net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90 #include <linux/export.h>
  91
  92 /* Set to 3 to get tracing... */
  93 #define ACONF_DEBUG 2
  94
  95 #if ACONF_DEBUG >= 3
  96 #define ADBG(x) printk x
  97 #else
  98 #define ADBG(x)
  99 #endif
 100
 101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 102
 103 static inline u32 cstamp_delta(unsigned long cstamp)
 104 {
 105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 106 }
 107
 108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 111
 112 #ifdef CONFIG_SYSCTL
 113 static void addrconf_sysctl_register(struct inet6_dev *idev);
 114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 115 #else
 116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 117 {
 118 }
 119
 120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 121 {
 122 }
 123 #endif
 124
 125 #ifdef CONFIG_IPV6_PRIVACY
 126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 128 static void ipv6_regen_rndid(unsigned long data);
 129 #endif
 130
 131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 132 static int ipv6_count_addresses(struct inet6_dev *idev);
 133
 134 /*
 135  *      Configured unicast address hash table
 136  */
 137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 138 static DEFINE_SPINLOCK(addrconf_hash_lock);
 139
 140 static void addrconf_verify(unsigned long);
 141
 142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 143 static DEFINE_SPINLOCK(addrconf_verify_lock);
 144
 145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 147
 148 static void addrconf_type_change(struct net_device *dev,
 149                                  unsigned long event);
 150 static int addrconf_ifdown(struct net_device *dev, int how);
 151
 152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 153 static void addrconf_dad_timer(unsigned long data);
 154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 155 static void addrconf_dad_run(struct inet6_dev *idev);
 156 static void addrconf_rs_timer(unsigned long data);
 157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159
 160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 161                                 struct prefix_info *pinfo);
 162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 163                                struct net_device *dev);
 164
 165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 166
 167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 168         .forwarding             = 0,
 169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 170         .mtu6                   = IPV6_MIN_MTU,
 171         .accept_ra              = 1,
 172         .accept_redirects       = 1,
 173         .autoconf               = 1,
 174         .force_mld_version      = 0,
 175         .dad_transmits          = 1,
 176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 179 #ifdef CONFIG_IPV6_PRIVACY
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185 #endif
 186         .max_addresses          = IPV6_MAX_ADDRESSES,
 187         .accept_ra_defrtr       = 1,
 188         .accept_ra_pinfo        = 1,
 189 #ifdef CONFIG_IPV6_ROUTER_PREF
 190         .accept_ra_rtr_pref     = 1,
 191         .rtr_probe_interval     = 60 * HZ,
 192 #ifdef CONFIG_IPV6_ROUTE_INFO
 193         .accept_ra_rt_info_max_plen = 0,
 194 #endif
 195 #endif
 196         .proxy_ndp              = 0,
 197         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 198         .disable_ipv6           = 0,
 199         .accept_dad             = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .dad_transmits          = 1,
 210         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 211         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 212         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 213 #ifdef CONFIG_IPV6_PRIVACY
 214         .use_tempaddr           = 0,
 215         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 216         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 217         .regen_max_retry        = REGEN_MAX_RETRY,
 218         .max_desync_factor      = MAX_DESYNC_FACTOR,
 219 #endif
 220         .max_addresses          = IPV6_MAX_ADDRESSES,
 221         .accept_ra_defrtr       = 1,
 222         .accept_ra_pinfo        = 1,
 223 #ifdef CONFIG_IPV6_ROUTER_PREF
 224         .accept_ra_rtr_pref     = 1,
 225         .rtr_probe_interval     = 60 * HZ,
 226 #ifdef CONFIG_IPV6_ROUTE_INFO
 227         .accept_ra_rt_info_max_plen = 0,
 228 #endif
 229 #endif
 230         .proxy_ndp              = 0,
 231         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 232         .disable_ipv6           = 0,
 233         .accept_dad             = 1,
 234 };
 235
 236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 241
 242 /* Check if a valid qdisc is available */
 243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 244 {
 245         return !qdisc_tx_is_noop(dev);
 246 }
 247
 248 /* Check if a route is valid prefix route */
 249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 250 {
 251         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 252 }
 253
 254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 255 {
 256         if (del_timer(&ifp->timer))
 257                 __in6_ifa_put(ifp);
 258 }
 259
 260 enum addrconf_timer_t {
 261         AC_NONE,
 262         AC_DAD,
 263         AC_RS,
 264 };
 265
 266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 267                                enum addrconf_timer_t what,
 268                                unsigned long when)
 269 {
 270         if (!del_timer(&ifp->timer))
 271                 in6_ifa_hold(ifp);
 272
 273         switch (what) {
 274         case AC_DAD:
 275                 ifp->timer.function = addrconf_dad_timer;
 276                 break;
 277         case AC_RS:
 278                 ifp->timer.function = addrconf_rs_timer;
 279                 break;
 280         default:
 281                 break;
 282         }
 283         ifp->timer.expires = jiffies + when;
 284         add_timer(&ifp->timer);
 285 }
 286
 287 static int snmp6_alloc_dev(struct inet6_dev *idev)
 288 {
 289         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 290                           sizeof(struct ipstats_mib),
 291                           __alignof__(struct ipstats_mib)) < 0)
 292                 goto err_ip;
 293         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 294                                         GFP_KERNEL);
 295         if (!idev->stats.icmpv6dev)
 296                 goto err_icmp;
 297         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 298                                            GFP_KERNEL);
 299         if (!idev->stats.icmpv6msgdev)
 300                 goto err_icmpmsg;
 301
 302         return 0;
 303
 304 err_icmpmsg:
 305         kfree(idev->stats.icmpv6dev);
 306 err_icmp:
 307         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 308 err_ip:
 309         return -ENOMEM;
 310 }
 311
 312 static void snmp6_free_dev(struct inet6_dev *idev)
 313 {
 314         kfree(idev->stats.icmpv6msgdev);
 315         kfree(idev->stats.icmpv6dev);
 316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 317 }
 318
 319 /* Nobody refers to this device, we may destroy it. */
 320
 321 void in6_dev_finish_destroy(struct inet6_dev *idev)
 322 {
 323         struct net_device *dev = idev->dev;
 324
 325         WARN_ON(!list_empty(&idev->addr_list));
 326         WARN_ON(idev->mc_list != NULL);
 327
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 pr_warning("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         kfree_rcu(idev, rcu);
 338 }
 339
 340 EXPORT_SYMBOL(in6_dev_finish_destroy);
 341
 342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 343 {
 344         struct inet6_dev *ndev;
 345
 346         ASSERT_RTNL();
 347
 348         if (dev->mtu < IPV6_MIN_MTU)
 349                 return NULL;
 350
 351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 352
 353         if (ndev == NULL)
 354                 return NULL;
 355
 356         rwlock_init(&ndev->lock);
 357         ndev->dev = dev;
 358         INIT_LIST_HEAD(&ndev->addr_list);
 359
 360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (ndev->nd_parms == NULL) {
 365                 kfree(ndev);
 366                 return NULL;
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 dev_put(dev);
 379                 kfree(ndev);
 380                 return NULL;
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG((KERN_WARNING
 385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name));
 387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 388                 ndev->dead = 1;
 389                 in6_dev_finish_destroy(ndev);
 390                 return NULL;
 391         }
 392
 393         /* One reference from device.  We must do this before
 394          * we invoke __ipv6_regen_rndid().
 395          */
 396         in6_dev_hold(ndev);
 397
 398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 399                 ndev->cnf.accept_dad = -1;
 400
 401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Multicast RS\n",
 405                        dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 ndev->cnf.use_tempaddr = -1;
 419         } else {
 420                 in6_dev_hold(ndev);
 421                 ipv6_regen_rndid((unsigned long) ndev);
 422         }
 423 #endif
 424
 425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 426                 ndev->if_flags |= IF_READY;
 427
 428         ipv6_mc_init_dev(ndev);
 429         ndev->tstamp = jiffies;
 430         addrconf_sysctl_register(ndev);
 431         /* protected by rtnl_lock */
 432         RCU_INIT_POINTER(dev->ip6_ptr, ndev);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         return ndev;
 438 }
 439
 440 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 441 {
 442         struct inet6_dev *idev;
 443
 444         ASSERT_RTNL();
 445
 446         idev = __in6_dev_get(dev);
 447         if (!idev) {
 448                 idev = ipv6_add_dev(dev);
 449                 if (!idev)
 450                         return NULL;
 451         }
 452
 453         if (dev->flags&IFF_UP)
 454                 ipv6_mc_up(idev);
 455         return idev;
 456 }
 457
 458 #ifdef CONFIG_SYSCTL
 459 static void dev_forward_change(struct inet6_dev *idev)
 460 {
 461         struct net_device *dev;
 462         struct inet6_ifaddr *ifa;
 463
 464         if (!idev)
 465                 return;
 466         dev = idev->dev;
 467         if (idev->cnf.forwarding)
 468                 dev_disable_lro(dev);
 469         if (dev && (dev->flags & IFF_MULTICAST)) {
 470                 if (idev->cnf.forwarding)
 471                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 472                 else
 473                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 474         }
 475
 476         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 477                 if (ifa->flags&IFA_F_TENTATIVE)
 478                         continue;
 479                 if (idev->cnf.forwarding)
 480                         addrconf_join_anycast(ifa);
 481                 else
 482                         addrconf_leave_anycast(ifa);
 483         }
 484 }
 485
 486
 487 static void addrconf_forward_change(struct net *net, __s32 newf)
 488 {
 489         struct net_device *dev;
 490         struct inet6_dev *idev;
 491
 492         rcu_read_lock();
 493         for_each_netdev_rcu(net, dev) {
 494                 idev = __in6_dev_get(dev);
 495                 if (idev) {
 496                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 497                         idev->cnf.forwarding = newf;
 498                         if (changed)
 499                                 dev_forward_change(idev);
 500                 }
 501         }
 502         rcu_read_unlock();
 503 }
 504
 505 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 506 {
 507         struct net *net;
 508
 509         net = (struct net *)table->extra2;
 510         if (p == &net->ipv6.devconf_dflt->forwarding)
 511                 return 0;
 512
 513         if (!rtnl_trylock()) {
 514                 /* Restore the original values before restarting */
 515                 *p = old;
 516                 return restart_syscall();
 517         }
 518
 519         if (p == &net->ipv6.devconf_all->forwarding) {
 520                 __s32 newf = net->ipv6.devconf_all->forwarding;
 521                 net->ipv6.devconf_dflt->forwarding = newf;
 522                 addrconf_forward_change(net, newf);
 523         } else if ((!*p) ^ (!old))
 524                 dev_forward_change((struct inet6_dev *)table->extra1);
 525         rtnl_unlock();
 526
 527         if (*p)
 528                 rt6_purge_dflt_routers(net);
 529         return 1;
 530 }
 531 #endif
 532
 533 /* Nobody refers to this ifaddr, destroy it */
 534 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 535 {
 536         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 537
 538 #ifdef NET_REFCNT_DEBUG
 539         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 540 #endif
 541
 542         in6_dev_put(ifp->idev);
 543
 544         if (del_timer(&ifp->timer))
 545                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 546
 547         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 548                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 549                 return;
 550         }
 551         dst_release(&ifp->rt->dst);
 552
 553         kfree_rcu(ifp, rcu);
 554 }
 555
 556 static void
 557 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 558 {
 559         struct list_head *p;
 560         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 561
 562         /*
 563          * Each device address list is sorted in order of scope -
 564          * global before linklocal.
 565          */
 566         list_for_each(p, &idev->addr_list) {
 567                 struct inet6_ifaddr *ifa
 568                         = list_entry(p, struct inet6_ifaddr, if_list);
 569                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 570                         break;
 571         }
 572
 573         list_add_tail(&ifp->if_list, p);
 574 }
 575
 576 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 577 {
 578         /*
 579          * We perform the hash function over the last 64 bits of the address
 580          * This will include the IEEE address token on links that support it.
 581          */
 582         return jhash_2words((__force u32)addr->s6_addr32[2],
 583                             (__force u32)addr->s6_addr32[3], 0)
 584                 & (IN6_ADDR_HSIZE - 1);
 585 }
 586
 587 /* On success it returns ifp with increased reference count */
 588
 589 static struct inet6_ifaddr *
 590 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 591               int scope, u32 flags)
 592 {
 593         struct inet6_ifaddr *ifa = NULL;
 594         struct rt6_info *rt;
 595         unsigned int hash;
 596         int err = 0;
 597         int addr_type = ipv6_addr_type(addr);
 598
 599         if (addr_type == IPV6_ADDR_ANY ||
 600             addr_type & IPV6_ADDR_MULTICAST ||
 601             (!(idev->dev->flags & IFF_LOOPBACK) &&
 602              addr_type & IPV6_ADDR_LOOPBACK))
 603                 return ERR_PTR(-EADDRNOTAVAIL);
 604
 605         rcu_read_lock_bh();
 606         if (idev->dead) {
 607                 err = -ENODEV;                  /*XXX*/
 608                 goto out2;
 609         }
 610
 611         if (idev->cnf.disable_ipv6) {
 612                 err = -EACCES;
 613                 goto out2;
 614         }
 615
 616         spin_lock(&addrconf_hash_lock);
 617
 618         /* Ignore adding duplicate addresses on an interface */
 619         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 620                 ADBG(("ipv6_add_addr: already assigned\n"));
 621                 err = -EEXIST;
 622                 goto out;
 623         }
 624
 625         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 626
 627         if (ifa == NULL) {
 628                 ADBG(("ipv6_add_addr: malloc failed\n"));
 629                 err = -ENOBUFS;
 630                 goto out;
 631         }
 632
 633         rt = addrconf_dst_alloc(idev, addr, 0);
 634         if (IS_ERR(rt)) {
 635                 err = PTR_ERR(rt);
 636                 goto out;
 637         }
 638
 639         ipv6_addr_copy(&ifa->addr, addr);
 640
 641         spin_lock_init(&ifa->lock);
 642         spin_lock_init(&ifa->state_lock);
 643         init_timer(&ifa->timer);
 644         INIT_HLIST_NODE(&ifa->addr_lst);
 645         ifa->timer.data = (unsigned long) ifa;
 646         ifa->scope = scope;
 647         ifa->prefix_len = pfxlen;
 648         ifa->flags = flags | IFA_F_TENTATIVE;
 649         ifa->cstamp = ifa->tstamp = jiffies;
 650
 651         ifa->rt = rt;
 652
 653         /*
 654          * part one of RFC 4429, section 3.3
 655          * We should not configure an address as
 656          * optimistic if we do not yet know the link
 657          * layer address of our nexhop router
 658          */
 659
 660         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 661                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 662
 663         ifa->idev = idev;
 664         in6_dev_hold(idev);
 665         /* For caller */
 666         in6_ifa_hold(ifa);
 667
 668         /* Add to big hash table */
 669         hash = ipv6_addr_hash(addr);
 670
 671         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 672         spin_unlock(&addrconf_hash_lock);
 673
 674         write_lock(&idev->lock);
 675         /* Add to inet6_dev unicast addr list. */
 676         ipv6_link_dev_addr(idev, ifa);
 677
 678 #ifdef CONFIG_IPV6_PRIVACY
 679         if (ifa->flags&IFA_F_TEMPORARY) {
 680                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 681                 in6_ifa_hold(ifa);
 682         }
 683 #endif
 684
 685         in6_ifa_hold(ifa);
 686         write_unlock(&idev->lock);
 687 out2:
 688         rcu_read_unlock_bh();
 689
 690         if (likely(err == 0))
 691                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 692         else {
 693                 kfree(ifa);
 694                 ifa = ERR_PTR(err);
 695         }
 696
 697         return ifa;
 698 out:
 699         spin_unlock(&addrconf_hash_lock);
 700         goto out2;
 701 }
 702
 703 /* This function wants to get referenced ifp and releases it before return */
 704
 705 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 706 {
 707         struct inet6_ifaddr *ifa, *ifn;
 708         struct inet6_dev *idev = ifp->idev;
 709         int state;
 710         int deleted = 0, onlink = 0;
 711         unsigned long expires = jiffies;
 712
 713         spin_lock_bh(&ifp->state_lock);
 714         state = ifp->state;
 715         ifp->state = INET6_IFADDR_STATE_DEAD;
 716         spin_unlock_bh(&ifp->state_lock);
 717
 718         if (state == INET6_IFADDR_STATE_DEAD)
 719                 goto out;
 720
 721         spin_lock_bh(&addrconf_hash_lock);
 722         hlist_del_init_rcu(&ifp->addr_lst);
 723         spin_unlock_bh(&addrconf_hash_lock);
 724
 725         write_lock_bh(&idev->lock);
 726 #ifdef CONFIG_IPV6_PRIVACY
 727         if (ifp->flags&IFA_F_TEMPORARY) {
 728                 list_del(&ifp->tmp_list);
 729                 if (ifp->ifpub) {
 730                         in6_ifa_put(ifp->ifpub);
 731                         ifp->ifpub = NULL;
 732                 }
 733                 __in6_ifa_put(ifp);
 734         }
 735 #endif
 736
 737         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 738                 if (ifa == ifp) {
 739                         list_del_init(&ifp->if_list);
 740                         __in6_ifa_put(ifp);
 741
 742                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 743                                 break;
 744                         deleted = 1;
 745                         continue;
 746                 } else if (ifp->flags & IFA_F_PERMANENT) {
 747                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 748                                               ifp->prefix_len)) {
 749                                 if (ifa->flags & IFA_F_PERMANENT) {
 750                                         onlink = 1;
 751                                         if (deleted)
 752                                                 break;
 753                                 } else {
 754                                         unsigned long lifetime;
 755
 756                                         if (!onlink)
 757                                                 onlink = -1;
 758
 759                                         spin_lock(&ifa->lock);
 760
 761                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 762                                         /*
 763                                          * Note: Because this address is
 764                                          * not permanent, lifetime <
 765                                          * LONG_MAX / HZ here.
