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         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 293                                         GFP_KERNEL);
 294         if (!idev->stats.icmpv6dev)
 295                 goto err_icmp;
 296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 297                                            GFP_KERNEL);
 298         if (!idev->stats.icmpv6msgdev)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         kfree(idev->stats.icmpv6dev);
 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         kfree(idev->stats.icmpv6msgdev);
 314         kfree(idev->stats.icmpv6dev);
 315         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 316 }
 317
 318 /* Nobody refers to this device, we may destroy it. */
 319
 320 void in6_dev_finish_destroy(struct inet6_dev *idev)
 321 {
 322         struct net_device *dev = idev->dev;
 323
 324         WARN_ON(!list_empty(&idev->addr_list));
 325         WARN_ON(idev->mc_list != NULL);
 326
 327 #ifdef NET_REFCNT_DEBUG
 328         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 329 #endif
 330         dev_put(dev);
 331         if (!idev->dead) {
 332                 pr_warning("Freeing alive inet6 device %p\n", idev);
 333                 return;
 334         }
 335         snmp6_free_dev(idev);
 336         kfree_rcu(idev, rcu);
 337 }
 338
 339 EXPORT_SYMBOL(in6_dev_finish_destroy);
 340
 341 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 342 {
 343         struct inet6_dev *ndev;
 344
 345         ASSERT_RTNL();
 346
 347         if (dev->mtu < IPV6_MIN_MTU)
 348                 return NULL;
 349
 350         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 351
 352         if (ndev == NULL)
 353                 return NULL;
 354
 355         rwlock_init(&ndev->lock);
 356         ndev->dev = dev;
 357         INIT_LIST_HEAD(&ndev->addr_list);
 358
 359         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 360         ndev->cnf.mtu6 = dev->mtu;
 361         ndev->cnf.sysctl = NULL;
 362         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 363         if (ndev->nd_parms == NULL) {
 364                 kfree(ndev);
 365                 return NULL;
 366         }
 367         if (ndev->cnf.forwarding)
 368                 dev_disable_lro(dev);
 369         /* We refer to the device */
 370         dev_hold(dev);
 371
 372         if (snmp6_alloc_dev(ndev) < 0) {
 373                 ADBG((KERN_WARNING
 374                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 375                         __func__, dev->name));
 376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 377                 dev_put(dev);
 378                 kfree(ndev);
 379                 return NULL;
 380         }
 381
 382         if (snmp6_register_dev(ndev) < 0) {
 383                 ADBG((KERN_WARNING
 384                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 385                         __func__, dev->name));
 386                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 387                 ndev->dead = 1;
 388                 in6_dev_finish_destroy(ndev);
 389                 return NULL;
 390         }
 391
 392         /* One reference from device.  We must do this before
 393          * we invoke __ipv6_regen_rndid().
 394          */
 395         in6_dev_hold(ndev);
 396
 397         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 398                 ndev->cnf.accept_dad = -1;
 399
 400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 401         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 402                 printk(KERN_INFO
 403                        "%s: Disabled Multicast RS\n",
 404                        dev->name);
 405                 ndev->cnf.rtr_solicits = 0;
 406         }
 407 #endif
 408
 409 #ifdef CONFIG_IPV6_PRIVACY
 410         INIT_LIST_HEAD(&ndev->tempaddr_list);
 411         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 412         if ((dev->flags&IFF_LOOPBACK) ||
 413             dev->type == ARPHRD_TUNNEL ||
 414             dev->type == ARPHRD_TUNNEL6 ||
 415             dev->type == ARPHRD_SIT ||
 416             dev->type == ARPHRD_NONE) {
 417                 ndev->cnf.use_tempaddr = -1;
 418         } else {
 419                 in6_dev_hold(ndev);
 420                 ipv6_regen_rndid((unsigned long) ndev);
 421         }
 422 #endif
 423
 424         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 425                 ndev->if_flags |= IF_READY;
 426
 427         ipv6_mc_init_dev(ndev);
 428         ndev->tstamp = jiffies;
 429         addrconf_sysctl_register(ndev);
 430         /* protected by rtnl_lock */
 431         rcu_assign_pointer(dev->ip6_ptr, ndev);
 432
 433         /* Join all-node multicast group */
 434         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 435
 436         return ndev;
 437 }
 438
 439 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 440 {
 441         struct inet6_dev *idev;
 442
 443         ASSERT_RTNL();
 444
 445         idev = __in6_dev_get(dev);
 446         if (!idev) {
 447                 idev = ipv6_add_dev(dev);
 448                 if (!idev)
 449                         return NULL;
 450         }
 451
 452         if (dev->flags&IFF_UP)
 453                 ipv6_mc_up(idev);
 454         return idev;
 455 }
 456
 457 #ifdef CONFIG_SYSCTL
 458 static void dev_forward_change(struct inet6_dev *idev)
 459 {
 460         struct net_device *dev;
 461         struct inet6_ifaddr *ifa;
 462
 463         if (!idev)
 464                 return;
 465         dev = idev->dev;
 466         if (idev->cnf.forwarding)
 467                 dev_disable_lro(dev);
 468         if (dev && (dev->flags & IFF_MULTICAST)) {
 469                 if (idev->cnf.forwarding)
 470                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 471                 else
 472                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 473         }
 474
 475         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 476                 if (ifa->flags&IFA_F_TENTATIVE)
 477                         continue;
 478                 if (idev->cnf.forwarding)
 479                         addrconf_join_anycast(ifa);
 480                 else
 481                         addrconf_leave_anycast(ifa);
 482         }
 483 }
 484
 485
 486 static void addrconf_forward_change(struct net *net, __s32 newf)
 487 {
 488         struct net_device *dev;
 489         struct inet6_dev *idev;
 490
 491         rcu_read_lock();
 492         for_each_netdev_rcu(net, dev) {
 493                 idev = __in6_dev_get(dev);
 494                 if (idev) {
 495                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 496                         idev->cnf.forwarding = newf;
 497                         if (changed)
 498                                 dev_forward_change(idev);
 499                 }
 500         }
 501         rcu_read_unlock();
 502 }
 503
 504 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 505 {
 506         struct net *net;
 507
 508         net = (struct net *)table->extra2;
 509         if (p == &net->ipv6.devconf_dflt->forwarding)
 510                 return 0;
 511
 512         if (!rtnl_trylock()) {
 513                 /* Restore the original values before restarting */
 514                 *p = old;
 515                 return restart_syscall();
 516         }
 517
 518         if (p == &net->ipv6.devconf_all->forwarding) {
 519                 __s32 newf = net->ipv6.devconf_all->forwarding;
 520                 net->ipv6.devconf_dflt->forwarding = newf;
 521                 addrconf_forward_change(net, newf);
 522         } else if ((!*p) ^ (!old))
 523                 dev_forward_change((struct inet6_dev *)table->extra1);
 524         rtnl_unlock();
 525
 526         if (*p)
 527                 rt6_purge_dflt_routers(net);
 528         return 1;
 529 }
 530 #endif
 531
 532 /* Nobody refers to this ifaddr, destroy it */
 533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 534 {
 535         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 536
 537 #ifdef NET_REFCNT_DEBUG
 538         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 539 #endif
 540
 541         in6_dev_put(ifp->idev);
 542
 543         if (del_timer(&ifp->timer))
 544                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 545
 546         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 547                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 548                 return;
 549         }
 550         dst_release(&ifp->rt->dst);
 551
 552         kfree_rcu(ifp, rcu);
 553 }
 554
 555 static void
 556 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 557 {
 558         struct list_head *p;
 559         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 560
 561         /*
 562          * Each device address list is sorted in order of scope -
 563          * global before linklocal.
 564          */
 565         list_for_each(p, &idev->addr_list) {
 566                 struct inet6_ifaddr *ifa
 567                         = list_entry(p, struct inet6_ifaddr, if_list);
 568                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 569                         break;
 570         }
 571
 572         list_add_tail(&ifp->if_list, p);
 573 }
 574
 575 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 576 {
 577         /*
 578          * We perform the hash function over the last 64 bits of the address
 579          * This will include the IEEE address token on links that support it.
 580          */
 581         return jhash_2words((__force u32)addr->s6_addr32[2],
 582                             (__force u32)addr->s6_addr32[3], 0)
 583                 & (IN6_ADDR_HSIZE - 1);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, 0);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ipv6_addr_copy(&ifa->addr, addr);
 639
 640         spin_lock_init(&ifa->lock);
 641         spin_lock_init(&ifa->state_lock);
 642         init_timer(&ifa->timer);
 643         INIT_HLIST_NODE(&ifa->addr_lst);
 644         ifa->timer.data = (unsigned long) ifa;
 645         ifa->scope = scope;
 646         ifa->prefix_len = pfxlen;
 647         ifa->flags = flags | IFA_F_TENTATIVE;
 648         ifa->cstamp = ifa->tstamp = jiffies;
 649
 650         ifa->rt = rt;
 651
 652         /*
 653          * part one of RFC 4429, section 3.3
 654          * We should not configure an address as
 655          * optimistic if we do not yet know the link
 656          * layer address of our nexhop router
 657          */
 658
 659         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 660                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 661
 662         ifa->idev = idev;
 663         in6_dev_hold(idev);
 664         /* For caller */
 665         in6_ifa_hold(ifa);
 666
 667         /* Add to big hash table */
 668         hash = ipv6_addr_hash(addr);
 669
 670         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 671         spin_unlock(&addrconf_hash_lock);
 672
 673         write_lock(&idev->lock);
 674         /* Add to inet6_dev unicast addr list. */
 675         ipv6_link_dev_addr(idev, ifa);
 676
 677 #ifdef CONFIG_IPV6_PRIVACY
 678         if (ifa->flags&IFA_F_TEMPORARY) {
 679                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 680                 in6_ifa_hold(ifa);
 681         }
 682 #endif
 683
 684         in6_ifa_hold(ifa);
 685         write_unlock(&idev->lock);
 686 out2:
 687         rcu_read_unlock_bh();
 688
 689         if (likely(err == 0))
 690                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 691         else {
 692                 kfree(ifa);
 693                 ifa = ERR_PTR(err);
 694         }
 695
 696         return ifa;
 697 out:
 698         spin_unlock(&addrconf_hash_lock);
 699         goto out2;
 700 }
 701
 702 /* This function wants to get referenced ifp and releases it before return */
 703
 704 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 705 {
 706         struct inet6_ifaddr *ifa, *ifn;
 707         struct inet6_dev *idev = ifp->idev;
 708         int state;
 709         int deleted = 0, onlink = 0;
 710         unsigned long expires = jiffies;
 711
 712         spin_lock_bh(&ifp->state_lock);
 713         state = ifp->state;
 714         ifp->state = INET6_IFADDR_STATE_DEAD;
 715         spin_unlock_bh(&ifp->state_lock);
 716
 717         if (state == INET6_IFADDR_STATE_DEAD)
 718                 goto out;
 719
 720         spin_lock_bh(&addrconf_hash_lock);
 721         hlist_del_init_rcu(&ifp->addr_lst);
 722         spin_unlock_bh(&addrconf_hash_lock);
 723
 724         write_lock_bh(&idev->lock);
 725 #ifdef CONFIG_IPV6_PRIVACY
 726         if (ifp->flags&IFA_F_TEMPORARY) {
 727                 list_del(&ifp->tmp_list);
 728                 if (ifp->ifpub) {
 729                         in6_ifa_put(ifp->ifpub);
 730                         ifp->ifpub = NULL;
 731                 }
 732                 __in6_ifa_put(ifp);
 733         }
 734 #endif
 735
 736         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 737                 if (ifa == ifp) {
 738                         list_del_init(&ifp->if_list);
 739                         __in6_ifa_put(ifp);
 740
 741                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 742                                 break;
 743                         deleted = 1;
 744                         continue;
 745                 } else if (ifp->flags & IFA_F_PERMANENT) {
 746                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 747                                               ifp->prefix_len)) {
 748                                 if (ifa->flags & IFA_F_PERMANENT) {
 749                                         onlink = 1;
 750                                         if (deleted)
 751                                                 break;
 752                                 } else {
 753                                         unsigned long lifetime;
 754
 755                                         if (!onlink)
 756                                                 onlink = -1;
 757
 758                                         spin_lock(&ifa->lock);
 759
 760                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 761                                         /*
 762                                          * Note: Because this address is
 763                                          * not permanent, lifetime <
 764                                          * LONG_MAX / HZ here.
 765                                          */
 766                                         if (time_before(expires,
 767                                                         ifa->tstamp + lifetime * HZ))
 768                                                 expires = ifa->tstamp + lifetime * HZ;
 769                                         spin_unlock(&ifa->lock);
 770                                 }
 771                         }
 772                 }
 773         }
 774         write_unlock_bh(&idev->lock);
 775
 776         addrconf_del_timer(ifp);
 777
 778         ipv6_ifa_notify(RTM_DELADDR, ifp);
 779
 780         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 781
 782         /*
 783          * Purge or update corresponding prefix
 784          *
 785          * 1) we don't purge prefix here if address was not permanent.
 786          *    prefix is managed by its own lifetime.
 787          * 2) if there're no addresses, delete prefix.
 788          * 3) if there're still other permanent address(es),
 789          *    corresponding prefix is still permanent.
 790          * 4) otherwise, update prefix lifetime to the
 791          *    longest valid lifetime among the corresponding
 792          *    addresses on the device.
 793          *    Note: subsequent RA will update lifetime.
 794          *
 795          * --yoshfuji
 796          */
 797         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 798                 struct in6_addr prefix;
 799                 struct rt6_info *rt;
 800                 struct net *net = dev_net(ifp->idev->dev);
 801                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 802                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 803
 804                 if (rt && addrconf_is_prefix_route(rt)) {
 805                         if (onlink == 0) {
 806                                 ip6_del_rt(rt);
 807                                 rt = NULL;
 808                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 809                                 rt->rt6i_expires = expires;
 810                                 rt->rt6i_flags |= RTF_EXPIRES;
 811                         }
 812                 }
 813                 dst_release(&rt->dst);
 814         }
 815
 816         /* clean up prefsrc entries */
 817         rt6_remove_prefsrc(ifp);
 818 out:
 819         in6_ifa_put(ifp);
 820 }
 821
 822 #ifdef CONFIG_IPV6_PRIVACY
 823 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 824 {
 825         struct inet6_dev *idev = ifp->idev;
 826         struct in6_addr addr, *tmpaddr;
 827         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 828         unsigned long regen_advance;
 829         int tmp_plen;
 830         int ret = 0;
 831         int max_addresses;
 832         u32 addr_flags;
 833
 834         write_lock(&idev->lock);
 835         if (ift) {
 836                 spin_lock_bh(&ift->lock);
 837                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 838                 spin_unlock_bh(&ift->lock);
 839                 tmpaddr = &addr;
 840         } else {
 841                 tmpaddr = NULL;
 842         }
 843 retry:
 844         in6_dev_hold(idev);
 845         if (idev->cnf.use_tempaddr <= 0) {
 846                 write_unlock(&idev->lock);
 847                 printk(KERN_INFO
 848                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 849                 in6_dev_put(idev);
 850                 ret = -1;
 851                 goto out;
 852         }
 853         spin_lock_bh(&ifp->lock);
 854         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 855                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 856                 spin_unlock_bh(&ifp->lock);
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_WARNING
 859                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         in6_ifa_hold(ifp);
 865         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 866         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 871                 in6_ifa_put(ifp);
 872                 in6_dev_put(idev);
 873                 ret = -1;
 874                 goto out;
 875         }
 876         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 877         age = (jiffies - ifp->tstamp) / HZ;
 878         tmp_valid_lft = min_t(__u32,
 879                               ifp->valid_lft,
 880                               idev->cnf.temp_valid_lft + age);
 881         tmp_prefered_lft = min_t(__u32,
 882                                  ifp->prefered_lft,
 883                                  idev->cnf.temp_prefered_lft + age -
 884                                  idev->cnf.max_desync_factor);
 885         tmp_plen = ifp->prefix_len;
 886         max_addresses = idev->cnf.max_addresses;
 887         tmp_cstamp = ifp->cstamp;
 888         tmp_tstamp = ifp->tstamp;
 889         spin_unlock_bh(&ifp->lock);
 890
 891         regen_advance = idev->cnf.regen_max_retry *
 892                         idev->cnf.dad_transmits *
 893                         idev->nd_parms->retrans_time / HZ;
 894         write_unlock(&idev->lock);
 895
 896         /* A temporary address is created only if this calculated Preferred
 897          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 898          * an implementation must not create a temporary address with a zero
 899          * Preferred Lifetime.
