  *      Choose an appropriate source address (RFC3484)
 959  */
 960 enum {
 961         IPV6_SADDR_RULE_INIT = 0,
 962         IPV6_SADDR_RULE_LOCAL,
 963         IPV6_SADDR_RULE_SCOPE,
 964         IPV6_SADDR_RULE_PREFERRED,
 965 #ifdef CONFIG_IPV6_MIP6
 966         IPV6_SADDR_RULE_HOA,
 967 #endif
 968         IPV6_SADDR_RULE_OIF,
 969         IPV6_SADDR_RULE_LABEL,
 970 #ifdef CONFIG_IPV6_PRIVACY
 971         IPV6_SADDR_RULE_PRIVACY,
 972 #endif
 973         IPV6_SADDR_RULE_ORCHID,
 974         IPV6_SADDR_RULE_PREFIX,
 975         IPV6_SADDR_RULE_MAX
 976 };
 977
 978 struct ipv6_saddr_score {
 979         int                     rule;
 980         int                     addr_type;
 981         struct inet6_ifaddr     *ifa;
 982         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 983         int                     scopedist;
 984         int                     matchlen;
 985 };
 986
 987 struct ipv6_saddr_dst {
 988         const struct in6_addr *addr;
 989         int ifindex;
 990         int scope;
 991         int label;
 992         unsigned int prefs;
 993 };
 994
 995 static inline int ipv6_saddr_preferred(int type)
 996 {
 997         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 998                 return 1;
 999         return 0;
1000 }
1001
1002 static int ipv6_get_saddr_eval(struct net *net,
1003                                struct ipv6_saddr_score *score,
1004                                struct ipv6_saddr_dst *dst,
1005                                int i)
1006 {
1007         int ret;
1008
1009         if (i <= score->rule) {
1010                 switch (i) {
1011                 case IPV6_SADDR_RULE_SCOPE:
1012                         ret = score->scopedist;
1013                         break;
1014                 case IPV6_SADDR_RULE_PREFIX:
1015                         ret = score->matchlen;
1016                         break;
1017                 default:
1018                         ret = !!test_bit(i, score->scorebits);
1019                 }
1020                 goto out;
1021         }
1022
1023         switch (i) {
1024         case IPV6_SADDR_RULE_INIT:
1025                 /* Rule 0: remember if hiscore is not ready yet */
1026                 ret = !!score->ifa;
1027                 break;
1028         case IPV6_SADDR_RULE_LOCAL:
1029                 /* Rule 1: Prefer same address */
1030                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1031                 break;
1032         case IPV6_SADDR_RULE_SCOPE:
1033                 /* Rule 2: Prefer appropriate scope
1034                  *
1035                  *      ret
1036                  *       ^
1037                  *    -1 |  d 15
1038                  *    ---+--+-+---> scope
1039                  *       |
1040                  *       |             d is scope of the destination.
1041                  *  B-d  |  \
1042                  *       |   \      <- smaller scope is better if
1043                  *  B-15 |    \        if scope is enough for destinaion.
1044                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1045                  * d-C-1 | /
1046                  *       |/         <- greater is better
1047                  *   -C  /             if scope is not enough for destination.
1048                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1049                  *
1050                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1051                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1052                  * Assume B = 0 and we get C > 29.
1053                  */
1054                 ret = __ipv6_addr_src_scope(score->addr_type);
1055                 if (ret >= dst->scope)
1056                         ret = -ret;
1057                 else
1058                         ret -= 128;     /* 30 is enough */
1059                 score->scopedist = ret;
1060                 break;
1061         case IPV6_SADDR_RULE_PREFERRED:
1062                 /* Rule 3: Avoid deprecated and optimistic addresses */
1063                 ret = ipv6_saddr_preferred(score->addr_type) ||
1064                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1065                 break;
1066 #ifdef CONFIG_IPV6_MIP6
1067         case IPV6_SADDR_RULE_HOA:
1068             {
1069                 /* Rule 4: Prefer home address */
1070                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1071                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1072                 break;
1073             }
1074 #endif
1075         case IPV6_SADDR_RULE_OIF:
1076                 /* Rule 5: Prefer outgoing interface */
1077                 ret = (!dst->ifindex ||
1078                        dst->ifindex == score->ifa->idev->dev->ifindex);
1079                 break;
1080         case IPV6_SADDR_RULE_LABEL:
1081                 /* Rule 6: Prefer matching label */
1082                 ret = ipv6_addr_label(net,
1083                                       &score->ifa->addr, score->addr_type,
1084                                       score->ifa->idev->dev->ifindex) == dst->label;
1085                 break;
1086 #ifdef CONFIG_IPV6_PRIVACY
1087         case IPV6_SADDR_RULE_PRIVACY:
1088             {
1089                 /* Rule 7: Prefer public address
1090                  * Note: prefer temprary address if use_tempaddr >= 2
1091                  */
1092                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1093                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1094                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1095                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1096                 break;
1097             }
1098 #endif
1099         case IPV6_SADDR_RULE_ORCHID:
1100                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1101                  *          non-ORCHID vs non-ORCHID
1102                  */
1103                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1104                         ipv6_addr_orchid(dst->addr));
1105                 break;
1106         case IPV6_SADDR_RULE_PREFIX:
1107                 /* Rule 8: Use longest matching prefix */
1108                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1109                                                        dst->addr);
1110                 break;
1111         default:
1112                 ret = 0;
1113         }
1114
1115         if (ret)
1116                 __set_bit(i, score->scorebits);
1117         score->rule = i;
1118 out:
1119         return ret;
1120 }
1121
1122 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1123                        const struct in6_addr *daddr, unsigned int prefs,
1124                        struct in6_addr *saddr)
1125 {
1126         struct ipv6_saddr_score scores[2],
1127                                 *score = &scores[0], *hiscore = &scores[1];
1128         struct ipv6_saddr_dst dst;
1129         struct net_device *dev;
1130         int dst_type;
1131
1132         dst_type = __ipv6_addr_type(daddr);
1133         dst.addr = daddr;
1134         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1135         dst.scope = __ipv6_addr_src_scope(dst_type);
1136         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1137         dst.prefs = prefs;
1138
1139         hiscore->rule = -1;
1140         hiscore->ifa = NULL;
1141
1142         rcu_read_lock();
1143
1144         for_each_netdev_rcu(net, dev) {
1145                 struct inet6_dev *idev;
1146
1147                 /* Candidate Source Address (section 4)
1148                  *  - multicast and link-local destination address,
1149                  *    the set of candidate source address MUST only
1150                  *    include addresses assigned to interfaces
1151                  *    belonging to the same link as the outgoing
1152                  *    interface.
1153                  * (- For site-local destination addresses, the
1154                  *    set of candidate source addresses MUST only
1155                  *    include addresses assigned to interfaces
1156                  *    belonging to the same site as the outgoing
1157                  *    interface.)
1158                  */
1159                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1160                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1161                     dst.ifindex && dev->ifindex != dst.ifindex)
1162                         continue;
1163
1164                 idev = __in6_dev_get(dev);
1165                 if (!idev)
1166                         continue;
1167
1168                 read_lock_bh(&idev->lock);
1169                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1170                         int i;
1171
1172                         /*
1173                          * - Tentative Address (RFC2462 section 5.4)
1174                          *  - A tentative address is not considered
1175                          *    "assigned to an interface" in the traditional
1176                          *    sense, unless it is also flagged as optimistic.
1177                          * - Candidate Source Address (section 4)
1178                          *  - In any case, anycast addresses, multicast
1179                          *    addresses, and the unspecified address MUST
1180                          *    NOT be included in a candidate set.
1181                          */
1182                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1183                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1184                                 continue;
1185
1186                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1187
1188                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1189                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1190                                 LIMIT_NETDEBUG(KERN_DEBUG
1191                                                "ADDRCONF: unspecified / multicast address "
1192                                                "assigned as unicast address on %s",
1193                                                dev->name);
1194                                 continue;
1195                         }
1196
1197                         score->rule = -1;
1198                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1199
1200                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1201                                 int minihiscore, miniscore;
1202
1203                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1204                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1205
1206                                 if (minihiscore > miniscore) {
1207                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1208                                             score->scopedist > 0) {
1209                                                 /*
1210                                                  * special case:
1211                                                  * each remaining entry
1212                                                  * has too small (not enough)
1213                                                  * scope, because ifa entries
1214                                                  * are sorted by their scope
1215                                                  * values.
1216                                                  */
1217                                                 goto try_nextdev;
1218                                         }
1219                                         break;
1220                                 } else if (minihiscore < miniscore) {
1221                                         if (hiscore->ifa)
1222                                                 in6_ifa_put(hiscore->ifa);
1223
1224                                         in6_ifa_hold(score->ifa);
1225
1226                                         swap(hiscore, score);
1227
1228                                         /* restore our iterator */
1229                                         score->ifa = hiscore->ifa;
1230
1231                                         break;
1232                                 }
1233                         }
1234                 }
1235 try_nextdev:
1236                 read_unlock_bh(&idev->lock);
1237         }
1238         rcu_read_unlock();
1239
1240         if (!hiscore->ifa)
1241                 return -EADDRNOTAVAIL;
1242
1243         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1244         in6_ifa_put(hiscore->ifa);
1245         return 0;
1246 }
1247 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1248
1249 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1250                     unsigned char banned_flags)
1251 {
1252         struct inet6_dev *idev;
1253         int err = -EADDRNOTAVAIL;
1254
1255         rcu_read_lock();
1256         idev = __in6_dev_get(dev);
1257         if (idev) {
1258                 struct inet6_ifaddr *ifp;
1259
1260                 read_lock_bh(&idev->lock);
1261                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1262                         if (ifp->scope == IFA_LINK &&
1263                             !(ifp->flags & banned_flags)) {
1264                                 ipv6_addr_copy(addr, &ifp->addr);
1265                                 err = 0;
1266                                 break;
1267                         }
1268                 }
1269                 read_unlock_bh(&idev->lock);
1270         }
1271         rcu_read_unlock();
1272         return err;
1273 }
1274
1275 static int ipv6_count_addresses(struct inet6_dev *idev)
1276 {
1277         int cnt = 0;
1278         struct inet6_ifaddr *ifp;
1279
1280         read_lock_bh(&idev->lock);
1281         list_for_each_entry(ifp, &idev->addr_list, if_list)
1282                 cnt++;
1283         read_unlock_bh(&idev->lock);
1284         return cnt;
1285 }
1286
1287 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1288                   struct net_device *dev, int strict)
1289 {
1290         struct inet6_ifaddr *ifp;
1291         struct hlist_node *node;
1292         unsigned int hash = ipv6_addr_hash(addr);
1293
1294         rcu_read_lock_bh();
1295         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1296                 if (!net_eq(dev_net(ifp->idev->dev), net))
1297                         continue;
1298                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1299                     !(ifp->flags&IFA_F_TENTATIVE) &&
1300                     (dev == NULL || ifp->idev->dev == dev ||
1301                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1302                         rcu_read_unlock_bh();
1303                         return 1;
1304                 }
1305         }
1306
1307         rcu_read_unlock_bh();
1308         return 0;
1309 }
1310 EXPORT_SYMBOL(ipv6_chk_addr);
1311
1312 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1313                                struct net_device *dev)
1314 {
1315         unsigned int hash = ipv6_addr_hash(addr);
1316         struct inet6_ifaddr *ifp;
1317         struct hlist_node *node;
1318
1319         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1320                 if (!net_eq(dev_net(ifp->idev->dev), net))
1321                         continue;
1322                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1323                         if (dev == NULL || ifp->idev->dev == dev)
1324                                 return true;
1325                 }
1326         }
1327         return false;
1328 }
1329
1330 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1331 {
1332         struct inet6_dev *idev;
1333         struct inet6_ifaddr *ifa;
1334         int     onlink;
1335
1336         onlink = 0;
1337         rcu_read_lock();
1338         idev = __in6_dev_get(dev);
1339         if (idev) {
1340                 read_lock_bh(&idev->lock);
1341                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1342                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1343                                                    ifa->prefix_len);
1344                         if (onlink)
1345                                 break;
1346                 }
1347                 read_unlock_bh(&idev->lock);
1348         }
1349         rcu_read_unlock();
1350         return onlink;
1351 }
1352
1353 EXPORT_SYMBOL(ipv6_chk_prefix);
1354
1355 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1356                                      struct net_device *dev, int strict)
1357 {
1358         struct inet6_ifaddr *ifp, *result = NULL;
1359         unsigned int hash = ipv6_addr_hash(addr);
1360         struct hlist_node *node;
1361
1362         rcu_read_lock_bh();
1363         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1364                 if (!net_eq(dev_net(ifp->idev->dev), net))
1365                         continue;
1366                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1367                         if (dev == NULL || ifp->idev->dev == dev ||
1368                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1369                                 result = ifp;
1370                                 in6_ifa_hold(ifp);
1371                                 break;
1372                         }
1373                 }
1374         }
1375         rcu_read_unlock_bh();
1376
1377         return result;
1378 }
1379
1380 /* Gets referenced address, destroys ifaddr */
1381
1382 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1383 {
1384         if (ifp->flags&IFA_F_PERMANENT) {
1385                 spin_lock_bh(&ifp->lock);
1386                 addrconf_del_timer(ifp);
1387                 ifp->flags |= IFA_F_TENTATIVE;
1388                 if (dad_failed)
1389                         ifp->flags |= IFA_F_DADFAILED;
1390                 spin_unlock_bh(&ifp->lock);
1391                 if (dad_failed)
1392                         ipv6_ifa_notify(0, ifp);
1393                 in6_ifa_put(ifp);
1394 #ifdef CONFIG_IPV6_PRIVACY
1395         } else if (ifp->flags&IFA_F_TEMPORARY) {
1396                 struct inet6_ifaddr *ifpub;
1397                 spin_lock_bh(&ifp->lock);
1398                 ifpub = ifp->ifpub;
1399                 if (ifpub) {
1400                         in6_ifa_hold(ifpub);
1401                         spin_unlock_bh(&ifp->lock);
1402                         ipv6_create_tempaddr(ifpub, ifp);
1403                         in6_ifa_put(ifpub);
1404                 } else {
1405                         spin_unlock_bh(&ifp->lock);
1406                 }
1407                 ipv6_del_addr(ifp);
1408 #endif
1409         } else
1410                 ipv6_del_addr(ifp);
1411 }
1412
1413 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1414 {
1415         int err = -ENOENT;
1416
1417         spin_lock(&ifp->state_lock);
1418         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1419                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1420                 err = 0;
1421         }
1422         spin_unlock(&ifp->state_lock);
1423
1424         return err;
1425 }
1426
1427 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1428 {
1429         struct inet6_dev *idev = ifp->idev;
1430
1431         if (addrconf_dad_end(ifp)) {
1432                 in6_ifa_put(ifp);
1433                 return;
1434         }
1435
1436         if (net_ratelimit())
1437                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1438                         ifp->idev->dev->name, &ifp->addr);
1439
1440         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1441                 struct in6_addr addr;
1442
1443                 addr.s6_addr32[0] = htonl(0xfe800000);
1444                 addr.s6_addr32[1] = 0;
1445
1446                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1447                     ipv6_addr_equal(&ifp->addr, &addr)) {
1448                         /* DAD failed for link-local based on MAC address */
1449                         idev->cnf.disable_ipv6 = 1;
1450
1451                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1452                                 ifp->idev->dev->name);
1453                 }
1454         }
1455
1456         addrconf_dad_stop(ifp, 1);
1457 }
1458
1459 /* Join to solicited addr multicast group. */
1460
1461 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1462 {
1463         struct in6_addr maddr;
1464
1465         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1466                 return;
1467
1468         addrconf_addr_solict_mult(addr, &maddr);
1469         ipv6_dev_mc_inc(dev, &maddr);
1470 }
1471
1472 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1473 {
1474         struct in6_addr maddr;
1475
1476         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1477                 return;
1478
1479         addrconf_addr_solict_mult(addr, &maddr);
1480         __ipv6_dev_mc_dec(idev, &maddr);
1481 }
1482
1483 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1484 {
1485         struct in6_addr addr;
1486         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1487         if (ipv6_addr_any(&addr))
1488                 return;
1489         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1490 }
1491
1492 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1493 {
1494         struct in6_addr addr;
1495         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1496         if (ipv6_addr_any(&addr))
1497                 return;
1498         __ipv6_dev_ac_dec(ifp->idev, &addr);
1499 }
1500
1501 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1502 {
1503         if (dev->addr_len != ETH_ALEN)
1504                 return -1;
1505         memcpy(eui, dev->dev_addr, 3);
1506         memcpy(eui + 5, dev->dev_addr + 3, 3);
1507
1508         /*
1509          * The zSeries OSA network cards can be shared among various
1510          * OS instances, but the OSA cards have only one MAC address.
