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 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/ipv6.h>
  74 #include <net/protocol.h>
  75 #include <net/ndisc.h>
  76 #include <net/ip6_route.h>
  77 #include <net/addrconf.h>
  78 #include <net/tcp.h>
  79 #include <net/ip.h>
  80 #include <net/netlink.h>
  81 #include <net/pkt_sched.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84
  85 #ifdef CONFIG_IPV6_PRIVACY
  86 #include <linux/random.h>
  87 #endif
  88
  89 #include <linux/uaccess.h>
  90 #include <asm/unaligned.h>
  91
  92 #include <linux/proc_fs.h>
  93 #include <linux/seq_file.h>
  94 #include <linux/export.h>
  95
  96 /* Set to 3 to get tracing... */
  97 #define ACONF_DEBUG 2
  98
  99 #if ACONF_DEBUG >= 3
 100 #define ADBG(x) printk x
 101 #else
 102 #define ADBG(x)
 103 #endif
 104
 105 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 106
 107 static inline u32 cstamp_delta(unsigned long cstamp)
 108 {
 109         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 110 }
 111
 112 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 113 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 114 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 115
 116 #ifdef CONFIG_SYSCTL
 117 static void addrconf_sysctl_register(struct inet6_dev *idev);
 118 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 119 #else
 120 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 121 {
 122 }
 123
 124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 125 {
 126 }
 127 #endif
 128
 129 #ifdef CONFIG_IPV6_PRIVACY
 130 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 131 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 132 static void ipv6_regen_rndid(unsigned long data);
 133 #endif
 134
 135 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 136 static int ipv6_count_addresses(struct inet6_dev *idev);
 137
 138 /*
 139  *      Configured unicast address hash table
 140  */
 141 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 142 static DEFINE_SPINLOCK(addrconf_hash_lock);
 143
 144 static void addrconf_verify(unsigned long);
 145
 146 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 147 static DEFINE_SPINLOCK(addrconf_verify_lock);
 148
 149 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 150 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 151
 152 static void addrconf_type_change(struct net_device *dev,
 153                                  unsigned long event);
 154 static int addrconf_ifdown(struct net_device *dev, int how);
 155
 156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 157 static void addrconf_dad_timer(unsigned long data);
 158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 159 static void addrconf_dad_run(struct inet6_dev *idev);
 160 static void addrconf_rs_timer(unsigned long data);
 161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 163
 164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 165                                 struct prefix_info *pinfo);
 166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 167                                struct net_device *dev);
 168
 169 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 170
 171 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 172         .forwarding             = 0,
 173         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 174         .mtu6                   = IPV6_MIN_MTU,
 175         .accept_ra              = 1,
 176         .accept_redirects       = 1,
 177         .autoconf               = 1,
 178         .force_mld_version      = 0,
 179         .dad_transmits          = 1,
 180         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 181         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 182         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 183 #ifdef CONFIG_IPV6_PRIVACY
 184         .use_tempaddr           = 0,
 185         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 186         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 187         .regen_max_retry        = REGEN_MAX_RETRY,
 188         .max_desync_factor      = MAX_DESYNC_FACTOR,
 189 #endif
 190         .max_addresses          = IPV6_MAX_ADDRESSES,
 191         .accept_ra_defrtr       = 1,
 192         .accept_ra_pinfo        = 1,
 193 #ifdef CONFIG_IPV6_ROUTER_PREF
 194         .accept_ra_rtr_pref     = 1,
 195         .rtr_probe_interval     = 60 * HZ,
 196 #ifdef CONFIG_IPV6_ROUTE_INFO
 197         .accept_ra_rt_info_max_plen = 0,
 198 #endif
 199 #endif
 200         .proxy_ndp              = 0,
 201         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 202         .disable_ipv6           = 0,
 203         .accept_dad             = 1,
 204 };
 205
 206 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 207         .forwarding             = 0,
 208         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 209         .mtu6                   = IPV6_MIN_MTU,