 900          */
 901         if (tmp_prefered_lft <= regen_advance) {
 902                 in6_ifa_put(ifp);
 903                 in6_dev_put(idev);
 904                 ret = -1;
 905                 goto out;
 906         }
 907
 908         addr_flags = IFA_F_TEMPORARY;
 909         /* set in addrconf_prefix_rcv() */
 910         if (ifp->flags & IFA_F_OPTIMISTIC)
 911                 addr_flags |= IFA_F_OPTIMISTIC;
 912
 913         ift = !max_addresses ||
 914               ipv6_count_addresses(idev) < max_addresses ?
 915                 ipv6_add_addr(idev, &addr, tmp_plen,
 916                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 917                               addr_flags) : NULL;
 918         if (!ift || IS_ERR(ift)) {
 919                 in6_ifa_put(ifp);
 920                 in6_dev_put(idev);
 921                 printk(KERN_INFO
 922                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 923                 tmpaddr = &addr;
 924                 write_lock(&idev->lock);
 925                 goto retry;
 926         }
 927
 928         spin_lock_bh(&ift->lock);
 929         ift->ifpub = ifp;
 930         ift->valid_lft = tmp_valid_lft;
 931         ift->prefered_lft = tmp_prefered_lft;
 932         ift->cstamp = tmp_cstamp;
 933         ift->tstamp = tmp_tstamp;
 934         spin_unlock_bh(&ift->lock);
 935
 936         addrconf_dad_start(ift, 0);
 937         in6_ifa_put(ift);
 938         in6_dev_put(idev);
 939 out:
 940         return ret;
 941 }
 942 #endif
 943
 944 /*
 945  *      Choose an appropriate source address (RFC3484)
 946  */
 947 enum {
 948         IPV6_SADDR_RULE_INIT = 0,
 949         IPV6_SADDR_RULE_LOCAL,
 950         IPV6_SADDR_RULE_SCOPE,
 951         IPV6_SADDR_RULE_PREFERRED,
 952 #ifdef CONFIG_IPV6_MIP6
 953         IPV6_SADDR_RULE_HOA,
 954 #endif
 955         IPV6_SADDR_RULE_OIF,
 956         IPV6_SADDR_RULE_LABEL,
 957 #ifdef CONFIG_IPV6_PRIVACY
 958         IPV6_SADDR_RULE_PRIVACY,
 959 #endif
 960         IPV6_SADDR_RULE_ORCHID,
 961         IPV6_SADDR_RULE_PREFIX,
 962         IPV6_SADDR_RULE_MAX
 963 };
 964
 965 struct ipv6_saddr_score {
 966         int                     rule;
 967         int                     addr_type;
 968         struct inet6_ifaddr     *ifa;
 969         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 970         int                     scopedist;
 971         int                     matchlen;
 972 };
 973
 974 struct ipv6_saddr_dst {
 975         const struct in6_addr *addr;
 976         int ifindex;
 977         int scope;
 978         int label;
 979         unsigned int prefs;
 980 };
 981
 982 static inline int ipv6_saddr_preferred(int type)
 983 {
 984         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 985                 return 1;
 986         return 0;
 987 }
 988
 989 static int ipv6_get_saddr_eval(struct net *net,
 990                                struct ipv6_saddr_score *score,
 991                                struct ipv6_saddr_dst *dst,
 992                                int i)
 993 {
 994         int ret;
 995
 996         if (i <= score->rule) {
 997                 switch (i) {
 998                 case IPV6_SADDR_RULE_SCOPE:
 999                         ret = score->scopedist;
1000                         break;
1001                 case IPV6_SADDR_RULE_PREFIX:
1002                         ret = score->matchlen;
1003                         break;
1004                 default:
1005                         ret = !!test_bit(i, score->scorebits);
1006                 }
1007                 goto out;
1008         }
1009
1010         switch (i) {
1011         case IPV6_SADDR_RULE_INIT:
1012                 /* Rule 0: remember if hiscore is not ready yet */
1013                 ret = !!score->ifa;
1014                 break;
1015         case IPV6_SADDR_RULE_LOCAL:
1016                 /* Rule 1: Prefer same address */
1017                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1018                 break;
1019         case IPV6_SADDR_RULE_SCOPE:
1020                 /* Rule 2: Prefer appropriate scope
1021                  *
1022                  *      ret
1023                  *       ^
1024                  *    -1 |  d 15
1025                  *    ---+--+-+---> scope
1026                  *       |
1027                  *       |             d is scope of the destination.
1028                  *  B-d  |  \
1029                  *       |   \      <- smaller scope is better if
1030                  *  B-15 |    \        if scope is enough for destinaion.
1031                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1032                  * d-C-1 | /
1033                  *       |/         <- greater is better
1034                  *   -C  /             if scope is not enough for destination.
1035                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1036                  *
1037                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1038                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1039                  * Assume B = 0 and we get C > 29.
1040                  */
1041                 ret = __ipv6_addr_src_scope(score->addr_type);
1042                 if (ret >= dst->scope)
1043                         ret = -ret;
1044                 else
1045                         ret -= 128;     /* 30 is enough */
1046                 score->scopedist = ret;
1047                 break;
1048         case IPV6_SADDR_RULE_PREFERRED:
1049                 /* Rule 3: Avoid deprecated and optimistic addresses */
1050                 ret = ipv6_saddr_preferred(score->addr_type) ||
1051                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1052                 break;
1053 #ifdef CONFIG_IPV6_MIP6
1054         case IPV6_SADDR_RULE_HOA:
1055             {
1056                 /* Rule 4: Prefer home address */
1057                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1058                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1059                 break;
1060             }
1061 #endif
1062         case IPV6_SADDR_RULE_OIF:
1063                 /* Rule 5: Prefer outgoing interface */
1064                 ret = (!dst->ifindex ||
1065                        dst->ifindex == score->ifa->idev->dev->ifindex);
1066                 break;
1067         case IPV6_SADDR_RULE_LABEL:
1068                 /* Rule 6: Prefer matching label */
1069                 ret = ipv6_addr_label(net,
1070                                       &score->ifa->addr, score->addr_type,
1071                                       score->ifa->idev->dev->ifindex) == dst->label;
1072                 break;
1073 #ifdef CONFIG_IPV6_PRIVACY
1074         case IPV6_SADDR_RULE_PRIVACY:
1075             {
1076                 /* Rule 7: Prefer public address
1077                  * Note: prefer temporary address if use_tempaddr >= 2
1078                  */
1079                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1080                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1081                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1082                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1083                 break;
1084             }
1085 #endif
1086         case IPV6_SADDR_RULE_ORCHID:
1087                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1088                  *          non-ORCHID vs non-ORCHID
1089                  */
1090                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1091                         ipv6_addr_orchid(dst->addr));
1092                 break;
1093         case IPV6_SADDR_RULE_PREFIX:
1094                 /* Rule 8: Use longest matching prefix */
1095                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1096                                                        dst->addr);
1097                 break;
1098         default:
1099                 ret = 0;
1100         }
1101
1102         if (ret)
1103                 __set_bit(i, score->scorebits);
1104         score->rule = i;
1105 out:
1106         return ret;
1107 }
1108
1109 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1110                        const struct in6_addr *daddr, unsigned int prefs,
1111                        struct in6_addr *saddr)
1112 {
1113         struct ipv6_saddr_score scores[2],
1114                                 *score = &scores[0], *hiscore = &scores[1];
1115         struct ipv6_saddr_dst dst;
1116         struct net_device *dev;
1117         int dst_type;
1118
1119         dst_type = __ipv6_addr_type(daddr);
1120         dst.addr = daddr;
1121         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1122         dst.scope = __ipv6_addr_src_scope(dst_type);
1123         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1124         dst.prefs = prefs;
1125
1126         hiscore->rule = -1;
1127         hiscore->ifa = NULL;
1128
1129         rcu_read_lock();
1130
1131         for_each_netdev_rcu(net, dev) {
1132                 struct inet6_dev *idev;
1133
1134                 /* Candidate Source Address (section 4)
1135                  *  - multicast and link-local destination address,
1136                  *    the set of candidate source address MUST only
1137                  *    include addresses assigned to interfaces
1138                  *    belonging to the same link as the outgoing
1139                  *    interface.
1140                  * (- For site-local destination addresses, the
1141                  *    set of candidate source addresses MUST only
1142                  *    include addresses assigned to interfaces
1143                  *    belonging to the same site as the outgoing
1144                  *    interface.)
1145                  */
1146                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1147                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1148                     dst.ifindex && dev->ifindex != dst.ifindex)
1149                         continue;
1150
1151                 idev = __in6_dev_get(dev);
1152                 if (!idev)
1153                         continue;
1154
1155                 read_lock_bh(&idev->lock);
1156                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1157                         int i;
1158
1159                         /*
1160                          * - Tentative Address (RFC2462 section 5.4)
1161                          *  - A tentative address is not considered
1162                          *    "assigned to an interface" in the traditional
1163                          *    sense, unless it is also flagged as optimistic.
1164                          * - Candidate Source Address (section 4)
1165                          *  - In any case, anycast addresses, multicast
1166                          *    addresses, and the unspecified address MUST
1167                          *    NOT be included in a candidate set.
1168                          */
1169                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1170                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1171                                 continue;
1172
1173                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1174
1175                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1176                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1177                                 LIMIT_NETDEBUG(KERN_DEBUG
1178                                                "ADDRCONF: unspecified / multicast address "
1179                                                "assigned as unicast address on %s",
1180                                                dev->name);
1181                                 continue;
1182                         }
1183
1184                         score->rule = -1;
1185                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1186
1187                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1188                                 int minihiscore, miniscore;
1189
1190                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1191                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1192
1193                                 if (minihiscore > miniscore) {
1194                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1195                                             score->scopedist > 0) {
1196                                                 /*
1197                                                  * special case:
1198                                                  * each remaining entry
1199                                                  * has too small (not enough)
1200                                                  * scope, because ifa entries
1201                                                  * are sorted by their scope
1202                                                  * values.
1203                                                  */
1204                                                 goto try_nextdev;
1205                                         }
1206                                         break;
1207                                 } else if (minihiscore < miniscore) {
1208                                         if (hiscore->ifa)
1209                                                 in6_ifa_put(hiscore->ifa);
1210
1211                                         in6_ifa_hold(score->ifa);
1212
1213                                         swap(hiscore, score);
1214
1215                                         /* restore our iterator */
1216                                         score->ifa = hiscore->ifa;
1217
1218                                         break;
1219                                 }
1220                         }
1221                 }
1222 try_nextdev:
1223                 read_unlock_bh(&idev->lock);
1224         }
1225         rcu_read_unlock();
1226
1227         if (!hiscore->ifa)
1228                 return -EADDRNOTAVAIL;
1229
1230         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1231         in6_ifa_put(hiscore->ifa);
1232         return 0;
1233 }
1234 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1235
1236 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1237                     unsigned char banned_flags)
1238 {
1239         struct inet6_dev *idev;
1240         int err = -EADDRNOTAVAIL;
1241
1242         rcu_read_lock();
1243         idev = __in6_dev_get(dev);
1244         if (idev) {
1245                 struct inet6_ifaddr *ifp;
1246
1247                 read_lock_bh(&idev->lock);
1248                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1249                         if (ifp->scope == IFA_LINK &&
1250                             !(ifp->flags & banned_flags)) {
1251                                 ipv6_addr_copy(addr, &ifp->addr);
1252                                 err = 0;
1253                                 break;
1254                         }
1255                 }
1256                 read_unlock_bh(&idev->lock);
1257         }
1258         rcu_read_unlock();
1259         return err;
1260 }
1261
1262 static int ipv6_count_addresses(struct inet6_dev *idev)
1263 {
1264         int cnt = 0;
1265         struct inet6_ifaddr *ifp;
1266
1267         read_lock_bh(&idev->lock);
1268         list_for_each_entry(ifp, &idev->addr_list, if_list)
1269                 cnt++;
1270         read_unlock_bh(&idev->lock);
1271         return cnt;
1272 }
1273
1274 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1275                   struct net_device *dev, int strict)
1276 {
1277         struct inet6_ifaddr *ifp;
1278         struct hlist_node *node;
1279         unsigned int hash = ipv6_addr_hash(addr);
1280
1281         rcu_read_lock_bh();
1282         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1283                 if (!net_eq(dev_net(ifp->idev->dev), net))
1284                         continue;
1285                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1286                     !(ifp->flags&IFA_F_TENTATIVE) &&
1287                     (dev == NULL || ifp->idev->dev == dev ||
1288                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1289                         rcu_read_unlock_bh();
1290                         return 1;
1291                 }
1292         }
1293
1294         rcu_read_unlock_bh();
1295         return 0;
1296 }
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1298
1299 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1300                                struct net_device *dev)
1301 {
1302         unsigned int hash = ipv6_addr_hash(addr);
1303         struct inet6_ifaddr *ifp;
1304         struct hlist_node *node;
1305
1306         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1307                 if (!net_eq(dev_net(ifp->idev->dev), net))
1308                         continue;
1309                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310                         if (dev == NULL || ifp->idev->dev == dev)
1311                                 return true;
1312                 }
1313         }
1314         return false;
1315 }
1316
1317 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1318 {
1319         struct inet6_dev *idev;
1320         struct inet6_ifaddr *ifa;
1321         int     onlink;
1322
1323         onlink = 0;
1324         rcu_read_lock();
1325         idev = __in6_dev_get(dev);
1326         if (idev) {
1327                 read_lock_bh(&idev->lock);
1328                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1329                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1330                                                    ifa->prefix_len);
1331                         if (onlink)
1332                                 break;
1333                 }
1334                 read_unlock_bh(&idev->lock);
1335         }
1336         rcu_read_unlock();
1337         return onlink;
1338 }
1339
1340 EXPORT_SYMBOL(ipv6_chk_prefix);
1341
1342 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343                                      struct net_device *dev, int strict)
1344 {
1345         struct inet6_ifaddr *ifp, *result = NULL;
1346         unsigned int hash = ipv6_addr_hash(addr);
1347         struct hlist_node *node;
1348
1349         rcu_read_lock_bh();
1350         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1351                 if (!net_eq(dev_net(ifp->idev->dev), net))
1352                         continue;
1353                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1354                         if (dev == NULL || ifp->idev->dev == dev ||
1355                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1356                                 result = ifp;
1357                                 in6_ifa_hold(ifp);
1358                                 break;
1359                         }
1360                 }
1361         }
1362         rcu_read_unlock_bh();
1363
1364         return result;
1365 }
1366
1367 /* Gets referenced address, destroys ifaddr */
1368
1369 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1370 {
1371         if (ifp->flags&IFA_F_PERMANENT) {
1372                 spin_lock_bh(&ifp->lock);
1373                 addrconf_del_timer(ifp);
1374                 ifp->flags |= IFA_F_TENTATIVE;
1375                 if (dad_failed)
1376                         ifp->flags |= IFA_F_DADFAILED;
1377                 spin_unlock_bh(&ifp->lock);
1378                 if (dad_failed)
1379                         ipv6_ifa_notify(0, ifp);
1380                 in6_ifa_put(ifp);
1381 #ifdef CONFIG_IPV6_PRIVACY
1382         } else if (ifp->flags&IFA_F_TEMPORARY) {
1383                 struct inet6_ifaddr *ifpub;
1384                 spin_lock_bh(&ifp->lock);
1385                 ifpub = ifp->ifpub;
1386                 if (ifpub) {
1387                         in6_ifa_hold(ifpub);
1388                         spin_unlock_bh(&ifp->lock);
1389                         ipv6_create_tempaddr(ifpub, ifp);
1390                         in6_ifa_put(ifpub);
1391                 } else {
1392                         spin_unlock_bh(&ifp->lock);
1393                 }
1394                 ipv6_del_addr(ifp);
1395 #endif
1396         } else
1397                 ipv6_del_addr(ifp);
1398 }
1399
1400 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1401 {
1402         int err = -ENOENT;
1403
1404         spin_lock(&ifp->state_lock);
1405         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1406                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1407                 err = 0;
1408         }
1409         spin_unlock(&ifp->state_lock);
1410
1411         return err;
1412 }
1413
1414 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1415 {
1416         struct inet6_dev *idev = ifp->idev;
1417
1418         if (addrconf_dad_end(ifp)) {
1419                 in6_ifa_put(ifp);
1420                 return;
1421         }
1422
1423         if (net_ratelimit())
1424                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1425                         ifp->idev->dev->name, &ifp->addr);
1426
1427         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1428                 struct in6_addr addr;
1429
1430                 addr.s6_addr32[0] = htonl(0xfe800000);
1431                 addr.s6_addr32[1] = 0;
1432
1433                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1434                     ipv6_addr_equal(&ifp->addr, &addr)) {
1435                         /* DAD failed for link-local based on MAC address */
1436                         idev->cnf.disable_ipv6 = 1;
1437
1438                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1439                                 ifp->idev->dev->name);
1440                 }
1441         }
1442
1443         addrconf_dad_stop(ifp, 1);
1444 }
1445
1446 /* Join to solicited addr multicast group. */
1447
1448 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1449 {
1450         struct in6_addr maddr;
1451
1452         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1453                 return;
1454
1455         addrconf_addr_solict_mult(addr, &maddr);
1456         ipv6_dev_mc_inc(dev, &maddr);
1457 }
1458
1459 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1460 {
1461         struct in6_addr maddr;
1462
1463         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1464                 return;
1465
1466         addrconf_addr_solict_mult(addr, &maddr);
1467         __ipv6_dev_mc_dec(idev, &maddr);
1468 }
1469
1470 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1471 {
1472         struct in6_addr addr;
1473         if (ifp->prefix_len == 127) /* RFC 6164 */
1474                 return;
1475         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1476         if (ipv6_addr_any(&addr))
1477                 return;
1478         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1479 }
1480
1481 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1482 {
1483         struct in6_addr addr;
1484         if (ifp->prefix_len == 127) /* RFC 6164 */
1485                 return;
1486         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1487         if (ipv6_addr_any(&addr))
1488                 return;
1489         __ipv6_dev_ac_dec(ifp->idev, &addr);
1490 }
1491
1492 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1493 {
1494         if (dev->addr_len != ETH_ALEN)
1495                 return -1;
1496         memcpy(eui, dev->dev_addr, 3);
1497         memcpy(eui + 5, dev->dev_addr + 3, 3);
1498
1499         /*
1500          * The zSeries OSA network cards can be shared among various
1501          * OS instances, but the OSA cards have only one MAC address.