 766                                          */
 767                                         if (time_before(expires,
 768                                                         ifa->tstamp + lifetime * HZ))
 769                                                 expires = ifa->tstamp + lifetime * HZ;
 770                                         spin_unlock(&ifa->lock);
 771                                 }
 772                         }
 773                 }
 774         }
 775         write_unlock_bh(&idev->lock);
 776
 777         addrconf_del_timer(ifp);
 778
 779         ipv6_ifa_notify(RTM_DELADDR, ifp);
 780
 781         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 782
 783         /*
 784          * Purge or update corresponding prefix
 785          *
 786          * 1) we don't purge prefix here if address was not permanent.
 787          *    prefix is managed by its own lifetime.
 788          * 2) if there're no addresses, delete prefix.
 789          * 3) if there're still other permanent address(es),
 790          *    corresponding prefix is still permanent.
 791          * 4) otherwise, update prefix lifetime to the
 792          *    longest valid lifetime among the corresponding
 793          *    addresses on the device.
 794          *    Note: subsequent RA will update lifetime.
 795          *
 796          * --yoshfuji
 797          */
 798         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 799                 struct in6_addr prefix;
 800                 struct rt6_info *rt;
 801                 struct net *net = dev_net(ifp->idev->dev);
 802                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 803                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 804
 805                 if (rt && addrconf_is_prefix_route(rt)) {
 806                         if (onlink == 0) {
 807                                 ip6_del_rt(rt);
 808                                 rt = NULL;
 809                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 810                                 rt->rt6i_expires = expires;
 811                                 rt->rt6i_flags |= RTF_EXPIRES;
 812                         }
 813                 }
 814                 dst_release(&rt->dst);
 815         }
 816
 817         /* clean up prefsrc entries */
 818         rt6_remove_prefsrc(ifp);
 819 out:
 820         in6_ifa_put(ifp);
 821 }
 822
 823 #ifdef CONFIG_IPV6_PRIVACY
 824 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 825 {
 826         struct inet6_dev *idev = ifp->idev;
 827         struct in6_addr addr, *tmpaddr;
 828         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 829         unsigned long regen_advance;
 830         int tmp_plen;
 831         int ret = 0;
 832         int max_addresses;
 833         u32 addr_flags;
 834         unsigned long now = jiffies;
 835
 836         write_lock(&idev->lock);
 837         if (ift) {
 838                 spin_lock_bh(&ift->lock);
 839                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 840                 spin_unlock_bh(&ift->lock);
 841                 tmpaddr = &addr;
 842         } else {
 843                 tmpaddr = NULL;
 844         }
 845 retry:
 846         in6_dev_hold(idev);
 847         if (idev->cnf.use_tempaddr <= 0) {
 848                 write_unlock(&idev->lock);
 849                 printk(KERN_INFO
 850                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 851                 in6_dev_put(idev);
 852                 ret = -1;
 853                 goto out;
 854         }
 855         spin_lock_bh(&ifp->lock);
 856         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 857                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 858                 spin_unlock_bh(&ifp->lock);
 859                 write_unlock(&idev->lock);
 860                 printk(KERN_WARNING
 861                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 862                 in6_dev_put(idev);
 863                 ret = -1;
 864                 goto out;
 865         }
 866         in6_ifa_hold(ifp);
 867         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 868         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 869                 spin_unlock_bh(&ifp->lock);
 870                 write_unlock(&idev->lock);
 871                 printk(KERN_WARNING
 872                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 873                 in6_ifa_put(ifp);
 874                 in6_dev_put(idev);
 875                 ret = -1;
 876                 goto out;
 877         }
 878         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 879         age = (now - ifp->tstamp) / HZ;
 880         tmp_valid_lft = min_t(__u32,
 881                               ifp->valid_lft,
 882                               idev->cnf.temp_valid_lft + age);
 883         tmp_prefered_lft = min_t(__u32,
 884                                  ifp->prefered_lft,
 885                                  idev->cnf.temp_prefered_lft + age -
 886                                  idev->cnf.max_desync_factor);
 887         tmp_plen = ifp->prefix_len;
 888         max_addresses = idev->cnf.max_addresses;
 889         tmp_tstamp = ifp->tstamp;
 890         spin_unlock_bh(&ifp->lock);
 891
 892         regen_advance = idev->cnf.regen_max_retry *
 893                         idev->cnf.dad_transmits *
 894                         idev->nd_parms->retrans_time / HZ;
 895         write_unlock(&idev->lock);
 896
 897         /* A temporary address is created only if this calculated Preferred
 898          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 899          * an implementation must not create a temporary address with a zero
 900          * Preferred Lifetime.
 901          */
 902         if (tmp_prefered_lft <= regen_advance) {
 903                 in6_ifa_put(ifp);
 904                 in6_dev_put(idev);
 905                 ret = -1;
 906                 goto out;
 907         }
 908
 909         addr_flags = IFA_F_TEMPORARY;
 910         /* set in addrconf_prefix_rcv() */
 911         if (ifp->flags & IFA_F_OPTIMISTIC)
 912                 addr_flags |= IFA_F_OPTIMISTIC;
 913
 914         ift = !max_addresses ||
 915               ipv6_count_addresses(idev) < max_addresses ?
 916                 ipv6_add_addr(idev, &addr, tmp_plen,
 917                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 918                               addr_flags) : NULL;
 919         if (!ift || IS_ERR(ift)) {
 920                 in6_ifa_put(ifp);
 921                 in6_dev_put(idev);
 922                 printk(KERN_INFO
 923                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 924                 tmpaddr = &addr;
 925                 write_lock(&idev->lock);
 926                 goto retry;
 927         }
 928
 929         spin_lock_bh(&ift->lock);
 930         ift->ifpub = ifp;
 931         ift->valid_lft = tmp_valid_lft;
 932         ift->prefered_lft = tmp_prefered_lft;
 933         ift->cstamp = now;
 934         ift->tstamp = tmp_tstamp;
 935         spin_unlock_bh(&ift->lock);
 936
 937         addrconf_dad_start(ift, 0);
 938         in6_ifa_put(ift);
 939         in6_dev_put(idev);
 940 out:
 941         return ret;
 942 }
 943 #endif
 944
 945 /*
 946  *      Choose an appropriate source address (RFC3484)
 947  */
 948 enum {
 949         IPV6_SADDR_RULE_INIT = 0,
 950         IPV6_SADDR_RULE_LOCAL,
 951         IPV6_SADDR_RULE_SCOPE,
 952         IPV6_SADDR_RULE_PREFERRED,
 953 #ifdef CONFIG_IPV6_MIP6
 954         IPV6_SADDR_RULE_HOA,
 955 #endif
 956         IPV6_SADDR_RULE_OIF,
 957         IPV6_SADDR_RULE_LABEL,
 958 #ifdef CONFIG_IPV6_PRIVACY
 959         IPV6_SADDR_RULE_PRIVACY,
 960 #endif
 961         IPV6_SADDR_RULE_ORCHID,
 962         IPV6_SADDR_RULE_PREFIX,
 963         IPV6_SADDR_RULE_MAX
 964 };
 965
 966 struct ipv6_saddr_score {
 967         int                     rule;
 968         int                     addr_type;
 969         struct inet6_ifaddr     *ifa;
 970         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 971         int                     scopedist;
 972         int                     matchlen;
 973 };
 974
 975 struct ipv6_saddr_dst {
 976         const struct in6_addr *addr;
 977         int ifindex;
 978         int scope;
 979         int label;
 980         unsigned int prefs;
 981 };
 982
 983 static inline int ipv6_saddr_preferred(int type)
 984 {
 985         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 986                 return 1;
 987         return 0;
 988 }
 989
 990 static int ipv6_get_saddr_eval(struct net *net,
 991                                struct ipv6_saddr_score *score,
 992                                struct ipv6_saddr_dst *dst,
 993                                int i)
 994 {
 995         int ret;
 996
 997         if (i <= score->rule) {
 998                 switch (i) {
 999                 case IPV6_SADDR_RULE_SCOPE:
1000                         ret = score->scopedist;
1001                         break;
1002                 case IPV6_SADDR_RULE_PREFIX:
1003                         ret = score->matchlen;
1004                         break;
1005                 default:
1006                         ret = !!test_bit(i, score->scorebits);
1007                 }
1008                 goto out;
1009         }
1010
1011         switch (i) {
1012         case IPV6_SADDR_RULE_INIT:
1013                 /* Rule 0: remember if hiscore is not ready yet */
1014                 ret = !!score->ifa;
1015                 break;
1016         case IPV6_SADDR_RULE_LOCAL:
1017                 /* Rule 1: Prefer same address */
1018                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1019                 break;
1020         case IPV6_SADDR_RULE_SCOPE:
1021                 /* Rule 2: Prefer appropriate scope
1022                  *
1023                  *      ret
1024                  *       ^
1025                  *    -1 |  d 15
1026                  *    ---+--+-+---> scope
1027                  *       |
1028                  *       |             d is scope of the destination.
1029                  *  B-d  |  \
1030                  *       |   \      <- smaller scope is better if
1031                  *  B-15 |    \        if scope is enough for destinaion.
1032                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1033                  * d-C-1 | /
1034                  *       |/         <- greater is better
1035                  *   -C  /             if scope is not enough for destination.
1036                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1037                  *
1038                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1039                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1040                  * Assume B = 0 and we get C > 29.
1041                  */
1042                 ret = __ipv6_addr_src_scope(score->addr_type);
1043                 if (ret >= dst->scope)
1044                         ret = -ret;
1045                 else
1046                         ret -= 128;     /* 30 is enough */
1047                 score->scopedist = ret;
1048                 break;
1049         case IPV6_SADDR_RULE_PREFERRED:
1050                 /* Rule 3: Avoid deprecated and optimistic addresses */
1051                 ret = ipv6_saddr_preferred(score->addr_type) ||
1052                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1053                 break;
1054 #ifdef CONFIG_IPV6_MIP6
1055         case IPV6_SADDR_RULE_HOA:
1056             {
1057                 /* Rule 4: Prefer home address */
1058                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1059                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1060                 break;
1061             }
1062 #endif
1063         case IPV6_SADDR_RULE_OIF:
1064                 /* Rule 5: Prefer outgoing interface */
1065                 ret = (!dst->ifindex ||
1066                        dst->ifindex == score->ifa->idev->dev->ifindex);
1067                 break;
1068         case IPV6_SADDR_RULE_LABEL:
1069                 /* Rule 6: Prefer matching label */
1070                 ret = ipv6_addr_label(net,
1071                                       &score->ifa->addr, score->addr_type,
1072                                       score->ifa->idev->dev->ifindex) == dst->label;
1073                 break;
1074 #ifdef CONFIG_IPV6_PRIVACY
1075         case IPV6_SADDR_RULE_PRIVACY:
1076             {
1077                 /* Rule 7: Prefer public address
1078                  * Note: prefer temporary address if use_tempaddr >= 2
1079                  */
1080                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1081                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1082                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1083                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1084                 break;
1085             }
1086 #endif
1087         case IPV6_SADDR_RULE_ORCHID:
1088                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1089                  *          non-ORCHID vs non-ORCHID
1090                  */
1091                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1092                         ipv6_addr_orchid(dst->addr));
1093                 break;
1094         case IPV6_SADDR_RULE_PREFIX:
1095                 /* Rule 8: Use longest matching prefix */
1096                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1097                                                        dst->addr);
1098                 break;
1099         default:
1100                 ret = 0;
1101         }
1102
1103         if (ret)
1104                 __set_bit(i, score->scorebits);
1105         score->rule = i;
1106 out:
1107         return ret;
1108 }
1109
1110 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1111                        const struct in6_addr *daddr, unsigned int prefs,
1112                        struct in6_addr *saddr)
1113 {
1114         struct ipv6_saddr_score scores[2],
1115                                 *score = &scores[0], *hiscore = &scores[1];
1116         struct ipv6_saddr_dst dst;
1117         struct net_device *dev;
1118         int dst_type;
1119
1120         dst_type = __ipv6_addr_type(daddr);
1121         dst.addr = daddr;
1122         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1123         dst.scope = __ipv6_addr_src_scope(dst_type);
1124         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1125         dst.prefs = prefs;
1126
1127         hiscore->rule = -1;
1128         hiscore->ifa = NULL;
1129
1130         rcu_read_lock();
1131
1132         for_each_netdev_rcu(net, dev) {
1133                 struct inet6_dev *idev;
1134
1135                 /* Candidate Source Address (section 4)
1136                  *  - multicast and link-local destination address,
1137                  *    the set of candidate source address MUST only
1138                  *    include addresses assigned to interfaces
1139                  *    belonging to the same link as the outgoing
1140                  *    interface.
1141                  * (- For site-local destination addresses, the
1142                  *    set of candidate source addresses MUST only
1143                  *    include addresses assigned to interfaces
1144                  *    belonging to the same site as the outgoing
1145                  *    interface.)
1146                  */
1147                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1148                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1149                     dst.ifindex && dev->ifindex != dst.ifindex)
1150                         continue;
1151
1152                 idev = __in6_dev_get(dev);
1153                 if (!idev)
1154                         continue;
1155
1156                 read_lock_bh(&idev->lock);
1157                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1158                         int i;
1159
1160                         /*
1161                          * - Tentative Address (RFC2462 section 5.4)
1162                          *  - A tentative address is not considered
1163                          *    "assigned to an interface" in the traditional
1164                          *    sense, unless it is also flagged as optimistic.
1165                          * - Candidate Source Address (section 4)
1166                          *  - In any case, anycast addresses, multicast
1167                          *    addresses, and the unspecified address MUST
1168                          *    NOT be included in a candidate set.
1169                          */
1170                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1171                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1172                                 continue;
1173
1174                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1175
1176                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1177                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1178                                 LIMIT_NETDEBUG(KERN_DEBUG
1179                                                "ADDRCONF: unspecified / multicast address "
1180                                                "assigned as unicast address on %s",
1181                                                dev->name);
1182                                 continue;
1183                         }
1184
1185                         score->rule = -1;
1186                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1187
1188                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1189                                 int minihiscore, miniscore;
1190
1191                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1192                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1193
1194                                 if (minihiscore > miniscore) {
1195                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1196                                             score->scopedist > 0) {
1197                                                 /*
1198                                                  * special case:
1199                                                  * each remaining entry
1200                                                  * has too small (not enough)
1201                                                  * scope, because ifa entries
1202                                                  * are sorted by their scope
1203                                                  * values.