1511          * This leads to duplicate address conflicts in conjunction
1512          * with IPv6 if more than one instance uses the same card.
1513          *
1514          * The driver for these cards can deliver a unique 16-bit
1515          * identifier for each instance sharing the same card.  It is
1516          * placed instead of 0xFFFE in the interface identifier.  The
1517          * "u" bit of the interface identifier is not inverted in this
1518          * case.  Hence the resulting interface identifier has local
1519          * scope according to RFC2373.
1520          */
1521         if (dev->dev_id) {
1522                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1523                 eui[4] = dev->dev_id & 0xFF;
1524         } else {
1525                 eui[3] = 0xFF;
1526                 eui[4] = 0xFE;
1527                 eui[0] ^= 2;
1528         }
1529         return 0;
1530 }
1531
1532 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1533 {
1534         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1535         if (dev->addr_len != ARCNET_ALEN)
1536                 return -1;
1537         memset(eui, 0, 7);
1538         eui[7] = *(u8*)dev->dev_addr;
1539         return 0;
1540 }
1541
1542 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1543 {
1544         if (dev->addr_len != INFINIBAND_ALEN)
1545                 return -1;
1546         memcpy(eui, dev->dev_addr + 12, 8);
1547         eui[0] |= 2;
1548         return 0;
1549 }
1550
1551 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1552 {
1553         if (addr == 0)
1554                 return -1;
1555         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1556                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1557                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1558                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1559                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1560                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1561         eui[1] = 0;
1562         eui[2] = 0x5E;
1563         eui[3] = 0xFE;
1564         memcpy(eui + 4, &addr, 4);
1565         return 0;
1566 }
1567
1568 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1569 {
1570         if (dev->priv_flags & IFF_ISATAP)
1571                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1572         return -1;
1573 }
1574
1575 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1576 {
1577         switch (dev->type) {
1578         case ARPHRD_ETHER:
1579         case ARPHRD_FDDI:
1580         case ARPHRD_IEEE802_TR:
1581                 return addrconf_ifid_eui48(eui, dev);
1582         case ARPHRD_ARCNET:
1583                 return addrconf_ifid_arcnet(eui, dev);
1584         case ARPHRD_INFINIBAND:
1585                 return addrconf_ifid_infiniband(eui, dev);
1586         case ARPHRD_SIT:
1587                 return addrconf_ifid_sit(eui, dev);
1588         }
1589         return -1;
1590 }
1591
1592 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1593 {
1594         int err = -1;
1595         struct inet6_ifaddr *ifp;
1596
1597         read_lock_bh(&idev->lock);
1598         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1599                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1600                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1601                         err = 0;
1602                         break;
1603                 }
1604         }
1605         read_unlock_bh(&idev->lock);
1606         return err;
1607 }
1608
1609 #ifdef CONFIG_IPV6_PRIVACY
1610 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1611 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1612 {
1613 regen:
1614         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1615         idev->rndid[0] &= ~0x02;
1616
1617         /*
1618          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1619          * check if generated address is not inappropriate
1620          *
1621          *  - Reserved subnet anycast (RFC 2526)
1622          *      11111101 11....11 1xxxxxxx
1623          *  - ISATAP (RFC4214) 6.1
1624          *      00-00-5E-FE-xx-xx-xx-xx
1625          *  - value 0
1626          *  - XXX: already assigned to an address on the device
1627          */
1628         if (idev->rndid[0] == 0xfd &&
1629             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1630             (idev->rndid[7]&0x80))
1631                 goto regen;
1632         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1633                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1634                         goto regen;
1635                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1636                         goto regen;
1637         }
1638
1639         return 0;
1640 }
1641
1642 static void ipv6_regen_rndid(unsigned long data)
1643 {
1644         struct inet6_dev *idev = (struct inet6_dev *) data;
1645         unsigned long expires;
1646
1647         rcu_read_lock_bh();
1648         write_lock_bh(&idev->lock);
1649
1650         if (idev->dead)
1651                 goto out;
1652
1653         if (__ipv6_regen_rndid(idev) < 0)
1654                 goto out;
1655
1656         expires = jiffies +
1657                 idev->cnf.temp_prefered_lft * HZ -
1658                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1659                 idev->cnf.max_desync_factor * HZ;
1660         if (time_before(expires, jiffies)) {
1661                 printk(KERN_WARNING
1662                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1663                         idev->dev->name);
1664                 goto out;
1665         }
1666
1667         if (!mod_timer(&idev->regen_timer, expires))
1668                 in6_dev_hold(idev);
1669
1670 out:
1671         write_unlock_bh(&idev->lock);
1672         rcu_read_unlock_bh();
1673         in6_dev_put(idev);
1674 }
1675
1676 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1677         int ret = 0;
1678
1679         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1680                 ret = __ipv6_regen_rndid(idev);
1681         return ret;
1682 }
1683 #endif
1684
1685 /*
1686  *      Add prefix route.
1687  */
1688
1689 static void
1690 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1691                       unsigned long expires, u32 flags)
1692 {
1693         struct fib6_config cfg = {
1694                 .fc_table = RT6_TABLE_PREFIX,
1695                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1696                 .fc_ifindex = dev->ifindex,
1697                 .fc_expires = expires,
1698                 .fc_dst_len = plen,
1699                 .fc_flags = RTF_UP | flags,
1700                 .fc_nlinfo.nl_net = dev_net(dev),
1701                 .fc_protocol = RTPROT_KERNEL,
1702         };
1703
1704         ipv6_addr_copy(&cfg.fc_dst, pfx);
1705
1706         /* Prevent useless cloning on PtP SIT.
1707            This thing is done here expecting that the whole
1708            class of non-broadcast devices need not cloning.
1709          */
1710 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1711         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1712                 cfg.fc_flags |= RTF_NONEXTHOP;
1713 #endif
1714
1715         ip6_route_add(&cfg);
1716 }
1717
1718 /* Create "default" multicast route to the interface */
1719
1720 static void addrconf_add_mroute(struct net_device *dev)
1721 {
1722         struct fib6_config cfg = {
1723                 .fc_table = RT6_TABLE_LOCAL,
1724                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1725                 .fc_ifindex = dev->ifindex,
1726                 .fc_dst_len = 8,
1727                 .fc_flags = RTF_UP,
1728                 .fc_nlinfo.nl_net = dev_net(dev),
1729         };
1730
1731         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1732
1733         ip6_route_add(&cfg);
1734 }
1735
1736 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1737 static void sit_route_add(struct net_device *dev)
1738 {
1739         struct fib6_config cfg = {
1740                 .fc_table = RT6_TABLE_MAIN,
1741                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1742                 .fc_ifindex = dev->ifindex,
1743                 .fc_dst_len = 96,
1744                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1745                 .fc_nlinfo.nl_net = dev_net(dev),
1746         };
1747
1748         /* prefix length - 96 bits "::d.d.d.d" */
1749         ip6_route_add(&cfg);
1750 }
1751 #endif
1752
1753 static void addrconf_add_lroute(struct net_device *dev)
1754 {
1755         struct in6_addr addr;
1756
1757         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1758         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1759 }
1760
1761 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1762 {
1763         struct inet6_dev *idev;
1764
1765         ASSERT_RTNL();
1766
1767         idev = ipv6_find_idev(dev);
1768         if (!idev)
1769                 return ERR_PTR(-ENOBUFS);
1770
1771         if (idev->cnf.disable_ipv6)
1772                 return ERR_PTR(-EACCES);
1773
1774         /* Add default multicast route */
1775         addrconf_add_mroute(dev);
1776
1777         /* Add link local route */
1778         addrconf_add_lroute(dev);
1779         return idev;
1780 }
1781
1782 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1783 {
1784         struct prefix_info *pinfo;
1785         __u32 valid_lft;
1786         __u32 prefered_lft;
1787         int addr_type;
1788         struct inet6_dev *in6_dev;
1789         struct net *net = dev_net(dev);
1790
1791         pinfo = (struct prefix_info *) opt;
1792
1793         if (len < sizeof(struct prefix_info)) {
1794                 ADBG(("addrconf: prefix option too short\n"));
1795                 return;
1796         }
1797
1798         /*
1799          *      Validation checks ([ADDRCONF], page 19)
1800          */
1801
1802         addr_type = ipv6_addr_type(&pinfo->prefix);
1803
1804         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1805                 return;
1806
1807         valid_lft = ntohl(pinfo->valid);
1808         prefered_lft = ntohl(pinfo->prefered);
1809
1810         if (prefered_lft > valid_lft) {
1811                 if (net_ratelimit())
1812                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1813                 return;
1814         }
1815
1816         in6_dev = in6_dev_get(dev);
1817
1818         if (in6_dev == NULL) {
1819                 if (net_ratelimit())
1820                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1821                 return;
1822         }
1823
1824         /*
1825          *      Two things going on here:
1826          *      1) Add routes for on-link prefixes
1827          *      2) Configure prefixes with the auto flag set
1828          */
1829
1830         if (pinfo->onlink) {
1831                 struct rt6_info *rt;
1832                 unsigned long rt_expires;
1833
1834                 /* Avoid arithmetic overflow. Really, we could
1835                  * save rt_expires in seconds, likely valid_lft,
1836                  * but it would require division in fib gc, that it
1837                  * not good.
1838                  */
1839                 if (HZ > USER_HZ)
1840                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1841                 else
1842                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1843
1844                 if (addrconf_finite_timeout(rt_expires))
1845                         rt_expires *= HZ;
1846
1847                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1848                                 dev->ifindex, 1);
1849
1850                 if (rt && addrconf_is_prefix_route(rt)) {
1851                         /* Autoconf prefix route */
1852                         if (valid_lft == 0) {
1853                                 ip6_del_rt(rt);
1854                                 rt = NULL;
1855                         } else if (addrconf_finite_timeout(rt_expires)) {
1856                                 /* not infinity */
1857                                 rt->rt6i_expires = jiffies + rt_expires;
1858                                 rt->rt6i_flags |= RTF_EXPIRES;
1859                         } else {
1860                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1861                                 rt->rt6i_expires = 0;
1862                         }
1863                 } else if (valid_lft) {
1864                         clock_t expires = 0;
1865                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1866                         if (addrconf_finite_timeout(rt_expires)) {
1867                                 /* not infinity */
1868                                 flags |= RTF_EXPIRES;
1869                                 expires = jiffies_to_clock_t(rt_expires);
1870                         }
1871                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1872                                               dev, expires, flags);
1873                 }
1874                 if (rt)
1875                         dst_release(&rt->dst);
1876         }
1877
1878         /* Try to figure out our local address for this prefix */
1879
1880         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1881                 struct inet6_ifaddr * ifp;
1882                 struct in6_addr addr;
1883                 int create = 0, update_lft = 0;
1884
1885                 if (pinfo->prefix_len == 64) {
1886                         memcpy(&addr, &pinfo->prefix, 8);
1887                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1888                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1889                                 in6_dev_put(in6_dev);
1890                                 return;
1891                         }
1892                         goto ok;
1893                 }
1894                 if (net_ratelimit())
1895                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1896                                pinfo->prefix_len);
1897                 in6_dev_put(in6_dev);
1898                 return;
1899
1900 ok:
1901
1902                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1903
1904                 if (ifp == NULL && valid_lft) {
1905                         int max_addresses = in6_dev->cnf.max_addresses;
1906                         u32 addr_flags = 0;
1907
1908 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1909                         if (in6_dev->cnf.optimistic_dad &&
1910                             !net->ipv6.devconf_all->forwarding)
1911                                 addr_flags = IFA_F_OPTIMISTIC;
1912 #endif
1913
1914                         /* Do not allow to create too much of autoconfigured
1915                          * addresses; this would be too easy way to crash kernel.