1502          * This leads to duplicate address conflicts in conjunction
1503          * with IPv6 if more than one instance uses the same card.
1504          *
1505          * The driver for these cards can deliver a unique 16-bit
1506          * identifier for each instance sharing the same card.  It is
1507          * placed instead of 0xFFFE in the interface identifier.  The
1508          * "u" bit of the interface identifier is not inverted in this
1509          * case.  Hence the resulting interface identifier has local
1510          * scope according to RFC2373.
1511          */
1512         if (dev->dev_id) {
1513                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1514                 eui[4] = dev->dev_id & 0xFF;
1515         } else {
1516                 eui[3] = 0xFF;
1517                 eui[4] = 0xFE;
1518                 eui[0] ^= 2;
1519         }
1520         return 0;
1521 }
1522
1523 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1524 {
1525         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1526         if (dev->addr_len != ARCNET_ALEN)
1527                 return -1;
1528         memset(eui, 0, 7);
1529         eui[7] = *(u8*)dev->dev_addr;
1530         return 0;
1531 }
1532
1533 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1534 {
1535         if (dev->addr_len != INFINIBAND_ALEN)
1536                 return -1;
1537         memcpy(eui, dev->dev_addr + 12, 8);
1538         eui[0] |= 2;
1539         return 0;
1540 }
1541
1542 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1543 {
1544         if (addr == 0)
1545                 return -1;
1546         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1547                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1548                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1549                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1550                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1551                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1552         eui[1] = 0;
1553         eui[2] = 0x5E;
1554         eui[3] = 0xFE;
1555         memcpy(eui + 4, &addr, 4);
1556         return 0;
1557 }
1558
1559 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1560 {
1561         if (dev->priv_flags & IFF_ISATAP)
1562                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1563         return -1;
1564 }
1565
1566 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1567 {
1568         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1569 }
1570
1571 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1572 {
1573         switch (dev->type) {
1574         case ARPHRD_ETHER:
1575         case ARPHRD_FDDI:
1576         case ARPHRD_IEEE802_TR:
1577                 return addrconf_ifid_eui48(eui, dev);
1578         case ARPHRD_ARCNET:
1579                 return addrconf_ifid_arcnet(eui, dev);
1580         case ARPHRD_INFINIBAND:
1581                 return addrconf_ifid_infiniband(eui, dev);
1582         case ARPHRD_SIT:
1583                 return addrconf_ifid_sit(eui, dev);
1584         case ARPHRD_IPGRE:
1585                 return addrconf_ifid_gre(eui, dev);
1586         }
1587         return -1;
1588 }
1589
1590 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1591 {
1592         int err = -1;
1593         struct inet6_ifaddr *ifp;
1594
1595         read_lock_bh(&idev->lock);
1596         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1597                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1598                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1599                         err = 0;
1600                         break;
1601                 }
1602         }
1603         read_unlock_bh(&idev->lock);
1604         return err;
1605 }
1606
1607 #ifdef CONFIG_IPV6_PRIVACY
1608 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1609 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1610 {
1611 regen:
1612         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1613         idev->rndid[0] &= ~0x02;
1614
1615         /*
1616          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1617          * check if generated address is not inappropriate
1618          *
1619          *  - Reserved subnet anycast (RFC 2526)
1620          *      11111101 11....11 1xxxxxxx
1621          *  - ISATAP (RFC4214) 6.1
1622          *      00-00-5E-FE-xx-xx-xx-xx
1623          *  - value 0
1624          *  - XXX: already assigned to an address on the device
1625          */
1626         if (idev->rndid[0] == 0xfd &&
1627             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1628             (idev->rndid[7]&0x80))
1629                 goto regen;
1630         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1631                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1632                         goto regen;
1633                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1634                         goto regen;
1635         }
1636
1637         return 0;
1638 }
1639
1640 static void ipv6_regen_rndid(unsigned long data)
1641 {
1642         struct inet6_dev *idev = (struct inet6_dev *) data;
1643         unsigned long expires;
1644
1645         rcu_read_lock_bh();
1646         write_lock_bh(&idev->lock);
1647
1648         if (idev->dead)
1649                 goto out;
1650
1651         if (__ipv6_regen_rndid(idev) < 0)
1652                 goto out;
1653
1654         expires = jiffies +
1655                 idev->cnf.temp_prefered_lft * HZ -
1656                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1657                 idev->cnf.max_desync_factor * HZ;
1658         if (time_before(expires, jiffies)) {
1659                 printk(KERN_WARNING
1660                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1661                         idev->dev->name);
1662                 goto out;
1663         }
1664
1665         if (!mod_timer(&idev->regen_timer, expires))
1666                 in6_dev_hold(idev);
1667
1668 out:
1669         write_unlock_bh(&idev->lock);
1670         rcu_read_unlock_bh();
1671         in6_dev_put(idev);
1672 }
1673
1674 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1675         int ret = 0;
1676
1677         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1678                 ret = __ipv6_regen_rndid(idev);
1679         return ret;
1680 }
1681 #endif
1682
1683 /*
1684  *      Add prefix route.
1685  */
1686
1687 static void
1688 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1689                       unsigned long expires, u32 flags)
1690 {
1691         struct fib6_config cfg = {
1692                 .fc_table = RT6_TABLE_PREFIX,
1693                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1694                 .fc_ifindex = dev->ifindex,
1695                 .fc_expires = expires,
1696                 .fc_dst_len = plen,
1697                 .fc_flags = RTF_UP | flags,
1698                 .fc_nlinfo.nl_net = dev_net(dev),
1699                 .fc_protocol = RTPROT_KERNEL,
1700         };
1701
1702         ipv6_addr_copy(&cfg.fc_dst, pfx);
1703
1704         /* Prevent useless cloning on PtP SIT.
1705            This thing is done here expecting that the whole
1706            class of non-broadcast devices need not cloning.
1707          */
1708 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1709         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1710                 cfg.fc_flags |= RTF_NONEXTHOP;
1711 #endif
1712
1713         ip6_route_add(&cfg);
1714 }
1715
1716 /* Create "default" multicast route to the interface */
1717
1718 static void addrconf_add_mroute(struct net_device *dev)
1719 {
1720         struct fib6_config cfg = {
1721                 .fc_table = RT6_TABLE_LOCAL,
1722                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1723                 .fc_ifindex = dev->ifindex,
1724                 .fc_dst_len = 8,
1725                 .fc_flags = RTF_UP,
1726                 .fc_nlinfo.nl_net = dev_net(dev),
1727         };
1728
1729         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1730
1731         ip6_route_add(&cfg);
1732 }
1733
1734 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1735 static void sit_route_add(struct net_device *dev)
1736 {
1737         struct fib6_config cfg = {
1738                 .fc_table = RT6_TABLE_MAIN,
1739                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1740                 .fc_ifindex = dev->ifindex,
1741                 .fc_dst_len = 96,
1742                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1743                 .fc_nlinfo.nl_net = dev_net(dev),
1744         };
1745
1746         /* prefix length - 96 bits "::d.d.d.d" */
1747         ip6_route_add(&cfg);
1748 }
1749 #endif
1750
1751 static void addrconf_add_lroute(struct net_device *dev)
1752 {
1753         struct in6_addr addr;
1754
1755         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1756         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1757 }
1758
1759 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1760 {
1761         struct inet6_dev *idev;
1762
1763         ASSERT_RTNL();
1764
1765         idev = ipv6_find_idev(dev);
1766         if (!idev)
1767                 return ERR_PTR(-ENOBUFS);
1768
1769         if (idev->cnf.disable_ipv6)
1770                 return ERR_PTR(-EACCES);
1771
1772         /* Add default multicast route */
1773         addrconf_add_mroute(dev);
1774
1775         /* Add link local route */
1776         addrconf_add_lroute(dev);
1777         return idev;
1778 }
1779
1780 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1781 {
1782         struct prefix_info *pinfo;
1783         __u32 valid_lft;
1784         __u32 prefered_lft;
1785         int addr_type;
1786         struct inet6_dev *in6_dev;
1787         struct net *net = dev_net(dev);
1788
1789         pinfo = (struct prefix_info *) opt;
1790
1791         if (len < sizeof(struct prefix_info)) {
1792                 ADBG(("addrconf: prefix option too short\n"));
1793                 return;
1794         }
1795
1796         /*
1797          *      Validation checks ([ADDRCONF], page 19)
1798          */
1799
1800         addr_type = ipv6_addr_type(&pinfo->prefix);
1801
1802         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1803                 return;
1804
1805         valid_lft = ntohl(pinfo->valid);
1806         prefered_lft = ntohl(pinfo->prefered);
1807
1808         if (prefered_lft > valid_lft) {
1809                 if (net_ratelimit())
1810                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1811                 return;
1812         }
1813
1814         in6_dev = in6_dev_get(dev);
1815
1816         if (in6_dev == NULL) {
1817                 if (net_ratelimit())
1818                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1819                 return;
1820         }
1821
1822         /*
1823          *      Two things going on here:
1824          *      1) Add routes for on-link prefixes
1825          *      2) Configure prefixes with the auto flag set
1826          */
1827
1828         if (pinfo->onlink) {
1829                 struct rt6_info *rt;
1830                 unsigned long rt_expires;
1831
1832                 /* Avoid arithmetic overflow. Really, we could
1833                  * save rt_expires in seconds, likely valid_lft,
1834                  * but it would require division in fib gc, that it
1835                  * not good.
1836                  */
1837                 if (HZ > USER_HZ)
1838                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1839                 else
1840                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1841
1842                 if (addrconf_finite_timeout(rt_expires))
1843                         rt_expires *= HZ;
1844
1845                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1846                                 dev->ifindex, 1);
1847
1848                 if (rt && addrconf_is_prefix_route(rt)) {
1849                         /* Autoconf prefix route */
1850                         if (valid_lft == 0) {
1851                                 ip6_del_rt(rt);
1852                                 rt = NULL;
1853                         } else if (addrconf_finite_timeout(rt_expires)) {
1854                                 /* not infinity */
1855                                 rt->rt6i_expires = jiffies + rt_expires;
1856                                 rt->rt6i_flags |= RTF_EXPIRES;
1857                         } else {
1858                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1859                                 rt->rt6i_expires = 0;
1860                         }
1861                 } else if (valid_lft) {
1862                         clock_t expires = 0;
1863                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1864                         if (addrconf_finite_timeout(rt_expires)) {
1865                                 /* not infinity */
1866                                 flags |= RTF_EXPIRES;
1867                                 expires = jiffies_to_clock_t(rt_expires);
1868                         }
1869                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1870                                               dev, expires, flags);
1871                 }
1872                 if (rt)
1873                         dst_release(&rt->dst);
1874         }
1875
1876         /* Try to figure out our local address for this prefix */
1877
1878         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1879                 struct inet6_ifaddr * ifp;
1880                 struct in6_addr addr;
1881                 int create = 0, update_lft = 0;
1882
1883                 if (pinfo->prefix_len == 64) {
1884                         memcpy(&addr, &pinfo->prefix, 8);
1885                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1886                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1887                                 in6_dev_put(in6_dev);
1888                                 return;
1889                         }
1890                         goto ok;
1891                 }
1892                 if (net_ratelimit())
1893                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1894                                pinfo->prefix_len);
1895                 in6_dev_put(in6_dev);
1896                 return;
1897
1898 ok:
1899
1900                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1901
1902                 if (ifp == NULL && valid_lft) {
1903                         int max_addresses = in6_dev->cnf.max_addresses;
1904                         u32 addr_flags = 0;
1905
1906 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1907                         if (in6_dev->cnf.optimistic_dad &&
1908                             !net->ipv6.devconf_all->forwarding)
1909                                 addr_flags = IFA_F_OPTIMISTIC;
1910 #endif
1911
1912                         /* Do not allow to create too much of autoconfigured
1913                          * addresses; this would be too easy way to crash kernel.
1914                          */
1915                         if (!max_addresses ||
1916                             ipv6_count_addresses(in6_dev) < max_addresses)
1917                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1918                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1919                                                     addr_flags);
1920
1921                         if (!ifp || IS_ERR(ifp)) {
1922                                 in6_dev_put(in6_dev);
1923                                 return;
1924                         }
1925
1926                         update_lft = create = 1;
1927                         ifp->cstamp = jiffies;
1928                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1929                 }
1930
1931                 if (ifp) {
1932                         int flags;
1933                         unsigned long now;
1934 #ifdef CONFIG_IPV6_PRIVACY
1935                         struct inet6_ifaddr *ift;
1936 #endif
1937                         u32 stored_lft;
1938
1939                         /* update lifetime (RFC2462 5.5.3 e) */
1940                         spin_lock(&ifp->lock);
1941                         now = jiffies;
1942                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1943                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1944                         else
1945                                 stored_lft = 0;
1946                         if (!update_lft && stored_lft) {
1947                                 if (valid_lft > MIN_VALID_LIFETIME ||
1948                                     valid_lft > stored_lft)
1949                                         update_lft = 1;
1950                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1951                                         /* valid_lft <= stored_lft is always true */
1952                                         /*
1953                                          * RFC 4862 Section 5.5.3e:
1954                                          * "Note that the preferred lifetime of
1955                                          *  the corresponding address is always
1956                                          *  reset to the Preferred Lifetime in
1957                                          *  the received Prefix Information
1958                                          *  option, regardless of whether the
1959                                          *  valid lifetime is also reset or
1960                                          *  ignored."
1961                                          *
1962                                          *  So if the preferred lifetime in
1963                                          *  this advertisement is different
1964                                          *  than what we have stored, but the
1965                                          *  valid lifetime is invalid, just
1966                                          *  reset prefered_lft.
1967                                          *
1968                                          *  We must set the valid lifetime
1969                                          *  to the stored lifetime since we'll
1970                                          *  be updating the timestamp below,
1971                                          *  else we'll set it back to the
1972                                          *  minimum.