1204                                                  */
1205                                                 goto try_nextdev;
1206                                         }
1207                                         break;
1208                                 } else if (minihiscore < miniscore) {
1209                                         if (hiscore->ifa)
1210                                                 in6_ifa_put(hiscore->ifa);
1211
1212                                         in6_ifa_hold(score->ifa);
1213
1214                                         swap(hiscore, score);
1215
1216                                         /* restore our iterator */
1217                                         score->ifa = hiscore->ifa;
1218
1219                                         break;
1220                                 }
1221                         }
1222                 }
1223 try_nextdev:
1224                 read_unlock_bh(&idev->lock);
1225         }
1226         rcu_read_unlock();
1227
1228         if (!hiscore->ifa)
1229                 return -EADDRNOTAVAIL;
1230
1231         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1232         in6_ifa_put(hiscore->ifa);
1233         return 0;
1234 }
1235 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1236
1237 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1238                     unsigned char banned_flags)
1239 {
1240         struct inet6_dev *idev;
1241         int err = -EADDRNOTAVAIL;
1242
1243         rcu_read_lock();
1244         idev = __in6_dev_get(dev);
1245         if (idev) {
1246                 struct inet6_ifaddr *ifp;
1247
1248                 read_lock_bh(&idev->lock);
1249                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1250                         if (ifp->scope == IFA_LINK &&
1251                             !(ifp->flags & banned_flags)) {
1252                                 ipv6_addr_copy(addr, &ifp->addr);
1253                                 err = 0;
1254                                 break;
1255                         }
1256                 }
1257                 read_unlock_bh(&idev->lock);
1258         }
1259         rcu_read_unlock();
1260         return err;
1261 }
1262
1263 static int ipv6_count_addresses(struct inet6_dev *idev)
1264 {
1265         int cnt = 0;
1266         struct inet6_ifaddr *ifp;
1267
1268         read_lock_bh(&idev->lock);
1269         list_for_each_entry(ifp, &idev->addr_list, if_list)
1270                 cnt++;
1271         read_unlock_bh(&idev->lock);
1272         return cnt;
1273 }
1274
1275 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1276                   struct net_device *dev, int strict)
1277 {
1278         struct inet6_ifaddr *ifp;
1279         struct hlist_node *node;
1280         unsigned int hash = ipv6_addr_hash(addr);
1281
1282         rcu_read_lock_bh();
1283         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1284                 if (!net_eq(dev_net(ifp->idev->dev), net))
1285                         continue;
1286                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1287                     !(ifp->flags&IFA_F_TENTATIVE) &&
1288                     (dev == NULL || ifp->idev->dev == dev ||
1289                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1290                         rcu_read_unlock_bh();
1291                         return 1;
1292                 }
1293         }
1294
1295         rcu_read_unlock_bh();
1296         return 0;
1297 }
1298 EXPORT_SYMBOL(ipv6_chk_addr);
1299
1300 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1301                                struct net_device *dev)
1302 {
1303         unsigned int hash = ipv6_addr_hash(addr);
1304         struct inet6_ifaddr *ifp;
1305         struct hlist_node *node;
1306
1307         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1308                 if (!net_eq(dev_net(ifp->idev->dev), net))
1309                         continue;
1310                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1311                         if (dev == NULL || ifp->idev->dev == dev)
1312                                 return true;
1313                 }
1314         }
1315         return false;
1316 }
1317
1318 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1319 {
1320         struct inet6_dev *idev;
1321         struct inet6_ifaddr *ifa;
1322         int     onlink;
1323
1324         onlink = 0;
1325         rcu_read_lock();
1326         idev = __in6_dev_get(dev);
1327         if (idev) {
1328                 read_lock_bh(&idev->lock);
1329                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1330                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1331                                                    ifa->prefix_len);
1332                         if (onlink)
1333                                 break;
1334                 }
1335                 read_unlock_bh(&idev->lock);
1336         }
1337         rcu_read_unlock();
1338         return onlink;
1339 }
1340
1341 EXPORT_SYMBOL(ipv6_chk_prefix);
1342
1343 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1344                                      struct net_device *dev, int strict)
1345 {
1346         struct inet6_ifaddr *ifp, *result = NULL;
1347         unsigned int hash = ipv6_addr_hash(addr);
1348         struct hlist_node *node;
1349
1350         rcu_read_lock_bh();
1351         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1352                 if (!net_eq(dev_net(ifp->idev->dev), net))
1353                         continue;
1354                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1355                         if (dev == NULL || ifp->idev->dev == dev ||
1356                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1357                                 result = ifp;
1358                                 in6_ifa_hold(ifp);
1359                                 break;
1360                         }
1361                 }
1362         }
1363         rcu_read_unlock_bh();
1364
1365         return result;
1366 }
1367
1368 /* Gets referenced address, destroys ifaddr */
1369
1370 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1371 {
1372         if (ifp->flags&IFA_F_PERMANENT) {
1373                 spin_lock_bh(&ifp->lock);
1374                 addrconf_del_timer(ifp);
1375                 ifp->flags |= IFA_F_TENTATIVE;
1376                 if (dad_failed)
1377                         ifp->flags |= IFA_F_DADFAILED;
1378                 spin_unlock_bh(&ifp->lock);
1379                 if (dad_failed)
1380                         ipv6_ifa_notify(0, ifp);
1381                 in6_ifa_put(ifp);
1382 #ifdef CONFIG_IPV6_PRIVACY
1383         } else if (ifp->flags&IFA_F_TEMPORARY) {
1384                 struct inet6_ifaddr *ifpub;
1385                 spin_lock_bh(&ifp->lock);
1386                 ifpub = ifp->ifpub;
1387                 if (ifpub) {
1388                         in6_ifa_hold(ifpub);
1389                         spin_unlock_bh(&ifp->lock);
1390                         ipv6_create_tempaddr(ifpub, ifp);
1391                         in6_ifa_put(ifpub);
1392                 } else {
1393                         spin_unlock_bh(&ifp->lock);
1394                 }
1395                 ipv6_del_addr(ifp);
1396 #endif
1397         } else
1398                 ipv6_del_addr(ifp);
1399 }
1400
1401 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1402 {
1403         int err = -ENOENT;
1404
1405         spin_lock(&ifp->state_lock);
1406         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1407                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1408                 err = 0;
1409         }
1410         spin_unlock(&ifp->state_lock);
1411
1412         return err;
1413 }
1414
1415 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1416 {
1417         struct inet6_dev *idev = ifp->idev;
1418
1419         if (addrconf_dad_end(ifp)) {
1420                 in6_ifa_put(ifp);
1421                 return;
1422         }
1423
1424         if (net_ratelimit())
1425                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1426                         ifp->idev->dev->name, &ifp->addr);
1427
1428         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1429                 struct in6_addr addr;
1430
1431                 addr.s6_addr32[0] = htonl(0xfe800000);
1432                 addr.s6_addr32[1] = 0;
1433
1434                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1435                     ipv6_addr_equal(&ifp->addr, &addr)) {
1436                         /* DAD failed for link-local based on MAC address */
1437                         idev->cnf.disable_ipv6 = 1;
1438
1439                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1440                                 ifp->idev->dev->name);
1441                 }
1442         }
1443
1444         addrconf_dad_stop(ifp, 1);
1445 }
1446
1447 /* Join to solicited addr multicast group. */
1448
1449 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1450 {
1451         struct in6_addr maddr;
1452
1453         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1454                 return;
1455
1456         addrconf_addr_solict_mult(addr, &maddr);
1457         ipv6_dev_mc_inc(dev, &maddr);
1458 }
1459
1460 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1461 {
1462         struct in6_addr maddr;
1463
1464         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1465                 return;
1466
1467         addrconf_addr_solict_mult(addr, &maddr);
1468         __ipv6_dev_mc_dec(idev, &maddr);
1469 }
1470
1471 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1472 {
1473         struct in6_addr addr;
1474         if (ifp->prefix_len == 127) /* RFC 6164 */
1475                 return;
1476         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1477         if (ipv6_addr_any(&addr))
1478                 return;
1479         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1480 }
1481
1482 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1483 {
1484         struct in6_addr addr;
1485         if (ifp->prefix_len == 127) /* RFC 6164 */
1486                 return;
1487         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1488         if (ipv6_addr_any(&addr))
1489                 return;
1490         __ipv6_dev_ac_dec(ifp->idev, &addr);
1491 }
1492
1493 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1494 {
1495         if (dev->addr_len != ETH_ALEN)
1496                 return -1;
1497         memcpy(eui, dev->dev_addr, 3);
1498         memcpy(eui + 5, dev->dev_addr + 3, 3);
1499
1500         /*
1501          * The zSeries OSA network cards can be shared among various
1502          * OS instances, but the OSA cards have only one MAC address.
1503          * This leads to duplicate address conflicts in conjunction
1504          * with IPv6 if more than one instance uses the same card.
1505          *
1506          * The driver for these cards can deliver a unique 16-bit
1507          * identifier for each instance sharing the same card.  It is
1508          * placed instead of 0xFFFE in the interface identifier.  The
1509          * "u" bit of the interface identifier is not inverted in this
1510          * case.  Hence the resulting interface identifier has local
1511          * scope according to RFC2373.
1512          */
1513         if (dev->dev_id) {
1514                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1515                 eui[4] = dev->dev_id & 0xFF;
1516         } else {
1517                 eui[3] = 0xFF;
1518                 eui[4] = 0xFE;
1519                 eui[0] ^= 2;
1520         }
1521         return 0;
1522 }
1523
1524 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1525 {
1526         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1527         if (dev->addr_len != ARCNET_ALEN)
1528                 return -1;
1529         memset(eui, 0, 7);
1530         eui[7] = *(u8*)dev->dev_addr;
1531         return 0;
1532 }
1533
1534 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1535 {
1536         if (dev->addr_len != INFINIBAND_ALEN)
1537                 return -1;
1538         memcpy(eui, dev->dev_addr + 12, 8);
1539         eui[0] |= 2;
1540         return 0;
1541 }
1542
1543 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1544 {
1545         if (addr == 0)
1546                 return -1;
1547         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1548                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1549                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1550                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1551                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1552                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1553         eui[1] = 0;
1554         eui[2] = 0x5E;
1555         eui[3] = 0xFE;
1556         memcpy(eui + 4, &addr, 4);
1557         return 0;
1558 }
1559
1560 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1561 {
1562         if (dev->priv_flags & IFF_ISATAP)
1563                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1564         return -1;
1565 }
1566
1567 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1568 {
1569         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1570 }
1571
1572 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1573 {
1574         switch (dev->type) {
1575         case ARPHRD_ETHER:
1576         case ARPHRD_FDDI:
1577         case ARPHRD_IEEE802_TR:
1578                 return addrconf_ifid_eui48(eui, dev);
1579         case ARPHRD_ARCNET:
1580                 return addrconf_ifid_arcnet(eui, dev);
1581         case ARPHRD_INFINIBAND:
1582                 return addrconf_ifid_infiniband(eui, dev);
1583         case ARPHRD_SIT:
1584                 return addrconf_ifid_sit(eui, dev);
1585         case ARPHRD_IPGRE:
1586                 return addrconf_ifid_gre(eui, dev);
1587         }
1588         return -1;
1589 }
1590
1591 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1592 {
1593         int err = -1;
1594         struct inet6_ifaddr *ifp;
1595
1596         read_lock_bh(&idev->lock);
1597         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1598                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1599                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1600                         err = 0;
1601                         break;
1602                 }
1603         }
1604         read_unlock_bh(&idev->lock);
1605         return err;
1606 }
1607
1608 #ifdef CONFIG_IPV6_PRIVACY
1609 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1610 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1611 {
1612 regen:
1613         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1614         idev->rndid[0] &= ~0x02;
1615
1616         /*
1617          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1618          * check if generated address is not inappropriate
1619          *
1620          *  - Reserved subnet anycast (RFC 2526)
1621          *      11111101 11....11 1xxxxxxx
1622          *  - ISATAP (RFC4214) 6.1
1623          *      00-00-5E-FE-xx-xx-xx-xx
1624          *  - value 0
1625          *  - XXX: already assigned to an address on the device
1626          */
1627         if (idev->rndid[0] == 0xfd &&
1628             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1629             (idev->rndid[7]&0x80))
1630                 goto regen;
1631         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1632                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1633                         goto regen;
1634                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1635                         goto regen;
1636         }
1637
1638         return 0;
1639 }
1640
1641 static void ipv6_regen_rndid(unsigned long data)
1642 {
1643         struct inet6_dev *idev = (struct inet6_dev *) data;
1644         unsigned long expires;
1645
1646         rcu_read_lock_bh();
1647         write_lock_bh(&idev->lock);
1648
1649         if (idev->dead)
1650                 goto out;
1651
1652         if (__ipv6_regen_rndid(idev) < 0)
1653                 goto out;
1654
1655         expires = jiffies +
1656                 idev->cnf.temp_prefered_lft * HZ -
1657                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1658                 idev->cnf.max_desync_factor * HZ;
1659         if (time_before(expires, jiffies)) {
1660                 printk(KERN_WARNING
1661                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1662                         idev->dev->name);
1663                 goto out;
1664         }
1665
1666         if (!mod_timer(&idev->regen_timer, expires))
1667                 in6_dev_hold(idev);
1668
1669 out:
1670         write_unlock_bh(&idev->lock);
1671         rcu_read_unlock_bh();
1672         in6_dev_put(idev);
1673 }
1674
1675 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1676         int ret = 0;
1677
1678         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1679                 ret = __ipv6_regen_rndid(idev);
1680         return ret;
1681 }
1682 #endif
1683
1684 /*
1685  *      Add prefix route.
1686  */
1687
1688 static void
1689 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1690                       unsigned long expires, u32 flags)
1691 {
1692         struct fib6_config cfg = {
1693                 .fc_table = RT6_TABLE_PREFIX,
1694                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1695                 .fc_ifindex = dev->ifindex,
1696                 .fc_expires = expires,
1697                 .fc_dst_len = plen,
1698                 .fc_flags = RTF_UP | flags,
1699                 .fc_nlinfo.nl_net = dev_net(dev),
1700                 .fc_protocol = RTPROT_KERNEL,
1701         };
1702
1703         ipv6_addr_copy(&cfg.fc_dst, pfx);
1704
1705         /* Prevent useless cloning on PtP SIT.
1706            This thing is done here expecting that the whole
1707            class of non-broadcast devices need not cloning.
1708          */
1709 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1710         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1711                 cfg.fc_flags |= RTF_NONEXTHOP;
1712 #endif
1713
1714         ip6_route_add(&cfg);
1715 }
1716
1717
1718 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1719                                                   int plen,
1720                                                   const struct net_device *dev,
1721                                                   u32 flags, u32 noflags)
1722 {
1723         struct fib6_node *fn;
1724         struct rt6_info *rt = NULL;
1725         struct fib6_table *table;
1726
1727         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1728         if (table == NULL)
1729                 return NULL;
1730
1731         write_lock_bh(&table->tb6_lock);
1732         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1733         if (!fn)
1734                 goto out;
1735         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1736                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1737                         continue;
1738                 if ((rt->rt6i_flags & flags) != flags)
1739                         continue;
1740                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1741                         continue;
1742                 dst_hold(&rt->dst);
1743                 break;
1744         }
1745 out:
1746         write_unlock_bh(&table->tb6_lock);
1747         return rt;
1748 }
1749
1750
1751 /* Create "default" multicast route to the interface */
1752
1753 static void addrconf_add_mroute(struct net_device *dev)
1754 {
1755         struct fib6_config cfg = {
1756                 .fc_table = RT6_TABLE_LOCAL,
1757                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1758                 .fc_ifindex = dev->ifindex,
1759                 .fc_dst_len = 8,
1760                 .fc_flags = RTF_UP,
1761                 .fc_nlinfo.nl_net = dev_net(dev),
1762         };
1763
1764         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1765
1766         ip6_route_add(&cfg);
1767 }
1768
1769 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1770 static void sit_route_add(struct net_device *dev)
1771 {
1772         struct fib6_config cfg = {
1773                 .fc_table = RT6_TABLE_MAIN,
1774                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1775                 .fc_ifindex = dev->ifindex,
1776                 .fc_dst_len = 96,
1777                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1778                 .fc_nlinfo.nl_net = dev_net(dev),
1779         };
1780
1781         /* prefix length - 96 bits "::d.d.d.d" */
1782         ip6_route_add(&cfg);
1783 }
1784 #endif
1785
1786 static void addrconf_add_lroute(struct net_device *dev)
1787 {
1788         struct in6_addr addr;
1789
1790         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1791         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1792 }
1793
1794 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1795 {
1796         struct inet6_dev *idev;
1797
1798         ASSERT_RTNL();
1799
1800         idev = ipv6_find_idev(dev);
1801         if (!idev)
1802                 return ERR_PTR(-ENOBUFS);
1803
1804         if (idev->cnf.disable_ipv6)
1805                 return ERR_PTR(-EACCES);
1806
1807         /* Add default multicast route */
1808         addrconf_add_mroute(dev);
1809
1810         /* Add link local route */
1811         addrconf_add_lroute(dev);
1812         return idev;
1813 }
1814
1815 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1816 {
1817         struct prefix_info *pinfo;
1818         __u32 valid_lft;
1819         __u32 prefered_lft;
1820         int addr_type;
1821         struct inet6_dev *in6_dev;
1822         struct net *net = dev_net(dev);
1823
1824         pinfo = (struct prefix_info *) opt;
1825
1826         if (len < sizeof(struct prefix_info)) {
1827                 ADBG(("addrconf: prefix option too short\n"));
1828                 return;
1829         }
1830
1831         /*
1832          *      Validation checks ([ADDRCONF], page 19)
1833          */
1834
1835         addr_type = ipv6_addr_type(&pinfo->prefix);
1836
1837         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1838                 return;
1839
1840         valid_lft = ntohl(pinfo->valid);
1841         prefered_lft = ntohl(pinfo->prefered);
1842
1843         if (prefered_lft > valid_lft) {
1844                 if (net_ratelimit())
1845                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1846                 return;
1847         }
1848
1849         in6_dev = in6_dev_get(dev);
1850
1851         if (in6_dev == NULL) {
1852                 if (net_ratelimit())
1853                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1854                 return;
1855         }
1856
1857         /*
1858          *      Two things going on here:
1859          *      1) Add routes for on-link prefixes
1860          *      2) Configure prefixes with the auto flag set
1861          */
1862
1863         if (pinfo->onlink) {
1864                 struct rt6_info *rt;
1865                 unsigned long rt_expires;
1866
1867                 /* Avoid arithmetic overflow. Really, we could
1868                  * save rt_expires in seconds, likely valid_lft,
1869                  * but it would require division in fib gc, that it
1870                  * not good.