1916                          */
1917                         if (!max_addresses ||
1918                             ipv6_count_addresses(in6_dev) < max_addresses)
1919                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1920                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1921                                                     addr_flags);
1922
1923                         if (!ifp || IS_ERR(ifp)) {
1924                                 in6_dev_put(in6_dev);
1925                                 return;
1926                         }
1927
1928                         update_lft = create = 1;
1929                         ifp->cstamp = jiffies;
1930                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1931                 }
1932
1933                 if (ifp) {
1934                         int flags;
1935                         unsigned long now;
1936 #ifdef CONFIG_IPV6_PRIVACY
1937                         struct inet6_ifaddr *ift;
1938 #endif
1939                         u32 stored_lft;
1940
1941                         /* update lifetime (RFC2462 5.5.3 e) */
1942                         spin_lock(&ifp->lock);
1943                         now = jiffies;
1944                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1945                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1946                         else
1947                                 stored_lft = 0;
1948                         if (!update_lft && stored_lft) {
1949                                 if (valid_lft > MIN_VALID_LIFETIME ||
1950                                     valid_lft > stored_lft)
1951                                         update_lft = 1;
1952                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1953                                         /* valid_lft <= stored_lft is always true */
1954                                         /*
1955                                          * RFC 4862 Section 5.5.3e:
1956                                          * "Note that the preferred lifetime of
1957                                          *  the corresponding address is always
1958                                          *  reset to the Preferred Lifetime in
1959                                          *  the received Prefix Information
1960                                          *  option, regardless of whether the
1961                                          *  valid lifetime is also reset or
1962                                          *  ignored."
1963                                          *
1964                                          *  So if the preferred lifetime in
1965                                          *  this advertisement is different
1966                                          *  than what we have stored, but the
1967                                          *  valid lifetime is invalid, just
1968                                          *  reset prefered_lft.
1969                                          *
1970                                          *  We must set the valid lifetime
1971                                          *  to the stored lifetime since we'll
1972                                          *  be updating the timestamp below,
1973                                          *  else we'll set it back to the
1974                                          *  minumum.
1975                                          */
1976                                         if (prefered_lft != ifp->prefered_lft) {
1977                                                 valid_lft = stored_lft;
1978                                                 update_lft = 1;
1979                                         }
1980                                 } else {
1981                                         valid_lft = MIN_VALID_LIFETIME;
1982                                         if (valid_lft < prefered_lft)
1983                                                 prefered_lft = valid_lft;
1984                                         update_lft = 1;
1985                                 }
1986                         }
1987
1988                         if (update_lft) {
1989                                 ifp->valid_lft = valid_lft;
1990                                 ifp->prefered_lft = prefered_lft;
1991                                 ifp->tstamp = now;
1992                                 flags = ifp->flags;
1993                                 ifp->flags &= ~IFA_F_DEPRECATED;
1994                                 spin_unlock(&ifp->lock);
1995
1996                                 if (!(flags&IFA_F_TENTATIVE))
1997                                         ipv6_ifa_notify(0, ifp);
1998                         } else
1999                                 spin_unlock(&ifp->lock);
2000
2001 #ifdef CONFIG_IPV6_PRIVACY
2002                         read_lock_bh(&in6_dev->lock);
2003                         /* update all temporary addresses in the list */
2004                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2005                                 /*
2006                                  * When adjusting the lifetimes of an existing
2007                                  * temporary address, only lower the lifetimes.
2008                                  * Implementations must not increase the
2009                                  * lifetimes of an existing temporary address
2010                                  * when processing a Prefix Information Option.
2011                                  */
2012                                 if (ifp != ift->ifpub)
2013                                         continue;
2014
2015                                 spin_lock(&ift->lock);
2016                                 flags = ift->flags;
2017                                 if (ift->valid_lft > valid_lft &&
2018                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2019                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2020                                 if (ift->prefered_lft > prefered_lft &&
2021                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2022                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2023                                 spin_unlock(&ift->lock);
2024                                 if (!(flags&IFA_F_TENTATIVE))
2025                                         ipv6_ifa_notify(0, ift);
2026                         }
2027
2028                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2029                                 /*
2030                                  * When a new public address is created as described in [ADDRCONF],
2031                                  * also create a new temporary address. Also create a temporary
2032                                  * address if it's enabled but no temporary address currently exists.
2033                                  */
2034                                 read_unlock_bh(&in6_dev->lock);
2035                                 ipv6_create_tempaddr(ifp, NULL);
2036                         } else {
2037                                 read_unlock_bh(&in6_dev->lock);
2038                         }
2039 #endif
2040                         in6_ifa_put(ifp);
2041                         addrconf_verify(0);
2042                 }
2043         }
2044         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2045         in6_dev_put(in6_dev);
2046 }
2047
2048 /*
2049  *      Set destination address.
2050  *      Special case for SIT interfaces where we create a new "virtual"
2051  *      device.
2052  */
2053 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2054 {
2055         struct in6_ifreq ireq;
2056         struct net_device *dev;
2057         int err = -EINVAL;
2058
2059         rtnl_lock();
2060
2061         err = -EFAULT;
2062         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2063                 goto err_exit;
2064
2065         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2066
2067         err = -ENODEV;
2068         if (dev == NULL)
2069                 goto err_exit;
2070
2071 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2072         if (dev->type == ARPHRD_SIT) {
2073                 const struct net_device_ops *ops = dev->netdev_ops;
2074                 struct ifreq ifr;
2075                 struct ip_tunnel_parm p;
2076
2077                 err = -EADDRNOTAVAIL;
2078                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2079                         goto err_exit;
2080
2081                 memset(&p, 0, sizeof(p));
2082                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2083                 p.iph.saddr = 0;
2084                 p.iph.version = 4;
2085                 p.iph.ihl = 5;
2086                 p.iph.protocol = IPPROTO_IPV6;
2087                 p.iph.ttl = 64;
2088                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2089
2090                 if (ops->ndo_do_ioctl) {
2091                         mm_segment_t oldfs = get_fs();
2092
2093                         set_fs(KERNEL_DS);
2094                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2095                         set_fs(oldfs);
2096                 } else
2097                         err = -EOPNOTSUPP;
2098
2099                 if (err == 0) {
2100                         err = -ENOBUFS;
2101                         dev = __dev_get_by_name(net, p.name);
2102                         if (!dev)
2103                                 goto err_exit;
2104                         err = dev_open(dev);
2105                 }
2106         }
2107 #endif
2108
2109 err_exit:
2110         rtnl_unlock();
2111         return err;
2112 }
2113
2114 /*
2115  *      Manual configuration of address on an interface
2116  */
2117 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2118                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2119                           __u32 valid_lft)
2120 {
2121         struct inet6_ifaddr *ifp;
2122         struct inet6_dev *idev;
2123         struct net_device *dev;
2124         int scope;
2125         u32 flags;
2126         clock_t expires;
2127         unsigned long timeout;
2128
2129         ASSERT_RTNL();
2130
2131         if (plen > 128)
2132                 return -EINVAL;
2133
2134         /* check the lifetime */
2135         if (!valid_lft || prefered_lft > valid_lft)
2136                 return -EINVAL;
2137
2138         dev = __dev_get_by_index(net, ifindex);
2139         if (!dev)
2140                 return -ENODEV;
2141
2142         idev = addrconf_add_dev(dev);
2143         if (IS_ERR(idev))
2144                 return PTR_ERR(idev);
2145
2146         scope = ipv6_addr_scope(pfx);
2147
2148         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2149         if (addrconf_finite_timeout(timeout)) {
2150                 expires = jiffies_to_clock_t(timeout * HZ);
2151                 valid_lft = timeout;
2152                 flags = RTF_EXPIRES;
2153         } else {
2154                 expires = 0;
2155                 flags = 0;
2156                 ifa_flags |= IFA_F_PERMANENT;
2157         }
2158
2159         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2160         if (addrconf_finite_timeout(timeout)) {
2161                 if (timeout == 0)
2162                         ifa_flags |= IFA_F_DEPRECATED;
2163                 prefered_lft = timeout;
2164         }
2165
2166         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2167
2168         if (!IS_ERR(ifp)) {
2169                 spin_lock_bh(&ifp->lock);
2170                 ifp->valid_lft = valid_lft;
2171                 ifp->prefered_lft = prefered_lft;
2172                 ifp->tstamp = jiffies;
2173                 spin_unlock_bh(&ifp->lock);
2174
2175                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2176                                       expires, flags);
2177                 /*
2178                  * Note that section 3.1 of RFC 4429 indicates
2179                  * that the Optimistic flag should not be set for
2180                  * manually configured addresses
2181                  */
2182                 addrconf_dad_start(ifp, 0);
2183                 in6_ifa_put(ifp);
2184                 addrconf_verify(0);
2185                 return 0;
2186         }
2187
2188         return PTR_ERR(ifp);
2189 }
2190
2191 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2192                           unsigned int plen)
2193 {
2194         struct inet6_ifaddr *ifp;
2195         struct inet6_dev *idev;
2196         struct net_device *dev;
2197
2198         if (plen > 128)
2199                 return -EINVAL;
2200
2201         dev = __dev_get_by_index(net, ifindex);
2202         if (!dev)
2203                 return -ENODEV;
2204
2205         if ((idev = __in6_dev_get(dev)) == NULL)
2206                 return -ENXIO;
2207
2208         read_lock_bh(&idev->lock);
2209         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2210                 if (ifp->prefix_len == plen &&
2211                     ipv6_addr_equal(pfx, &ifp->addr)) {
2212                         in6_ifa_hold(ifp);
2213                         read_unlock_bh(&idev->lock);
2214
2215                         ipv6_del_addr(ifp);
2216
2217                         /* If the last address is deleted administratively,
2218                            disable IPv6 on this interface.