1973                                          */
1974                                         if (prefered_lft != ifp->prefered_lft) {
1975                                                 valid_lft = stored_lft;
1976                                                 update_lft = 1;
1977                                         }
1978                                 } else {
1979                                         valid_lft = MIN_VALID_LIFETIME;
1980                                         if (valid_lft < prefered_lft)
1981                                                 prefered_lft = valid_lft;
1982                                         update_lft = 1;
1983                                 }
1984                         }
1985
1986                         if (update_lft) {
1987                                 ifp->valid_lft = valid_lft;
1988                                 ifp->prefered_lft = prefered_lft;
1989                                 ifp->tstamp = now;
1990                                 flags = ifp->flags;
1991                                 ifp->flags &= ~IFA_F_DEPRECATED;
1992                                 spin_unlock(&ifp->lock);
1993
1994                                 if (!(flags&IFA_F_TENTATIVE))
1995                                         ipv6_ifa_notify(0, ifp);
1996                         } else
1997                                 spin_unlock(&ifp->lock);
1998
1999 #ifdef CONFIG_IPV6_PRIVACY
2000                         read_lock_bh(&in6_dev->lock);
2001                         /* update all temporary addresses in the list */
2002                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2003                                 /*
2004                                  * When adjusting the lifetimes of an existing
2005                                  * temporary address, only lower the lifetimes.
2006                                  * Implementations must not increase the
2007                                  * lifetimes of an existing temporary address
2008                                  * when processing a Prefix Information Option.
2009                                  */
2010                                 if (ifp != ift->ifpub)
2011                                         continue;
2012
2013                                 spin_lock(&ift->lock);
2014                                 flags = ift->flags;
2015                                 if (ift->valid_lft > valid_lft &&
2016                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2017                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2018                                 if (ift->prefered_lft > prefered_lft &&
2019                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2020                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2021                                 spin_unlock(&ift->lock);
2022                                 if (!(flags&IFA_F_TENTATIVE))
2023                                         ipv6_ifa_notify(0, ift);
2024                         }
2025
2026                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2027                                 /*
2028                                  * When a new public address is created as described in [ADDRCONF],
2029                                  * also create a new temporary address. Also create a temporary
2030                                  * address if it's enabled but no temporary address currently exists.
2031                                  */
2032                                 read_unlock_bh(&in6_dev->lock);
2033                                 ipv6_create_tempaddr(ifp, NULL);
2034                         } else {
2035                                 read_unlock_bh(&in6_dev->lock);
2036                         }
2037 #endif
2038                         in6_ifa_put(ifp);
2039                         addrconf_verify(0);
2040                 }
2041         }
2042         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2043         in6_dev_put(in6_dev);
2044 }
2045
2046 /*
2047  *      Set destination address.
2048  *      Special case for SIT interfaces where we create a new "virtual"
2049  *      device.
2050  */
2051 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2052 {
2053         struct in6_ifreq ireq;
2054         struct net_device *dev;
2055         int err = -EINVAL;
2056
2057         rtnl_lock();
2058
2059         err = -EFAULT;
2060         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2061                 goto err_exit;
2062
2063         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2064
2065         err = -ENODEV;
2066         if (dev == NULL)
2067                 goto err_exit;
2068
2069 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2070         if (dev->type == ARPHRD_SIT) {
2071                 const struct net_device_ops *ops = dev->netdev_ops;
2072                 struct ifreq ifr;
2073                 struct ip_tunnel_parm p;
2074
2075                 err = -EADDRNOTAVAIL;
2076                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2077                         goto err_exit;
2078
2079                 memset(&p, 0, sizeof(p));
2080                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2081                 p.iph.saddr = 0;
2082                 p.iph.version = 4;
2083                 p.iph.ihl = 5;
2084                 p.iph.protocol = IPPROTO_IPV6;
2085                 p.iph.ttl = 64;
2086                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2087
2088                 if (ops->ndo_do_ioctl) {
2089                         mm_segment_t oldfs = get_fs();
2090
2091                         set_fs(KERNEL_DS);
2092                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2093                         set_fs(oldfs);
2094                 } else
2095                         err = -EOPNOTSUPP;
2096
2097                 if (err == 0) {
2098                         err = -ENOBUFS;
2099                         dev = __dev_get_by_name(net, p.name);
2100                         if (!dev)
2101                                 goto err_exit;
2102                         err = dev_open(dev);
2103                 }
2104         }
2105 #endif
2106
2107 err_exit:
2108         rtnl_unlock();
2109         return err;
2110 }
2111
2112 /*
2113  *      Manual configuration of address on an interface
2114  */
2115 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2116                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2117                           __u32 valid_lft)
2118 {
2119         struct inet6_ifaddr *ifp;
2120         struct inet6_dev *idev;
2121         struct net_device *dev;
2122         int scope;
2123         u32 flags;
2124         clock_t expires;
2125         unsigned long timeout;
2126
2127         ASSERT_RTNL();
2128
2129         if (plen > 128)
2130                 return -EINVAL;
2131
2132         /* check the lifetime */
2133         if (!valid_lft || prefered_lft > valid_lft)
2134                 return -EINVAL;
2135
2136         dev = __dev_get_by_index(net, ifindex);
2137         if (!dev)
2138                 return -ENODEV;
2139
2140         idev = addrconf_add_dev(dev);
2141         if (IS_ERR(idev))
2142                 return PTR_ERR(idev);
2143
2144         scope = ipv6_addr_scope(pfx);
2145
2146         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2147         if (addrconf_finite_timeout(timeout)) {
2148                 expires = jiffies_to_clock_t(timeout * HZ);
2149                 valid_lft = timeout;
2150                 flags = RTF_EXPIRES;
2151         } else {
2152                 expires = 0;
2153                 flags = 0;
2154                 ifa_flags |= IFA_F_PERMANENT;
2155         }
2156
2157         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2158         if (addrconf_finite_timeout(timeout)) {
2159                 if (timeout == 0)
2160                         ifa_flags |= IFA_F_DEPRECATED;
2161                 prefered_lft = timeout;
2162         }
2163
2164         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2165
2166         if (!IS_ERR(ifp)) {
2167                 spin_lock_bh(&ifp->lock);
2168                 ifp->valid_lft = valid_lft;
2169                 ifp->prefered_lft = prefered_lft;
2170                 ifp->tstamp = jiffies;
2171                 spin_unlock_bh(&ifp->lock);
2172
2173                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2174                                       expires, flags);
2175                 /*
2176                  * Note that section 3.1 of RFC 4429 indicates
2177                  * that the Optimistic flag should not be set for
2178                  * manually configured addresses
2179                  */
2180                 addrconf_dad_start(ifp, 0);
2181                 in6_ifa_put(ifp);
2182                 addrconf_verify(0);
2183                 return 0;
2184         }
2185
2186         return PTR_ERR(ifp);
2187 }
2188
2189 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2190                           unsigned int plen)
2191 {
2192         struct inet6_ifaddr *ifp;
2193         struct inet6_dev *idev;
2194         struct net_device *dev;
2195
2196         if (plen > 128)
2197                 return -EINVAL;
2198
2199         dev = __dev_get_by_index(net, ifindex);
2200         if (!dev)
2201                 return -ENODEV;
2202
2203         if ((idev = __in6_dev_get(dev)) == NULL)
2204                 return -ENXIO;
2205
2206         read_lock_bh(&idev->lock);
2207         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2208                 if (ifp->prefix_len == plen &&
2209                     ipv6_addr_equal(pfx, &ifp->addr)) {
2210                         in6_ifa_hold(ifp);
2211                         read_unlock_bh(&idev->lock);
2212
2213                         ipv6_del_addr(ifp);
2214
2215                         /* If the last address is deleted administratively,
2216                            disable IPv6 on this interface.
2217                          */
2218                         if (list_empty(&idev->addr_list))
2219                                 addrconf_ifdown(idev->dev, 1);
2220                         return 0;
2221                 }
2222         }
2223         read_unlock_bh(&idev->lock);
2224         return -EADDRNOTAVAIL;
2225 }
2226
2227
2228 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2229 {
2230         struct in6_ifreq ireq;
2231         int err;
2232
2233         if (!capable(CAP_NET_ADMIN))
2234                 return -EPERM;
2235
2236         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2237                 return -EFAULT;
2238
2239         rtnl_lock();
2240         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2241                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2242                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2243         rtnl_unlock();
2244         return err;
2245 }
2246
2247 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2248 {
2249         struct in6_ifreq ireq;
2250         int err;
2251
2252         if (!capable(CAP_NET_ADMIN))
2253                 return -EPERM;
2254
2255         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2256                 return -EFAULT;
2257
2258         rtnl_lock();
2259         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2260                              ireq.ifr6_prefixlen);
2261         rtnl_unlock();
2262         return err;
2263 }
2264
2265 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2266                      int plen, int scope)
2267 {
2268         struct inet6_ifaddr *ifp;
2269
2270         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2271         if (!IS_ERR(ifp)) {
2272                 spin_lock_bh(&ifp->lock);
2273                 ifp->flags &= ~IFA_F_TENTATIVE;
2274                 spin_unlock_bh(&ifp->lock);
2275                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2276                 in6_ifa_put(ifp);
2277         }
2278 }
2279
2280 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2281 static void sit_add_v4_addrs(struct inet6_dev *idev)
2282 {
2283         struct in6_addr addr;
2284         struct net_device *dev;
2285         struct net *net = dev_net(idev->dev);
2286         int scope;
2287
2288         ASSERT_RTNL();
2289
2290         memset(&addr, 0, sizeof(struct in6_addr));
2291         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2292
2293         if (idev->dev->flags&IFF_POINTOPOINT) {
2294                 addr.s6_addr32[0] = htonl(0xfe800000);
2295                 scope = IFA_LINK;
2296         } else {
2297                 scope = IPV6_ADDR_COMPATv4;
2298         }
2299
2300         if (addr.s6_addr32[3]) {
2301                 add_addr(idev, &addr, 128, scope);
2302                 return;
2303         }
2304
2305         for_each_netdev(net, dev) {
2306                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2307                 if (in_dev && (dev->flags & IFF_UP)) {
2308                         struct in_ifaddr * ifa;
2309
2310                         int flag = scope;
2311
2312                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2313                                 int plen;
2314
2315                                 addr.s6_addr32[3] = ifa->ifa_local;
2316
2317                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2318                                         continue;
2319                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2320                                         if (idev->dev->flags&IFF_POINTOPOINT)
2321                                                 continue;
2322                                         flag |= IFA_HOST;
2323                                 }
2324                                 if (idev->dev->flags&IFF_POINTOPOINT)
2325                                         plen = 64;
2326                                 else
2327                                         plen = 96;
2328
2329                                 add_addr(idev, &addr, plen, flag);
2330                         }
2331                 }
2332         }
2333 }
2334 #endif
2335
2336 static void init_loopback(struct net_device *dev)
2337 {
2338         struct inet6_dev  *idev;
2339
2340         /* ::1 */
2341
2342         ASSERT_RTNL();
2343
2344         if ((idev = ipv6_find_idev(dev)) == NULL) {
2345                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2346                 return;
2347         }
2348
2349         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2350 }
2351
2352 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2353 {
2354         struct inet6_ifaddr * ifp;
2355         u32 addr_flags = IFA_F_PERMANENT;
2356
2357 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2358         if (idev->cnf.optimistic_dad &&
2359             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2360                 addr_flags |= IFA_F_OPTIMISTIC;
2361 #endif
2362
2363
2364         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2365         if (!IS_ERR(ifp)) {
2366                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2367                 addrconf_dad_start(ifp, 0);
2368                 in6_ifa_put(ifp);
2369         }
2370 }
2371
2372 static void addrconf_dev_config(struct net_device *dev)
2373 {
2374         struct in6_addr addr;
2375         struct inet6_dev    * idev;
2376
2377         ASSERT_RTNL();
2378
2379         if ((dev->type != ARPHRD_ETHER) &&
2380             (dev->type != ARPHRD_FDDI) &&
2381             (dev->type != ARPHRD_IEEE802_TR) &&
2382             (dev->type != ARPHRD_ARCNET) &&
2383             (dev->type != ARPHRD_INFINIBAND)) {
2384                 /* Alas, we support only Ethernet autoconfiguration. */
2385                 return;
2386         }
2387
2388         idev = addrconf_add_dev(dev);
2389         if (IS_ERR(idev))
2390                 return;
2391
2392         memset(&addr, 0, sizeof(struct in6_addr));
2393         addr.s6_addr32[0] = htonl(0xFE800000);
2394
2395         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2396                 addrconf_add_linklocal(idev, &addr);
2397 }
2398
2399 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2400 static void addrconf_sit_config(struct net_device *dev)
2401 {
2402         struct inet6_dev *idev;
2403
2404         ASSERT_RTNL();
2405
2406         /*
2407          * Configure the tunnel with one of our IPv4
2408          * addresses... we should configure all of
2409          * our v4 addrs in the tunnel
2410          */
2411
2412         if ((idev = ipv6_find_idev(dev)) == NULL) {
2413                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2414                 return;
2415         }
2416
2417         if (dev->priv_flags & IFF_ISATAP) {
2418                 struct in6_addr addr;
2419
2420                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2421                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2422                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2423                         addrconf_add_linklocal(idev, &addr);
2424                 return;
2425         }
2426
2427         sit_add_v4_addrs(idev);
2428
2429         if (dev->flags&IFF_POINTOPOINT) {
2430                 addrconf_add_mroute(dev);
2431                 addrconf_add_lroute(dev);
2432         } else
2433                 sit_route_add(dev);
2434 }
2435 #endif
2436
2437 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2438 static void addrconf_gre_config(struct net_device *dev)
2439 {
2440         struct inet6_dev *idev;
2441         struct in6_addr addr;
2442
2443         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2444
2445         ASSERT_RTNL();
2446
2447         if ((idev = ipv6_find_idev(dev)) == NULL) {
2448                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2449                 return;
2450         }
2451
2452         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2453         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2454
2455         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2456                 addrconf_add_linklocal(idev, &addr);
2457 }
2458 #endif
2459
2460 static inline int
2461 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2462 {
2463         struct in6_addr lladdr;
2464
2465         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2466                 addrconf_add_linklocal(idev, &lladdr);
2467                 return 0;
2468         }
2469         return -1;
2470 }
2471
2472 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2473 {
2474         struct net_device *link_dev;
2475         struct net *net = dev_net(idev->dev);
2476
2477         /* first try to inherit the link-local address from the link device */
2478         if (idev->dev->iflink &&
2479             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2480                 if (!ipv6_inherit_linklocal(idev, link_dev))
2481                         return;
2482         }
2483         /* then try to inherit it from any device */
2484         for_each_netdev(net, link_dev) {
2485                 if (!ipv6_inherit_linklocal(idev, link_dev))
2486                         return;
2487         }
2488         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2489 }
2490
2491 /*
2492  * Autoconfigure tunnel with a link-local address so routing protocols,
2493  * DHCPv6, MLD etc. can be run over the virtual link
2494  */
2495
2496 static void addrconf_ip6_tnl_config(struct net_device *dev)
2497 {
2498         struct inet6_dev *idev;
2499
2500         ASSERT_RTNL();
2501
2502         idev = addrconf_add_dev(dev);
2503         if (IS_ERR(idev)) {
2504                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2505                 return;
2506         }
2507         ip6_tnl_add_linklocal(idev);
2508 }
2509
2510 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2511                            void * data)
2512 {
2513         struct net_device *dev = (struct net_device *) data;
2514         struct inet6_dev *idev = __in6_dev_get(dev);
2515         int run_pending = 0;
2516         int err;
2517
2518         switch (event) {
2519         case NETDEV_REGISTER:
2520                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2521                         idev = ipv6_add_dev(dev);
2522                         if (!idev)
2523                                 return notifier_from_errno(-ENOMEM);
2524                 }
2525                 break;
2526
2527         case NETDEV_UP:
2528         case NETDEV_CHANGE:
2529                 if (dev->flags & IFF_SLAVE)
2530                         break;
2531
2532                 if (event == NETDEV_UP) {
2533                         if (!addrconf_qdisc_ok(dev)) {
2534                                 /* device is not ready yet. */
2535                                 printk(KERN_INFO
2536                                         "ADDRCONF(NETDEV_UP): %s: "
2537                                         "link is not ready\n",
2538                                         dev->name);
2539                                 break;
2540                         }
2541
2542                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2543                                 idev = ipv6_add_dev(dev);
2544
2545                         if (idev) {
2546                                 idev->if_flags |= IF_READY;
2547                                 run_pending = 1;
2548                         }
2549                 } else {
2550                         if (!addrconf_qdisc_ok(dev)) {
2551                                 /* device is still not ready. */
2552                                 break;
2553                         }
2554
2555                         if (idev) {
2556                                 if (idev->if_flags & IF_READY)
2557                                         /* device is already configured. */
2558                                         break;
2559                                 idev->if_flags |= IF_READY;
2560                         }
2561
2562                         printk(KERN_INFO
2563                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2564                                         "link becomes ready\n",
2565                                         dev->name);
2566
2567                         run_pending = 1;
2568                 }
2569
2570                 switch (dev->type) {
2571 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2572                 case ARPHRD_SIT:
2573                         addrconf_sit_config(dev);
2574                         break;
2575 #endif
2576 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2577                 case ARPHRD_IPGRE:
2578                         addrconf_gre_config(dev);
2579                         break;
2580 #endif
2581                 case ARPHRD_TUNNEL6:
2582                         addrconf_ip6_tnl_config(dev);
2583                         break;
2584                 case ARPHRD_LOOPBACK:
2585                         init_loopback(dev);
2586                         break;
2587
2588                 default:
2589                         addrconf_dev_config(dev);
2590                         break;
2591                 }
2592
2593                 if (idev) {
2594                         if (run_pending)
2595                                 addrconf_dad_run(idev);
2596
2597                         /*
2598                          * If the MTU changed during the interface down,
2599                          * when the interface up, the changed MTU must be
2600                          * reflected in the idev as well as routers.