1871                  */
1872                 if (HZ > USER_HZ)
1873                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1874                 else
1875                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1876
1877                 if (addrconf_finite_timeout(rt_expires))
1878                         rt_expires *= HZ;
1879
1880                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1881                                                pinfo->prefix_len,
1882                                                dev,
1883                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1884                                                RTF_GATEWAY | RTF_DEFAULT);
1885
1886                 if (rt) {
1887                         /* Autoconf prefix route */
1888                         if (valid_lft == 0) {
1889                                 ip6_del_rt(rt);
1890                                 rt = NULL;
1891                         } else if (addrconf_finite_timeout(rt_expires)) {
1892                                 /* not infinity */
1893                                 rt->rt6i_expires = jiffies + rt_expires;
1894                                 rt->rt6i_flags |= RTF_EXPIRES;
1895                         } else {
1896                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1897                                 rt->rt6i_expires = 0;
1898                         }
1899                 } else if (valid_lft) {
1900                         clock_t expires = 0;
1901                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1902                         if (addrconf_finite_timeout(rt_expires)) {
1903                                 /* not infinity */
1904                                 flags |= RTF_EXPIRES;
1905                                 expires = jiffies_to_clock_t(rt_expires);
1906                         }
1907                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1908                                               dev, expires, flags);
1909                 }
1910                 if (rt)
1911                         dst_release(&rt->dst);
1912         }
1913
1914         /* Try to figure out our local address for this prefix */
1915
1916         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1917                 struct inet6_ifaddr * ifp;
1918                 struct in6_addr addr;
1919                 int create = 0, update_lft = 0;
1920
1921                 if (pinfo->prefix_len == 64) {
1922                         memcpy(&addr, &pinfo->prefix, 8);
1923                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1924                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1925                                 in6_dev_put(in6_dev);
1926                                 return;
1927                         }
1928                         goto ok;
1929                 }
1930                 if (net_ratelimit())
1931                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1932                                pinfo->prefix_len);
1933                 in6_dev_put(in6_dev);
1934                 return;
1935
1936 ok:
1937
1938                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1939
1940                 if (ifp == NULL && valid_lft) {
1941                         int max_addresses = in6_dev->cnf.max_addresses;
1942                         u32 addr_flags = 0;
1943
1944 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1945                         if (in6_dev->cnf.optimistic_dad &&
1946                             !net->ipv6.devconf_all->forwarding)
1947                                 addr_flags = IFA_F_OPTIMISTIC;
1948 #endif
1949
1950                         /* Do not allow to create too much of autoconfigured
1951                          * addresses; this would be too easy way to crash kernel.
1952                          */
1953                         if (!max_addresses ||
1954                             ipv6_count_addresses(in6_dev) < max_addresses)
1955                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1956                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1957                                                     addr_flags);
1958
1959                         if (!ifp || IS_ERR(ifp)) {
1960                                 in6_dev_put(in6_dev);
1961                                 return;
1962                         }
1963
1964                         update_lft = create = 1;
1965                         ifp->cstamp = jiffies;
1966                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1967                 }
1968
1969                 if (ifp) {
1970                         int flags;
1971                         unsigned long now;
1972 #ifdef CONFIG_IPV6_PRIVACY
1973                         struct inet6_ifaddr *ift;
1974 #endif
1975                         u32 stored_lft;
1976
1977                         /* update lifetime (RFC2462 5.5.3 e) */
1978                         spin_lock(&ifp->lock);
1979                         now = jiffies;
1980                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1981                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1982                         else
1983                                 stored_lft = 0;
1984                         if (!update_lft && stored_lft) {
1985                                 if (valid_lft > MIN_VALID_LIFETIME ||
1986                                     valid_lft > stored_lft)
1987                                         update_lft = 1;
1988                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1989                                         /* valid_lft <= stored_lft is always true */
1990                                         /*
1991                                          * RFC 4862 Section 5.5.3e:
1992                                          * "Note that the preferred lifetime of
1993                                          *  the corresponding address is always
1994                                          *  reset to the Preferred Lifetime in
1995                                          *  the received Prefix Information
1996                                          *  option, regardless of whether the
1997                                          *  valid lifetime is also reset or
1998                                          *  ignored."
1999                                          *
2000                                          *  So if the preferred lifetime in
2001                                          *  this advertisement is different
2002                                          *  than what we have stored, but the
2003                                          *  valid lifetime is invalid, just
2004                                          *  reset prefered_lft.
2005                                          *
2006                                          *  We must set the valid lifetime
2007                                          *  to the stored lifetime since we'll
2008                                          *  be updating the timestamp below,
2009                                          *  else we'll set it back to the
2010                                          *  minimum.
2011                                          */
2012                                         if (prefered_lft != ifp->prefered_lft) {
2013                                                 valid_lft = stored_lft;
2014                                                 update_lft = 1;
2015                                         }
2016                                 } else {
2017                                         valid_lft = MIN_VALID_LIFETIME;
2018                                         if (valid_lft < prefered_lft)
2019                                                 prefered_lft = valid_lft;
2020                                         update_lft = 1;
2021                                 }
2022                         }
2023
2024                         if (update_lft) {
2025                                 ifp->valid_lft = valid_lft;
2026                                 ifp->prefered_lft = prefered_lft;
2027                                 ifp->tstamp = now;
2028                                 flags = ifp->flags;
2029                                 ifp->flags &= ~IFA_F_DEPRECATED;
2030                                 spin_unlock(&ifp->lock);
2031
2032                                 if (!(flags&IFA_F_TENTATIVE))
2033                                         ipv6_ifa_notify(0, ifp);
2034                         } else
2035                                 spin_unlock(&ifp->lock);
2036
2037 #ifdef CONFIG_IPV6_PRIVACY
2038                         read_lock_bh(&in6_dev->lock);
2039                         /* update all temporary addresses in the list */
2040                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2041                                             tmp_list) {
2042                                 int age, max_valid, max_prefered;
2043
2044                                 if (ifp != ift->ifpub)
2045                                         continue;
2046
2047                                 /*
2048                                  * RFC 4941 section 3.3:
2049                                  * If a received option will extend the lifetime
2050                                  * of a public address, the lifetimes of
2051                                  * temporary addresses should be extended,
2052                                  * subject to the overall constraint that no
2053                                  * temporary addresses should ever remain
2054                                  * "valid" or "preferred" for a time longer than
2055                                  * (TEMP_VALID_LIFETIME) or
2056                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2057                                  * respectively.
2058                                  */
2059                                 age = (now - ift->cstamp) / HZ;
2060                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2061                                 if (max_valid < 0)
2062                                         max_valid = 0;
2063
2064                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2065                                                in6_dev->cnf.max_desync_factor -
2066                                                age;
2067                                 if (max_prefered < 0)
2068                                         max_prefered = 0;
2069
2070                                 if (valid_lft > max_valid)
2071                                         valid_lft = max_valid;
2072
2073                                 if (prefered_lft > max_prefered)
2074                                         prefered_lft = max_prefered;
2075
2076                                 spin_lock(&ift->lock);
2077                                 flags = ift->flags;
2078                                 ift->valid_lft = valid_lft;
2079                                 ift->prefered_lft = prefered_lft;
2080                                 ift->tstamp = now;
2081                                 if (prefered_lft > 0)
2082                                         ift->flags &= ~IFA_F_DEPRECATED;
2083
2084                                 spin_unlock(&ift->lock);
2085                                 if (!(flags&IFA_F_TENTATIVE))
2086                                         ipv6_ifa_notify(0, ift);
2087                         }
2088
2089                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2090                                 /*
2091                                  * When a new public address is created as
2092                                  * described in [ADDRCONF], also create a new
2093                                  * temporary address. Also create a temporary
2094                                  * address if it's enabled but no temporary
2095                                  * address currently exists.
2096                                  */
2097                                 read_unlock_bh(&in6_dev->lock);
2098                                 ipv6_create_tempaddr(ifp, NULL);
2099                         } else {
2100                                 read_unlock_bh(&in6_dev->lock);
2101                         }
2102 #endif
2103                         in6_ifa_put(ifp);
2104                         addrconf_verify(0);
2105                 }
2106         }
2107         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2108         in6_dev_put(in6_dev);
2109 }
2110
2111 /*
2112  *      Set destination address.
2113  *      Special case for SIT interfaces where we create a new "virtual"
2114  *      device.
2115  */
2116 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2117 {
2118         struct in6_ifreq ireq;
2119         struct net_device *dev;
2120         int err = -EINVAL;
2121
2122         rtnl_lock();
2123
2124         err = -EFAULT;
2125         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2126                 goto err_exit;
2127
2128         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2129
2130         err = -ENODEV;
2131         if (dev == NULL)
2132                 goto err_exit;
2133
2134 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2135         if (dev->type == ARPHRD_SIT) {
2136                 const struct net_device_ops *ops = dev->netdev_ops;
2137                 struct ifreq ifr;
2138                 struct ip_tunnel_parm p;
2139
2140                 err = -EADDRNOTAVAIL;
2141                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2142                         goto err_exit;
2143
2144                 memset(&p, 0, sizeof(p));
2145                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2146                 p.iph.saddr = 0;
2147                 p.iph.version = 4;
2148                 p.iph.ihl = 5;
2149                 p.iph.protocol = IPPROTO_IPV6;
2150                 p.iph.ttl = 64;
2151                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2152
2153                 if (ops->ndo_do_ioctl) {
2154                         mm_segment_t oldfs = get_fs();
2155
2156                         set_fs(KERNEL_DS);
2157                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2158                         set_fs(oldfs);
2159                 } else
2160                         err = -EOPNOTSUPP;
2161
2162                 if (err == 0) {
2163                         err = -ENOBUFS;
2164                         dev = __dev_get_by_name(net, p.name);
2165                         if (!dev)
2166                                 goto err_exit;
2167                         err = dev_open(dev);
2168                 }
2169         }
2170 #endif
2171
2172 err_exit:
2173         rtnl_unlock();
2174         return err;
2175 }
2176
2177 /*
2178  *      Manual configuration of address on an interface
2179  */
2180 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2181                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2182                           __u32 valid_lft)
2183 {
2184         struct inet6_ifaddr *ifp;
2185         struct inet6_dev *idev;
2186         struct net_device *dev;
2187         int scope;
2188         u32 flags;
2189         clock_t expires;
2190         unsigned long timeout;
2191
2192         ASSERT_RTNL();
2193
2194         if (plen > 128)
2195                 return -EINVAL;
2196
2197         /* check the lifetime */
2198         if (!valid_lft || prefered_lft > valid_lft)
2199                 return -EINVAL;
2200
2201         dev = __dev_get_by_index(net, ifindex);
2202         if (!dev)
2203                 return -ENODEV;
2204
2205         idev = addrconf_add_dev(dev);
2206         if (IS_ERR(idev))
2207                 return PTR_ERR(idev);
2208
2209         scope = ipv6_addr_scope(pfx);
2210
2211         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2212         if (addrconf_finite_timeout(timeout)) {
2213                 expires = jiffies_to_clock_t(timeout * HZ);
2214                 valid_lft = timeout;
2215                 flags = RTF_EXPIRES;
2216         } else {
2217                 expires = 0;
2218                 flags = 0;
2219                 ifa_flags |= IFA_F_PERMANENT;
2220         }
2221
2222         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2223         if (addrconf_finite_timeout(timeout)) {
2224                 if (timeout == 0)
2225                         ifa_flags |= IFA_F_DEPRECATED;
2226                 prefered_lft = timeout;
2227         }
2228
2229         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2230
2231         if (!IS_ERR(ifp)) {
2232                 spin_lock_bh(&ifp->lock);
2233                 ifp->valid_lft = valid_lft;
2234                 ifp->prefered_lft = prefered_lft;
2235                 ifp->tstamp = jiffies;
2236                 spin_unlock_bh(&ifp->lock);
2237
2238                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2239                                       expires, flags);
2240                 /*
2241                  * Note that section 3.1 of RFC 4429 indicates
2242                  * that the Optimistic flag should not be set for
2243                  * manually configured addresses
2244                  */
2245                 addrconf_dad_start(ifp, 0);
2246                 in6_ifa_put(ifp);
2247                 addrconf_verify(0);
2248                 return 0;
2249         }
2250
2251         return PTR_ERR(ifp);
2252 }
2253
2254 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2255                           unsigned int plen)
2256 {
2257         struct inet6_ifaddr *ifp;
2258         struct inet6_dev *idev;
2259         struct net_device *dev;
2260
2261         if (plen > 128)
2262                 return -EINVAL;
2263
2264         dev = __dev_get_by_index(net, ifindex);
2265         if (!dev)
2266                 return -ENODEV;
2267
2268         if ((idev = __in6_dev_get(dev)) == NULL)
2269                 return -ENXIO;
2270
2271         read_lock_bh(&idev->lock);
2272         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2273                 if (ifp->prefix_len == plen &&
2274                     ipv6_addr_equal(pfx, &ifp->addr)) {
2275                         in6_ifa_hold(ifp);
2276                         read_unlock_bh(&idev->lock);
2277
2278                         ipv6_del_addr(ifp);
2279
2280                         /* If the last address is deleted administratively,
2281                            disable IPv6 on this interface.
2282                          */
2283                         if (list_empty(&idev->addr_list))
2284                                 addrconf_ifdown(idev->dev, 1);
2285                         return 0;
2286                 }
2287         }
2288         read_unlock_bh(&idev->lock);
2289         return -EADDRNOTAVAIL;
2290 }
2291
2292
2293 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2294 {
2295         struct in6_ifreq ireq;
2296         int err;
2297
2298         if (!capable(CAP_NET_ADMIN))
2299                 return -EPERM;
2300
2301         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2302                 return -EFAULT;
2303
2304         rtnl_lock();
2305         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2306                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2307                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2308         rtnl_unlock();
2309         return err;
2310 }
2311
2312 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2313 {
2314         struct in6_ifreq ireq;
2315         int err;
2316
2317         if (!capable(CAP_NET_ADMIN))
2318                 return -EPERM;
2319
2320         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2321                 return -EFAULT;
2322
2323         rtnl_lock();
2324         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2325                              ireq.ifr6_prefixlen);
2326         rtnl_unlock();
2327         return err;
2328 }
2329
2330 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2331                      int plen, int scope)
2332 {
2333         struct inet6_ifaddr *ifp;
2334
2335         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2336         if (!IS_ERR(ifp)) {
2337                 spin_lock_bh(&ifp->lock);
2338                 ifp->flags &= ~IFA_F_TENTATIVE;
2339                 spin_unlock_bh(&ifp->lock);
2340                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2341                 in6_ifa_put(ifp);
2342         }
2343 }
2344
2345 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2346 static void sit_add_v4_addrs(struct inet6_dev *idev)
2347 {
2348         struct in6_addr addr;
2349         struct net_device *dev;
2350         struct net *net = dev_net(idev->dev);
2351         int scope;
2352
2353         ASSERT_RTNL();
2354
2355         memset(&addr, 0, sizeof(struct in6_addr));
2356         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2357
2358         if (idev->dev->flags&IFF_POINTOPOINT) {
2359                 addr.s6_addr32[0] = htonl(0xfe800000);
2360                 scope = IFA_LINK;
2361         } else {
2362                 scope = IPV6_ADDR_COMPATv4;
2363         }
2364
2365         if (addr.s6_addr32[3]) {
2366                 add_addr(idev, &addr, 128, scope);
2367                 return;
2368         }
2369
2370         for_each_netdev(net, dev) {
2371                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2372                 if (in_dev && (dev->flags & IFF_UP)) {
2373                         struct in_ifaddr * ifa;
2374
2375                         int flag = scope;
2376
2377                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2378                                 int plen;
2379
2380                                 addr.s6_addr32[3] = ifa->ifa_local;
2381
2382                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2383                                         continue;
2384                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2385                                         if (idev->dev->flags&IFF_POINTOPOINT)
2386                                                 continue;
2387                                         flag |= IFA_HOST;
2388                                 }
2389                                 if (idev->dev->flags&IFF_POINTOPOINT)
2390                                         plen = 64;
2391                                 else
2392                                         plen = 96;
2393
2394                                 add_addr(idev, &addr, plen, flag);
2395                         }
2396                 }
2397         }
2398 }
2399 #endif
2400
2401 static void init_loopback(struct net_device *dev)
2402 {
2403         struct inet6_dev  *idev;
2404
2405         /* ::1 */
2406
2407         ASSERT_RTNL();
2408
2409         if ((idev = ipv6_find_idev(dev)) == NULL) {
2410                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2411                 return;
2412         }
2413
2414         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2415 }
2416
2417 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2418 {
2419         struct inet6_ifaddr * ifp;
2420         u32 addr_flags = IFA_F_PERMANENT;
2421
2422 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2423         if (idev->cnf.optimistic_dad &&
2424             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2425                 addr_flags |= IFA_F_OPTIMISTIC;
2426 #endif
2427
2428
2429         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2430         if (!IS_ERR(ifp)) {
2431                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2432                 addrconf_dad_start(ifp, 0);
2433                 in6_ifa_put(ifp);
2434         }
2435 }
2436
2437 static void addrconf_dev_config(struct net_device *dev)
2438 {
2439         struct in6_addr addr;
2440         struct inet6_dev    * idev;
2441
2442         ASSERT_RTNL();
2443
2444         if ((dev->type != ARPHRD_ETHER) &&
2445             (dev->type != ARPHRD_FDDI) &&
2446             (dev->type != ARPHRD_IEEE802_TR) &&
2447             (dev->type != ARPHRD_ARCNET) &&
2448             (dev->type != ARPHRD_INFINIBAND)) {
2449                 /* Alas, we support only Ethernet autoconfiguration. */
2450                 return;
2451         }
2452
2453         idev = addrconf_add_dev(dev);
2454         if (IS_ERR(idev))
2455                 return;
2456
2457         memset(&addr, 0, sizeof(struct in6_addr));
2458         addr.s6_addr32[0] = htonl(0xFE800000);
2459
2460         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2461                 addrconf_add_linklocal(idev, &addr);
2462 }
2463
2464 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2465 static void addrconf_sit_config(struct net_device *dev)
2466 {
2467         struct inet6_dev *idev;
2468
2469         ASSERT_RTNL();
2470
2471         /*
2472          * Configure the tunnel with one of our IPv4
2473          * addresses... we should configure all of
2474          * our v4 addrs in the tunnel
2475          */
2476
2477         if ((idev = ipv6_find_idev(dev)) == NULL) {
2478                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2479                 return;
2480         }
2481
2482         if (dev->priv_flags & IFF_ISATAP) {
2483                 struct in6_addr addr;
2484
2485                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2486                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2487                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2488                         addrconf_add_linklocal(idev, &addr);
2489                 return;
2490         }
2491
2492         sit_add_v4_addrs(idev);
2493
2494         if (dev->flags&IFF_POINTOPOINT) {
2495                 addrconf_add_mroute(dev);
2496                 addrconf_add_lroute(dev);
2497         } else
2498                 sit_route_add(dev);
2499 }
2500 #endif
2501
2502 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2503 static void addrconf_gre_config(struct net_device *dev)
2504 {
2505         struct inet6_dev *idev;
2506         struct in6_addr addr;
2507
2508         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2509
2510         ASSERT_RTNL();
2511
2512         if ((idev = ipv6_find_idev(dev)) == NULL) {
2513                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2514                 return;
2515         }
2516
2517         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2518         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2519
2520         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2521                 addrconf_add_linklocal(idev, &addr);
2522 }
2523 #endif
2524
2525 static inline int
2526 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2527 {
2528         struct in6_addr lladdr;
2529
2530         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2531                 addrconf_add_linklocal(idev, &lladdr);
2532                 return 0;
2533         }
2534         return -1;
2535 }
2536
2537 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2538 {
2539         struct net_device *link_dev;
2540         struct net *net = dev_net(idev->dev);
2541
2542         /* first try to inherit the link-local address from the link device */
2543         if (idev->dev->iflink &&
2544             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2545                 if (!ipv6_inherit_linklocal(idev, link_dev))
2546                         return;
2547         }
2548         /* then try to inherit it from any device */
2549         for_each_netdev(net, link_dev) {
2550                 if (!ipv6_inherit_linklocal(idev, link_dev))
2551                         return;
2552         }
2553         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2554 }
2555
2556 /*
2557  * Autoconfigure tunnel with a link-local address so routing protocols,
2558  * DHCPv6, MLD etc. can be run over the virtual link
2559  */
2560
2561 static void addrconf_ip6_tnl_config(struct net_device *dev)
2562 {
2563         struct inet6_dev *idev;
2564
2565         ASSERT_RTNL();
2566
2567         idev = addrconf_add_dev(dev);
2568         if (IS_ERR(idev)) {
2569                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2570                 return;
2571         }
2572         ip6_tnl_add_linklocal(idev);
2573 }
2574
2575 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2576                            void * data)
2577 {
2578         struct net_device *dev = (struct net_device *) data;
2579         struct inet6_dev *idev = __in6_dev_get(dev);
2580         int run_pending = 0;
2581         int err;
2582
2583         switch (event) {
2584         case NETDEV_REGISTER:
2585                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2586                         idev = ipv6_add_dev(dev);
2587                         if (!idev)
2588                                 return notifier_from_errno(-ENOMEM);
2589                 }
2590                 break;
2591
2592         case NETDEV_UP:
2593         case NETDEV_CHANGE:
2594                 if (dev->flags & IFF_SLAVE)
2595                         break;
2596
2597                 if (event == NETDEV_UP) {
2598                         if (!addrconf_qdisc_ok(dev)) {
2599                                 /* device is not ready yet. */
2600                                 printk(KERN_INFO
2601                                         "ADDRCONF(NETDEV_UP): %s: "
2602                                         "link is not ready\n",
2603                                         dev->name);
2604                                 break;
2605                         }
2606
2607                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2608                                 idev = ipv6_add_dev(dev);
2609
2610                         if (idev) {
2611                                 idev->if_flags |= IF_READY;
2612                                 run_pending = 1;
2613                         }
2614                 } else {
2615                         if (!addrconf_qdisc_ok(dev)) {
2616                                 /* device is still not ready. */
2617                                 break;
2618                         }
2619
2620                         if (idev) {
2621                                 if (idev->if_flags & IF_READY)
2622                                         /* device is already configured. */
2623                                         break;
2624                                 idev->if_flags |= IF_READY;
2625                         }
2626
2627                         printk(KERN_INFO
2628                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2629                                         "link becomes ready\n",
2630                                         dev->name);
2631
2632                         run_pending = 1;
2633                 }
2634
2635                 switch (dev->type) {
2636 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2637                 case ARPHRD_SIT:
2638                         addrconf_sit_config(dev);
2639                         break;
2640 #endif
2641 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2642                 case ARPHRD_IPGRE:
2643                         addrconf_gre_config(dev);
2644                         break;
2645 #endif
2646                 case ARPHRD_TUNNEL6:
2647                         addrconf_ip6_tnl_config(dev);
2648                         break;
2649                 case ARPHRD_LOOPBACK:
2650                         init_loopback(dev);
2651                         break;
2652
2653                 default:
2654                         addrconf_dev_config(dev);
2655                         break;
2656                 }
2657
2658                 if (idev) {
2659                         if (run_pending)
2660                                 addrconf_dad_run(idev);
2661
2662                         /*
2663                          * If the MTU changed during the interface down,
2664                          * when the interface up, the changed MTU must be
2665                          * reflected in the idev as well as routers.