2219                          */
2220                         if (list_empty(&idev->addr_list))
2221                                 addrconf_ifdown(idev->dev, 1);
2222                         return 0;
2223                 }
2224         }
2225         read_unlock_bh(&idev->lock);
2226         return -EADDRNOTAVAIL;
2227 }
2228
2229
2230 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2231 {
2232         struct in6_ifreq ireq;
2233         int err;
2234
2235         if (!capable(CAP_NET_ADMIN))
2236                 return -EPERM;
2237
2238         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2239                 return -EFAULT;
2240
2241         rtnl_lock();
2242         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2243                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2244                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2245         rtnl_unlock();
2246         return err;
2247 }
2248
2249 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2250 {
2251         struct in6_ifreq ireq;
2252         int err;
2253
2254         if (!capable(CAP_NET_ADMIN))
2255                 return -EPERM;
2256
2257         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2258                 return -EFAULT;
2259
2260         rtnl_lock();
2261         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2262                              ireq.ifr6_prefixlen);
2263         rtnl_unlock();
2264         return err;
2265 }
2266
2267 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2268                      int plen, int scope)
2269 {
2270         struct inet6_ifaddr *ifp;
2271
2272         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2273         if (!IS_ERR(ifp)) {
2274                 spin_lock_bh(&ifp->lock);
2275                 ifp->flags &= ~IFA_F_TENTATIVE;
2276                 spin_unlock_bh(&ifp->lock);
2277                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2278                 in6_ifa_put(ifp);
2279         }
2280 }
2281
2282 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2283 static void sit_add_v4_addrs(struct inet6_dev *idev)
2284 {
2285         struct in6_addr addr;
2286         struct net_device *dev;
2287         struct net *net = dev_net(idev->dev);
2288         int scope;
2289
2290         ASSERT_RTNL();
2291
2292         memset(&addr, 0, sizeof(struct in6_addr));
2293         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2294
2295         if (idev->dev->flags&IFF_POINTOPOINT) {
2296                 addr.s6_addr32[0] = htonl(0xfe800000);
2297                 scope = IFA_LINK;
2298         } else {
2299                 scope = IPV6_ADDR_COMPATv4;
2300         }
2301
2302         if (addr.s6_addr32[3]) {
2303                 add_addr(idev, &addr, 128, scope);
2304                 return;
2305         }
2306
2307         for_each_netdev(net, dev) {
2308                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2309                 if (in_dev && (dev->flags & IFF_UP)) {
2310                         struct in_ifaddr * ifa;
2311
2312                         int flag = scope;
2313
2314                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2315                                 int plen;
2316
2317                                 addr.s6_addr32[3] = ifa->ifa_local;
2318
2319                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2320                                         continue;
2321                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2322                                         if (idev->dev->flags&IFF_POINTOPOINT)
2323                                                 continue;
2324                                         flag |= IFA_HOST;
2325                                 }
2326                                 if (idev->dev->flags&IFF_POINTOPOINT)
2327                                         plen = 64;
2328                                 else
2329                                         plen = 96;
2330
2331                                 add_addr(idev, &addr, plen, flag);
2332                         }
2333                 }
2334         }
2335 }
2336 #endif
2337
2338 static void init_loopback(struct net_device *dev)
2339 {
2340         struct inet6_dev  *idev;
2341
2342         /* ::1 */
2343
2344         ASSERT_RTNL();
2345
2346         if ((idev = ipv6_find_idev(dev)) == NULL) {
2347                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2348                 return;
2349         }
2350
2351         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2352 }
2353
2354 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2355 {
2356         struct inet6_ifaddr * ifp;
2357         u32 addr_flags = IFA_F_PERMANENT;
2358
2359 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2360         if (idev->cnf.optimistic_dad &&
2361             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2362                 addr_flags |= IFA_F_OPTIMISTIC;
2363 #endif
2364
2365
2366         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2367         if (!IS_ERR(ifp)) {
2368                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2369                 addrconf_dad_start(ifp, 0);
2370                 in6_ifa_put(ifp);
2371         }
2372 }
2373
2374 static void addrconf_dev_config(struct net_device *dev)
2375 {
2376         struct in6_addr addr;
2377         struct inet6_dev    * idev;
2378
2379         ASSERT_RTNL();
2380
2381         if ((dev->type != ARPHRD_ETHER) &&
2382             (dev->type != ARPHRD_FDDI) &&
2383             (dev->type != ARPHRD_IEEE802_TR) &&
2384             (dev->type != ARPHRD_ARCNET) &&
2385             (dev->type != ARPHRD_INFINIBAND)) {
2386                 /* Alas, we support only Ethernet autoconfiguration. */
2387                 return;
2388         }
2389
2390         idev = addrconf_add_dev(dev);
2391         if (IS_ERR(idev))
2392                 return;
2393
2394         memset(&addr, 0, sizeof(struct in6_addr));
2395         addr.s6_addr32[0] = htonl(0xFE800000);
2396
2397         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2398                 addrconf_add_linklocal(idev, &addr);
2399 }
2400
2401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2402 static void addrconf_sit_config(struct net_device *dev)
2403 {
2404         struct inet6_dev *idev;
2405
2406         ASSERT_RTNL();
2407
2408         /*
2409          * Configure the tunnel with one of our IPv4
2410          * addresses... we should configure all of
2411          * our v4 addrs in the tunnel
2412          */
2413
2414         if ((idev = ipv6_find_idev(dev)) == NULL) {
2415                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2416                 return;
2417         }
2418
2419         if (dev->priv_flags & IFF_ISATAP) {
2420                 struct in6_addr addr;
2421
2422                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2423                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2424                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2425                         addrconf_add_linklocal(idev, &addr);
2426                 return;
2427         }
2428
2429         sit_add_v4_addrs(idev);
2430
2431         if (dev->flags&IFF_POINTOPOINT) {
2432                 addrconf_add_mroute(dev);
2433                 addrconf_add_lroute(dev);
2434         } else
2435                 sit_route_add(dev);
2436 }
2437 #endif
2438
2439 static inline int
2440 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2441 {
2442         struct in6_addr lladdr;
2443
2444         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2445                 addrconf_add_linklocal(idev, &lladdr);
2446                 return 0;
2447         }
2448         return -1;
2449 }
2450
2451 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2452 {
2453         struct net_device *link_dev;
2454         struct net *net = dev_net(idev->dev);
2455
2456         /* first try to inherit the link-local address from the link device */
2457         if (idev->dev->iflink &&
2458             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2459                 if (!ipv6_inherit_linklocal(idev, link_dev))
2460                         return;
2461         }
2462         /* then try to inherit it from any device */
2463         for_each_netdev(net, link_dev) {
2464                 if (!ipv6_inherit_linklocal(idev, link_dev))
2465                         return;
2466         }
2467         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2468 }
2469
2470 /*
2471  * Autoconfigure tunnel with a link-local address so routing protocols,
2472  * DHCPv6, MLD etc. can be run over the virtual link
2473  */
2474
2475 static void addrconf_ip6_tnl_config(struct net_device *dev)
2476 {
2477         struct inet6_dev *idev;
2478
2479         ASSERT_RTNL();
2480
2481         idev = addrconf_add_dev(dev);
2482         if (IS_ERR(idev)) {
2483                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2484                 return;
2485         }
2486         ip6_tnl_add_linklocal(idev);
2487 }
2488
2489 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2490                            void * data)
2491 {
2492         struct net_device *dev = (struct net_device *) data;
2493         struct inet6_dev *idev = __in6_dev_get(dev);
2494         int run_pending = 0;
2495         int err;
2496
2497         switch (event) {
2498         case NETDEV_REGISTER:
2499                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2500                         idev = ipv6_add_dev(dev);
2501                         if (!idev)
2502                                 return notifier_from_errno(-ENOMEM);
2503                 }
2504                 break;
2505
2506         case NETDEV_UP:
2507         case NETDEV_CHANGE:
2508                 if (dev->flags & IFF_SLAVE)
2509                         break;
2510
2511                 if (event == NETDEV_UP) {
2512                         if (!addrconf_qdisc_ok(dev)) {
2513                                 /* device is not ready yet. */
2514                                 printk(KERN_INFO
2515                                         "ADDRCONF(NETDEV_UP): %s: "
2516                                         "link is not ready\n",
2517                                         dev->name);
2518                                 break;
2519                         }
2520
2521                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2522                                 idev = ipv6_add_dev(dev);
2523
2524                         if (idev) {
2525                                 idev->if_flags |= IF_READY;
2526                                 run_pending = 1;
2527                         }
2528                 } else {
2529                         if (!addrconf_qdisc_ok(dev)) {
2530                                 /* device is still not ready. */
2531                                 break;
2532                         }
2533
2534                         if (idev) {
2535                                 if (idev->if_flags & IF_READY)
2536                                         /* device is already configured. */
2537                                         break;
2538                                 idev->if_flags |= IF_READY;
2539                         }
2540
2541                         printk(KERN_INFO
2542                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2543                                         "link becomes ready\n",
2544                                         dev->name);
2545
2546                         run_pending = 1;
2547                 }
2548
2549                 switch (dev->type) {
2550 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2551                 case ARPHRD_SIT:
2552                         addrconf_sit_config(dev);
2553                         break;
2554 #endif
2555                 case ARPHRD_TUNNEL6:
2556                         addrconf_ip6_tnl_config(dev);
2557                         break;
2558                 case ARPHRD_LOOPBACK:
2559                         init_loopback(dev);
2560                         break;
2561
2562                 default:
2563                         addrconf_dev_config(dev);
2564                         break;
2565                 }
2566
2567                 if (idev) {
2568                         if (run_pending)
2569                                 addrconf_dad_run(idev);
2570
2571                         /*
2572                          * If the MTU changed during the interface down,
2573                          * when the interface up, the changed MTU must be
2574                          * reflected in the idev as well as routers.
2575                          */
2576                         if (idev->cnf.mtu6 != dev->mtu &&
2577                             dev->mtu >= IPV6_MIN_MTU) {
2578                                 rt6_mtu_change(dev, dev->mtu);
2579                                 idev->cnf.mtu6 = dev->mtu;
2580                         }
2581                         idev->tstamp = jiffies;
2582                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2583
2584                         /*
2585                          * If the changed mtu during down is lower than
2586                          * IPV6_MIN_MTU stop IPv6 on this interface.
2587                          */
2588                         if (dev->mtu < IPV6_MIN_MTU)
2589                                 addrconf_ifdown(dev, 1);
2590                 }
2591                 break;
2592
2593         case NETDEV_CHANGEMTU:
2594                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2595                         rt6_mtu_change(dev, dev->mtu);
2596                         idev->cnf.mtu6 = dev->mtu;
2597                         break;
2598                 }
2599
2600                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2601                         idev = ipv6_add_dev(dev);
2602                         if (idev)
2603                                 break;
2604                 }
2605
2606                 /*
2607                  * MTU falled under IPV6_MIN_MTU.
2608                  * Stop IPv6 on this interface.
2609                  */
2610
2611         case NETDEV_DOWN:
2612         case NETDEV_UNREGISTER:
2613                 /*
2614                  *      Remove all addresses from this interface.
2615                  */
2616                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2617                 break;
2618
2619         case NETDEV_CHANGENAME:
2620                 if (idev) {
2621                         snmp6_unregister_dev(idev);
2622                         addrconf_sysctl_unregister(idev);
2623                         addrconf_sysctl_register(idev);
2624                         err = snmp6_register_dev(idev);
2625                         if (err)
2626                                 return notifier_from_errno(err);
2627                 }
2628                 break;
2629
2630         case NETDEV_PRE_TYPE_CHANGE:
2631         case NETDEV_POST_TYPE_CHANGE:
2632                 addrconf_type_change(dev, event);
2633                 break;
2634         }
2635
2636         return NOTIFY_OK;
2637 }
2638
2639 /*
2640  *      addrconf module should be notified of a device going up
2641  */
2642 static struct notifier_block ipv6_dev_notf = {
2643         .notifier_call = addrconf_notify,
2644 };
2645
2646 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2647 {
2648         struct inet6_dev *idev;
2649         ASSERT_RTNL();
2650
2651         idev = __in6_dev_get(dev);
2652
2653         if (event == NETDEV_POST_TYPE_CHANGE)
2654                 ipv6_mc_remap(idev);
2655         else if (event == NETDEV_PRE_TYPE_CHANGE)
2656                 ipv6_mc_unmap(idev);
2657 }
2658
2659 static int addrconf_ifdown(struct net_device *dev, int how)
2660 {
2661         struct net *net = dev_net(dev);
2662         struct inet6_dev *idev;
2663         struct inet6_ifaddr *ifa;
2664         int state, i;
2665
2666         ASSERT_RTNL();
2667
2668         rt6_ifdown(net, dev);
2669         neigh_ifdown(&nd_tbl, dev);
2670
2671         idev = __in6_dev_get(dev);
2672         if (idev == NULL)
2673                 return -ENODEV;
2674
2675         /*
2676          * Step 1: remove reference to ipv6 device from parent device.
2677          *         Do not dev_put!
2678          */
2679         if (how) {
2680                 idev->dead = 1;
2681
2682                 /* protected by rtnl_lock */
2683                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2684
2685                 /* Step 1.5: remove snmp6 entry */
2686                 snmp6_unregister_dev(idev);
2687
2688         }
2689
2690         /* Step 2: clear hash table */
2691         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2692                 struct hlist_head *h = &inet6_addr_lst[i];
2693                 struct hlist_node *n;
2694
2695                 spin_lock_bh(&addrconf_hash_lock);
2696         restart:
2697                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2698                         if (ifa->idev == idev) {
2699                                 hlist_del_init_rcu(&ifa->addr_lst);
2700                                 addrconf_del_timer(ifa);
2701                                 goto restart;
2702                         }
2703                 }
2704                 spin_unlock_bh(&addrconf_hash_lock);
2705         }
2706
2707         write_lock_bh(&idev->lock);
2708
2709         /* Step 2: clear flags for stateless addrconf */
2710         if (!how)
2711                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2712
2713 #ifdef CONFIG_IPV6_PRIVACY
2714         if (how && del_timer(&idev->regen_timer))
2715                 in6_dev_put(idev);
2716
2717         /* Step 3: clear tempaddr list */
2718         while (!list_empty(&idev->tempaddr_list)) {
2719                 ifa = list_first_entry(&idev->tempaddr_list,
2720                                        struct inet6_ifaddr, tmp_list);
2721                 list_del(&ifa->tmp_list);
2722                 write_unlock_bh(&idev->lock);
2723                 spin_lock_bh(&ifa->lock);
2724
2725                 if (ifa->ifpub) {
2726                         in6_ifa_put(ifa->ifpub);
2727                         ifa->ifpub = NULL;
2728                 }
2729                 spin_unlock_bh(&ifa->lock);
2730                 in6_ifa_put(ifa);
2731                 write_lock_bh(&idev->lock);
2732         }
2733 #endif
2734
2735         while (!list_empty(&idev->addr_list)) {
2736                 ifa = list_first_entry(&idev->addr_list,
2737                                        struct inet6_ifaddr, if_list);
2738                 addrconf_del_timer(ifa);
2739
2740                 list_del(&ifa->if_list);
2741
2742                 write_unlock_bh(&idev->lock);
2743
2744                 spin_lock_bh(&ifa->state_lock);
2745                 state = ifa->state;
2746                 ifa->state = INET6_IFADDR_STATE_DEAD;
2747                 spin_unlock_bh(&ifa->state_lock);
2748
2749                 if (state != INET6_IFADDR_STATE_DEAD) {
2750                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2751                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2752                 }
2753                 in6_ifa_put(ifa);
2754
2755                 write_lock_bh(&idev->lock);
2756         }
2757
2758         write_unlock_bh(&idev->lock);
2759
2760         /* Step 5: Discard multicast list */
2761         if (how)
2762                 ipv6_mc_destroy_dev(idev);
2763         else
2764                 ipv6_mc_down(idev);
2765
2766         idev->tstamp = jiffies;
2767
2768         /* Last: Shot the device (if unregistered) */
2769         if (how) {
2770                 addrconf_sysctl_unregister(idev);
2771                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2772                 neigh_ifdown(&nd_tbl, dev);
2773                 in6_dev_put(idev);
2774         }
2775         return 0;
2776 }
2777
2778 static void addrconf_rs_timer(unsigned long data)
2779 {
2780         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2781         struct inet6_dev *idev = ifp->idev;
2782
2783         read_lock(&idev->lock);
2784         if (idev->dead || !(idev->if_flags & IF_READY))
2785                 goto out;
2786
2787         if (idev->cnf.forwarding)
2788                 goto out;
2789
2790         /* Announcement received after solicitation was sent */
2791         if (idev->if_flags & IF_RA_RCVD)
2792                 goto out;
2793
2794         spin_lock(&ifp->lock);
2795         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2796                 /* The wait after the last probe can be shorter */
2797                 addrconf_mod_timer(ifp, AC_RS,
2798                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2799                                    idev->cnf.rtr_solicit_delay :
2800                                    idev->cnf.rtr_solicit_interval);
2801                 spin_unlock(&ifp->lock);
2802
2803                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2804         } else {
2805                 spin_unlock(&ifp->lock);
2806                 /*
2807                  * Note: we do not support deprecated "all on-link"
2808                  * assumption any longer.
2809                  */
2810                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2811                        idev->dev->name);
2812         }
2813
2814 out:
2815         read_unlock(&idev->lock);
2816         in6_ifa_put(ifp);
2817 }
2818
2819 /*
2820  *      Duplicate Address Detection
2821  */
2822 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2823 {
2824         unsigned long rand_num;
2825         struct inet6_dev *idev = ifp->idev;
2826
2827         if (ifp->flags & IFA_F_OPTIMISTIC)
2828                 rand_num = 0;
2829         else
2830                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2831
2832         ifp->probes = idev->cnf.dad_transmits;
2833         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2834 }
2835
2836 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2837 {
2838         struct inet6_dev *idev = ifp->idev;
2839         struct net_device *dev = idev->dev;
2840
2841         addrconf_join_solict(dev, &ifp->addr);
2842
2843         net_srandom(ifp->addr.s6_addr32[3]);
2844
2845         read_lock_bh(&idev->lock);
2846         spin_lock(&ifp->lock);
2847         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2848                 goto out;
2849
2850         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2851             idev->cnf.accept_dad < 1 ||
2852             !(ifp->flags&IFA_F_TENTATIVE) ||
2853             ifp->flags & IFA_F_NODAD) {
2854                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2855                 spin_unlock(&ifp->lock);
2856                 read_unlock_bh(&idev->lock);
2857
2858                 addrconf_dad_completed(ifp);
2859                 return;
2860         }
2861
2862         if (!(idev->if_flags & IF_READY)) {
2863                 spin_unlock(&ifp->lock);
2864                 read_unlock_bh(&idev->lock);
2865                 /*
2866                  * If the device is not ready:
2867                  * - keep it tentative if it is a permanent address.