2601                          */
2602                         if (idev->cnf.mtu6 != dev->mtu &&
2603                             dev->mtu >= IPV6_MIN_MTU) {
2604                                 rt6_mtu_change(dev, dev->mtu);
2605                                 idev->cnf.mtu6 = dev->mtu;
2606                         }
2607                         idev->tstamp = jiffies;
2608                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2609
2610                         /*
2611                          * If the changed mtu during down is lower than
2612                          * IPV6_MIN_MTU stop IPv6 on this interface.
2613                          */
2614                         if (dev->mtu < IPV6_MIN_MTU)
2615                                 addrconf_ifdown(dev, 1);
2616                 }
2617                 break;
2618
2619         case NETDEV_CHANGEMTU:
2620                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2621                         rt6_mtu_change(dev, dev->mtu);
2622                         idev->cnf.mtu6 = dev->mtu;
2623                         break;
2624                 }
2625
2626                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2627                         idev = ipv6_add_dev(dev);
2628                         if (idev)
2629                                 break;
2630                 }
2631
2632                 /*
2633                  * MTU falled under IPV6_MIN_MTU.
2634                  * Stop IPv6 on this interface.
2635                  */
2636
2637         case NETDEV_DOWN:
2638         case NETDEV_UNREGISTER:
2639                 /*
2640                  *      Remove all addresses from this interface.
2641                  */
2642                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2643                 break;
2644
2645         case NETDEV_CHANGENAME:
2646                 if (idev) {
2647                         snmp6_unregister_dev(idev);
2648                         addrconf_sysctl_unregister(idev);
2649                         addrconf_sysctl_register(idev);
2650                         err = snmp6_register_dev(idev);
2651                         if (err)
2652                                 return notifier_from_errno(err);
2653                 }
2654                 break;
2655
2656         case NETDEV_PRE_TYPE_CHANGE:
2657         case NETDEV_POST_TYPE_CHANGE:
2658                 addrconf_type_change(dev, event);
2659                 break;
2660         }
2661
2662         return NOTIFY_OK;
2663 }
2664
2665 /*
2666  *      addrconf module should be notified of a device going up
2667  */
2668 static struct notifier_block ipv6_dev_notf = {
2669         .notifier_call = addrconf_notify,
2670 };
2671
2672 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2673 {
2674         struct inet6_dev *idev;
2675         ASSERT_RTNL();
2676
2677         idev = __in6_dev_get(dev);
2678
2679         if (event == NETDEV_POST_TYPE_CHANGE)
2680                 ipv6_mc_remap(idev);
2681         else if (event == NETDEV_PRE_TYPE_CHANGE)
2682                 ipv6_mc_unmap(idev);
2683 }
2684
2685 static int addrconf_ifdown(struct net_device *dev, int how)
2686 {
2687         struct net *net = dev_net(dev);
2688         struct inet6_dev *idev;
2689         struct inet6_ifaddr *ifa;
2690         int state, i;
2691
2692         ASSERT_RTNL();
2693
2694         rt6_ifdown(net, dev);
2695         neigh_ifdown(&nd_tbl, dev);
2696
2697         idev = __in6_dev_get(dev);
2698         if (idev == NULL)
2699                 return -ENODEV;
2700
2701         /*
2702          * Step 1: remove reference to ipv6 device from parent device.
2703          *         Do not dev_put!
2704          */
2705         if (how) {
2706                 idev->dead = 1;
2707
2708                 /* protected by rtnl_lock */
2709                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2710
2711                 /* Step 1.5: remove snmp6 entry */
2712                 snmp6_unregister_dev(idev);
2713
2714         }
2715
2716         /* Step 2: clear hash table */
2717         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2718                 struct hlist_head *h = &inet6_addr_lst[i];
2719                 struct hlist_node *n;
2720
2721                 spin_lock_bh(&addrconf_hash_lock);
2722         restart:
2723                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2724                         if (ifa->idev == idev) {
2725                                 hlist_del_init_rcu(&ifa->addr_lst);
2726                                 addrconf_del_timer(ifa);
2727                                 goto restart;
2728                         }
2729                 }
2730                 spin_unlock_bh(&addrconf_hash_lock);
2731         }
2732
2733         write_lock_bh(&idev->lock);
2734
2735         /* Step 2: clear flags for stateless addrconf */
2736         if (!how)
2737                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2738
2739 #ifdef CONFIG_IPV6_PRIVACY
2740         if (how && del_timer(&idev->regen_timer))
2741                 in6_dev_put(idev);
2742
2743         /* Step 3: clear tempaddr list */
2744         while (!list_empty(&idev->tempaddr_list)) {
2745                 ifa = list_first_entry(&idev->tempaddr_list,
2746                                        struct inet6_ifaddr, tmp_list);
2747                 list_del(&ifa->tmp_list);
2748                 write_unlock_bh(&idev->lock);
2749                 spin_lock_bh(&ifa->lock);
2750
2751                 if (ifa->ifpub) {
2752                         in6_ifa_put(ifa->ifpub);
2753                         ifa->ifpub = NULL;
2754                 }
2755                 spin_unlock_bh(&ifa->lock);
2756                 in6_ifa_put(ifa);
2757                 write_lock_bh(&idev->lock);
2758         }
2759 #endif
2760
2761         while (!list_empty(&idev->addr_list)) {
2762                 ifa = list_first_entry(&idev->addr_list,
2763                                        struct inet6_ifaddr, if_list);
2764                 addrconf_del_timer(ifa);
2765
2766                 list_del(&ifa->if_list);
2767
2768                 write_unlock_bh(&idev->lock);
2769
2770                 spin_lock_bh(&ifa->state_lock);
2771                 state = ifa->state;
2772                 ifa->state = INET6_IFADDR_STATE_DEAD;
2773                 spin_unlock_bh(&ifa->state_lock);
2774
2775                 if (state != INET6_IFADDR_STATE_DEAD) {
2776                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2777                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2778                 }
2779                 in6_ifa_put(ifa);
2780
2781                 write_lock_bh(&idev->lock);
2782         }
2783
2784         write_unlock_bh(&idev->lock);
2785
2786         /* Step 5: Discard multicast list */
2787         if (how)
2788                 ipv6_mc_destroy_dev(idev);
2789         else
2790                 ipv6_mc_down(idev);
2791
2792         idev->tstamp = jiffies;
2793
2794         /* Last: Shot the device (if unregistered) */
2795         if (how) {
2796                 addrconf_sysctl_unregister(idev);
2797                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2798                 neigh_ifdown(&nd_tbl, dev);
2799                 in6_dev_put(idev);
2800         }
2801         return 0;
2802 }
2803
2804 static void addrconf_rs_timer(unsigned long data)
2805 {
2806         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2807         struct inet6_dev *idev = ifp->idev;
2808
2809         read_lock(&idev->lock);
2810         if (idev->dead || !(idev->if_flags & IF_READY))
2811                 goto out;
2812
2813         if (idev->cnf.forwarding)
2814                 goto out;
2815
2816         /* Announcement received after solicitation was sent */
2817         if (idev->if_flags & IF_RA_RCVD)
2818                 goto out;
2819
2820         spin_lock(&ifp->lock);
2821         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2822                 /* The wait after the last probe can be shorter */
2823                 addrconf_mod_timer(ifp, AC_RS,
2824                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2825                                    idev->cnf.rtr_solicit_delay :
2826                                    idev->cnf.rtr_solicit_interval);
2827                 spin_unlock(&ifp->lock);
2828
2829                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2830         } else {
2831                 spin_unlock(&ifp->lock);
2832                 /*
2833                  * Note: we do not support deprecated "all on-link"
2834                  * assumption any longer.
2835                  */
2836                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2837                        idev->dev->name);
2838         }
2839
2840 out:
2841         read_unlock(&idev->lock);
2842         in6_ifa_put(ifp);
2843 }
2844
2845 /*
2846  *      Duplicate Address Detection
2847  */
2848 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2849 {
2850         unsigned long rand_num;
2851         struct inet6_dev *idev = ifp->idev;
2852
2853         if (ifp->flags & IFA_F_OPTIMISTIC)
2854                 rand_num = 0;
2855         else
2856                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2857
2858         ifp->probes = idev->cnf.dad_transmits;
2859         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2860 }
2861
2862 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2863 {
2864         struct inet6_dev *idev = ifp->idev;
2865         struct net_device *dev = idev->dev;
2866
2867         addrconf_join_solict(dev, &ifp->addr);
2868
2869         net_srandom(ifp->addr.s6_addr32[3]);
2870
2871         read_lock_bh(&idev->lock);
2872         spin_lock(&ifp->lock);
2873         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2874                 goto out;
2875
2876         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2877             idev->cnf.accept_dad < 1 ||
2878             !(ifp->flags&IFA_F_TENTATIVE) ||
2879             ifp->flags & IFA_F_NODAD) {
2880                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2881                 spin_unlock(&ifp->lock);
2882                 read_unlock_bh(&idev->lock);
2883
2884                 addrconf_dad_completed(ifp);
2885                 return;
2886         }
2887
2888         if (!(idev->if_flags & IF_READY)) {
2889                 spin_unlock(&ifp->lock);
2890                 read_unlock_bh(&idev->lock);
2891                 /*
2892                  * If the device is not ready:
2893                  * - keep it tentative if it is a permanent address.
2894                  * - otherwise, kill it.
2895                  */
2896                 in6_ifa_hold(ifp);
2897                 addrconf_dad_stop(ifp, 0);
2898                 return;
2899         }
2900
2901         /*
2902          * Optimistic nodes can start receiving
2903          * Frames right away
2904          */
2905         if (ifp->flags & IFA_F_OPTIMISTIC)
2906                 ip6_ins_rt(ifp->rt);
2907
2908         addrconf_dad_kick(ifp);
2909 out:
2910         spin_unlock(&ifp->lock);
2911         read_unlock_bh(&idev->lock);
2912 }
2913
2914 static void addrconf_dad_timer(unsigned long data)
2915 {
2916         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2917         struct inet6_dev *idev = ifp->idev;
2918         struct in6_addr mcaddr;
2919
2920         if (!ifp->probes && addrconf_dad_end(ifp))
2921                 goto out;
2922
2923         read_lock(&idev->lock);
2924         if (idev->dead || !(idev->if_flags & IF_READY)) {
2925                 read_unlock(&idev->lock);
2926                 goto out;
2927         }
2928
2929         spin_lock(&ifp->lock);
2930         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2931                 spin_unlock(&ifp->lock);
2932                 read_unlock(&idev->lock);
2933                 goto out;
2934         }
2935
2936         if (ifp->probes == 0) {
2937                 /*
2938                  * DAD was successful
2939                  */
2940
2941                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2942                 spin_unlock(&ifp->lock);
2943                 read_unlock(&idev->lock);
2944
2945                 addrconf_dad_completed(ifp);
2946
2947                 goto out;
2948         }
2949
2950         ifp->probes--;
2951         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2952         spin_unlock(&ifp->lock);
2953         read_unlock(&idev->lock);
2954
2955         /* send a neighbour solicitation for our addr */
2956         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2957         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2958 out:
2959         in6_ifa_put(ifp);
2960 }
2961
2962 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2963 {
2964         struct net_device *dev = ifp->idev->dev;
2965
2966         /*
2967          *      Configure the address for reception. Now it is valid.
2968          */
2969
2970         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2971
2972         /* If added prefix is link local and forwarding is off,
2973            start sending router solicitations.