2666                          */
2667                         if (idev->cnf.mtu6 != dev->mtu &&
2668                             dev->mtu >= IPV6_MIN_MTU) {
2669                                 rt6_mtu_change(dev, dev->mtu);
2670                                 idev->cnf.mtu6 = dev->mtu;
2671                         }
2672                         idev->tstamp = jiffies;
2673                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2674
2675                         /*
2676                          * If the changed mtu during down is lower than
2677                          * IPV6_MIN_MTU stop IPv6 on this interface.
2678                          */
2679                         if (dev->mtu < IPV6_MIN_MTU)
2680                                 addrconf_ifdown(dev, 1);
2681                 }
2682                 break;
2683
2684         case NETDEV_CHANGEMTU:
2685                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2686                         rt6_mtu_change(dev, dev->mtu);
2687                         idev->cnf.mtu6 = dev->mtu;
2688                         break;
2689                 }
2690
2691                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2692                         idev = ipv6_add_dev(dev);
2693                         if (idev)
2694                                 break;
2695                 }
2696
2697                 /*
2698                  * MTU falled under IPV6_MIN_MTU.
2699                  * Stop IPv6 on this interface.
2700                  */
2701
2702         case NETDEV_DOWN:
2703         case NETDEV_UNREGISTER:
2704                 /*
2705                  *      Remove all addresses from this interface.
2706                  */
2707                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2708                 break;
2709
2710         case NETDEV_CHANGENAME:
2711                 if (idev) {
2712                         snmp6_unregister_dev(idev);
2713                         addrconf_sysctl_unregister(idev);
2714                         addrconf_sysctl_register(idev);
2715                         err = snmp6_register_dev(idev);
2716                         if (err)
2717                                 return notifier_from_errno(err);
2718                 }
2719                 break;
2720
2721         case NETDEV_PRE_TYPE_CHANGE:
2722         case NETDEV_POST_TYPE_CHANGE:
2723                 addrconf_type_change(dev, event);
2724                 break;
2725         }
2726
2727         return NOTIFY_OK;
2728 }
2729
2730 /*
2731  *      addrconf module should be notified of a device going up
2732  */
2733 static struct notifier_block ipv6_dev_notf = {
2734         .notifier_call = addrconf_notify,
2735 };
2736
2737 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2738 {
2739         struct inet6_dev *idev;
2740         ASSERT_RTNL();
2741
2742         idev = __in6_dev_get(dev);
2743
2744         if (event == NETDEV_POST_TYPE_CHANGE)
2745                 ipv6_mc_remap(idev);
2746         else if (event == NETDEV_PRE_TYPE_CHANGE)
2747                 ipv6_mc_unmap(idev);
2748 }
2749
2750 static int addrconf_ifdown(struct net_device *dev, int how)
2751 {
2752         struct net *net = dev_net(dev);
2753         struct inet6_dev *idev;
2754         struct inet6_ifaddr *ifa;
2755         int state, i;
2756
2757         ASSERT_RTNL();
2758
2759         rt6_ifdown(net, dev);
2760         neigh_ifdown(&nd_tbl, dev);
2761
2762         idev = __in6_dev_get(dev);
2763         if (idev == NULL)
2764                 return -ENODEV;
2765
2766         /*
2767          * Step 1: remove reference to ipv6 device from parent device.
2768          *         Do not dev_put!
2769          */
2770         if (how) {
2771                 idev->dead = 1;
2772
2773                 /* protected by rtnl_lock */
2774                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2775
2776                 /* Step 1.5: remove snmp6 entry */
2777                 snmp6_unregister_dev(idev);
2778
2779         }
2780
2781         /* Step 2: clear hash table */
2782         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2783                 struct hlist_head *h = &inet6_addr_lst[i];
2784                 struct hlist_node *n;
2785
2786                 spin_lock_bh(&addrconf_hash_lock);
2787         restart:
2788                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2789                         if (ifa->idev == idev) {
2790                                 hlist_del_init_rcu(&ifa->addr_lst);
2791                                 addrconf_del_timer(ifa);
2792                                 goto restart;
2793                         }
2794                 }
2795                 spin_unlock_bh(&addrconf_hash_lock);
2796         }
2797
2798         write_lock_bh(&idev->lock);
2799
2800         /* Step 2: clear flags for stateless addrconf */
2801         if (!how)
2802                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2803
2804 #ifdef CONFIG_IPV6_PRIVACY
2805         if (how && del_timer(&idev->regen_timer))
2806                 in6_dev_put(idev);
2807
2808         /* Step 3: clear tempaddr list */
2809         while (!list_empty(&idev->tempaddr_list)) {
2810                 ifa = list_first_entry(&idev->tempaddr_list,
2811                                        struct inet6_ifaddr, tmp_list);
2812                 list_del(&ifa->tmp_list);
2813                 write_unlock_bh(&idev->lock);
2814                 spin_lock_bh(&ifa->lock);
2815
2816                 if (ifa->ifpub) {
2817                         in6_ifa_put(ifa->ifpub);
2818                         ifa->ifpub = NULL;
2819                 }
2820                 spin_unlock_bh(&ifa->lock);
2821                 in6_ifa_put(ifa);
2822                 write_lock_bh(&idev->lock);
2823         }
2824 #endif
2825
2826         while (!list_empty(&idev->addr_list)) {
2827                 ifa = list_first_entry(&idev->addr_list,
2828                                        struct inet6_ifaddr, if_list);
2829                 addrconf_del_timer(ifa);
2830
2831                 list_del(&ifa->if_list);
2832
2833                 write_unlock_bh(&idev->lock);
2834
2835                 spin_lock_bh(&ifa->state_lock);
2836                 state = ifa->state;
2837                 ifa->state = INET6_IFADDR_STATE_DEAD;
2838                 spin_unlock_bh(&ifa->state_lock);
2839
2840                 if (state != INET6_IFADDR_STATE_DEAD) {
2841                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2842                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2843                 }
2844                 in6_ifa_put(ifa);
2845
2846                 write_lock_bh(&idev->lock);
2847         }
2848
2849         write_unlock_bh(&idev->lock);
2850
2851         /* Step 5: Discard multicast list */
2852         if (how)
2853                 ipv6_mc_destroy_dev(idev);
2854         else
2855                 ipv6_mc_down(idev);
2856
2857         idev->tstamp = jiffies;
2858
2859         /* Last: Shot the device (if unregistered) */
2860         if (how) {
2861                 addrconf_sysctl_unregister(idev);
2862                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2863                 neigh_ifdown(&nd_tbl, dev);
2864                 in6_dev_put(idev);
2865         }
2866         return 0;
2867 }
2868
2869 static void addrconf_rs_timer(unsigned long data)
2870 {
2871         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2872         struct inet6_dev *idev = ifp->idev;
2873
2874         read_lock(&idev->lock);
2875         if (idev->dead || !(idev->if_flags & IF_READY))
2876                 goto out;
2877
2878         if (idev->cnf.forwarding)
2879                 goto out;
2880
2881         /* Announcement received after solicitation was sent */
2882         if (idev->if_flags & IF_RA_RCVD)
2883                 goto out;
2884
2885         spin_lock(&ifp->lock);
2886         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2887                 /* The wait after the last probe can be shorter */
2888                 addrconf_mod_timer(ifp, AC_RS,
2889                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2890                                    idev->cnf.rtr_solicit_delay :
2891                                    idev->cnf.rtr_solicit_interval);
2892                 spin_unlock(&ifp->lock);
2893
2894                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2895         } else {
2896                 spin_unlock(&ifp->lock);
2897                 /*
2898                  * Note: we do not support deprecated "all on-link"
2899                  * assumption any longer.
2900                  */
2901                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2902                        idev->dev->name);
2903         }
2904
2905 out:
2906         read_unlock(&idev->lock);
2907         in6_ifa_put(ifp);
2908 }
2909
2910 /*
2911  *      Duplicate Address Detection
2912  */
2913 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2914 {
2915         unsigned long rand_num;
2916         struct inet6_dev *idev = ifp->idev;
2917
2918         if (ifp->flags & IFA_F_OPTIMISTIC)
2919                 rand_num = 0;
2920         else
2921                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2922
2923         ifp->probes = idev->cnf.dad_transmits;
2924         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2925 }
2926
2927 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2928 {
2929         struct inet6_dev *idev = ifp->idev;
2930         struct net_device *dev = idev->dev;
2931
2932         addrconf_join_solict(dev, &ifp->addr);
2933
2934         net_srandom(ifp->addr.s6_addr32[3]);
2935
2936         read_lock_bh(&idev->lock);
2937         spin_lock(&ifp->lock);
2938         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2939                 goto out;
2940
2941         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2942             idev->cnf.accept_dad < 1 ||
2943             !(ifp->flags&IFA_F_TENTATIVE) ||
2944             ifp->flags & IFA_F_NODAD) {
2945                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2946                 spin_unlock(&ifp->lock);
2947                 read_unlock_bh(&idev->lock);
2948
2949                 addrconf_dad_completed(ifp);
2950                 return;
2951         }
2952
2953         if (!(idev->if_flags & IF_READY)) {
2954                 spin_unlock(&ifp->lock);
2955                 read_unlock_bh(&idev->lock);
2956                 /*
2957                  * If the device is not ready:
2958                  * - keep it tentative if it is a permanent address.
2959                  * - otherwise, kill it.
2960                  */
2961                 in6_ifa_hold(ifp);
2962                 addrconf_dad_stop(ifp, 0);
2963                 return;
2964         }
2965
2966         /*
2967          * Optimistic nodes can start receiving
2968          * Frames right away
2969          */
2970         if (ifp->flags & IFA_F_OPTIMISTIC)
2971                 ip6_ins_rt(ifp->rt);
2972
2973         addrconf_dad_kick(ifp);
2974 out:
2975         spin_unlock(&ifp->lock);
2976         read_unlock_bh(&idev->lock);
2977 }
2978
2979 static void addrconf_dad_timer(unsigned long data)
2980 {
2981         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2982         struct inet6_dev *idev = ifp->idev;
2983         struct in6_addr mcaddr;
2984
2985         if (!ifp->probes && addrconf_dad_end(ifp))
2986                 goto out;
2987
2988         read_lock(&idev->lock);
2989         if (idev->dead || !(idev->if_flags & IF_READY)) {
2990                 read_unlock(&idev->lock);
2991                 goto out;
2992         }
2993
2994         spin_lock(&ifp->lock);
2995         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2996                 spin_unlock(&ifp->lock);
2997                 read_unlock(&idev->lock);
2998                 goto out;
2999         }
3000
3001         if (ifp->probes == 0) {
3002                 /*
3003                  * DAD was successful
3004                  */
3005
3006                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3007                 spin_unlock(&ifp->lock);
3008                 read_unlock(&idev->lock);
3009
3010                 addrconf_dad_completed(ifp);
3011
3012                 goto out;
3013         }
3014
3015         ifp->probes--;
3016         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3017         spin_unlock(&ifp->lock);
3018         read_unlock(&idev->lock);
3019
3020         /* send a neighbour solicitation for our addr */
3021         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3022         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3023 out:
3024         in6_ifa_put(ifp);
3025 }
3026
3027 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3028 {
3029         struct net_device *dev = ifp->idev->dev;
3030
3031         /*
3032          *      Configure the address for reception. Now it is valid.
3033          */
3034
3035         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3036
3037         /* If added prefix is link local and we are prepared to process
3038            router advertisements, start sending router solicitations.