2868                  * - otherwise, kill it.
2869                  */
2870                 in6_ifa_hold(ifp);
2871                 addrconf_dad_stop(ifp, 0);
2872                 return;
2873         }
2874
2875         /*
2876          * Optimistic nodes can start receiving
2877          * Frames right away
2878          */
2879         if (ifp->flags & IFA_F_OPTIMISTIC)
2880                 ip6_ins_rt(ifp->rt);
2881
2882         addrconf_dad_kick(ifp);
2883 out:
2884         spin_unlock(&ifp->lock);
2885         read_unlock_bh(&idev->lock);
2886 }
2887
2888 static void addrconf_dad_timer(unsigned long data)
2889 {
2890         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2891         struct inet6_dev *idev = ifp->idev;
2892         struct in6_addr mcaddr;
2893
2894         if (!ifp->probes && addrconf_dad_end(ifp))
2895                 goto out;
2896
2897         read_lock(&idev->lock);
2898         if (idev->dead || !(idev->if_flags & IF_READY)) {
2899                 read_unlock(&idev->lock);
2900                 goto out;
2901         }
2902
2903         spin_lock(&ifp->lock);
2904         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2905                 spin_unlock(&ifp->lock);
2906                 read_unlock(&idev->lock);
2907                 goto out;
2908         }
2909
2910         if (ifp->probes == 0) {
2911                 /*
2912                  * DAD was successful
2913                  */
2914
2915                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2916                 spin_unlock(&ifp->lock);
2917                 read_unlock(&idev->lock);
2918
2919                 addrconf_dad_completed(ifp);
2920
2921                 goto out;
2922         }
2923
2924         ifp->probes--;
2925         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2926         spin_unlock(&ifp->lock);
2927         read_unlock(&idev->lock);
2928
2929         /* send a neighbour solicitation for our addr */
2930         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2931         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2932 out:
2933         in6_ifa_put(ifp);
2934 }
2935
2936 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2937 {
2938         struct net_device *dev = ifp->idev->dev;
2939
2940         /*
2941          *      Configure the address for reception. Now it is valid.
2942          */
2943
2944         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2945
2946         /* If added prefix is link local and forwarding is off,
2947            start sending router solicitations.
2948          */
2949
2950         if ((ifp->idev->cnf.forwarding == 0 ||
2951              ifp->idev->cnf.forwarding == 2) &&
2952             ifp->idev->cnf.rtr_solicits > 0 &&
2953             (dev->flags&IFF_LOOPBACK) == 0 &&
2954             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2955                 /*
2956                  *      If a host as already performed a random delay
2957                  *      [...] as part of DAD [...] there is no need
2958                  *      to delay again before sending the first RS
2959                  */
2960                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2961
2962                 spin_lock_bh(&ifp->lock);
2963                 ifp->probes = 1;
2964                 ifp->idev->if_flags |= IF_RS_SENT;
2965                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2966                 spin_unlock_bh(&ifp->lock);
2967         }
2968 }
2969
2970 static void addrconf_dad_run(struct inet6_dev *idev)
2971 {
2972         struct inet6_ifaddr *ifp;
2973
2974         read_lock_bh(&idev->lock);
2975         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2976                 spin_lock(&ifp->lock);
2977                 if (ifp->flags & IFA_F_TENTATIVE &&
2978                     ifp->state == INET6_IFADDR_STATE_DAD)
2979                         addrconf_dad_kick(ifp);
2980                 spin_unlock(&ifp->lock);
2981         }
2982         read_unlock_bh(&idev->lock);
2983 }
2984
2985 #ifdef CONFIG_PROC_FS
2986 struct if6_iter_state {
2987         struct seq_net_private p;
2988         int bucket;
2989 };
2990
2991 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2992 {
2993         struct inet6_ifaddr *ifa = NULL;
2994         struct if6_iter_state *state = seq->private;
2995         struct net *net = seq_file_net(seq);
2996
2997         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2998                 struct hlist_node *n;
2999                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3000                                          addr_lst)
3001                         if (net_eq(dev_net(ifa->idev->dev), net))
3002                                 return ifa;
3003         }
3004         return NULL;
3005 }
3006
3007 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3008                                          struct inet6_ifaddr *ifa)
3009 {
3010         struct if6_iter_state *state = seq->private;
3011         struct net *net = seq_file_net(seq);
3012         struct hlist_node *n = &ifa->addr_lst;
3013
3014         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3015                 if (net_eq(dev_net(ifa->idev->dev), net))
3016                         return ifa;
3017
3018         while (++state->bucket < IN6_ADDR_HSIZE) {
3019                 hlist_for_each_entry_rcu_bh(ifa, n,
3020                                      &inet6_addr_lst[state->bucket], addr_lst) {
3021                         if (net_eq(dev_net(ifa->idev->dev), net))
3022                                 return ifa;
3023                 }
3024         }
3025
3026         return NULL;
3027 }
3028
3029 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3030 {
3031         struct inet6_ifaddr *ifa = if6_get_first(seq);
3032
3033         if (ifa)
3034                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3035                         --pos;
3036         return pos ? NULL : ifa;
3037 }
3038
3039 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3040         __acquires(rcu_bh)
3041 {
3042         rcu_read_lock_bh();
3043         return if6_get_idx(seq, *pos);
3044 }
3045
3046 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3047 {
3048         struct inet6_ifaddr *ifa;
3049
3050         ifa = if6_get_next(seq, v);
3051         ++*pos;
3052         return ifa;
3053 }
3054
3055 static void if6_seq_stop(struct seq_file *seq, void *v)
3056         __releases(rcu_bh)
3057 {
3058         rcu_read_unlock_bh();
3059 }
3060
3061 static int if6_seq_show(struct seq_file *seq, void *v)
3062 {
3063         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3064         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3065                    &ifp->addr,
3066                    ifp->idev->dev->ifindex,
3067                    ifp->prefix_len,
3068                    ifp->scope,
3069                    ifp->flags,
3070                    ifp->idev->dev->name);
3071         return 0;
3072 }
3073
3074 static const struct seq_operations if6_seq_ops = {
3075         .start  = if6_seq_start,
3076         .next   = if6_seq_next,
3077         .show   = if6_seq_show,
3078         .stop   = if6_seq_stop,
3079 };
3080
3081 static int if6_seq_open(struct inode *inode, struct file *file)
3082 {
3083         return seq_open_net(inode, file, &if6_seq_ops,
3084                             sizeof(struct if6_iter_state));
3085 }
3086
3087 static const struct file_operations if6_fops = {
3088         .owner          = THIS_MODULE,
3089         .open           = if6_seq_open,
3090         .read           = seq_read,
3091         .llseek         = seq_lseek,
3092         .release        = seq_release_net,
3093 };
3094
3095 static int __net_init if6_proc_net_init(struct net *net)
3096 {
3097         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3098                 return -ENOMEM;
3099         return 0;
3100 }
3101
3102 static void __net_exit if6_proc_net_exit(struct net *net)
3103 {
3104        proc_net_remove(net, "if_inet6");
3105 }
3106
3107 static struct pernet_operations if6_proc_net_ops = {
3108        .init = if6_proc_net_init,
3109        .exit = if6_proc_net_exit,
3110 };
3111
3112 int __init if6_proc_init(void)
3113 {
3114         return register_pernet_subsys(&if6_proc_net_ops);
3115 }
3116
3117 void if6_proc_exit(void)
3118 {
3119         unregister_pernet_subsys(&if6_proc_net_ops);
3120 }
3121 #endif  /* CONFIG_PROC_FS */
3122
3123 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3124 /* Check if address is a home address configured on any interface. */
3125 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3126 {
3127         int ret = 0;
3128         struct inet6_ifaddr *ifp = NULL;
3129         struct hlist_node *n;
3130         unsigned int hash = ipv6_addr_hash(addr);
3131
3132         rcu_read_lock_bh();
3133         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3134                 if (!net_eq(dev_net(ifp->idev->dev), net))
3135                         continue;
3136                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3137                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3138                         ret = 1;
3139                         break;
3140                 }
3141         }
3142         rcu_read_unlock_bh();
3143         return ret;
3144 }
3145 #endif
3146
3147 /*
3148  *      Periodic address status verification
3149  */
3150
3151 static void addrconf_verify(unsigned long foo)
3152 {
3153         unsigned long now, next, next_sec, next_sched;
3154         struct inet6_ifaddr *ifp;
3155         struct hlist_node *node;
3156         int i;
3157
3158         rcu_read_lock_bh();
3159         spin_lock(&addrconf_verify_lock);
3160         now = jiffies;
3161         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3162
3163         del_timer(&addr_chk_timer);
3164
3165         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3166 restart:
3167                 hlist_for_each_entry_rcu_bh(ifp, node,
3168                                          &inet6_addr_lst[i], addr_lst) {
3169                         unsigned long age;
3170
3171                         if (ifp->flags & IFA_F_PERMANENT)
3172                                 continue;
3173
3174                         spin_lock(&ifp->lock);
3175                         /* We try to batch several events at once. */
3176                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3177
3178                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3179                             age >= ifp->valid_lft) {
3180                                 spin_unlock(&ifp->lock);
3181                                 in6_ifa_hold(ifp);
3182                                 ipv6_del_addr(ifp);
3183                                 goto restart;
3184                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3185                                 spin_unlock(&ifp->lock);
3186                                 continue;
3187                         } else if (age >= ifp->prefered_lft) {
3188                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3189                                 int deprecate = 0;
3190
3191                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3192                                         deprecate = 1;
3193                                         ifp->flags |= IFA_F_DEPRECATED;
3194                                 }
3195
3196                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3197                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3198
3199                                 spin_unlock(&ifp->lock);
3200
3201                                 if (deprecate) {
3202                                         in6_ifa_hold(ifp);
3203
3204                                         ipv6_ifa_notify(0, ifp);
3205                                         in6_ifa_put(ifp);
3206                                         goto restart;
3207                                 }
3208 #ifdef CONFIG_IPV6_PRIVACY
3209                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3210                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3211                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3212                                         ifp->idev->cnf.dad_transmits *
3213                                         ifp->idev->nd_parms->retrans_time / HZ;
3214
3215                                 if (age >= ifp->prefered_lft - regen_advance) {
3216                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3217                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3218                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3219                                         if (!ifp->regen_count && ifpub) {
3220                                                 ifp->regen_count++;
3221                                                 in6_ifa_hold(ifp);
3222                                                 in6_ifa_hold(ifpub);
3223                                                 spin_unlock(&ifp->lock);
3224
3225                                                 spin_lock(&ifpub->lock);
3226                                                 ifpub->regen_count = 0;
3227                                                 spin_unlock(&ifpub->lock);
3228                                                 ipv6_create_tempaddr(ifpub, ifp);
3229                                                 in6_ifa_put(ifpub);
3230                                                 in6_ifa_put(ifp);
3231                                                 goto restart;
3232                                         }
3233                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3234                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3235                                 spin_unlock(&ifp->lock);
3236 #endif
3237                         } else {
3238                                 /* ifp->prefered_lft <= ifp->valid_lft */
3239                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3240                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3241                                 spin_unlock(&ifp->lock);
3242                         }
3243                 }
3244         }
3245
3246         next_sec = round_jiffies_up(next);
3247         next_sched = next;
3248
3249         /* If rounded timeout is accurate enough, accept it. */
3250         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3251                 next_sched = next_sec;
3252
3253         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3254         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3255                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3256
3257         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3258               now, next, next_sec, next_sched));
3259
3260         addr_chk_timer.expires = next_sched;
3261         add_timer(&addr_chk_timer);
3262         spin_unlock(&addrconf_verify_lock);
3263         rcu_read_unlock_bh();
3264 }
3265
3266 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3267 {
3268         struct in6_addr *pfx = NULL;
3269
3270         if (addr)
3271                 pfx = nla_data(addr);
3272
3273         if (local) {
3274                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3275                         pfx = NULL;
3276                 else
3277                         pfx = nla_data(local);
3278         }
3279
3280         return pfx;
3281 }
3282
3283 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3284         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3285         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3286         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3287 };
3288
3289 static int
3290 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3291 {
3292         struct net *net = sock_net(skb->sk);
3293         struct ifaddrmsg *ifm;
3294         struct nlattr *tb[IFA_MAX+1];
3295         struct in6_addr *pfx;
3296         int err;
3297
3298         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3299         if (err < 0)
3300                 return err;
3301
3302         ifm = nlmsg_data(nlh);
3303         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3304         if (pfx == NULL)
3305                 return -EINVAL;
3306
3307         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3308 }
3309
3310 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3311                              u32 prefered_lft, u32 valid_lft)
3312 {
3313         u32 flags;
3314         clock_t expires;
3315         unsigned long timeout;
3316
3317         if (!valid_lft || (prefered_lft > valid_lft))
3318                 return -EINVAL;
3319
3320         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3321         if (addrconf_finite_timeout(timeout)) {
3322                 expires = jiffies_to_clock_t(timeout * HZ);
3323                 valid_lft = timeout;
3324                 flags = RTF_EXPIRES;
3325         } else {
3326                 expires = 0;
3327                 flags = 0;
3328                 ifa_flags |= IFA_F_PERMANENT;
3329         }
3330
3331         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3332         if (addrconf_finite_timeout(timeout)) {
3333                 if (timeout == 0)
3334                         ifa_flags |= IFA_F_DEPRECATED;
3335                 prefered_lft = timeout;
3336         }
3337
3338         spin_lock_bh(&ifp->lock);
3339         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3340         ifp->tstamp = jiffies;
3341         ifp->valid_lft = valid_lft;
3342         ifp->prefered_lft = prefered_lft;
3343
3344         spin_unlock_bh(&ifp->lock);
3345         if (!(ifp->flags&IFA_F_TENTATIVE))
3346                 ipv6_ifa_notify(0, ifp);
3347
3348         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3349                               expires, flags);
3350         addrconf_verify(0);
3351
3352         return 0;
3353 }
3354
3355 static int
3356 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3357 {
3358         struct net *net = sock_net(skb->sk);
3359         struct ifaddrmsg *ifm;
3360         struct nlattr *tb[IFA_MAX+1];
3361         struct in6_addr *pfx;
3362         struct inet6_ifaddr *ifa;
3363         struct net_device *dev;
3364         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3365         u8 ifa_flags;
3366         int err;
3367
3368         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3369         if (err < 0)
3370                 return err;
3371
3372         ifm = nlmsg_data(nlh);
3373         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3374         if (pfx == NULL)
3375                 return -EINVAL;
3376
3377         if (tb[IFA_CACHEINFO]) {
3378                 struct ifa_cacheinfo *ci;
3379
3380                 ci = nla_data(tb[IFA_CACHEINFO]);
3381                 valid_lft = ci->ifa_valid;
3382                 preferred_lft = ci->ifa_prefered;
3383         } else {
3384                 preferred_lft = INFINITY_LIFE_TIME;
3385                 valid_lft = INFINITY_LIFE_TIME;
3386         }
3387
3388         dev =  __dev_get_by_index(net, ifm->ifa_index);
3389         if (dev == NULL)
3390                 return -ENODEV;
3391
3392         /* We ignore other flags so far. */
3393         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3394
3395         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3396         if (ifa == NULL) {
3397                 /*
3398                  * It would be best to check for !NLM_F_CREATE here but
3399                  * userspace alreay relies on not having to provide this.