2974          */
2975
2976         if ((ifp->idev->cnf.forwarding == 0 ||
2977              ifp->idev->cnf.forwarding == 2) &&
2978             ifp->idev->cnf.rtr_solicits > 0 &&
2979             (dev->flags&IFF_LOOPBACK) == 0 &&
2980             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2981                 /*
2982                  *      If a host as already performed a random delay
2983                  *      [...] as part of DAD [...] there is no need
2984                  *      to delay again before sending the first RS
2985                  */
2986                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2987
2988                 spin_lock_bh(&ifp->lock);
2989                 ifp->probes = 1;
2990                 ifp->idev->if_flags |= IF_RS_SENT;
2991                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2992                 spin_unlock_bh(&ifp->lock);
2993         }
2994 }
2995
2996 static void addrconf_dad_run(struct inet6_dev *idev)
2997 {
2998         struct inet6_ifaddr *ifp;
2999
3000         read_lock_bh(&idev->lock);
3001         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3002                 spin_lock(&ifp->lock);
3003                 if (ifp->flags & IFA_F_TENTATIVE &&
3004                     ifp->state == INET6_IFADDR_STATE_DAD)
3005                         addrconf_dad_kick(ifp);
3006                 spin_unlock(&ifp->lock);
3007         }
3008         read_unlock_bh(&idev->lock);
3009 }
3010
3011 #ifdef CONFIG_PROC_FS
3012 struct if6_iter_state {
3013         struct seq_net_private p;
3014         int bucket;
3015 };
3016
3017 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3018 {
3019         struct inet6_ifaddr *ifa = NULL;
3020         struct if6_iter_state *state = seq->private;
3021         struct net *net = seq_file_net(seq);
3022
3023         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3024                 struct hlist_node *n;
3025                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3026                                          addr_lst)
3027                         if (net_eq(dev_net(ifa->idev->dev), net))
3028                                 return ifa;
3029         }
3030         return NULL;
3031 }
3032
3033 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3034                                          struct inet6_ifaddr *ifa)
3035 {
3036         struct if6_iter_state *state = seq->private;
3037         struct net *net = seq_file_net(seq);
3038         struct hlist_node *n = &ifa->addr_lst;
3039
3040         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3041                 if (net_eq(dev_net(ifa->idev->dev), net))
3042                         return ifa;
3043
3044         while (++state->bucket < IN6_ADDR_HSIZE) {
3045                 hlist_for_each_entry_rcu_bh(ifa, n,
3046                                      &inet6_addr_lst[state->bucket], addr_lst) {
3047                         if (net_eq(dev_net(ifa->idev->dev), net))
3048                                 return ifa;
3049                 }
3050         }
3051
3052         return NULL;
3053 }
3054
3055 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3056 {
3057         struct inet6_ifaddr *ifa = if6_get_first(seq);
3058
3059         if (ifa)
3060                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3061                         --pos;
3062         return pos ? NULL : ifa;
3063 }
3064
3065 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3066         __acquires(rcu_bh)
3067 {
3068         rcu_read_lock_bh();
3069         return if6_get_idx(seq, *pos);
3070 }
3071
3072 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3073 {
3074         struct inet6_ifaddr *ifa;
3075
3076         ifa = if6_get_next(seq, v);
3077         ++*pos;
3078         return ifa;
3079 }
3080
3081 static void if6_seq_stop(struct seq_file *seq, void *v)
3082         __releases(rcu_bh)
3083 {
3084         rcu_read_unlock_bh();
3085 }
3086
3087 static int if6_seq_show(struct seq_file *seq, void *v)
3088 {
3089         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3090         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3091                    &ifp->addr,
3092                    ifp->idev->dev->ifindex,
3093                    ifp->prefix_len,
3094                    ifp->scope,
3095                    ifp->flags,
3096                    ifp->idev->dev->name);
3097         return 0;
3098 }
3099
3100 static const struct seq_operations if6_seq_ops = {
3101         .start  = if6_seq_start,
3102         .next   = if6_seq_next,
3103         .show   = if6_seq_show,
3104         .stop   = if6_seq_stop,
3105 };
3106
3107 static int if6_seq_open(struct inode *inode, struct file *file)
3108 {
3109         return seq_open_net(inode, file, &if6_seq_ops,
3110                             sizeof(struct if6_iter_state));
3111 }
3112
3113 static const struct file_operations if6_fops = {
3114         .owner          = THIS_MODULE,
3115         .open           = if6_seq_open,
3116         .read           = seq_read,
3117         .llseek         = seq_lseek,
3118         .release        = seq_release_net,
3119 };
3120
3121 static int __net_init if6_proc_net_init(struct net *net)
3122 {
3123         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3124                 return -ENOMEM;
3125         return 0;
3126 }
3127
3128 static void __net_exit if6_proc_net_exit(struct net *net)
3129 {
3130        proc_net_remove(net, "if_inet6");
3131 }
3132
3133 static struct pernet_operations if6_proc_net_ops = {
3134        .init = if6_proc_net_init,
3135        .exit = if6_proc_net_exit,
3136 };
3137
3138 int __init if6_proc_init(void)
3139 {
3140         return register_pernet_subsys(&if6_proc_net_ops);
3141 }
3142
3143 void if6_proc_exit(void)
3144 {
3145         unregister_pernet_subsys(&if6_proc_net_ops);
3146 }
3147 #endif  /* CONFIG_PROC_FS */
3148
3149 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3150 /* Check if address is a home address configured on any interface. */
3151 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3152 {
3153         int ret = 0;
3154         struct inet6_ifaddr *ifp = NULL;
3155         struct hlist_node *n;
3156         unsigned int hash = ipv6_addr_hash(addr);
3157
3158         rcu_read_lock_bh();
3159         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3160                 if (!net_eq(dev_net(ifp->idev->dev), net))
3161                         continue;
3162                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3163                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3164                         ret = 1;
3165                         break;
3166                 }
3167         }
3168         rcu_read_unlock_bh();
3169         return ret;
3170 }
3171 #endif
3172
3173 /*
3174  *      Periodic address status verification
3175  */
3176
3177 static void addrconf_verify(unsigned long foo)
3178 {
3179         unsigned long now, next, next_sec, next_sched;
3180         struct inet6_ifaddr *ifp;
3181         struct hlist_node *node;
3182         int i;
3183
3184         rcu_read_lock_bh();
3185         spin_lock(&addrconf_verify_lock);
3186         now = jiffies;
3187         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3188
3189         del_timer(&addr_chk_timer);
3190
3191         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3192 restart:
3193                 hlist_for_each_entry_rcu_bh(ifp, node,
3194                                          &inet6_addr_lst[i], addr_lst) {
3195                         unsigned long age;
3196
3197                         if (ifp->flags & IFA_F_PERMANENT)
3198                                 continue;
3199
3200                         spin_lock(&ifp->lock);
3201                         /* We try to batch several events at once. */
3202                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3203
3204                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3205                             age >= ifp->valid_lft) {
3206                                 spin_unlock(&ifp->lock);
3207                                 in6_ifa_hold(ifp);
3208                                 ipv6_del_addr(ifp);
3209                                 goto restart;
3210                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3211                                 spin_unlock(&ifp->lock);
3212                                 continue;
3213                         } else if (age >= ifp->prefered_lft) {
3214                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3215                                 int deprecate = 0;
3216
3217                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3218                                         deprecate = 1;
3219                                         ifp->flags |= IFA_F_DEPRECATED;
3220                                 }
3221
3222                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3223                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3224
3225                                 spin_unlock(&ifp->lock);
3226
3227                                 if (deprecate) {
3228                                         in6_ifa_hold(ifp);
3229
3230                                         ipv6_ifa_notify(0, ifp);
3231                                         in6_ifa_put(ifp);
3232                                         goto restart;
3233                                 }
3234 #ifdef CONFIG_IPV6_PRIVACY
3235                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3236                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3237                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3238                                         ifp->idev->cnf.dad_transmits *
3239                                         ifp->idev->nd_parms->retrans_time / HZ;
3240
3241                                 if (age >= ifp->prefered_lft - regen_advance) {
3242                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3243                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3244                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3245                                         if (!ifp->regen_count && ifpub) {
3246                                                 ifp->regen_count++;
3247                                                 in6_ifa_hold(ifp);
3248                                                 in6_ifa_hold(ifpub);
3249                                                 spin_unlock(&ifp->lock);
3250
3251                                                 spin_lock(&ifpub->lock);
3252                                                 ifpub->regen_count = 0;
3253                                                 spin_unlock(&ifpub->lock);
3254                                                 ipv6_create_tempaddr(ifpub, ifp);
3255                                                 in6_ifa_put(ifpub);
3256                                                 in6_ifa_put(ifp);
3257                                                 goto restart;
3258                                         }
3259                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3260                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3261                                 spin_unlock(&ifp->lock);
3262 #endif
3263                         } else {
3264                                 /* ifp->prefered_lft <= ifp->valid_lft */
3265                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3266                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3267                                 spin_unlock(&ifp->lock);
3268                         }
3269                 }
3270         }
3271
3272         next_sec = round_jiffies_up(next);
3273         next_sched = next;
3274
3275         /* If rounded timeout is accurate enough, accept it. */
3276         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3277                 next_sched = next_sec;
3278
3279         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3280         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3281                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3282
3283         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3284               now, next, next_sec, next_sched));
3285
3286         addr_chk_timer.expires = next_sched;
3287         add_timer(&addr_chk_timer);
3288         spin_unlock(&addrconf_verify_lock);
3289         rcu_read_unlock_bh();
3290 }
3291
3292 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3293 {
3294         struct in6_addr *pfx = NULL;
3295
3296         if (addr)
3297                 pfx = nla_data(addr);
3298
3299         if (local) {
3300                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3301                         pfx = NULL;
3302                 else
3303                         pfx = nla_data(local);
3304         }
3305
3306         return pfx;
3307 }
3308
3309 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3310         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3311         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3312         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3313 };
3314
3315 static int
3316 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3317 {
3318         struct net *net = sock_net(skb->sk);
3319         struct ifaddrmsg *ifm;
3320         struct nlattr *tb[IFA_MAX+1];
3321         struct in6_addr *pfx;
3322         int err;
3323
3324         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3325         if (err < 0)
3326                 return err;
3327
3328         ifm = nlmsg_data(nlh);
3329         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3330         if (pfx == NULL)
3331                 return -EINVAL;
3332
3333         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3334 }
3335
3336 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3337                              u32 prefered_lft, u32 valid_lft)
3338 {
3339         u32 flags;
3340         clock_t expires;
3341         unsigned long timeout;
3342
3343         if (!valid_lft || (prefered_lft > valid_lft))
3344                 return -EINVAL;
3345
3346         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3347         if (addrconf_finite_timeout(timeout)) {
3348                 expires = jiffies_to_clock_t(timeout * HZ);
3349                 valid_lft = timeout;
3350                 flags = RTF_EXPIRES;
3351         } else {
3352                 expires = 0;
3353                 flags = 0;
3354                 ifa_flags |= IFA_F_PERMANENT;
3355         }
3356
3357         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3358         if (addrconf_finite_timeout(timeout)) {
3359                 if (timeout == 0)
3360                         ifa_flags |= IFA_F_DEPRECATED;
3361                 prefered_lft = timeout;
3362         }
3363
3364         spin_lock_bh(&ifp->lock);
3365         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3366         ifp->tstamp = jiffies;
3367         ifp->valid_lft = valid_lft;
3368         ifp->prefered_lft = prefered_lft;
3369
3370         spin_unlock_bh(&ifp->lock);
3371         if (!(ifp->flags&IFA_F_TENTATIVE))
3372                 ipv6_ifa_notify(0, ifp);
3373
3374         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3375                               expires, flags);
3376         addrconf_verify(0);
3377
3378         return 0;
3379 }
3380
3381 static int
3382 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3383 {
3384         struct net *net = sock_net(skb->sk);
3385         struct ifaddrmsg *ifm;
3386         struct nlattr *tb[IFA_MAX+1];
3387         struct in6_addr *pfx;
3388         struct inet6_ifaddr *ifa;
3389         struct net_device *dev;
3390         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3391         u8 ifa_flags;
3392         int err;
3393
3394         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3395         if (err < 0)
3396                 return err;
3397
3398         ifm = nlmsg_data(nlh);
3399         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3400         if (pfx == NULL)
3401                 return -EINVAL;
3402
3403         if (tb[IFA_CACHEINFO]) {
3404                 struct ifa_cacheinfo *ci;
3405
3406                 ci = nla_data(tb[IFA_CACHEINFO]);
3407                 valid_lft = ci->ifa_valid;
3408                 preferred_lft = ci->ifa_prefered;
3409         } else {
3410                 preferred_lft = INFINITY_LIFE_TIME;
3411                 valid_lft = INFINITY_LIFE_TIME;
3412         }
3413
3414         dev =  __dev_get_by_index(net, ifm->ifa_index);
3415         if (dev == NULL)
3416                 return -ENODEV;
3417
3418         /* We ignore other flags so far. */
3419         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3420
3421         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3422         if (ifa == NULL) {
3423                 /*
3424                  * It would be best to check for !NLM_F_CREATE here but
3425                  * userspace alreay relies on not having to provide this.
3426                  */
3427                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3428                                       ifm->ifa_prefixlen, ifa_flags,
3429                                       preferred_lft, valid_lft);
3430         }
3431
3432         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3433             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3434                 err = -EEXIST;
3435         else
3436                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3437
3438         in6_ifa_put(ifa);
3439
3440         return err;
3441 }
3442
3443 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3444                           u8 scope, int ifindex)
3445 {
3446         struct ifaddrmsg *ifm;
3447
3448         ifm = nlmsg_data(nlh);
3449         ifm->ifa_family = AF_INET6;
3450         ifm->ifa_prefixlen = prefixlen;
3451         ifm->ifa_flags = flags;
3452         ifm->ifa_scope = scope;
3453         ifm->ifa_index = ifindex;
3454 }
3455
3456 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3457                          unsigned long tstamp, u32 preferred, u32 valid)
3458 {
3459         struct ifa_cacheinfo ci;
3460
3461         ci.cstamp = cstamp_delta(cstamp);
3462         ci.tstamp = cstamp_delta(tstamp);
3463         ci.ifa_prefered = preferred;
3464         ci.ifa_valid = valid;
3465
3466         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3467 }
3468
3469 static inline int rt_scope(int ifa_scope)
3470 {
3471         if (ifa_scope & IFA_HOST)
3472                 return RT_SCOPE_HOST;
3473         else if (ifa_scope & IFA_LINK)
3474                 return RT_SCOPE_LINK;
3475         else if (ifa_scope & IFA_SITE)
3476                 return RT_SCOPE_SITE;
3477         else
3478                 return RT_SCOPE_UNIVERSE;
3479 }
3480
3481 static inline int inet6_ifaddr_msgsize(void)
3482 {
3483         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3484                + nla_total_size(16) /* IFA_ADDRESS */
3485                + nla_total_size(sizeof(struct ifa_cacheinfo));
3486 }
3487
3488 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3489                              u32 pid, u32 seq, int event, unsigned int flags)
3490 {
3491         struct nlmsghdr  *nlh;
3492         u32 preferred, valid;
3493
3494         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3495         if (nlh == NULL)
3496                 return -EMSGSIZE;
3497
3498         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3499                       ifa->idev->dev->ifindex);
3500
3501         if (!(ifa->flags&IFA_F_PERMANENT)) {
3502                 preferred = ifa->prefered_lft;
3503                 valid = ifa->valid_lft;
3504                 if (preferred != INFINITY_LIFE_TIME) {
3505                         long tval = (jiffies - ifa->tstamp)/HZ;
3506                         if (preferred > tval)
3507                                 preferred -= tval;
3508                         else
3509                                 preferred = 0;
3510                         if (valid != INFINITY_LIFE_TIME) {
3511                                 if (valid > tval)
3512                                         valid -= tval;
3513                                 else
3514                                         valid = 0;
3515                         }
3516                 }
3517         } else {
3518                 preferred = INFINITY_LIFE_TIME;
3519                 valid = INFINITY_LIFE_TIME;
3520         }
3521
3522         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3523             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3524                 nlmsg_cancel(skb, nlh);
3525                 return -EMSGSIZE;
3526         }
3527
3528         return nlmsg_end(skb, nlh);
3529 }
3530
3531 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3532                                 u32 pid, u32 seq, int event, u16 flags)
3533 {
3534         struct nlmsghdr  *nlh;
3535         u8 scope = RT_SCOPE_UNIVERSE;
3536         int ifindex = ifmca->idev->dev->ifindex;
3537
3538         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3539                 scope = RT_SCOPE_SITE;
3540
3541         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3542         if (nlh == NULL)
3543                 return -EMSGSIZE;
3544
3545         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3546         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3547             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3548                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3549                 nlmsg_cancel(skb, nlh);
3550                 return -EMSGSIZE;
3551         }
3552
3553         return nlmsg_end(skb, nlh);
3554 }
3555
3556 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3557                                 u32 pid, u32 seq, int event, unsigned int flags)
3558 {
3559         struct nlmsghdr  *nlh;
3560         u8 scope = RT_SCOPE_UNIVERSE;
3561         int ifindex = ifaca->aca_idev->dev->ifindex;
3562
3563         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3564                 scope = RT_SCOPE_SITE;
3565
3566         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3567         if (nlh == NULL)
3568                 return -EMSGSIZE;
3569
3570         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3571         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3572             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3573                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3574                 nlmsg_cancel(skb, nlh);
3575                 return -EMSGSIZE;
3576         }
3577
3578         return nlmsg_end(skb, nlh);
3579 }
3580
3581 enum addr_type_t {
3582         UNICAST_ADDR,
3583         MULTICAST_ADDR,
3584         ANYCAST_ADDR,
3585 };
3586
3587 /* called with rcu_read_lock() */
3588 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3589                           struct netlink_callback *cb, enum addr_type_t type,