3039          */
3040
3041         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3042              ifp->idev->cnf.accept_ra == 2) &&
3043             ifp->idev->cnf.rtr_solicits > 0 &&
3044             (dev->flags&IFF_LOOPBACK) == 0 &&
3045             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3046                 /*
3047                  *      If a host as already performed a random delay
3048                  *      [...] as part of DAD [...] there is no need
3049                  *      to delay again before sending the first RS
3050                  */
3051                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3052
3053                 spin_lock_bh(&ifp->lock);
3054                 ifp->probes = 1;
3055                 ifp->idev->if_flags |= IF_RS_SENT;
3056                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3057                 spin_unlock_bh(&ifp->lock);
3058         }
3059 }
3060
3061 static void addrconf_dad_run(struct inet6_dev *idev)
3062 {
3063         struct inet6_ifaddr *ifp;
3064
3065         read_lock_bh(&idev->lock);
3066         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3067                 spin_lock(&ifp->lock);
3068                 if (ifp->flags & IFA_F_TENTATIVE &&
3069                     ifp->state == INET6_IFADDR_STATE_DAD)
3070                         addrconf_dad_kick(ifp);
3071                 spin_unlock(&ifp->lock);
3072         }
3073         read_unlock_bh(&idev->lock);
3074 }
3075
3076 #ifdef CONFIG_PROC_FS
3077 struct if6_iter_state {
3078         struct seq_net_private p;
3079         int bucket;
3080 };
3081
3082 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3083 {
3084         struct inet6_ifaddr *ifa = NULL;
3085         struct if6_iter_state *state = seq->private;
3086         struct net *net = seq_file_net(seq);
3087
3088         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3089                 struct hlist_node *n;
3090                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3091                                          addr_lst)
3092                         if (net_eq(dev_net(ifa->idev->dev), net))
3093                                 return ifa;
3094         }
3095         return NULL;
3096 }
3097
3098 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3099                                          struct inet6_ifaddr *ifa)
3100 {
3101         struct if6_iter_state *state = seq->private;
3102         struct net *net = seq_file_net(seq);
3103         struct hlist_node *n = &ifa->addr_lst;
3104
3105         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3106                 if (net_eq(dev_net(ifa->idev->dev), net))
3107                         return ifa;
3108
3109         while (++state->bucket < IN6_ADDR_HSIZE) {
3110                 hlist_for_each_entry_rcu_bh(ifa, n,
3111                                      &inet6_addr_lst[state->bucket], addr_lst) {
3112                         if (net_eq(dev_net(ifa->idev->dev), net))
3113                                 return ifa;
3114                 }
3115         }
3116
3117         return NULL;
3118 }
3119
3120 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3121 {
3122         struct inet6_ifaddr *ifa = if6_get_first(seq);
3123
3124         if (ifa)
3125                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3126                         --pos;
3127         return pos ? NULL : ifa;
3128 }
3129
3130 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3131         __acquires(rcu_bh)
3132 {
3133         rcu_read_lock_bh();
3134         return if6_get_idx(seq, *pos);
3135 }
3136
3137 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3138 {
3139         struct inet6_ifaddr *ifa;
3140
3141         ifa = if6_get_next(seq, v);
3142         ++*pos;
3143         return ifa;
3144 }
3145
3146 static void if6_seq_stop(struct seq_file *seq, void *v)
3147         __releases(rcu_bh)
3148 {
3149         rcu_read_unlock_bh();
3150 }
3151
3152 static int if6_seq_show(struct seq_file *seq, void *v)
3153 {
3154         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3155         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3156                    &ifp->addr,
3157                    ifp->idev->dev->ifindex,
3158                    ifp->prefix_len,
3159                    ifp->scope,
3160                    ifp->flags,
3161                    ifp->idev->dev->name);
3162         return 0;
3163 }
3164
3165 static const struct seq_operations if6_seq_ops = {
3166         .start  = if6_seq_start,
3167         .next   = if6_seq_next,
3168         .show   = if6_seq_show,
3169         .stop   = if6_seq_stop,
3170 };
3171
3172 static int if6_seq_open(struct inode *inode, struct file *file)
3173 {
3174         return seq_open_net(inode, file, &if6_seq_ops,
3175                             sizeof(struct if6_iter_state));
3176 }
3177
3178 static const struct file_operations if6_fops = {
3179         .owner          = THIS_MODULE,
3180         .open           = if6_seq_open,
3181         .read           = seq_read,
3182         .llseek         = seq_lseek,
3183         .release        = seq_release_net,
3184 };
3185
3186 static int __net_init if6_proc_net_init(struct net *net)
3187 {
3188         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3189                 return -ENOMEM;
3190         return 0;
3191 }
3192
3193 static void __net_exit if6_proc_net_exit(struct net *net)
3194 {
3195        proc_net_remove(net, "if_inet6");
3196 }
3197
3198 static struct pernet_operations if6_proc_net_ops = {
3199        .init = if6_proc_net_init,
3200        .exit = if6_proc_net_exit,
3201 };
3202
3203 int __init if6_proc_init(void)
3204 {
3205         return register_pernet_subsys(&if6_proc_net_ops);
3206 }
3207
3208 void if6_proc_exit(void)
3209 {
3210         unregister_pernet_subsys(&if6_proc_net_ops);
3211 }
3212 #endif  /* CONFIG_PROC_FS */
3213
3214 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3215 /* Check if address is a home address configured on any interface. */
3216 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3217 {
3218         int ret = 0;
3219         struct inet6_ifaddr *ifp = NULL;
3220         struct hlist_node *n;
3221         unsigned int hash = ipv6_addr_hash(addr);
3222
3223         rcu_read_lock_bh();
3224         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3225                 if (!net_eq(dev_net(ifp->idev->dev), net))
3226                         continue;
3227                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3228                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3229                         ret = 1;
3230                         break;
3231                 }
3232         }
3233         rcu_read_unlock_bh();
3234         return ret;
3235 }
3236 #endif
3237
3238 /*
3239  *      Periodic address status verification
3240  */
3241
3242 static void addrconf_verify(unsigned long foo)
3243 {
3244         unsigned long now, next, next_sec, next_sched;
3245         struct inet6_ifaddr *ifp;
3246         struct hlist_node *node;
3247         int i;
3248
3249         rcu_read_lock_bh();
3250         spin_lock(&addrconf_verify_lock);
3251         now = jiffies;
3252         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3253
3254         del_timer(&addr_chk_timer);
3255
3256         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3257 restart:
3258                 hlist_for_each_entry_rcu_bh(ifp, node,
3259                                          &inet6_addr_lst[i], addr_lst) {
3260                         unsigned long age;
3261
3262                         if (ifp->flags & IFA_F_PERMANENT)
3263                                 continue;
3264
3265                         spin_lock(&ifp->lock);
3266                         /* We try to batch several events at once. */
3267                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3268
3269                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3270                             age >= ifp->valid_lft) {
3271                                 spin_unlock(&ifp->lock);
3272                                 in6_ifa_hold(ifp);
3273                                 ipv6_del_addr(ifp);
3274                                 goto restart;
3275                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3276                                 spin_unlock(&ifp->lock);
3277                                 continue;
3278                         } else if (age >= ifp->prefered_lft) {
3279                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3280                                 int deprecate = 0;
3281
3282                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3283                                         deprecate = 1;
3284                                         ifp->flags |= IFA_F_DEPRECATED;
3285                                 }
3286
3287                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3288                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3289
3290                                 spin_unlock(&ifp->lock);
3291
3292                                 if (deprecate) {
3293                                         in6_ifa_hold(ifp);
3294
3295                                         ipv6_ifa_notify(0, ifp);
3296                                         in6_ifa_put(ifp);
3297                                         goto restart;
3298                                 }
3299 #ifdef CONFIG_IPV6_PRIVACY
3300                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3301                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3302                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3303                                         ifp->idev->cnf.dad_transmits *
3304                                         ifp->idev->nd_parms->retrans_time / HZ;
3305
3306                                 if (age >= ifp->prefered_lft - regen_advance) {
3307                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3308                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3309                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3310                                         if (!ifp->regen_count && ifpub) {
3311                                                 ifp->regen_count++;
3312                                                 in6_ifa_hold(ifp);
3313                                                 in6_ifa_hold(ifpub);
3314                                                 spin_unlock(&ifp->lock);
3315
3316                                                 spin_lock(&ifpub->lock);
3317                                                 ifpub->regen_count = 0;
3318                                                 spin_unlock(&ifpub->lock);
3319                                                 ipv6_create_tempaddr(ifpub, ifp);
3320                                                 in6_ifa_put(ifpub);
3321                                                 in6_ifa_put(ifp);
3322                                                 goto restart;
3323                                         }
3324                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3325                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3326                                 spin_unlock(&ifp->lock);
3327 #endif
3328                         } else {
3329                                 /* ifp->prefered_lft <= ifp->valid_lft */
3330                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3331                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3332                                 spin_unlock(&ifp->lock);
3333                         }
3334                 }
3335         }
3336
3337         next_sec = round_jiffies_up(next);
3338         next_sched = next;
3339
3340         /* If rounded timeout is accurate enough, accept it. */
3341         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3342                 next_sched = next_sec;
3343
3344         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3345         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3346                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3347
3348         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3349               now, next, next_sec, next_sched));
3350
3351         addr_chk_timer.expires = next_sched;
3352         add_timer(&addr_chk_timer);
3353         spin_unlock(&addrconf_verify_lock);
3354         rcu_read_unlock_bh();
3355 }
3356
3357 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3358 {
3359         struct in6_addr *pfx = NULL;
3360
3361         if (addr)
3362                 pfx = nla_data(addr);
3363
3364         if (local) {
3365                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3366                         pfx = NULL;
3367                 else
3368                         pfx = nla_data(local);
3369         }
3370
3371         return pfx;
3372 }
3373
3374 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3375         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3376         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3377         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3378 };
3379
3380 static int
3381 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3382 {
3383         struct net *net = sock_net(skb->sk);
3384         struct ifaddrmsg *ifm;
3385         struct nlattr *tb[IFA_MAX+1];
3386         struct in6_addr *pfx;
3387         int err;
3388
3389         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3390         if (err < 0)
3391                 return err;
3392
3393         ifm = nlmsg_data(nlh);
3394         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3395         if (pfx == NULL)
3396                 return -EINVAL;
3397
3398         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3399 }
3400
3401 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3402                              u32 prefered_lft, u32 valid_lft)
3403 {
3404         u32 flags;
3405         clock_t expires;
3406         unsigned long timeout;
3407
3408         if (!valid_lft || (prefered_lft > valid_lft))
3409                 return -EINVAL;
3410
3411         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3412         if (addrconf_finite_timeout(timeout)) {
3413                 expires = jiffies_to_clock_t(timeout * HZ);
3414                 valid_lft = timeout;
3415                 flags = RTF_EXPIRES;
3416         } else {
3417                 expires = 0;
3418                 flags = 0;
3419                 ifa_flags |= IFA_F_PERMANENT;
3420         }
3421
3422         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3423         if (addrconf_finite_timeout(timeout)) {
3424                 if (timeout == 0)
3425                         ifa_flags |= IFA_F_DEPRECATED;
3426                 prefered_lft = timeout;
3427         }
3428
3429         spin_lock_bh(&ifp->lock);
3430         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3431         ifp->tstamp = jiffies;
3432         ifp->valid_lft = valid_lft;
3433         ifp->prefered_lft = prefered_lft;
3434
3435         spin_unlock_bh(&ifp->lock);
3436         if (!(ifp->flags&IFA_F_TENTATIVE))
3437                 ipv6_ifa_notify(0, ifp);
3438
3439         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3440                               expires, flags);
3441         addrconf_verify(0);
3442
3443         return 0;
3444 }
3445
3446 static int
3447 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3448 {
3449         struct net *net = sock_net(skb->sk);
3450         struct ifaddrmsg *ifm;
3451         struct nlattr *tb[IFA_MAX+1];
3452         struct in6_addr *pfx;
3453         struct inet6_ifaddr *ifa;
3454         struct net_device *dev;
3455         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3456         u8 ifa_flags;
3457         int err;
3458
3459         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3460         if (err < 0)
3461                 return err;
3462
3463         ifm = nlmsg_data(nlh);
3464         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3465         if (pfx == NULL)
3466                 return -EINVAL;
3467
3468         if (tb[IFA_CACHEINFO]) {
3469                 struct ifa_cacheinfo *ci;
3470
3471                 ci = nla_data(tb[IFA_CACHEINFO]);
3472                 valid_lft = ci->ifa_valid;
3473                 preferred_lft = ci->ifa_prefered;
3474         } else {
3475                 preferred_lft = INFINITY_LIFE_TIME;
3476                 valid_lft = INFINITY_LIFE_TIME;
3477         }
3478
3479         dev =  __dev_get_by_index(net, ifm->ifa_index);
3480         if (dev == NULL)
3481                 return -ENODEV;
3482
3483         /* We ignore other flags so far. */
3484         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3485
3486         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3487         if (ifa == NULL) {
3488                 /*
3489                  * It would be best to check for !NLM_F_CREATE here but
3490                  * userspace alreay relies on not having to provide this.
3491                  */
3492                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3493                                       ifm->ifa_prefixlen, ifa_flags,
3494                                       preferred_lft, valid_lft);
3495         }
3496
3497         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3498             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3499                 err = -EEXIST;
3500         else
3501                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3502
3503         in6_ifa_put(ifa);
3504
3505         return err;
3506 }
3507
3508 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3509                           u8 scope, int ifindex)
3510 {
3511         struct ifaddrmsg *ifm;
3512
3513         ifm = nlmsg_data(nlh);
3514         ifm->ifa_family = AF_INET6;
3515         ifm->ifa_prefixlen = prefixlen;
3516         ifm->ifa_flags = flags;
3517         ifm->ifa_scope = scope;
3518         ifm->ifa_index = ifindex;
3519 }
3520
3521 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3522                          unsigned long tstamp, u32 preferred, u32 valid)
3523 {
3524         struct ifa_cacheinfo ci;
3525
3526         ci.cstamp = cstamp_delta(cstamp);
3527         ci.tstamp = cstamp_delta(tstamp);
3528         ci.ifa_prefered = preferred;
3529         ci.ifa_valid = valid;
3530
3531         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3532 }
3533
3534 static inline int rt_scope(int ifa_scope)
3535 {
3536         if (ifa_scope & IFA_HOST)
3537                 return RT_SCOPE_HOST;
3538         else if (ifa_scope & IFA_LINK)
3539                 return RT_SCOPE_LINK;
3540         else if (ifa_scope & IFA_SITE)
3541                 return RT_SCOPE_SITE;
3542         else
3543                 return RT_SCOPE_UNIVERSE;
3544 }
3545
3546 static inline int inet6_ifaddr_msgsize(void)
3547 {
3548         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3549                + nla_total_size(16) /* IFA_ADDRESS */
3550                + nla_total_size(sizeof(struct ifa_cacheinfo));
3551 }
3552
3553 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3554                              u32 pid, u32 seq, int event, unsigned int flags)
3555 {
3556         struct nlmsghdr  *nlh;
3557         u32 preferred, valid;
3558
3559         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3560         if (nlh == NULL)
3561                 return -EMSGSIZE;
3562
3563         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3564                       ifa->idev->dev->ifindex);
3565
3566         if (!(ifa->flags&IFA_F_PERMANENT)) {
3567                 preferred = ifa->prefered_lft;
3568                 valid = ifa->valid_lft;
3569                 if (preferred != INFINITY_LIFE_TIME) {
3570                         long tval = (jiffies - ifa->tstamp)/HZ;
3571                         if (preferred > tval)
3572                                 preferred -= tval;
3573                         else
3574                                 preferred = 0;
3575                         if (valid != INFINITY_LIFE_TIME) {
3576                                 if (valid > tval)
3577                                         valid -= tval;
3578                                 else
3579                                         valid = 0;
3580                         }
3581                 }
3582         } else {
3583                 preferred = INFINITY_LIFE_TIME;
3584                 valid = INFINITY_LIFE_TIME;
3585         }
3586
3587         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3588             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3589                 nlmsg_cancel(skb, nlh);
3590                 return -EMSGSIZE;
3591         }
3592
3593         return nlmsg_end(skb, nlh);
3594 }
3595
3596 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3597                                 u32 pid, u32 seq, int event, u16 flags)
3598 {
3599         struct nlmsghdr  *nlh;
3600         u8 scope = RT_SCOPE_UNIVERSE;
3601         int ifindex = ifmca->idev->dev->ifindex;
3602
3603         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3604                 scope = RT_SCOPE_SITE;
3605
3606         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3607         if (nlh == NULL)
3608                 return -EMSGSIZE;
3609
3610         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3611         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3612             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3613                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3614                 nlmsg_cancel(skb, nlh);
3615                 return -EMSGSIZE;
3616         }
3617
3618         return nlmsg_end(skb, nlh);
3619 }
3620
3621 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3622                                 u32 pid, u32 seq, int event, unsigned int flags)
3623 {
3624         struct nlmsghdr  *nlh;
3625         u8 scope = RT_SCOPE_UNIVERSE;
3626         int ifindex = ifaca->aca_idev->dev->ifindex;
3627
3628         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3629                 scope = RT_SCOPE_SITE;
3630
3631         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3632         if (nlh == NULL)
3633                 return -EMSGSIZE;
3634
3635         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3636         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3637             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3638                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3639                 nlmsg_cancel(skb, nlh);
3640                 return -EMSGSIZE;
3641         }
3642
3643         return nlmsg_end(skb, nlh);
3644 }
3645
3646 enum addr_type_t {
3647         UNICAST_ADDR,
3648         MULTICAST_ADDR,
3649         ANYCAST_ADDR,
3650 };
3651
3652 /* called with rcu_read_lock() */
3653 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3654                           struct netlink_callback *cb, enum addr_type_t type,