3400                  */
3401                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3402                                       ifm->ifa_prefixlen, ifa_flags,
3403                                       preferred_lft, valid_lft);
3404         }
3405
3406         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3407             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3408                 err = -EEXIST;
3409         else
3410                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3411
3412         in6_ifa_put(ifa);
3413
3414         return err;
3415 }
3416
3417 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3418                           u8 scope, int ifindex)
3419 {
3420         struct ifaddrmsg *ifm;
3421
3422         ifm = nlmsg_data(nlh);
3423         ifm->ifa_family = AF_INET6;
3424         ifm->ifa_prefixlen = prefixlen;
3425         ifm->ifa_flags = flags;
3426         ifm->ifa_scope = scope;
3427         ifm->ifa_index = ifindex;
3428 }
3429
3430 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3431                          unsigned long tstamp, u32 preferred, u32 valid)
3432 {
3433         struct ifa_cacheinfo ci;
3434
3435         ci.cstamp = cstamp_delta(cstamp);
3436         ci.tstamp = cstamp_delta(tstamp);
3437         ci.ifa_prefered = preferred;
3438         ci.ifa_valid = valid;
3439
3440         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3441 }
3442
3443 static inline int rt_scope(int ifa_scope)
3444 {
3445         if (ifa_scope & IFA_HOST)
3446                 return RT_SCOPE_HOST;
3447         else if (ifa_scope & IFA_LINK)
3448                 return RT_SCOPE_LINK;
3449         else if (ifa_scope & IFA_SITE)
3450                 return RT_SCOPE_SITE;
3451         else
3452                 return RT_SCOPE_UNIVERSE;
3453 }
3454
3455 static inline int inet6_ifaddr_msgsize(void)
3456 {
3457         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3458                + nla_total_size(16) /* IFA_ADDRESS */
3459                + nla_total_size(sizeof(struct ifa_cacheinfo));
3460 }
3461
3462 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3463                              u32 pid, u32 seq, int event, unsigned int flags)
3464 {
3465         struct nlmsghdr  *nlh;
3466         u32 preferred, valid;
3467
3468         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3469         if (nlh == NULL)
3470                 return -EMSGSIZE;
3471
3472         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3473                       ifa->idev->dev->ifindex);
3474
3475         if (!(ifa->flags&IFA_F_PERMANENT)) {
3476                 preferred = ifa->prefered_lft;
3477                 valid = ifa->valid_lft;
3478                 if (preferred != INFINITY_LIFE_TIME) {
3479                         long tval = (jiffies - ifa->tstamp)/HZ;
3480                         if (preferred > tval)
3481                                 preferred -= tval;
3482                         else
3483                                 preferred = 0;
3484                         if (valid != INFINITY_LIFE_TIME) {
3485                                 if (valid > tval)
3486                                         valid -= tval;
3487                                 else
3488                                         valid = 0;
3489                         }
3490                 }
3491         } else {
3492                 preferred = INFINITY_LIFE_TIME;
3493                 valid = INFINITY_LIFE_TIME;
3494         }
3495
3496         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3497             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3498                 nlmsg_cancel(skb, nlh);
3499                 return -EMSGSIZE;
3500         }
3501
3502         return nlmsg_end(skb, nlh);
3503 }
3504
3505 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3506                                 u32 pid, u32 seq, int event, u16 flags)
3507 {
3508         struct nlmsghdr  *nlh;
3509         u8 scope = RT_SCOPE_UNIVERSE;
3510         int ifindex = ifmca->idev->dev->ifindex;
3511
3512         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3513                 scope = RT_SCOPE_SITE;
3514
3515         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3516         if (nlh == NULL)
3517                 return -EMSGSIZE;
3518
3519         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3520         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3521             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3522                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3523                 nlmsg_cancel(skb, nlh);
3524                 return -EMSGSIZE;
3525         }
3526
3527         return nlmsg_end(skb, nlh);
3528 }
3529
3530 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3531                                 u32 pid, u32 seq, int event, unsigned int flags)
3532 {
3533         struct nlmsghdr  *nlh;
3534         u8 scope = RT_SCOPE_UNIVERSE;
3535         int ifindex = ifaca->aca_idev->dev->ifindex;
3536
3537         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3538                 scope = RT_SCOPE_SITE;
3539
3540         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3541         if (nlh == NULL)
3542                 return -EMSGSIZE;
3543
3544         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3545         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3546             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3547                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3548                 nlmsg_cancel(skb, nlh);
3549                 return -EMSGSIZE;
3550         }
3551
3552         return nlmsg_end(skb, nlh);
3553 }
3554
3555 enum addr_type_t {
3556         UNICAST_ADDR,
3557         MULTICAST_ADDR,
3558         ANYCAST_ADDR,
3559 };
3560
3561 /* called with rcu_read_lock() */
3562 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3563                           struct netlink_callback *cb, enum addr_type_t type,
3564                           int s_ip_idx, int *p_ip_idx)
3565 {
3566         struct ifmcaddr6 *ifmca;
3567         struct ifacaddr6 *ifaca;
3568         int err = 1;
3569         int ip_idx = *p_ip_idx;
3570
3571         read_lock_bh(&idev->lock);
3572         switch (type) {
3573         case UNICAST_ADDR: {
3574                 struct inet6_ifaddr *ifa;
3575
3576                 /* unicast address incl. temp addr */
3577                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3578                         if (++ip_idx < s_ip_idx)
3579                                 continue;
3580                         err = inet6_fill_ifaddr(skb, ifa,
3581                                                 NETLINK_CB(cb->skb).pid,
3582                                                 cb->nlh->nlmsg_seq,
3583                                                 RTM_NEWADDR,
3584                                                 NLM_F_MULTI);
3585                         if (err <= 0)
3586                                 break;
3587                 }
3588                 break;
3589         }
3590         case MULTICAST_ADDR:
3591                 /* multicast address */
3592                 for (ifmca = idev->mc_list; ifmca;
3593                      ifmca = ifmca->next, ip_idx++) {
3594                         if (ip_idx < s_ip_idx)
3595                                 continue;
3596                         err = inet6_fill_ifmcaddr(skb, ifmca,
3597                                                   NETLINK_CB(cb->skb).pid,
3598                                                   cb->nlh->nlmsg_seq,
3599                                                   RTM_GETMULTICAST,
3600                                                   NLM_F_MULTI);
3601                         if (err <= 0)
3602                                 break;
3603                 }
3604                 break;
3605         case ANYCAST_ADDR:
3606                 /* anycast address */
3607                 for (ifaca = idev->ac_list; ifaca;
3608                      ifaca = ifaca->aca_next, ip_idx++) {
3609                         if (ip_idx < s_ip_idx)
3610                                 continue;
3611                         err = inet6_fill_ifacaddr(skb, ifaca,
3612                                                   NETLINK_CB(cb->skb).pid,
3613                                                   cb->nlh->nlmsg_seq,
3614                                                   RTM_GETANYCAST,
3615                                                   NLM_F_MULTI);
3616                         if (err <= 0)
3617                                 break;
3618                 }
3619                 break;
3620         default:
3621                 break;
3622         }
3623         read_unlock_bh(&idev->lock);
3624         *p_ip_idx = ip_idx;
3625         return err;
3626 }
3627
3628 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3629                            enum addr_type_t type)
3630 {
3631         struct net *net = sock_net(skb->sk);
3632         int h, s_h;
3633         int idx, ip_idx;
3634         int s_idx, s_ip_idx;
3635         struct net_device *dev;
3636         struct inet6_dev *idev;
3637         struct hlist_head *head;
3638         struct hlist_node *node;
3639
3640         s_h = cb->args[0];
3641         s_idx = idx = cb->args[1];
3642         s_ip_idx = ip_idx = cb->args[2];
3643
3644         rcu_read_lock();
3645         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3646                 idx = 0;
3647                 head = &net->dev_index_head[h];
3648                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3649                         if (idx < s_idx)
3650                                 goto cont;
3651                         if (h > s_h || idx > s_idx)
3652                                 s_ip_idx = 0;
3653                         ip_idx = 0;
3654                         idev = __in6_dev_get(dev);
3655                         if (!idev)
3656                                 goto cont;
3657
3658                         if (in6_dump_addrs(idev, skb, cb, type,
3659                                            s_ip_idx, &ip_idx) <= 0)
3660                                 goto done;
3661 cont:
3662                         idx++;
3663                 }
3664         }
3665 done:
3666         rcu_read_unlock();
3667         cb->args[0] = h;
3668         cb->args[1] = idx;
3669         cb->args[2] = ip_idx;
3670
3671         return skb->len;
3672 }
3673
3674 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3675 {
3676         enum addr_type_t type = UNICAST_ADDR;
3677
3678         return inet6_dump_addr(skb, cb, type);
3679 }
3680
3681 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3682 {
3683         enum addr_type_t type = MULTICAST_ADDR;
3684
3685         return inet6_dump_addr(skb, cb, type);
3686 }
3687
3688
3689 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3690 {
3691         enum addr_type_t type = ANYCAST_ADDR;
3692
3693         return inet6_dump_addr(skb, cb, type);
3694 }
3695
3696 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3697                              void *arg)
3698 {
3699         struct net *net = sock_net(in_skb->sk);
3700         struct ifaddrmsg *ifm;
3701         struct nlattr *tb[IFA_MAX+1];
3702         struct in6_addr *addr = NULL;
3703         struct net_device *dev = NULL;
3704         struct inet6_ifaddr *ifa;
3705         struct sk_buff *skb;
3706         int err;
3707
3708         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3709         if (err < 0)
3710                 goto errout;
3711
3712         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3713         if (addr == NULL) {
3714                 err = -EINVAL;
3715                 goto errout;
3716         }
3717
3718         ifm = nlmsg_data(nlh);
3719         if (ifm->ifa_index)
3720                 dev = __dev_get_by_index(net, ifm->ifa_index);
3721
3722         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3723         if (!ifa) {
3724                 err = -EADDRNOTAVAIL;
3725                 goto errout;
3726         }
3727
3728         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3729         if (!skb) {
3730                 err = -ENOBUFS;
3731                 goto errout_ifa;
3732         }
3733
3734         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3735                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3736         if (err < 0) {
3737                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3738                 WARN_ON(err == -EMSGSIZE);
3739                 kfree_skb(skb);
3740                 goto errout_ifa;
3741         }
3742         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3743 errout_ifa:
3744         in6_ifa_put(ifa);
3745 errout:
3746         return err;
3747 }
3748
3749 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3750 {
3751         struct sk_buff *skb;
3752         struct net *net = dev_net(ifa->idev->dev);
3753         int err = -ENOBUFS;
3754
3755         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3756         if (skb == NULL)
3757                 goto errout;
3758
3759         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3760         if (err < 0) {
3761                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3762                 WARN_ON(err == -EMSGSIZE);
3763                 kfree_skb(skb);
3764                 goto errout;
3765         }
3766         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3767         return;
3768 errout:
3769         if (err < 0)
3770                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3771 }
3772
3773 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3774                                 __s32 *array, int bytes)
3775 {
3776         BUG_ON(bytes < (DEVCONF_MAX * 4));
3777
3778         memset(array, 0, bytes);
3779         array[DEVCONF_FORWARDING] = cnf->forwarding;
3780         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3781         array[DEVCONF_MTU6] = cnf->mtu6;
3782         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3783         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3784         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3785         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3786         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3787         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3788                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3789         array[DEVCONF_RTR_SOLICIT_DELAY] =
3790                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3791         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3792 #ifdef CONFIG_IPV6_PRIVACY
3793         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3794         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3795         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3796         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3797         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3798 #endif
3799         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3800         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3801         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3802 #ifdef CONFIG_IPV6_ROUTER_PREF
3803         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3804         array[DEVCONF_RTR_PROBE_INTERVAL] =
3805                 jiffies_to_msecs(cnf->rtr_probe_interval);
3806 #ifdef CONFIG_IPV6_ROUTE_INFO
3807         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3808 #endif
3809 #endif
3810         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3811         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3812 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3813         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3814 #endif
3815 #ifdef CONFIG_IPV6_MROUTE
3816         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3817 #endif
3818         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3819         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3820         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3821 }
3822
3823 static inline size_t inet6_ifla6_size(void)
3824 {
3825         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3826              + nla_total_size(sizeof(struct ifla_cacheinfo))
3827              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3828              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3829              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3830 }
3831
3832 static inline size_t inet6_if_nlmsg_size(void)
3833 {
3834         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3835                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3836                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3837                + nla_total_size(4) /* IFLA_MTU */
3838                + nla_total_size(4) /* IFLA_LINK */
3839                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3840 }
3841
3842 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3843                                       int items, int bytes)
3844 {
3845         int i;
3846         int pad = bytes - sizeof(u64) * items;
3847         BUG_ON(pad < 0);
3848