3590                           int s_ip_idx, int *p_ip_idx)
3591 {
3592         struct ifmcaddr6 *ifmca;
3593         struct ifacaddr6 *ifaca;
3594         int err = 1;
3595         int ip_idx = *p_ip_idx;
3596
3597         read_lock_bh(&idev->lock);
3598         switch (type) {
3599         case UNICAST_ADDR: {
3600                 struct inet6_ifaddr *ifa;
3601
3602                 /* unicast address incl. temp addr */
3603                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3604                         if (++ip_idx < s_ip_idx)
3605                                 continue;
3606                         err = inet6_fill_ifaddr(skb, ifa,
3607                                                 NETLINK_CB(cb->skb).pid,
3608                                                 cb->nlh->nlmsg_seq,
3609                                                 RTM_NEWADDR,
3610                                                 NLM_F_MULTI);
3611                         if (err <= 0)
3612                                 break;
3613                 }
3614                 break;
3615         }
3616         case MULTICAST_ADDR:
3617                 /* multicast address */
3618                 for (ifmca = idev->mc_list; ifmca;
3619                      ifmca = ifmca->next, ip_idx++) {
3620                         if (ip_idx < s_ip_idx)
3621                                 continue;
3622                         err = inet6_fill_ifmcaddr(skb, ifmca,
3623                                                   NETLINK_CB(cb->skb).pid,
3624                                                   cb->nlh->nlmsg_seq,
3625                                                   RTM_GETMULTICAST,
3626                                                   NLM_F_MULTI);
3627                         if (err <= 0)
3628                                 break;
3629                 }
3630                 break;
3631         case ANYCAST_ADDR:
3632                 /* anycast address */
3633                 for (ifaca = idev->ac_list; ifaca;
3634                      ifaca = ifaca->aca_next, ip_idx++) {
3635                         if (ip_idx < s_ip_idx)
3636                                 continue;
3637                         err = inet6_fill_ifacaddr(skb, ifaca,
3638                                                   NETLINK_CB(cb->skb).pid,
3639                                                   cb->nlh->nlmsg_seq,
3640                                                   RTM_GETANYCAST,
3641                                                   NLM_F_MULTI);
3642                         if (err <= 0)
3643                                 break;
3644                 }
3645                 break;
3646         default:
3647                 break;
3648         }
3649         read_unlock_bh(&idev->lock);
3650         *p_ip_idx = ip_idx;
3651         return err;
3652 }
3653
3654 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3655                            enum addr_type_t type)
3656 {
3657         struct net *net = sock_net(skb->sk);
3658         int h, s_h;
3659         int idx, ip_idx;
3660         int s_idx, s_ip_idx;
3661         struct net_device *dev;
3662         struct inet6_dev *idev;
3663         struct hlist_head *head;
3664         struct hlist_node *node;
3665
3666         s_h = cb->args[0];
3667         s_idx = idx = cb->args[1];
3668         s_ip_idx = ip_idx = cb->args[2];
3669
3670         rcu_read_lock();
3671         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3672                 idx = 0;
3673                 head = &net->dev_index_head[h];
3674                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3675                         if (idx < s_idx)
3676                                 goto cont;
3677                         if (h > s_h || idx > s_idx)
3678                                 s_ip_idx = 0;
3679                         ip_idx = 0;
3680                         idev = __in6_dev_get(dev);
3681                         if (!idev)
3682                                 goto cont;
3683
3684                         if (in6_dump_addrs(idev, skb, cb, type,
3685                                            s_ip_idx, &ip_idx) <= 0)
3686                                 goto done;
3687 cont:
3688                         idx++;
3689                 }
3690         }
3691 done:
3692         rcu_read_unlock();
3693         cb->args[0] = h;
3694         cb->args[1] = idx;
3695         cb->args[2] = ip_idx;
3696
3697         return skb->len;
3698 }
3699
3700 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3701 {
3702         enum addr_type_t type = UNICAST_ADDR;
3703
3704         return inet6_dump_addr(skb, cb, type);
3705 }
3706
3707 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3708 {
3709         enum addr_type_t type = MULTICAST_ADDR;
3710
3711         return inet6_dump_addr(skb, cb, type);
3712 }
3713
3714
3715 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3716 {
3717         enum addr_type_t type = ANYCAST_ADDR;
3718
3719         return inet6_dump_addr(skb, cb, type);
3720 }
3721
3722 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3723                              void *arg)
3724 {
3725         struct net *net = sock_net(in_skb->sk);
3726         struct ifaddrmsg *ifm;
3727         struct nlattr *tb[IFA_MAX+1];
3728         struct in6_addr *addr = NULL;
3729         struct net_device *dev = NULL;
3730         struct inet6_ifaddr *ifa;
3731         struct sk_buff *skb;
3732         int err;
3733
3734         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3735         if (err < 0)
3736                 goto errout;
3737
3738         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3739         if (addr == NULL) {
3740                 err = -EINVAL;
3741                 goto errout;
3742         }
3743
3744         ifm = nlmsg_data(nlh);
3745         if (ifm->ifa_index)
3746                 dev = __dev_get_by_index(net, ifm->ifa_index);
3747
3748         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3749         if (!ifa) {
3750                 err = -EADDRNOTAVAIL;
3751                 goto errout;
3752         }
3753
3754         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3755         if (!skb) {
3756                 err = -ENOBUFS;
3757                 goto errout_ifa;
3758         }
3759
3760         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3761                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3762         if (err < 0) {
3763                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3764                 WARN_ON(err == -EMSGSIZE);
3765                 kfree_skb(skb);
3766                 goto errout_ifa;
3767         }
3768         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3769 errout_ifa:
3770         in6_ifa_put(ifa);
3771 errout:
3772         return err;
3773 }
3774
3775 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3776 {
3777         struct sk_buff *skb;
3778         struct net *net = dev_net(ifa->idev->dev);
3779         int err = -ENOBUFS;
3780
3781         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3782         if (skb == NULL)
3783                 goto errout;
3784
3785         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3786         if (err < 0) {
3787                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3788                 WARN_ON(err == -EMSGSIZE);
3789                 kfree_skb(skb);
3790                 goto errout;
3791         }
3792         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3793         return;
3794 errout:
3795         if (err < 0)
3796                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3797 }
3798
3799 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3800                                 __s32 *array, int bytes)
3801 {
3802         BUG_ON(bytes < (DEVCONF_MAX * 4));
3803
3804         memset(array, 0, bytes);
3805         array[DEVCONF_FORWARDING] = cnf->forwarding;
3806         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3807         array[DEVCONF_MTU6] = cnf->mtu6;
3808         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3809         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3810         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3811         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3812         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3813         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3814                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3815         array[DEVCONF_RTR_SOLICIT_DELAY] =
3816                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3817         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3818 #ifdef CONFIG_IPV6_PRIVACY
3819         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3820         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3821         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3822         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3823         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3824 #endif
3825         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3826         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3827         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3828 #ifdef CONFIG_IPV6_ROUTER_PREF
3829         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3830         array[DEVCONF_RTR_PROBE_INTERVAL] =
3831                 jiffies_to_msecs(cnf->rtr_probe_interval);
3832 #ifdef CONFIG_IPV6_ROUTE_INFO
3833         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3834 #endif
3835 #endif
3836         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3837         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3838 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3839         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3840 #endif
3841 #ifdef CONFIG_IPV6_MROUTE
3842         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3843 #endif
3844         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3845         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3846         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3847 }
3848
3849 static inline size_t inet6_ifla6_size(void)
3850 {
3851         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3852              + nla_total_size(sizeof(struct ifla_cacheinfo))
3853              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3854              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3855              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3856 }
3857
3858 static inline size_t inet6_if_nlmsg_size(void)
3859 {
3860         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3861                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3862                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3863                + nla_total_size(4) /* IFLA_MTU */
3864                + nla_total_size(4) /* IFLA_LINK */
3865                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3866 }
3867
3868 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3869                                       int items, int bytes)
3870 {
3871         int i;
3872         int pad = bytes - sizeof(u64) * items;
3873         BUG_ON(pad < 0);
3874
3875         /* Use put_unaligned() because stats may not be aligned for u64. */
3876         put_unaligned(items, &stats[0]);
3877         for (i = 1; i < items; i++)
3878                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3879
3880         memset(&stats[items], 0, pad);
3881 }
3882
3883 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3884                                       int items, int bytes, size_t syncpoff)
3885 {
3886         int i;
3887         int pad = bytes - sizeof(u64) * items;
3888         BUG_ON(pad < 0);
3889
3890         /* Use put_unaligned() because stats may not be aligned for u64. */
3891         put_unaligned(items, &stats[0]);
3892         for (i = 1; i < items; i++)
3893                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3894
3895         memset(&stats[items], 0, pad);
3896 }
3897
3898 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3899                              int bytes)
3900 {
3901         switch (attrtype) {
3902         case IFLA_INET6_STATS:
3903                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3904                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3905                 break;
3906         case IFLA_INET6_ICMP6STATS:
3907                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3908                 break;
3909         }
3910 }
3911
3912 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3913 {
3914         struct nlattr *nla;
3915         struct ifla_cacheinfo ci;
3916
3917         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3918
3919         ci.max_reasm_len = IPV6_MAXPLEN;
3920         ci.tstamp = cstamp_delta(idev->tstamp);
3921         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3922         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3923         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3924
3925         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3926         if (nla == NULL)
3927                 goto nla_put_failure;
3928         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3929
3930         /* XXX - MC not implemented */
3931
3932         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3933         if (nla == NULL)
3934                 goto nla_put_failure;
3935         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3936
3937         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3938         if (nla == NULL)
3939                 goto nla_put_failure;
3940         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3941
3942         return 0;
3943
3944 nla_put_failure:
3945         return -EMSGSIZE;
3946 }
3947
3948 static size_t inet6_get_link_af_size(const struct net_device *dev)
3949 {
3950         if (!__in6_dev_get(dev))
3951                 return 0;
3952
3953         return inet6_ifla6_size();
3954 }
3955
3956 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3957 {
3958         struct inet6_dev *idev = __in6_dev_get(dev);
3959
3960         if (!idev)
3961                 return -ENODATA;
3962
3963         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3964                 return -EMSGSIZE;
3965
3966         return 0;
3967 }
3968
3969 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3970                              u32 pid, u32 seq, int event, unsigned int flags)
3971 {
3972         struct net_device *dev = idev->dev;
3973         struct ifinfomsg *hdr;
3974         struct nlmsghdr *nlh;
3975         void *protoinfo;
3976
3977         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3978         if (nlh == NULL)
3979                 return -EMSGSIZE;
3980
3981         hdr = nlmsg_data(nlh);
3982         hdr->ifi_family = AF_INET6;
3983         hdr->__ifi_pad = 0;
3984         hdr->ifi_type = dev->type;
3985         hdr->ifi_index = dev->ifindex;
3986         hdr->ifi_flags = dev_get_flags(dev);
3987         hdr->ifi_change = 0;
3988
3989         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3990
3991         if (dev->addr_len)
3992                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3993
3994         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3995         if (dev->ifindex != dev->iflink)
3996                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3997
3998         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3999         if (protoinfo == NULL)
4000                 goto nla_put_failure;
4001
4002         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4003                 goto nla_put_failure;
4004
4005         nla_nest_end(skb, protoinfo);
4006         return nlmsg_end(skb, nlh);
4007
4008 nla_put_failure:
4009         nlmsg_cancel(skb, nlh);
4010         return -EMSGSIZE;
4011 }
4012
4013 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4014 {
4015         struct net *net = sock_net(skb->sk);
4016         int h, s_h;
4017         int idx = 0, s_idx;
4018         struct net_device *dev;
4019         struct inet6_dev *idev;
4020         struct hlist_head *head;
4021         struct hlist_node *node;
4022
4023         s_h = cb->args[0];
4024         s_idx = cb->args[1];
4025
4026         rcu_read_lock();
4027         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4028                 idx = 0;
4029                 head = &net->dev_index_head[h];
4030                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4031                         if (idx < s_idx)
4032                                 goto cont;
4033                         idev = __in6_dev_get(dev);
4034                         if (!idev)
4035                                 goto cont;
4036                         if (inet6_fill_ifinfo(skb, idev,
4037                                               NETLINK_CB(cb->skb).pid,
4038                                               cb->nlh->nlmsg_seq,
4039                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4040                                 goto out;
4041 cont:
4042                         idx++;
4043                 }
4044         }
4045 out:
4046         rcu_read_unlock();
4047         cb->args[1] = idx;
4048         cb->args[0] = h;
4049
4050         return skb->len;
4051 }
4052
4053 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4054 {
4055         struct sk_buff *skb;
4056         struct net *net = dev_net(idev->dev);
4057         int err = -ENOBUFS;
4058
4059         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4060         if (skb == NULL)
4061                 goto errout;
4062
4063         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4064         if (err < 0) {
4065                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4066                 WARN_ON(err == -EMSGSIZE);
4067                 kfree_skb(skb);
4068                 goto errout;
4069         }
4070         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4071         return;
4072 errout:
4073         if (err < 0)
4074                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4075 }
4076
4077 static inline size_t inet6_prefix_nlmsg_size(void)
4078 {
4079         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4080                + nla_total_size(sizeof(struct in6_addr))
4081                + nla_total_size(sizeof(struct prefix_cacheinfo));
4082 }
4083
4084 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4085                              struct prefix_info *pinfo, u32 pid, u32 seq,
4086                              int event, unsigned int flags)
4087 {
4088         struct prefixmsg *pmsg;
4089         struct nlmsghdr *nlh;
4090         struct prefix_cacheinfo ci;
4091
4092         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4093         if (nlh == NULL)
4094                 return -EMSGSIZE;
4095
4096         pmsg = nlmsg_data(nlh);
4097         pmsg->prefix_family = AF_INET6;
4098         pmsg->prefix_pad1 = 0;
4099         pmsg->prefix_pad2 = 0;
4100         pmsg->prefix_ifindex = idev->dev->ifindex;
4101         pmsg->prefix_len = pinfo->prefix_len;
4102         pmsg->prefix_type = pinfo->type;
4103         pmsg->prefix_pad3 = 0;
4104         pmsg->prefix_flags = 0;
4105         if (pinfo->onlink)
4106                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4107         if (pinfo->autoconf)
4108                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4109
4110         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4111
4112         ci.preferred_time = ntohl(pinfo->prefered);
4113         ci.valid_time = ntohl(pinfo->valid);
4114         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4115
4116         return nlmsg_end(skb, nlh);
4117
4118 nla_put_failure:
4119         nlmsg_cancel(skb, nlh);
4120         return -EMSGSIZE;
4121 }
4122
4123 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4124                          struct prefix_info *pinfo)
4125 {
4126         struct sk_buff *skb;
4127         struct net *net = dev_net(idev->dev);
4128         int err = -ENOBUFS;
4129
4130         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4131         if (skb == NULL)
4132                 goto errout;
4133
4134         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4135         if (err < 0) {
4136                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4137                 WARN_ON(err == -EMSGSIZE);
4138                 kfree_skb(skb);
4139                 goto errout;
4140         }
4141         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4142         return;
4143 errout:
4144         if (err < 0)
4145                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4146 }
4147
4148 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4149 {
4150         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4151
4152         switch (event) {
4153         case RTM_NEWADDR:
4154                 /*
4155                  * If the address was optimistic
4156                  * we inserted the route at the start of
4157                  * our DAD process, so we don't need
4158                  * to do it again
4159                  */
4160                 if (!(ifp->rt->rt6i_node))
4161                         ip6_ins_rt(ifp->rt);
4162                 if (ifp->idev->cnf.forwarding)
4163                         addrconf_join_anycast(ifp);
4164                 break;
4165         case RTM_DELADDR:
4166                 if (ifp->idev->cnf.forwarding)
4167                         addrconf_leave_anycast(ifp);
4168                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4169                 dst_hold(&ifp->rt->dst);
4170
4171                 if (ip6_del_rt(ifp->rt))
4172                         dst_free(&ifp->rt->dst);
4173                 break;
4174         }
4175 }
4176
4177 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4178 {
4179         rcu_read_lock_bh();
4180         if (likely(ifp->idev->dead == 0))
4181                 __ipv6_ifa_notify(event, ifp);
4182         rcu_read_unlock_bh();
4183 }
4184
4185 #ifdef CONFIG_SYSCTL
4186
4187 static
4188 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4189                            void __user *buffer, size_t *lenp, loff_t *ppos)
4190 {
4191         int *valp = ctl->data;
4192         int val = *valp;
4193         loff_t pos = *ppos;
4194         int ret;
4195
4196         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4197
4198         if (write)
4199                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4200         if (ret)
4201                 *ppos = pos;
4202         return ret;
4203 }
4204
4205 static void dev_disable_change(struct inet6_dev *idev)
4206 {
4207         if (!idev || !idev->dev)
4208                 return;
4209
4210         if (idev->cnf.disable_ipv6)