3655                           int s_ip_idx, int *p_ip_idx)
3656 {
3657         struct ifmcaddr6 *ifmca;
3658         struct ifacaddr6 *ifaca;
3659         int err = 1;
3660         int ip_idx = *p_ip_idx;
3661
3662         read_lock_bh(&idev->lock);
3663         switch (type) {
3664         case UNICAST_ADDR: {
3665                 struct inet6_ifaddr *ifa;
3666
3667                 /* unicast address incl. temp addr */
3668                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3669                         if (++ip_idx < s_ip_idx)
3670                                 continue;
3671                         err = inet6_fill_ifaddr(skb, ifa,
3672                                                 NETLINK_CB(cb->skb).pid,
3673                                                 cb->nlh->nlmsg_seq,
3674                                                 RTM_NEWADDR,
3675                                                 NLM_F_MULTI);
3676                         if (err <= 0)
3677                                 break;
3678                 }
3679                 break;
3680         }
3681         case MULTICAST_ADDR:
3682                 /* multicast address */
3683                 for (ifmca = idev->mc_list; ifmca;
3684                      ifmca = ifmca->next, ip_idx++) {
3685                         if (ip_idx < s_ip_idx)
3686                                 continue;
3687                         err = inet6_fill_ifmcaddr(skb, ifmca,
3688                                                   NETLINK_CB(cb->skb).pid,
3689                                                   cb->nlh->nlmsg_seq,
3690                                                   RTM_GETMULTICAST,
3691                                                   NLM_F_MULTI);
3692                         if (err <= 0)
3693                                 break;
3694                 }
3695                 break;
3696         case ANYCAST_ADDR:
3697                 /* anycast address */
3698                 for (ifaca = idev->ac_list; ifaca;
3699                      ifaca = ifaca->aca_next, ip_idx++) {
3700                         if (ip_idx < s_ip_idx)
3701                                 continue;
3702                         err = inet6_fill_ifacaddr(skb, ifaca,
3703                                                   NETLINK_CB(cb->skb).pid,
3704                                                   cb->nlh->nlmsg_seq,
3705                                                   RTM_GETANYCAST,
3706                                                   NLM_F_MULTI);
3707                         if (err <= 0)
3708                                 break;
3709                 }
3710                 break;
3711         default:
3712                 break;
3713         }
3714         read_unlock_bh(&idev->lock);
3715         *p_ip_idx = ip_idx;
3716         return err;
3717 }
3718
3719 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3720                            enum addr_type_t type)
3721 {
3722         struct net *net = sock_net(skb->sk);
3723         int h, s_h;
3724         int idx, ip_idx;
3725         int s_idx, s_ip_idx;
3726         struct net_device *dev;
3727         struct inet6_dev *idev;
3728         struct hlist_head *head;
3729         struct hlist_node *node;
3730
3731         s_h = cb->args[0];
3732         s_idx = idx = cb->args[1];
3733         s_ip_idx = ip_idx = cb->args[2];
3734
3735         rcu_read_lock();
3736         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3737                 idx = 0;
3738                 head = &net->dev_index_head[h];
3739                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3740                         if (idx < s_idx)
3741                                 goto cont;
3742                         if (h > s_h || idx > s_idx)
3743                                 s_ip_idx = 0;
3744                         ip_idx = 0;
3745                         idev = __in6_dev_get(dev);
3746                         if (!idev)
3747                                 goto cont;
3748
3749                         if (in6_dump_addrs(idev, skb, cb, type,
3750                                            s_ip_idx, &ip_idx) <= 0)
3751                                 goto done;
3752 cont:
3753                         idx++;
3754                 }
3755         }
3756 done:
3757         rcu_read_unlock();
3758         cb->args[0] = h;
3759         cb->args[1] = idx;
3760         cb->args[2] = ip_idx;
3761
3762         return skb->len;
3763 }
3764
3765 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3766 {
3767         enum addr_type_t type = UNICAST_ADDR;
3768
3769         return inet6_dump_addr(skb, cb, type);
3770 }
3771
3772 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3773 {
3774         enum addr_type_t type = MULTICAST_ADDR;
3775
3776         return inet6_dump_addr(skb, cb, type);
3777 }
3778
3779
3780 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3781 {
3782         enum addr_type_t type = ANYCAST_ADDR;
3783
3784         return inet6_dump_addr(skb, cb, type);
3785 }
3786
3787 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3788                              void *arg)
3789 {
3790         struct net *net = sock_net(in_skb->sk);
3791         struct ifaddrmsg *ifm;
3792         struct nlattr *tb[IFA_MAX+1];
3793         struct in6_addr *addr = NULL;
3794         struct net_device *dev = NULL;
3795         struct inet6_ifaddr *ifa;
3796         struct sk_buff *skb;
3797         int err;
3798
3799         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3800         if (err < 0)
3801                 goto errout;
3802
3803         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3804         if (addr == NULL) {
3805                 err = -EINVAL;
3806                 goto errout;
3807         }
3808
3809         ifm = nlmsg_data(nlh);
3810         if (ifm->ifa_index)
3811                 dev = __dev_get_by_index(net, ifm->ifa_index);
3812
3813         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3814         if (!ifa) {
3815                 err = -EADDRNOTAVAIL;
3816                 goto errout;
3817         }
3818
3819         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3820         if (!skb) {
3821                 err = -ENOBUFS;
3822                 goto errout_ifa;
3823         }
3824
3825         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3826                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3827         if (err < 0) {
3828                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3829                 WARN_ON(err == -EMSGSIZE);
3830                 kfree_skb(skb);
3831                 goto errout_ifa;
3832         }
3833         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3834 errout_ifa:
3835         in6_ifa_put(ifa);
3836 errout:
3837         return err;
3838 }
3839
3840 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3841 {
3842         struct sk_buff *skb;
3843         struct net *net = dev_net(ifa->idev->dev);
3844         int err = -ENOBUFS;
3845
3846         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3847         if (skb == NULL)
3848                 goto errout;
3849
3850         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3851         if (err < 0) {
3852                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3853                 WARN_ON(err == -EMSGSIZE);
3854                 kfree_skb(skb);
3855                 goto errout;
3856         }
3857         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3858         return;
3859 errout:
3860         if (err < 0)
3861                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3862 }
3863
3864 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3865                                 __s32 *array, int bytes)
3866 {
3867         BUG_ON(bytes < (DEVCONF_MAX * 4));
3868
3869         memset(array, 0, bytes);
3870         array[DEVCONF_FORWARDING] = cnf->forwarding;
3871         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3872         array[DEVCONF_MTU6] = cnf->mtu6;
3873         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3874         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3875         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3876         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3877         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3878         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3879                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3880         array[DEVCONF_RTR_SOLICIT_DELAY] =
3881                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3882         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3883 #ifdef CONFIG_IPV6_PRIVACY
3884         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3885         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3886         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3887         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3888         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3889 #endif
3890         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3891         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3892         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3893 #ifdef CONFIG_IPV6_ROUTER_PREF
3894         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3895         array[DEVCONF_RTR_PROBE_INTERVAL] =
3896                 jiffies_to_msecs(cnf->rtr_probe_interval);
3897 #ifdef CONFIG_IPV6_ROUTE_INFO
3898         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3899 #endif
3900 #endif
3901         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3902         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3903 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3904         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3905 #endif
3906 #ifdef CONFIG_IPV6_MROUTE
3907         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3908 #endif
3909         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3910         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3911         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3912 }
3913
3914 static inline size_t inet6_ifla6_size(void)
3915 {
3916         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3917              + nla_total_size(sizeof(struct ifla_cacheinfo))
3918              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3919              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3920              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3921 }
3922
3923 static inline size_t inet6_if_nlmsg_size(void)
3924 {
3925         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3926                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3927                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3928                + nla_total_size(4) /* IFLA_MTU */
3929                + nla_total_size(4) /* IFLA_LINK */
3930                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3931 }
3932
3933 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3934                                       int items, int bytes)
3935 {
3936         int i;
3937         int pad = bytes - sizeof(u64) * items;
3938         BUG_ON(pad < 0);
3939
3940         /* Use put_unaligned() because stats may not be aligned for u64. */
3941         put_unaligned(items, &stats[0]);
3942         for (i = 1; i < items; i++)
3943                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3944
3945         memset(&stats[items], 0, pad);
3946 }
3947
3948 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3949                                       int items, int bytes, size_t syncpoff)
3950 {
3951         int i;
3952         int pad = bytes - sizeof(u64) * items;
3953         BUG_ON(pad < 0);
3954
3955         /* Use put_unaligned() because stats may not be aligned for u64. */
3956         put_unaligned(items, &stats[0]);
3957         for (i = 1; i < items; i++)
3958                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3959
3960         memset(&stats[items], 0, pad);
3961 }
3962
3963 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3964                              int bytes)
3965 {
3966         switch (attrtype) {
3967         case IFLA_INET6_STATS:
3968                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3969                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3970                 break;
3971         case IFLA_INET6_ICMP6STATS:
3972                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3973                 break;
3974         }
3975 }
3976
3977 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3978 {
3979         struct nlattr *nla;
3980         struct ifla_cacheinfo ci;
3981
3982         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3983
3984         ci.max_reasm_len = IPV6_MAXPLEN;
3985         ci.tstamp = cstamp_delta(idev->tstamp);
3986         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3987         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3988         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3989
3990         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3991         if (nla == NULL)
3992                 goto nla_put_failure;
3993         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3994
3995         /* XXX - MC not implemented */
3996
3997         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3998         if (nla == NULL)
3999                 goto nla_put_failure;
4000         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4001
4002         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4003         if (nla == NULL)
4004                 goto nla_put_failure;
4005         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4006
4007         return 0;
4008
4009 nla_put_failure:
4010         return -EMSGSIZE;
4011 }
4012
4013 static size_t inet6_get_link_af_size(const struct net_device *dev)
4014 {
4015         if (!__in6_dev_get(dev))
4016                 return 0;
4017
4018         return inet6_ifla6_size();
4019 }
4020
4021 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4022 {
4023         struct inet6_dev *idev = __in6_dev_get(dev);
4024
4025         if (!idev)
4026                 return -ENODATA;
4027
4028         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4029                 return -EMSGSIZE;
4030
4031         return 0;
4032 }
4033
4034 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4035                              u32 pid, u32 seq, int event, unsigned int flags)
4036 {
4037         struct net_device *dev = idev->dev;
4038         struct ifinfomsg *hdr;
4039         struct nlmsghdr *nlh;
4040         void *protoinfo;
4041
4042         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4043         if (nlh == NULL)
4044                 return -EMSGSIZE;
4045
4046         hdr = nlmsg_data(nlh);
4047         hdr->ifi_family = AF_INET6;
4048         hdr->__ifi_pad = 0;
4049         hdr->ifi_type = dev->type;
4050         hdr->ifi_index = dev->ifindex;
4051         hdr->ifi_flags = dev_get_flags(dev);
4052         hdr->ifi_change = 0;
4053
4054         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4055
4056         if (dev->addr_len)
4057                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4058
4059         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4060         if (dev->ifindex != dev->iflink)
4061                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4062
4063         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4064         if (protoinfo == NULL)
4065                 goto nla_put_failure;
4066
4067         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4068                 goto nla_put_failure;
4069
4070         nla_nest_end(skb, protoinfo);
4071         return nlmsg_end(skb, nlh);
4072
4073 nla_put_failure:
4074         nlmsg_cancel(skb, nlh);
4075         return -EMSGSIZE;
4076 }
4077
4078 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4079 {
4080         struct net *net = sock_net(skb->sk);
4081         int h, s_h;
4082         int idx = 0, s_idx;
4083         struct net_device *dev;
4084         struct inet6_dev *idev;
4085         struct hlist_head *head;
4086         struct hlist_node *node;
4087
4088         s_h = cb->args[0];
4089         s_idx = cb->args[1];
4090
4091         rcu_read_lock();
4092         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4093                 idx = 0;
4094                 head = &net->dev_index_head[h];
4095                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4096                         if (idx < s_idx)
4097                                 goto cont;
4098                         idev = __in6_dev_get(dev);
4099                         if (!idev)
4100                                 goto cont;
4101                         if (inet6_fill_ifinfo(skb, idev,
4102                                               NETLINK_CB(cb->skb).pid,
4103                                               cb->nlh->nlmsg_seq,
4104                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4105                                 goto out;
4106 cont:
4107                         idx++;
4108                 }
4109         }
4110 out:
4111         rcu_read_unlock();
4112         cb->args[1] = idx;
4113         cb->args[0] = h;
4114
4115         return skb->len;
4116 }
4117
4118 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4119 {
4120         struct sk_buff *skb;
4121         struct net *net = dev_net(idev->dev);
4122         int err = -ENOBUFS;
4123
4124         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4125         if (skb == NULL)
4126                 goto errout;
4127
4128         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4129         if (err < 0) {
4130                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4131                 WARN_ON(err == -EMSGSIZE);
4132                 kfree_skb(skb);
4133                 goto errout;
4134         }
4135         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4136         return;
4137 errout:
4138         if (err < 0)
4139                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4140 }
4141
4142 static inline size_t inet6_prefix_nlmsg_size(void)
4143 {
4144         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4145                + nla_total_size(sizeof(struct in6_addr))
4146                + nla_total_size(sizeof(struct prefix_cacheinfo));
4147 }
4148
4149 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4150                              struct prefix_info *pinfo, u32 pid, u32 seq,
4151                              int event, unsigned int flags)
4152 {
4153         struct prefixmsg *pmsg;
4154         struct nlmsghdr *nlh;
4155         struct prefix_cacheinfo ci;
4156
4157         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4158         if (nlh == NULL)
4159                 return -EMSGSIZE;
4160
4161         pmsg = nlmsg_data(nlh);
4162         pmsg->prefix_family = AF_INET6;
4163         pmsg->prefix_pad1 = 0;
4164         pmsg->prefix_pad2 = 0;
4165         pmsg->prefix_ifindex = idev->dev->ifindex;
4166         pmsg->prefix_len = pinfo->prefix_len;
4167         pmsg->prefix_type = pinfo->type;
4168         pmsg->prefix_pad3 = 0;
4169         pmsg->prefix_flags = 0;
4170         if (pinfo->onlink)
4171                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4172         if (pinfo->autoconf)
4173                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4174
4175         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4176
4177         ci.preferred_time = ntohl(pinfo->prefered);
4178         ci.valid_time = ntohl(pinfo->valid);
4179         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4180
4181         return nlmsg_end(skb, nlh);
4182
4183 nla_put_failure:
4184         nlmsg_cancel(skb, nlh);
4185         return -EMSGSIZE;
4186 }
4187
4188 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4189                          struct prefix_info *pinfo)
4190 {
4191         struct sk_buff *skb;
4192         struct net *net = dev_net(idev->dev);
4193         int err = -ENOBUFS;
4194
4195         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4196         if (skb == NULL)
4197                 goto errout;
4198
4199         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4200         if (err < 0) {
4201                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4202                 WARN_ON(err == -EMSGSIZE);
4203                 kfree_skb(skb);
4204                 goto errout;
4205         }
4206         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4207         return;
4208 errout:
4209         if (err < 0)
4210                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4211 }
4212
4213 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4214 {
4215         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4216
4217         switch (event) {
4218         case RTM_NEWADDR:
4219                 /*
4220                  * If the address was optimistic
4221                  * we inserted the route at the start of
4222                  * our DAD process, so we don't need
4223                  * to do it again
4224                  */
4225                 if (!(ifp->rt->rt6i_node))
4226                         ip6_ins_rt(ifp->rt);
4227                 if (ifp->idev->cnf.forwarding)
4228                         addrconf_join_anycast(ifp);
4229                 break;
4230         case RTM_DELADDR:
4231                 if (ifp->idev->cnf.forwarding)
4232                         addrconf_leave_anycast(ifp);
4233                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4234                 dst_hold(&ifp->rt->dst);
4235
4236                 if (ip6_del_rt(ifp->rt))
4237                         dst_free(&ifp->rt->dst);
4238                 break;
4239         }
4240 }
4241
4242 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4243 {
4244         rcu_read_lock_bh();
4245         if (likely(ifp->idev->dead == 0))
4246                 __ipv6_ifa_notify(event, ifp);
4247         rcu_read_unlock_bh();
4248 }
4249
4250 #ifdef CONFIG_SYSCTL
4251
4252 static
4253 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4254                            void __user *buffer, size_t *lenp, loff_t *ppos)
4255 {
4256         int *valp = ctl->data;
4257         int val = *valp;
4258         loff_t pos = *ppos;
4259         int ret;
4260
4261         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4262
4263         if (write)
4264                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4265         if (ret)
4266                 *ppos = pos;
4267         return ret;
4268 }
4269
4270 static void dev_disable_change(struct inet6_dev *idev)
4271 {
4272         if (!idev || !idev->dev)
4273                 return;
4274
4275         if (idev->cnf.disable_ipv6)