3849         /* Use put_unaligned() because stats may not be aligned for u64. */
3850         put_unaligned(items, &stats[0]);
3851         for (i = 1; i < items; i++)
3852                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3853
3854         memset(&stats[items], 0, pad);
3855 }
3856
3857 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3858                                       int items, int bytes, size_t syncpoff)
3859 {
3860         int i;
3861         int pad = bytes - sizeof(u64) * items;
3862         BUG_ON(pad < 0);
3863
3864         /* Use put_unaligned() because stats may not be aligned for u64. */
3865         put_unaligned(items, &stats[0]);
3866         for (i = 1; i < items; i++)
3867                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3868
3869         memset(&stats[items], 0, pad);
3870 }
3871
3872 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3873                              int bytes)
3874 {
3875         switch (attrtype) {
3876         case IFLA_INET6_STATS:
3877                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3878                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3879                 break;
3880         case IFLA_INET6_ICMP6STATS:
3881                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3882                 break;
3883         }
3884 }
3885
3886 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3887 {
3888         struct nlattr *nla;
3889         struct ifla_cacheinfo ci;
3890
3891         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3892
3893         ci.max_reasm_len = IPV6_MAXPLEN;
3894         ci.tstamp = cstamp_delta(idev->tstamp);
3895         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3896         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3897         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3898
3899         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3900         if (nla == NULL)
3901                 goto nla_put_failure;
3902         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3903
3904         /* XXX - MC not implemented */
3905
3906         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3907         if (nla == NULL)
3908                 goto nla_put_failure;
3909         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3910
3911         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3912         if (nla == NULL)
3913                 goto nla_put_failure;
3914         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3915
3916         return 0;
3917
3918 nla_put_failure:
3919         return -EMSGSIZE;
3920 }
3921
3922 static size_t inet6_get_link_af_size(const struct net_device *dev)
3923 {
3924         if (!__in6_dev_get(dev))
3925                 return 0;
3926
3927         return inet6_ifla6_size();
3928 }
3929
3930 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3931 {
3932         struct inet6_dev *idev = __in6_dev_get(dev);
3933
3934         if (!idev)
3935                 return -ENODATA;
3936
3937         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3938                 return -EMSGSIZE;
3939
3940         return 0;
3941 }
3942
3943 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3944                              u32 pid, u32 seq, int event, unsigned int flags)
3945 {
3946         struct net_device *dev = idev->dev;
3947         struct ifinfomsg *hdr;
3948         struct nlmsghdr *nlh;
3949         void *protoinfo;
3950
3951         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3952         if (nlh == NULL)
3953                 return -EMSGSIZE;
3954
3955         hdr = nlmsg_data(nlh);
3956         hdr->ifi_family = AF_INET6;
3957         hdr->__ifi_pad = 0;
3958         hdr->ifi_type = dev->type;
3959         hdr->ifi_index = dev->ifindex;
3960         hdr->ifi_flags = dev_get_flags(dev);
3961         hdr->ifi_change = 0;
3962
3963         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3964
3965         if (dev->addr_len)
3966                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3967
3968         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3969         if (dev->ifindex != dev->iflink)
3970                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3971
3972         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3973         if (protoinfo == NULL)
3974                 goto nla_put_failure;
3975
3976         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3977                 goto nla_put_failure;
3978
3979         nla_nest_end(skb, protoinfo);
3980         return nlmsg_end(skb, nlh);
3981
3982 nla_put_failure:
3983         nlmsg_cancel(skb, nlh);
3984         return -EMSGSIZE;
3985 }
3986
3987 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3988 {
3989         struct net *net = sock_net(skb->sk);
3990         int h, s_h;
3991         int idx = 0, s_idx;
3992         struct net_device *dev;
3993         struct inet6_dev *idev;
3994         struct hlist_head *head;
3995         struct hlist_node *node;
3996
3997         s_h = cb->args[0];
3998         s_idx = cb->args[1];
3999
4000         rcu_read_lock();
4001         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4002                 idx = 0;
4003                 head = &net->dev_index_head[h];
4004                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4005                         if (idx < s_idx)
4006                                 goto cont;
4007                         idev = __in6_dev_get(dev);
4008                         if (!idev)
4009                                 goto cont;
4010                         if (inet6_fill_ifinfo(skb, idev,
4011                                               NETLINK_CB(cb->skb).pid,
4012                                               cb->nlh->nlmsg_seq,
4013                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4014                                 goto out;
4015 cont:
4016                         idx++;
4017                 }
4018         }
4019 out:
4020         rcu_read_unlock();
4021         cb->args[1] = idx;
4022         cb->args[0] = h;
4023
4024         return skb->len;
4025 }
4026
4027 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4028 {
4029         struct sk_buff *skb;
4030         struct net *net = dev_net(idev->dev);
4031         int err = -ENOBUFS;
4032
4033         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4034         if (skb == NULL)
4035                 goto errout;
4036
4037         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4038         if (err < 0) {
4039                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4040                 WARN_ON(err == -EMSGSIZE);
4041                 kfree_skb(skb);
4042                 goto errout;
4043         }
4044         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4045         return;
4046 errout:
4047         if (err < 0)
4048                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4049 }
4050
4051 static inline size_t inet6_prefix_nlmsg_size(void)
4052 {
4053         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4054                + nla_total_size(sizeof(struct in6_addr))
4055                + nla_total_size(sizeof(struct prefix_cacheinfo));
4056 }
4057
4058 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4059                              struct prefix_info *pinfo, u32 pid, u32 seq,
4060                              int event, unsigned int flags)
4061 {
4062         struct prefixmsg *pmsg;
4063         struct nlmsghdr *nlh;
4064         struct prefix_cacheinfo ci;
4065
4066         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4067         if (nlh == NULL)
4068                 return -EMSGSIZE;
4069
4070         pmsg = nlmsg_data(nlh);
4071         pmsg->prefix_family = AF_INET6;
4072         pmsg->prefix_pad1 = 0;
4073         pmsg->prefix_pad2 = 0;
4074         pmsg->prefix_ifindex = idev->dev->ifindex;
4075         pmsg->prefix_len = pinfo->prefix_len;
4076         pmsg->prefix_type = pinfo->type;
4077         pmsg->prefix_pad3 = 0;
4078         pmsg->prefix_flags = 0;
4079         if (pinfo->onlink)
4080                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4081         if (pinfo->autoconf)
4082                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4083
4084         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4085
4086         ci.preferred_time = ntohl(pinfo->prefered);
4087         ci.valid_time = ntohl(pinfo->valid);
4088         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4089
4090         return nlmsg_end(skb, nlh);
4091
4092 nla_put_failure:
4093         nlmsg_cancel(skb, nlh);
4094         return -EMSGSIZE;
4095 }
4096
4097 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4098                          struct prefix_info *pinfo)
4099 {
4100         struct sk_buff *skb;
4101         struct net *net = dev_net(idev->dev);
4102         int err = -ENOBUFS;
4103
4104         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4105         if (skb == NULL)
4106                 goto errout;
4107
4108         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4109         if (err < 0) {
4110                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4111                 WARN_ON(err == -EMSGSIZE);
4112                 kfree_skb(skb);
4113                 goto errout;
4114         }
4115         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4116         return;
4117 errout:
4118         if (err < 0)
4119                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4120 }
4121
4122 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4123 {
4124         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4125
4126         switch (event) {
4127         case RTM_NEWADDR:
4128                 /*
4129                  * If the address was optimistic
4130                  * we inserted the route at the start of
4131                  * our DAD process, so we don't need
4132                  * to do it again
4133                  */
4134                 if (!(ifp->rt->rt6i_node))
4135                         ip6_ins_rt(ifp->rt);
4136                 if (ifp->idev->cnf.forwarding)
4137                         addrconf_join_anycast(ifp);
4138                 break;
4139         case RTM_DELADDR:
4140                 if (ifp->idev->cnf.forwarding)
4141                         addrconf_leave_anycast(ifp);
4142                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4143                 dst_hold(&ifp->rt->dst);
4144
4145                 if (ip6_del_rt(ifp->rt))
4146                         dst_free(&ifp->rt->dst);
4147                 break;
4148         }
4149 }
4150
4151 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4152 {
4153         rcu_read_lock_bh();
4154         if (likely(ifp->idev->dead == 0))
4155                 __ipv6_ifa_notify(event, ifp);
4156         rcu_read_unlock_bh();
4157 }
4158
4159 #ifdef CONFIG_SYSCTL
4160
4161 static
4162 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4163                            void __user *buffer, size_t *lenp, loff_t *ppos)
4164 {
4165         int *valp = ctl->data;
4166         int val = *valp;
4167         loff_t pos = *ppos;
4168         int ret;
4169
4170         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4171
4172         if (write)
4173                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4174         if (ret)
4175                 *ppos = pos;
4176         return ret;
4177 }
4178
4179 static void dev_disable_change(struct inet6_dev *idev)
4180 {
4181         if (!idev || !idev->dev)
4182                 return;
4183
4184         if (idev->cnf.disable_ipv6)
4185                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4186         else
4187                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4188 }
4189
4190 static void addrconf_disable_change(struct net *net, __s32 newf)
4191 {
4192         struct net_device *dev;
4193         struct inet6_dev *idev;
4194
4195         rcu_read_lock();
4196         for_each_netdev_rcu(net, dev) {
4197                 idev = __in6_dev_get(dev);
4198                 if (idev) {
4199                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4200                         idev->cnf.disable_ipv6 = newf;
4201                         if (changed)
4202                                 dev_disable_change(idev);
4203                 }
4204         }
4205         rcu_read_unlock();
4206 }
4207
4208 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4209 {
4210         struct net *net;
4211
4212         net = (struct net *)table->extra2;
4213
4214         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4215                 return 0;
4216
4217         if (!rtnl_trylock()) {
4218                 /* Restore the original values before restarting */
4219                 *p = old;
4220                 return restart_syscall();
4221         }
4222
4223         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4224                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4225                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4226                 addrconf_disable_change(net, newf);
4227         } else if ((!*p) ^ (!old))
4228                 dev_disable_change((struct inet6_dev *)table->extra1);
4229
4230         rtnl_unlock();
4231         return 0;
4232 }
4233
4234 static
4235 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4236                             void __user *buffer, size_t *lenp, loff_t *ppos)
4237 {
4238         int *valp = ctl->data;
4239         int val = *valp;
4240         loff_t pos = *ppos;
4241         int ret;
4242
4243         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4244
4245         if (write)
4246                 ret = addrconf_disable_ipv6(ctl, valp, val);
4247         if (ret)
4248                 *ppos = pos;
4249         return ret;
4250 }
4251
4252 static struct addrconf_sysctl_table
4253 {
4254         struct ctl_table_header *sysctl_header;
4255         ctl_table addrconf_vars[DEVCONF_MAX+1];
4256         char *dev_name;
4257 } addrconf_sysctl __read_mostly = {
4258         .sysctl_header = NULL,
4259         .addrconf_vars = {
4260                 {
4261                         .procname       = "forwarding",
4262                         .data           = &ipv6_devconf.forwarding,
4263                         .maxlen         = sizeof(int),
4264                         .mode           = 0644,
4265                         .proc_handler   = addrconf_sysctl_forward,
4266                 },
4267                 {
4268                         .procname       = "hop_limit",
4269                         .data           = &ipv6_devconf.hop_limit,
4270                         .maxlen         = sizeof(int),
4271                         .mode           = 0644,
4272                         .proc_handler   = proc_dointvec,
4273                 },
4274                 {
4275                         .procname       = "mtu",
4276                         .data           = &ipv6_devconf.mtu6,
4277                         .maxlen         = sizeof(int),
4278                         .mode           = 0644,
4279                         .proc_handler   = proc_dointvec,
4280                 },
4281                 {
4282                         .procname       = "accept_ra",
4283                         .data           = &ipv6_devconf.accept_ra,
4284                         .maxlen         = sizeof(int),
4285                         .mode           = 0644,
4286                         .proc_handler   = proc_dointvec,
4287                 },
4288                 {
4289                         .procname       = "accept_redirects",
4290                         .data           = &ipv6_devconf.accept_redirects,
4291                         .maxlen         = sizeof(int),
4292                         .mode           = 0644,
4293                         .proc_handler   = proc_dointvec,
4294                 },
4295                 {
4296                         .procname       = "autoconf",
4297                         .data           = &ipv6_devconf.autoconf,
4298                         .maxlen         = sizeof(int),
4299                         .mode           = 0644,
4300                         .proc_handler   = proc_dointvec,
4301                 },
4302                 {
4303                         .procname       = "dad_transmits",
4304                         .data           = &ipv6_devconf.dad_transmits,
4305                         .maxlen         = sizeof(int),
4306                         .mode           = 0644,
4307                         .proc_handler   = proc_dointvec,
4308                 },
4309                 {
4310                         .procname       = "router_solicitations",
4311                         .data           = &ipv6_devconf.rtr_solicits,
4312                         .maxlen         = sizeof(int),
4313                         .mode           = 0644,
4314                         .proc_handler   = proc_dointvec,
4315                 },
4316                 {
4317                         .procname       = "router_solicitation_interval",
4318                         .data           = &ipv6_devconf.rtr_solicit_interval,
4319                         .maxlen         = sizeof(int),