4211                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4212         else
4213                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4214 }
4215
4216 static void addrconf_disable_change(struct net *net, __s32 newf)
4217 {
4218         struct net_device *dev;
4219         struct inet6_dev *idev;
4220
4221         rcu_read_lock();
4222         for_each_netdev_rcu(net, dev) {
4223                 idev = __in6_dev_get(dev);
4224                 if (idev) {
4225                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4226                         idev->cnf.disable_ipv6 = newf;
4227                         if (changed)
4228                                 dev_disable_change(idev);
4229                 }
4230         }
4231         rcu_read_unlock();
4232 }
4233
4234 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4235 {
4236         struct net *net;
4237
4238         net = (struct net *)table->extra2;
4239
4240         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4241                 return 0;
4242
4243         if (!rtnl_trylock()) {
4244                 /* Restore the original values before restarting */
4245                 *p = old;
4246                 return restart_syscall();
4247         }
4248
4249         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4250                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4251                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4252                 addrconf_disable_change(net, newf);
4253         } else if ((!*p) ^ (!old))
4254                 dev_disable_change((struct inet6_dev *)table->extra1);
4255
4256         rtnl_unlock();
4257         return 0;
4258 }
4259
4260 static
4261 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4262                             void __user *buffer, size_t *lenp, loff_t *ppos)
4263 {
4264         int *valp = ctl->data;
4265         int val = *valp;
4266         loff_t pos = *ppos;
4267         int ret;
4268
4269         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4270
4271         if (write)
4272                 ret = addrconf_disable_ipv6(ctl, valp, val);
4273         if (ret)
4274                 *ppos = pos;
4275         return ret;
4276 }
4277
4278 static struct addrconf_sysctl_table
4279 {
4280         struct ctl_table_header *sysctl_header;
4281         ctl_table addrconf_vars[DEVCONF_MAX+1];
4282         char *dev_name;
4283 } addrconf_sysctl __read_mostly = {
4284         .sysctl_header = NULL,
4285         .addrconf_vars = {
4286                 {
4287                         .procname       = "forwarding",
4288                         .data           = &ipv6_devconf.forwarding,
4289                         .maxlen         = sizeof(int),
4290                         .mode           = 0644,
4291                         .proc_handler   = addrconf_sysctl_forward,
4292                 },
4293                 {
4294                         .procname       = "hop_limit",
4295                         .data           = &ipv6_devconf.hop_limit,
4296                         .maxlen         = sizeof(int),
4297                         .mode           = 0644,
4298                         .proc_handler   = proc_dointvec,
4299                 },
4300                 {
4301                         .procname       = "mtu",
4302                         .data           = &ipv6_devconf.mtu6,
4303                         .maxlen         = sizeof(int),
4304                         .mode           = 0644,
4305                         .proc_handler   = proc_dointvec,
4306                 },
4307                 {
4308                         .procname       = "accept_ra",
4309                         .data           = &ipv6_devconf.accept_ra,
4310                         .maxlen         = sizeof(int),
4311                         .mode           = 0644,
4312                         .proc_handler   = proc_dointvec,
4313                 },
4314                 {
4315                         .procname       = "accept_redirects",
4316                         .data           = &ipv6_devconf.accept_redirects,
4317                         .maxlen         = sizeof(int),
4318                         .mode           = 0644,
4319                         .proc_handler   = proc_dointvec,
4320                 },
4321                 {
4322                         .procname       = "autoconf",
4323                         .data           = &ipv6_devconf.autoconf,
4324                         .maxlen         = sizeof(int),
4325                         .mode           = 0644,
4326                         .proc_handler   = proc_dointvec,
4327                 },
4328                 {
4329                         .procname       = "dad_transmits",
4330                         .data           = &ipv6_devconf.dad_transmits,
4331                         .maxlen         = sizeof(int),
4332                         .mode           = 0644,
4333                         .proc_handler   = proc_dointvec,
4334                 },
4335                 {
4336                         .procname       = "router_solicitations",
4337                         .data           = &ipv6_devconf.rtr_solicits,
4338                         .maxlen         = sizeof(int),
4339                         .mode           = 0644,
4340                         .proc_handler   = proc_dointvec,
4341                 },
4342                 {
4343                         .procname       = "router_solicitation_interval",
4344                         .data           = &ipv6_devconf.rtr_solicit_interval,
4345                         .maxlen         = sizeof(int),
4346                         .mode           = 0644,
4347                         .proc_handler   = proc_dointvec_jiffies,
4348                 },
4349                 {
4350                         .procname       = "router_solicitation_delay",
4351                         .data           = &ipv6_devconf.rtr_solicit_delay,
4352                         .maxlen         = sizeof(int),
4353                         .mode           = 0644,
4354                         .proc_handler   = proc_dointvec_jiffies,
4355                 },
4356                 {
4357                         .procname       = "force_mld_version",
4358                         .data           = &ipv6_devconf.force_mld_version,
4359                         .maxlen         = sizeof(int),
4360                         .mode           = 0644,
4361                         .proc_handler   = proc_dointvec,
4362                 },
4363 #ifdef CONFIG_IPV6_PRIVACY
4364                 {
4365                         .procname       = "use_tempaddr",
4366                         .data           = &ipv6_devconf.use_tempaddr,
4367                         .maxlen         = sizeof(int),
4368                         .mode           = 0644,
4369                         .proc_handler   = proc_dointvec,
4370                 },
4371                 {
4372                         .procname       = "temp_valid_lft",
4373                         .data           = &ipv6_devconf.temp_valid_lft,
4374                         .maxlen         = sizeof(int),
4375                         .mode           = 0644,
4376                         .proc_handler   = proc_dointvec,
4377                 },
4378                 {
4379                         .procname       = "temp_prefered_lft",
4380                         .data           = &ipv6_devconf.temp_prefered_lft,
4381                         .maxlen         = sizeof(int),
4382                         .mode           = 0644,
4383                         .proc_handler   = proc_dointvec,
4384                 },
4385                 {
4386                         .procname       = "regen_max_retry",
4387                         .data           = &ipv6_devconf.regen_max_retry,
4388                         .maxlen         = sizeof(int),
4389                         .mode           = 0644,
4390                         .proc_handler   = proc_dointvec,
4391                 },
4392                 {
4393                         .procname       = "max_desync_factor",
4394                         .data           = &ipv6_devconf.max_desync_factor,
4395                         .maxlen         = sizeof(int),
4396                         .mode           = 0644,
4397                         .proc_handler   = proc_dointvec,
4398                 },
4399 #endif
4400                 {
4401                         .procname       = "max_addresses",
4402                         .data           = &ipv6_devconf.max_addresses,
4403                         .maxlen         = sizeof(int),
4404                         .mode           = 0644,
4405                         .proc_handler   = proc_dointvec,
4406                 },
4407                 {
4408                         .procname       = "accept_ra_defrtr",
4409                         .data           = &ipv6_devconf.accept_ra_defrtr,
4410                         .maxlen         = sizeof(int),
4411                         .mode           = 0644,
4412                         .proc_handler   = proc_dointvec,
4413                 },
4414                 {
4415                         .procname       = "accept_ra_pinfo",
4416                         .data           = &ipv6_devconf.accept_ra_pinfo,
4417                         .maxlen         = sizeof(int),
4418                         .mode           = 0644,
4419                         .proc_handler   = proc_dointvec,
4420                 },
4421 #ifdef CONFIG_IPV6_ROUTER_PREF
4422                 {
4423                         .procname       = "accept_ra_rtr_pref",
4424                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4425                         .maxlen         = sizeof(int),
4426                         .mode           = 0644,
4427                         .proc_handler   = proc_dointvec,
4428                 },
4429                 {
4430                         .procname       = "router_probe_interval",
4431                         .data           = &ipv6_devconf.rtr_probe_interval,
4432                         .maxlen         = sizeof(int),
4433                         .mode           = 0644,
4434                         .proc_handler   = proc_dointvec_jiffies,
4435                 },
4436 #ifdef CONFIG_IPV6_ROUTE_INFO
4437                 {
4438                         .procname       = "accept_ra_rt_info_max_plen",
4439                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4440                         .maxlen         = sizeof(int),
4441                         .mode           = 0644,
4442                         .proc_handler   = proc_dointvec,
4443                 },
4444 #endif
4445 #endif
4446                 {
4447                         .procname       = "proxy_ndp",
4448                         .data           = &ipv6_devconf.proxy_ndp,
4449                         .maxlen         = sizeof(int),
4450                         .mode           = 0644,
4451                         .proc_handler   = proc_dointvec,
4452                 },
4453                 {
4454                         .procname       = "accept_source_route",
4455                         .data           = &ipv6_devconf.accept_source_route,
4456                         .maxlen         = sizeof(int),
4457                         .mode           = 0644,
4458                         .proc_handler   = proc_dointvec,
4459                 },
4460 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4461                 {
4462                         .procname       = "optimistic_dad",
4463                         .data           = &ipv6_devconf.optimistic_dad,
4464                         .maxlen         = sizeof(int),
4465                         .mode           = 0644,
4466                         .proc_handler   = proc_dointvec,
4467
4468                 },
4469 #endif
4470 #ifdef CONFIG_IPV6_MROUTE
4471                 {
4472                         .procname       = "mc_forwarding",
4473                         .data           = &ipv6_devconf.mc_forwarding,
4474                         .maxlen         = sizeof(int),
4475                         .mode           = 0444,
4476                         .proc_handler   = proc_dointvec,
4477                 },
4478 #endif
4479                 {
4480                         .procname       = "disable_ipv6",
4481                         .data           = &ipv6_devconf.disable_ipv6,
4482                         .maxlen         = sizeof(int),
4483                         .mode           = 0644,
4484                         .proc_handler   = addrconf_sysctl_disable,
4485                 },
4486                 {
4487                         .procname       = "accept_dad",
4488                         .data           = &ipv6_devconf.accept_dad,
4489                         .maxlen         = sizeof(int),
4490                         .mode           = 0644,
4491                         .proc_handler   = proc_dointvec,
4492                 },
4493                 {
4494                         .procname       = "force_tllao",
4495                         .data           = &ipv6_devconf.force_tllao,
4496                         .maxlen         = sizeof(int),
4497                         .mode           = 0644,
4498                         .proc_handler   = proc_dointvec
4499                 },
4500                 {
4501                         /* sentinel */
4502                 }
4503         },
4504 };
4505
4506 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4507                 struct inet6_dev *idev, struct ipv6_devconf *p)
4508 {
4509         int i;
4510         struct addrconf_sysctl_table *t;
4511
4512 #define ADDRCONF_CTL_PATH_DEV   3
4513
4514         struct ctl_path addrconf_ctl_path[] = {
4515                 { .procname = "net", },
4516                 { .procname = "ipv6", },
4517                 { .procname = "conf", },
4518                 { /* to be set */ },
4519                 { },
4520         };
4521
4522
4523         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4524         if (t == NULL)
4525                 goto out;
4526
4527         for (i = 0; t->addrconf_vars[i].data; i++) {
4528                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4529                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4530                 t->addrconf_vars[i].extra2 = net;
4531         }
4532
4533         /*
4534          * Make a copy of dev_name, because '.procname' is regarded as const
4535          * by sysctl and we wouldn't want anyone to change it under our feet
4536          * (see SIOCSIFNAME).
4537          */
4538         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4539         if (!t->dev_name)
4540                 goto free;
4541
4542         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4543
4544         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4545                         t->addrconf_vars);
4546         if (t->sysctl_header == NULL)
4547                 goto free_procname;
4548
4549         p->sysctl = t;
4550         return 0;
4551
4552 free_procname:
4553         kfree(t->dev_name);
4554 free:
4555         kfree(t);
4556 out:
4557         return -ENOBUFS;
4558 }
4559
4560 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4561 {
4562         struct addrconf_sysctl_table *t;
4563
4564         if (p->sysctl == NULL)
4565                 return;
4566
4567         t = p->sysctl;
4568         p->sysctl = NULL;
4569         unregister_net_sysctl_table(t->sysctl_header);
4570         kfree(t->dev_name);
4571         kfree(t);
4572 }
4573
4574 static void addrconf_sysctl_register(struct inet6_dev *idev)
4575 {
4576         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4577                               &ndisc_ifinfo_sysctl_change);
4578         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4579                                         idev, &idev->cnf);
4580 }
4581
4582 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4583 {
4584         __addrconf_sysctl_unregister(&idev->cnf);
4585         neigh_sysctl_unregister(idev->nd_parms);
4586 }
4587
4588
4589 #endif
4590
4591 static int __net_init addrconf_init_net(struct net *net)
4592 {
4593         int err;
4594         struct ipv6_devconf *all, *dflt;
4595
4596         err = -ENOMEM;
4597         all = &ipv6_devconf;
4598         dflt = &ipv6_devconf_dflt;
4599
4600         if (!net_eq(net, &init_net)) {
4601                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4602                 if (all == NULL)
4603                         goto err_alloc_all;
4604
4605                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4606                 if (dflt == NULL)
4607                         goto err_alloc_dflt;
4608         } else {
4609                 /* these will be inherited by all namespaces */
4610                 dflt->autoconf = ipv6_defaults.autoconf;
4611                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4612         }
4613
4614         net->ipv6.devconf_all = all;
4615         net->ipv6.devconf_dflt = dflt;
4616
4617 #ifdef CONFIG_SYSCTL
4618         err = __addrconf_sysctl_register(net, "all", NULL, all);
4619         if (err < 0)
4620                 goto err_reg_all;
4621
4622         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4623         if (err < 0)
4624                 goto err_reg_dflt;
4625 #endif
4626         return 0;
4627
4628 #ifdef CONFIG_SYSCTL
4629 err_reg_dflt:
4630         __addrconf_sysctl_unregister(all);
4631 err_reg_all:
4632         kfree(dflt);
4633 #endif
4634 err_alloc_dflt:
4635         kfree(all);
4636 err_alloc_all:
4637         return err;
4638 }
4639
4640 static void __net_exit addrconf_exit_net(struct net *net)
4641 {
4642 #ifdef CONFIG_SYSCTL
4643         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4644         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4645 #endif
4646         if (!net_eq(net, &init_net)) {
4647                 kfree(net->ipv6.devconf_dflt);
4648                 kfree(net->ipv6.devconf_all);
4649         }
4650 }
4651
4652 static struct pernet_operations addrconf_ops = {
4653         .init = addrconf_init_net,
4654         .exit = addrconf_exit_net,
4655 };
4656
4657 /*
4658  *      Device notifier
4659  */
4660
4661 int register_inet6addr_notifier(struct notifier_block *nb)
4662 {
4663         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4664 }
4665 EXPORT_SYMBOL(register_inet6addr_notifier);
4666
4667 int unregister_inet6addr_notifier(struct notifier_block *nb)
4668 {
4669         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4670 }
4671 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4672
4673 static struct rtnl_af_ops inet6_ops = {
4674         .family           = AF_INET6,
4675         .fill_link_af     = inet6_fill_link_af,
4676         .get_link_af_size = inet6_get_link_af_size,
4677 };
4678
4679 /*
4680  *      Init / cleanup code
4681  */
4682
4683 int __init addrconf_init(void)
4684 {
4685         int i, err;
4686
4687         err = ipv6_addr_label_init();
4688         if (err < 0) {
4689                 printk(KERN_CRIT "IPv6 Addrconf:"
4690                        " cannot initialize default policy table: %d.\n", err);
4691                 goto out;
4692         }
4693
4694         err = register_pernet_subsys(&addrconf_ops);
4695         if (err < 0)
4696                 goto out_addrlabel;
4697
4698         /* The addrconf netdev notifier requires that loopback_dev
4699          * has it's ipv6 private information allocated and setup
4700          * before it can bring up and give link-local addresses
4701          * to other devices which are up.
4702          *
4703          * Unfortunately, loopback_dev is not necessarily the first
4704          * entry in the global dev_base list of net devices.  In fact,
4705          * it is likely to be the very last entry on that list.
4706          * So this causes the notifier registry below to try and
4707          * give link-local addresses to all devices besides loopback_dev
4708          * first, then loopback_dev, which cases all the non-loopback_dev
4709          * devices to fail to get a link-local address.
4710          *
4711          * So, as a temporary fix, allocate the ipv6 structure for
4712          * loopback_dev first by hand.
4713          * Longer term, all of the dependencies ipv6 has upon the loopback
4714          * device and it being up should be removed.
4715          */
4716         rtnl_lock();
4717         if (!ipv6_add_dev(init_net.loopback_dev))
4718                 err = -ENOMEM;
4719         rtnl_unlock();
4720         if (err)
4721                 goto errlo;
4722
4723         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4724                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4725
4726         register_netdevice_notifier(&ipv6_dev_notf);
4727
4728         addrconf_verify(0);
4729
4730         err = rtnl_af_register(&inet6_ops);
4731         if (err < 0)
4732                 goto errout_af;
4733
4734         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4735                               NULL);
4736         if (err < 0)
4737                 goto errout;
4738
4739         /* Only the first call to __rtnl_register can fail */
4740         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4741         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4742         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4743                         inet6_dump_ifaddr, NULL);
4744         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4745                         inet6_dump_ifmcaddr, NULL);
4746         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4747                         inet6_dump_ifacaddr, NULL);
4748
4749         ipv6_addr_label_rtnl_register();
4750
4751         return 0;
4752 errout:
4753         rtnl_af_unregister(&inet6_ops);
4754 errout_af:
4755         unregister_netdevice_notifier(&ipv6_dev_notf);
4756 errlo:
4757         unregister_pernet_subsys(&addrconf_ops);
4758 out_addrlabel:
4759         ipv6_addr_label_cleanup();
4760 out:
4761         return err;
4762 }
4763
4764 void addrconf_cleanup(void)
4765 {
4766         struct net_device *dev;
4767         int i;
4768
4769         unregister_netdevice_notifier(&ipv6_dev_notf);
4770         unregister_pernet_subsys(&addrconf_ops);
4771         ipv6_addr_label_cleanup();
4772
4773         rtnl_lock();
4774
4775         __rtnl_af_unregister(&inet6_ops);
4776
4777         /* clean dev list */
4778         for_each_netdev(&init_net, dev) {
4779                 if (__in6_dev_get(dev) == NULL)
4780                         continue;
4781                 addrconf_ifdown(dev, 1);
4782         }
4783         addrconf_ifdown(init_net.loopback_dev, 2);
4784
4785         /*
4786          *      Check hash table.
4787          */
4788         spin_lock_bh(&addrconf_hash_lock);
4789         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4790                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4791         spin_unlock_bh(&addrconf_hash_lock);
4792
4793         del_timer(&addr_chk_timer);
4794         rtnl_unlock();
4795 }