4276                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4277         else
4278                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4279 }
4280
4281 static void addrconf_disable_change(struct net *net, __s32 newf)
4282 {
4283         struct net_device *dev;
4284         struct inet6_dev *idev;
4285
4286         rcu_read_lock();
4287         for_each_netdev_rcu(net, dev) {
4288                 idev = __in6_dev_get(dev);
4289                 if (idev) {
4290                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4291                         idev->cnf.disable_ipv6 = newf;
4292                         if (changed)
4293                                 dev_disable_change(idev);
4294                 }
4295         }
4296         rcu_read_unlock();
4297 }
4298
4299 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4300 {
4301         struct net *net;
4302
4303         net = (struct net *)table->extra2;
4304
4305         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4306                 return 0;
4307
4308         if (!rtnl_trylock()) {
4309                 /* Restore the original values before restarting */
4310                 *p = old;
4311                 return restart_syscall();
4312         }
4313
4314         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4315                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4316                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4317                 addrconf_disable_change(net, newf);
4318         } else if ((!*p) ^ (!old))
4319                 dev_disable_change((struct inet6_dev *)table->extra1);
4320
4321         rtnl_unlock();
4322         return 0;
4323 }
4324
4325 static
4326 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4327                             void __user *buffer, size_t *lenp, loff_t *ppos)
4328 {
4329         int *valp = ctl->data;
4330         int val = *valp;
4331         loff_t pos = *ppos;
4332         int ret;
4333
4334         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4335
4336         if (write)
4337                 ret = addrconf_disable_ipv6(ctl, valp, val);
4338         if (ret)
4339                 *ppos = pos;
4340         return ret;
4341 }
4342
4343 static struct addrconf_sysctl_table
4344 {
4345         struct ctl_table_header *sysctl_header;
4346         ctl_table addrconf_vars[DEVCONF_MAX+1];
4347         char *dev_name;
4348 } addrconf_sysctl __read_mostly = {
4349         .sysctl_header = NULL,
4350         .addrconf_vars = {
4351                 {
4352                         .procname       = "forwarding",
4353                         .data           = &ipv6_devconf.forwarding,
4354                         .maxlen         = sizeof(int),
4355                         .mode           = 0644,
4356                         .proc_handler   = addrconf_sysctl_forward,
4357                 },
4358                 {
4359                         .procname       = "hop_limit",
4360                         .data           = &ipv6_devconf.hop_limit,
4361                         .maxlen         = sizeof(int),
4362                         .mode           = 0644,
4363                         .proc_handler   = proc_dointvec,
4364                 },
4365                 {
4366                         .procname       = "mtu",
4367                         .data           = &ipv6_devconf.mtu6,
4368                         .maxlen         = sizeof(int),
4369                         .mode           = 0644,
4370                         .proc_handler   = proc_dointvec,
4371                 },
4372                 {
4373                         .procname       = "accept_ra",
4374                         .data           = &ipv6_devconf.accept_ra,
4375                         .maxlen         = sizeof(int),
4376                         .mode           = 0644,
4377                         .proc_handler   = proc_dointvec,
4378                 },
4379                 {
4380                         .procname       = "accept_redirects",
4381                         .data           = &ipv6_devconf.accept_redirects,
4382                         .maxlen         = sizeof(int),
4383                         .mode           = 0644,
4384                         .proc_handler   = proc_dointvec,
4385                 },
4386                 {
4387                         .procname       = "autoconf",
4388                         .data           = &ipv6_devconf.autoconf,
4389                         .maxlen         = sizeof(int),
4390                         .mode           = 0644,
4391                         .proc_handler   = proc_dointvec,
4392                 },
4393                 {
4394                         .procname       = "dad_transmits",
4395                         .data           = &ipv6_devconf.dad_transmits,
4396                         .maxlen         = sizeof(int),
4397                         .mode           = 0644,
4398                         .proc_handler   = proc_dointvec,
4399                 },
4400                 {
4401                         .procname       = "router_solicitations",
4402                         .data           = &ipv6_devconf.rtr_solicits,
4403                         .maxlen         = sizeof(int),
4404                         .mode           = 0644,
4405                         .proc_handler   = proc_dointvec,
4406                 },
4407                 {
4408                         .procname       = "router_solicitation_interval",
4409                         .data           = &ipv6_devconf.rtr_solicit_interval,
4410                         .maxlen         = sizeof(int),
4411                         .mode           = 0644,
4412                         .proc_handler   = proc_dointvec_jiffies,
4413                 },
4414                 {
4415                         .procname       = "router_solicitation_delay",
4416                         .data           = &ipv6_devconf.rtr_solicit_delay,
4417                         .maxlen         = sizeof(int),
4418                         .mode           = 0644,
4419                         .proc_handler   = proc_dointvec_jiffies,
4420                 },
4421                 {
4422                         .procname       = "force_mld_version",
4423                         .data           = &ipv6_devconf.force_mld_version,
4424                         .maxlen         = sizeof(int),
4425                         .mode           = 0644,
4426                         .proc_handler   = proc_dointvec,
4427                 },
4428 #ifdef CONFIG_IPV6_PRIVACY
4429                 {
4430                         .procname       = "use_tempaddr",
4431                         .data           = &ipv6_devconf.use_tempaddr,
4432                         .maxlen         = sizeof(int),
4433                         .mode           = 0644,
4434                         .proc_handler   = proc_dointvec,
4435                 },
4436                 {
4437                         .procname       = "temp_valid_lft",
4438                         .data           = &ipv6_devconf.temp_valid_lft,
4439                         .maxlen         = sizeof(int),
4440                         .mode           = 0644,
4441                         .proc_handler   = proc_dointvec,
4442                 },
4443                 {
4444                         .procname       = "temp_prefered_lft",
4445                         .data           = &ipv6_devconf.temp_prefered_lft,
4446                         .maxlen         = sizeof(int),
4447                         .mode           = 0644,
4448                         .proc_handler   = proc_dointvec,
4449                 },
4450                 {
4451                         .procname       = "regen_max_retry",
4452                         .data           = &ipv6_devconf.regen_max_retry,
4453                         .maxlen         = sizeof(int),
4454                         .mode           = 0644,
4455                         .proc_handler   = proc_dointvec,
4456                 },
4457                 {
4458                         .procname       = "max_desync_factor",
4459                         .data           = &ipv6_devconf.max_desync_factor,
4460                         .maxlen         = sizeof(int),
4461                         .mode           = 0644,
4462                         .proc_handler   = proc_dointvec,
4463                 },
4464 #endif
4465                 {
4466                         .procname       = "max_addresses",
4467                         .data           = &ipv6_devconf.max_addresses,
4468                         .maxlen         = sizeof(int),
4469                         .mode           = 0644,
4470                         .proc_handler   = proc_dointvec,
4471                 },
4472                 {
4473                         .procname       = "accept_ra_defrtr",
4474                         .data           = &ipv6_devconf.accept_ra_defrtr,
4475                         .maxlen         = sizeof(int),
4476                         .mode           = 0644,
4477                         .proc_handler   = proc_dointvec,
4478                 },
4479                 {
4480                         .procname       = "accept_ra_pinfo",
4481                         .data           = &ipv6_devconf.accept_ra_pinfo,
4482                         .maxlen         = sizeof(int),
4483                         .mode           = 0644,
4484                         .proc_handler   = proc_dointvec,
4485                 },
4486 #ifdef CONFIG_IPV6_ROUTER_PREF
4487                 {
4488                         .procname       = "accept_ra_rtr_pref",
4489                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4490                         .maxlen         = sizeof(int),
4491                         .mode           = 0644,
4492                         .proc_handler   = proc_dointvec,
4493                 },
4494                 {
4495                         .procname       = "router_probe_interval",
4496                         .data           = &ipv6_devconf.rtr_probe_interval,
4497                         .maxlen         = sizeof(int),
4498                         .mode           = 0644,
4499                         .proc_handler   = proc_dointvec_jiffies,
4500                 },
4501 #ifdef CONFIG_IPV6_ROUTE_INFO
4502                 {
4503                         .procname       = "accept_ra_rt_info_max_plen",
4504                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4505                         .maxlen         = sizeof(int),
4506                         .mode           = 0644,
4507                         .proc_handler   = proc_dointvec,
4508                 },
4509 #endif
4510 #endif
4511                 {
4512                         .procname       = "proxy_ndp",
4513                         .data           = &ipv6_devconf.proxy_ndp,
4514                         .maxlen         = sizeof(int),
4515                         .mode           = 0644,
4516                         .proc_handler   = proc_dointvec,
4517                 },
4518                 {
4519                         .procname       = "accept_source_route",
4520                         .data           = &ipv6_devconf.accept_source_route,
4521                         .maxlen         = sizeof(int),
4522                         .mode           = 0644,
4523                         .proc_handler   = proc_dointvec,
4524                 },
4525 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4526                 {
4527                         .procname       = "optimistic_dad",
4528                         .data           = &ipv6_devconf.optimistic_dad,
4529                         .maxlen         = sizeof(int),
4530                         .mode           = 0644,
4531                         .proc_handler   = proc_dointvec,
4532
4533                 },
4534 #endif
4535 #ifdef CONFIG_IPV6_MROUTE
4536                 {
4537                         .procname       = "mc_forwarding",
4538                         .data           = &ipv6_devconf.mc_forwarding,
4539                         .maxlen         = sizeof(int),
4540                         .mode           = 0444,
4541                         .proc_handler   = proc_dointvec,
4542                 },
4543 #endif
4544                 {
4545                         .procname       = "disable_ipv6",
4546                         .data           = &ipv6_devconf.disable_ipv6,
4547                         .maxlen         = sizeof(int),
4548                         .mode           = 0644,
4549                         .proc_handler   = addrconf_sysctl_disable,
4550                 },
4551                 {
4552                         .procname       = "accept_dad",
4553                         .data           = &ipv6_devconf.accept_dad,
4554                         .maxlen         = sizeof(int),
4555                         .mode           = 0644,
4556                         .proc_handler   = proc_dointvec,
4557                 },
4558                 {
4559                         .procname       = "force_tllao",
4560                         .data           = &ipv6_devconf.force_tllao,
4561                         .maxlen         = sizeof(int),
4562                         .mode           = 0644,
4563                         .proc_handler   = proc_dointvec
4564                 },
4565                 {
4566                         /* sentinel */
4567                 }
4568         },
4569 };
4570
4571 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4572                 struct inet6_dev *idev, struct ipv6_devconf *p)
4573 {
4574         int i;
4575         struct addrconf_sysctl_table *t;
4576
4577 #define ADDRCONF_CTL_PATH_DEV   3
4578
4579         struct ctl_path addrconf_ctl_path[] = {
4580                 { .procname = "net", },
4581                 { .procname = "ipv6", },
4582                 { .procname = "conf", },
4583                 { /* to be set */ },
4584                 { },
4585         };
4586
4587
4588         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4589         if (t == NULL)
4590                 goto out;
4591
4592         for (i = 0; t->addrconf_vars[i].data; i++) {
4593                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4594                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4595                 t->addrconf_vars[i].extra2 = net;
4596         }
4597
4598         /*
4599          * Make a copy of dev_name, because '.procname' is regarded as const
4600          * by sysctl and we wouldn't want anyone to change it under our feet
4601          * (see SIOCSIFNAME).
4602          */
4603         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4604         if (!t->dev_name)
4605                 goto free;
4606
4607         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4608
4609         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4610                         t->addrconf_vars);
4611         if (t->sysctl_header == NULL)
4612                 goto free_procname;
4613
4614         p->sysctl = t;
4615         return 0;
4616
4617 free_procname:
4618         kfree(t->dev_name);
4619 free:
4620         kfree(t);
4621 out:
4622         return -ENOBUFS;
4623 }
4624
4625 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4626 {
4627         struct addrconf_sysctl_table *t;
4628
4629         if (p->sysctl == NULL)
4630                 return;
4631
4632         t = p->sysctl;
4633         p->sysctl = NULL;
4634         unregister_net_sysctl_table(t->sysctl_header);
4635         kfree(t->dev_name);
4636         kfree(t);
4637 }
4638
4639 static void addrconf_sysctl_register(struct inet6_dev *idev)
4640 {
4641         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4642                               &ndisc_ifinfo_sysctl_change);
4643         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4644                                         idev, &idev->cnf);
4645 }
4646
4647 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4648 {
4649         __addrconf_sysctl_unregister(&idev->cnf);
4650         neigh_sysctl_unregister(idev->nd_parms);
4651 }
4652
4653
4654 #endif
4655
4656 static int __net_init addrconf_init_net(struct net *net)
4657 {
4658         int err;
4659         struct ipv6_devconf *all, *dflt;
4660
4661         err = -ENOMEM;
4662         all = &ipv6_devconf;
4663         dflt = &ipv6_devconf_dflt;
4664
4665         if (!net_eq(net, &init_net)) {
4666                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4667                 if (all == NULL)
4668                         goto err_alloc_all;
4669
4670                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4671                 if (dflt == NULL)
4672                         goto err_alloc_dflt;
4673         } else {
4674                 /* these will be inherited by all namespaces */
4675                 dflt->autoconf = ipv6_defaults.autoconf;
4676                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4677         }
4678
4679         net->ipv6.devconf_all = all;
4680         net->ipv6.devconf_dflt = dflt;
4681
4682 #ifdef CONFIG_SYSCTL
4683         err = __addrconf_sysctl_register(net, "all", NULL, all);
4684         if (err < 0)
4685                 goto err_reg_all;
4686
4687         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4688         if (err < 0)
4689                 goto err_reg_dflt;
4690 #endif
4691         return 0;
4692
4693 #ifdef CONFIG_SYSCTL
4694 err_reg_dflt:
4695         __addrconf_sysctl_unregister(all);
4696 err_reg_all:
4697         kfree(dflt);
4698 #endif
4699 err_alloc_dflt:
4700         kfree(all);
4701 err_alloc_all:
4702         return err;
4703 }
4704
4705 static void __net_exit addrconf_exit_net(struct net *net)
4706 {
4707 #ifdef CONFIG_SYSCTL
4708         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4709         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4710 #endif
4711         if (!net_eq(net, &init_net)) {
4712                 kfree(net->ipv6.devconf_dflt);
4713                 kfree(net->ipv6.devconf_all);
4714         }
4715 }
4716
4717 static struct pernet_operations addrconf_ops = {
4718         .init = addrconf_init_net,
4719         .exit = addrconf_exit_net,
4720 };
4721
4722 /*
4723  *      Device notifier
4724  */
4725
4726 int register_inet6addr_notifier(struct notifier_block *nb)
4727 {
4728         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4729 }
4730 EXPORT_SYMBOL(register_inet6addr_notifier);
4731
4732 int unregister_inet6addr_notifier(struct notifier_block *nb)
4733 {
4734         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4735 }
4736 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4737
4738 static struct rtnl_af_ops inet6_ops = {
4739         .family           = AF_INET6,
4740         .fill_link_af     = inet6_fill_link_af,
4741         .get_link_af_size = inet6_get_link_af_size,
4742 };
4743
4744 /*
4745  *      Init / cleanup code
4746  */
4747
4748 int __init addrconf_init(void)
4749 {
4750         int i, err;
4751
4752         err = ipv6_addr_label_init();
4753         if (err < 0) {
4754                 printk(KERN_CRIT "IPv6 Addrconf:"
4755                        " cannot initialize default policy table: %d.\n", err);
4756                 goto out;
4757         }
4758
4759         err = register_pernet_subsys(&addrconf_ops);
4760         if (err < 0)
4761                 goto out_addrlabel;
4762
4763         /* The addrconf netdev notifier requires that loopback_dev
4764          * has it's ipv6 private information allocated and setup
4765          * before it can bring up and give link-local addresses
4766          * to other devices which are up.
4767          *
4768          * Unfortunately, loopback_dev is not necessarily the first
4769          * entry in the global dev_base list of net devices.  In fact,
4770          * it is likely to be the very last entry on that list.
4771          * So this causes the notifier registry below to try and
4772          * give link-local addresses to all devices besides loopback_dev
4773          * first, then loopback_dev, which cases all the non-loopback_dev
4774          * devices to fail to get a link-local address.
4775          *
4776          * So, as a temporary fix, allocate the ipv6 structure for
4777          * loopback_dev first by hand.
4778          * Longer term, all of the dependencies ipv6 has upon the loopback
4779          * device and it being up should be removed.
4780          */
4781         rtnl_lock();
4782         if (!ipv6_add_dev(init_net.loopback_dev))
4783                 err = -ENOMEM;
4784         rtnl_unlock();
4785         if (err)
4786                 goto errlo;
4787
4788         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4789                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4790
4791         register_netdevice_notifier(&ipv6_dev_notf);
4792
4793         addrconf_verify(0);
4794
4795         err = rtnl_af_register(&inet6_ops);
4796         if (err < 0)
4797                 goto errout_af;
4798
4799         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4800                               NULL);
4801         if (err < 0)
4802                 goto errout;
4803
4804         /* Only the first call to __rtnl_register can fail */
4805         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4806         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4807         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4808                         inet6_dump_ifaddr, NULL);
4809         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4810                         inet6_dump_ifmcaddr, NULL);
4811         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4812                         inet6_dump_ifacaddr, NULL);
4813
4814         ipv6_addr_label_rtnl_register();
4815
4816         return 0;
4817 errout:
4818         rtnl_af_unregister(&inet6_ops);
4819 errout_af:
4820         unregister_netdevice_notifier(&ipv6_dev_notf);
4821 errlo:
4822         unregister_pernet_subsys(&addrconf_ops);
4823 out_addrlabel:
4824         ipv6_addr_label_cleanup();
4825 out:
4826         return err;
4827 }
4828
4829 void addrconf_cleanup(void)
4830 {
4831         struct net_device *dev;
4832         int i;
4833
4834         unregister_netdevice_notifier(&ipv6_dev_notf);
4835         unregister_pernet_subsys(&addrconf_ops);
4836         ipv6_addr_label_cleanup();
4837
4838         rtnl_lock();
4839
4840         __rtnl_af_unregister(&inet6_ops);
4841
4842         /* clean dev list */
4843         for_each_netdev(&init_net, dev) {
4844                 if (__in6_dev_get(dev) == NULL)
4845                         continue;
4846                 addrconf_ifdown(dev, 1);
4847         }
4848         addrconf_ifdown(init_net.loopback_dev, 2);
4849
4850         /*
4851          *      Check hash table.
4852          */
4853         spin_lock_bh(&addrconf_hash_lock);
4854         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4855                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4856         spin_unlock_bh(&addrconf_hash_lock);
4857
4858         del_timer(&addr_chk_timer);
4859         rtnl_unlock();
4860 }