4320                         .mode           = 0644,
4321                         .proc_handler   = proc_dointvec_jiffies,
4322                 },
4323                 {
4324                         .procname       = "router_solicitation_delay",
4325                         .data           = &ipv6_devconf.rtr_solicit_delay,
4326                         .maxlen         = sizeof(int),
4327                         .mode           = 0644,
4328                         .proc_handler   = proc_dointvec_jiffies,
4329                 },
4330                 {
4331                         .procname       = "force_mld_version",
4332                         .data           = &ipv6_devconf.force_mld_version,
4333                         .maxlen         = sizeof(int),
4334                         .mode           = 0644,
4335                         .proc_handler   = proc_dointvec,
4336                 },
4337 #ifdef CONFIG_IPV6_PRIVACY
4338                 {
4339                         .procname       = "use_tempaddr",
4340                         .data           = &ipv6_devconf.use_tempaddr,
4341                         .maxlen         = sizeof(int),
4342                         .mode           = 0644,
4343                         .proc_handler   = proc_dointvec,
4344                 },
4345                 {
4346                         .procname       = "temp_valid_lft",
4347                         .data           = &ipv6_devconf.temp_valid_lft,
4348                         .maxlen         = sizeof(int),
4349                         .mode           = 0644,
4350                         .proc_handler   = proc_dointvec,
4351                 },
4352                 {
4353                         .procname       = "temp_prefered_lft",
4354                         .data           = &ipv6_devconf.temp_prefered_lft,
4355                         .maxlen         = sizeof(int),
4356                         .mode           = 0644,
4357                         .proc_handler   = proc_dointvec,
4358                 },
4359                 {
4360                         .procname       = "regen_max_retry",
4361                         .data           = &ipv6_devconf.regen_max_retry,
4362                         .maxlen         = sizeof(int),
4363                         .mode           = 0644,
4364                         .proc_handler   = proc_dointvec,
4365                 },
4366                 {
4367                         .procname       = "max_desync_factor",
4368                         .data           = &ipv6_devconf.max_desync_factor,
4369                         .maxlen         = sizeof(int),
4370                         .mode           = 0644,
4371                         .proc_handler   = proc_dointvec,
4372                 },
4373 #endif
4374                 {
4375                         .procname       = "max_addresses",
4376                         .data           = &ipv6_devconf.max_addresses,
4377                         .maxlen         = sizeof(int),
4378                         .mode           = 0644,
4379                         .proc_handler   = proc_dointvec,
4380                 },
4381                 {
4382                         .procname       = "accept_ra_defrtr",
4383                         .data           = &ipv6_devconf.accept_ra_defrtr,
4384                         .maxlen         = sizeof(int),
4385                         .mode           = 0644,
4386                         .proc_handler   = proc_dointvec,
4387                 },
4388                 {
4389                         .procname       = "accept_ra_pinfo",
4390                         .data           = &ipv6_devconf.accept_ra_pinfo,
4391                         .maxlen         = sizeof(int),
4392                         .mode           = 0644,
4393                         .proc_handler   = proc_dointvec,
4394                 },
4395 #ifdef CONFIG_IPV6_ROUTER_PREF
4396                 {
4397                         .procname       = "accept_ra_rtr_pref",
4398                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4399                         .maxlen         = sizeof(int),
4400                         .mode           = 0644,
4401                         .proc_handler   = proc_dointvec,
4402                 },
4403                 {
4404                         .procname       = "router_probe_interval",
4405                         .data           = &ipv6_devconf.rtr_probe_interval,
4406                         .maxlen         = sizeof(int),
4407                         .mode           = 0644,
4408                         .proc_handler   = proc_dointvec_jiffies,
4409                 },
4410 #ifdef CONFIG_IPV6_ROUTE_INFO
4411                 {
4412                         .procname       = "accept_ra_rt_info_max_plen",
4413                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4414                         .maxlen         = sizeof(int),
4415                         .mode           = 0644,
4416                         .proc_handler   = proc_dointvec,
4417                 },
4418 #endif
4419 #endif
4420                 {
4421                         .procname       = "proxy_ndp",
4422                         .data           = &ipv6_devconf.proxy_ndp,
4423                         .maxlen         = sizeof(int),
4424                         .mode           = 0644,
4425                         .proc_handler   = proc_dointvec,
4426                 },
4427                 {
4428                         .procname       = "accept_source_route",
4429                         .data           = &ipv6_devconf.accept_source_route,
4430                         .maxlen         = sizeof(int),
4431                         .mode           = 0644,
4432                         .proc_handler   = proc_dointvec,
4433                 },
4434 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4435                 {
4436                         .procname       = "optimistic_dad",
4437                         .data           = &ipv6_devconf.optimistic_dad,
4438                         .maxlen         = sizeof(int),
4439                         .mode           = 0644,
4440                         .proc_handler   = proc_dointvec,
4441
4442                 },
4443 #endif
4444 #ifdef CONFIG_IPV6_MROUTE
4445                 {
4446                         .procname       = "mc_forwarding",
4447                         .data           = &ipv6_devconf.mc_forwarding,
4448                         .maxlen         = sizeof(int),
4449                         .mode           = 0444,
4450                         .proc_handler   = proc_dointvec,
4451                 },
4452 #endif
4453                 {
4454                         .procname       = "disable_ipv6",
4455                         .data           = &ipv6_devconf.disable_ipv6,
4456                         .maxlen         = sizeof(int),
4457                         .mode           = 0644,
4458                         .proc_handler   = addrconf_sysctl_disable,
4459                 },
4460                 {
4461                         .procname       = "accept_dad",
4462                         .data           = &ipv6_devconf.accept_dad,
4463                         .maxlen         = sizeof(int),
4464                         .mode           = 0644,
4465                         .proc_handler   = proc_dointvec,
4466                 },
4467                 {
4468                         .procname       = "force_tllao",
4469                         .data           = &ipv6_devconf.force_tllao,
4470                         .maxlen         = sizeof(int),
4471                         .mode           = 0644,
4472                         .proc_handler   = proc_dointvec
4473                 },
4474                 {
4475                         /* sentinel */
4476                 }
4477         },
4478 };
4479
4480 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4481                 struct inet6_dev *idev, struct ipv6_devconf *p)
4482 {
4483         int i;
4484         struct addrconf_sysctl_table *t;
4485
4486 #define ADDRCONF_CTL_PATH_DEV   3
4487
4488         struct ctl_path addrconf_ctl_path[] = {
4489                 { .procname = "net", },
4490                 { .procname = "ipv6", },
4491                 { .procname = "conf", },
4492                 { /* to be set */ },
4493                 { },
4494         };
4495
4496
4497         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4498         if (t == NULL)
4499                 goto out;
4500
4501         for (i = 0; t->addrconf_vars[i].data; i++) {
4502                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4503                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4504                 t->addrconf_vars[i].extra2 = net;
4505         }
4506
4507         /*
4508          * Make a copy of dev_name, because '.procname' is regarded as const
4509          * by sysctl and we wouldn't want anyone to change it under our feet
4510          * (see SIOCSIFNAME).
4511          */
4512         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4513         if (!t->dev_name)
4514                 goto free;
4515
4516         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4517
4518         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4519                         t->addrconf_vars);
4520         if (t->sysctl_header == NULL)
4521                 goto free_procname;
4522
4523         p->sysctl = t;
4524         return 0;
4525
4526 free_procname:
4527         kfree(t->dev_name);
4528 free:
4529         kfree(t);
4530 out:
4531         return -ENOBUFS;
4532 }
4533
4534 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4535 {
4536         struct addrconf_sysctl_table *t;
4537
4538         if (p->sysctl == NULL)
4539                 return;
4540
4541         t = p->sysctl;
4542         p->sysctl = NULL;
4543         unregister_sysctl_table(t->sysctl_header);
4544         kfree(t->dev_name);
4545         kfree(t);
4546 }
4547
4548 static void addrconf_sysctl_register(struct inet6_dev *idev)
4549 {
4550         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4551                               &ndisc_ifinfo_sysctl_change);
4552         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4553                                         idev, &idev->cnf);
4554 }
4555
4556 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4557 {
4558         __addrconf_sysctl_unregister(&idev->cnf);
4559         neigh_sysctl_unregister(idev->nd_parms);
4560 }
4561
4562
4563 #endif
4564
4565 static int __net_init addrconf_init_net(struct net *net)
4566 {
4567         int err;
4568         struct ipv6_devconf *all, *dflt;
4569
4570         err = -ENOMEM;
4571         all = &ipv6_devconf;
4572         dflt = &ipv6_devconf_dflt;
4573
4574         if (!net_eq(net, &init_net)) {
4575                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4576                 if (all == NULL)
4577                         goto err_alloc_all;
4578
4579                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4580                 if (dflt == NULL)
4581                         goto err_alloc_dflt;
4582         } else {
4583                 /* these will be inherited by all namespaces */
4584                 dflt->autoconf = ipv6_defaults.autoconf;
4585                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4586         }
4587
4588         net->ipv6.devconf_all = all;
4589         net->ipv6.devconf_dflt = dflt;
4590
4591 #ifdef CONFIG_SYSCTL
4592         err = __addrconf_sysctl_register(net, "all", NULL, all);
4593         if (err < 0)
4594                 goto err_reg_all;
4595
4596         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4597         if (err < 0)
4598                 goto err_reg_dflt;
4599 #endif
4600         return 0;
4601
4602 #ifdef CONFIG_SYSCTL
4603 err_reg_dflt:
4604         __addrconf_sysctl_unregister(all);
4605 err_reg_all:
4606         kfree(dflt);
4607 #endif
4608 err_alloc_dflt:
4609         kfree(all);
4610 err_alloc_all:
4611         return err;
4612 }
4613
4614 static void __net_exit addrconf_exit_net(struct net *net)
4615 {
4616 #ifdef CONFIG_SYSCTL
4617         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4618         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4619 #endif
4620         if (!net_eq(net, &init_net)) {
4621                 kfree(net->ipv6.devconf_dflt);
4622                 kfree(net->ipv6.devconf_all);
4623         }
4624 }
4625
4626 static struct pernet_operations addrconf_ops = {
4627         .init = addrconf_init_net,
4628         .exit = addrconf_exit_net,
4629 };
4630
4631 /*
4632  *      Device notifier
4633  */
4634
4635 int register_inet6addr_notifier(struct notifier_block *nb)
4636 {
4637         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4638 }
4639 EXPORT_SYMBOL(register_inet6addr_notifier);
4640
4641 int unregister_inet6addr_notifier(struct notifier_block *nb)
4642 {
4643         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4644 }
4645 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4646
4647 static struct rtnl_af_ops inet6_ops = {
4648         .family           = AF_INET6,
4649         .fill_link_af     = inet6_fill_link_af,
4650         .get_link_af_size = inet6_get_link_af_size,
4651 };
4652
4653 /*
4654  *      Init / cleanup code
4655  */
4656
4657 int __init addrconf_init(void)
4658 {
4659         int i, err;
4660
4661         err = ipv6_addr_label_init();
4662         if (err < 0) {
4663                 printk(KERN_CRIT "IPv6 Addrconf:"
4664                        " cannot initialize default policy table: %d.\n", err);
4665                 goto out;
4666         }
4667
4668         err = register_pernet_subsys(&addrconf_ops);
4669         if (err < 0)
4670                 goto out_addrlabel;
4671
4672         /* The addrconf netdev notifier requires that loopback_dev
4673          * has it's ipv6 private information allocated and setup
4674          * before it can bring up and give link-local addresses
4675          * to other devices which are up.
4676          *
4677          * Unfortunately, loopback_dev is not necessarily the first
4678          * entry in the global dev_base list of net devices.  In fact,
4679          * it is likely to be the very last entry on that list.
4680          * So this causes the notifier registry below to try and
4681          * give link-local addresses to all devices besides loopback_dev
4682          * first, then loopback_dev, which cases all the non-loopback_dev
4683          * devices to fail to get a link-local address.
4684          *
4685          * So, as a temporary fix, allocate the ipv6 structure for
4686          * loopback_dev first by hand.
4687          * Longer term, all of the dependencies ipv6 has upon the loopback
4688          * device and it being up should be removed.
4689          */
4690         rtnl_lock();
4691         if (!ipv6_add_dev(init_net.loopback_dev))
4692                 err = -ENOMEM;
4693         rtnl_unlock();
4694         if (err)
4695                 goto errlo;
4696
4697         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4698                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4699
4700         register_netdevice_notifier(&ipv6_dev_notf);
4701
4702         addrconf_verify(0);
4703
4704         err = rtnl_af_register(&inet6_ops);
4705         if (err < 0)
4706                 goto errout_af;
4707
4708         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4709         if (err < 0)
4710                 goto errout;
4711
4712         /* Only the first call to __rtnl_register can fail */
4713         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4714         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4715         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4716         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4717         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4718
4719         ipv6_addr_label_rtnl_register();
4720
4721         return 0;
4722 errout:
4723         rtnl_af_unregister(&inet6_ops);
4724 errout_af:
4725         unregister_netdevice_notifier(&ipv6_dev_notf);
4726 errlo:
4727         unregister_pernet_subsys(&addrconf_ops);
4728 out_addrlabel:
4729         ipv6_addr_label_cleanup();
4730 out:
4731         return err;
4732 }
4733
4734 void addrconf_cleanup(void)
4735 {
4736         struct net_device *dev;
4737         int i;
4738
4739         unregister_netdevice_notifier(&ipv6_dev_notf);
4740         unregister_pernet_subsys(&addrconf_ops);
4741         ipv6_addr_label_cleanup();
4742
4743         rtnl_lock();
4744
4745         __rtnl_af_unregister(&inet6_ops);
4746
4747         /* clean dev list */
4748         for_each_netdev(&init_net, dev) {
4749                 if (__in6_dev_get(dev) == NULL)
4750                         continue;
4751                 addrconf_ifdown(dev, 1);
4752         }
4753         addrconf_ifdown(init_net.loopback_dev, 2);
4754
4755         /*
4756          *      Check hash table.
4757          */
4758         spin_lock_bh(&addrconf_hash_lock);
4759         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4760                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4761         spin_unlock_bh(&addrconf_hash_lock);
4762
4763         del_timer(&addr_chk_timer);
4764         rtnl_unlock();
4765 }