 210         .accept_ra              = 1,
 211         .accept_redirects       = 1,
 212         .autoconf               = 1,
 213         .dad_transmits          = 1,
 214         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 215         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 216         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 217 #ifdef CONFIG_IPV6_PRIVACY
 218         .use_tempaddr           = 0,
 219         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 220         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 221         .regen_max_retry        = REGEN_MAX_RETRY,
 222         .max_desync_factor      = MAX_DESYNC_FACTOR,
 223 #endif
 224         .max_addresses          = IPV6_MAX_ADDRESSES,
 225         .accept_ra_defrtr       = 1,
 226         .accept_ra_pinfo        = 1,
 227 #ifdef CONFIG_IPV6_ROUTER_PREF
 228         .accept_ra_rtr_pref     = 1,
 229         .rtr_probe_interval     = 60 * HZ,
 230 #ifdef CONFIG_IPV6_ROUTE_INFO
 231         .accept_ra_rt_info_max_plen = 0,
 232 #endif
 233 #endif
 234         .proxy_ndp              = 0,
 235         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 236         .disable_ipv6           = 0,
 237         .accept_dad             = 1,
 238 };
 239
 240 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 241 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 242 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 243 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 244 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 245
 246 /* Check if a valid qdisc is available */
 247 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 248 {
 249         return !qdisc_tx_is_noop(dev);
 250 }
 251
 252 /* Check if a route is valid prefix route */
 253 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 254 {
 255         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 256 }
 257
 258 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 259 {
 260         if (del_timer(&ifp->timer))
 261                 __in6_ifa_put(ifp);
 262 }
 263
 264 enum addrconf_timer_t {
 265         AC_NONE,
 266         AC_DAD,
 267         AC_RS,
 268 };
 269
 270 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 271                                enum addrconf_timer_t what,
 272                                unsigned long when)
 273 {
 274         if (!del_timer(&ifp->timer))
 275                 in6_ifa_hold(ifp);
 276
 277         switch (what) {
 278         case AC_DAD:
 279                 ifp->timer.function = addrconf_dad_timer;
 280                 break;
 281         case AC_RS:
 282                 ifp->timer.function = addrconf_rs_timer;
 283                 break;
 284         default:
 285                 break;
 286         }
 287         ifp->timer.expires = jiffies + when;
 288         add_timer(&ifp->timer);
 289 }
 290
 291 static int snmp6_alloc_dev(struct inet6_dev *idev)
 292 {
 293         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 294                           sizeof(struct ipstats_mib),
 295                           __alignof__(struct ipstats_mib)) < 0)
 296                 goto err_ip;
 297         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 298                                         GFP_KERNEL);
 299         if (!idev->stats.icmpv6dev)
 300                 goto err_icmp;
 301         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 302                                            GFP_KERNEL);
 303         if (!idev->stats.icmpv6msgdev)
 304                 goto err_icmpmsg;
 305
 306         return 0;
 307
 308 err_icmpmsg:
 309         kfree(idev->stats.icmpv6dev);
 310 err_icmp:
 311         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 312 err_ip:
 313         return -ENOMEM;
 314 }
 315
 316 static void snmp6_free_dev(struct inet6_dev *idev)
 317 {
 318         kfree(idev->stats.icmpv6msgdev);
 319         kfree(idev->stats.icmpv6dev);
 320         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 321 }
 322
 323 /* Nobody refers to this device, we may destroy it. */
 324
 325 void in6_dev_finish_destroy(struct inet6_dev *idev)
 326 {
 327         struct net_device *dev = idev->dev;
 328
 329         WARN_ON(!list_empty(&idev->addr_list));
 330         WARN_ON(idev->mc_list != NULL);
 331
 332 #ifdef NET_REFCNT_DEBUG
 333         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 334 #endif
 335         dev_put(dev);
 336         if (!idev->dead) {
 337                 pr_warn("Freeing alive inet6 device %p\n", idev);
 338                 return;
 339         }
 340         snmp6_free_dev(idev);
 341         kfree_rcu(idev, rcu);
 342 }
 343 EXPORT_SYMBOL(in6_dev_finish_destroy);
 344
 345 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 346 {
 347         struct inet6_dev *ndev;
 348
 349         ASSERT_RTNL();
 350
 351         if (dev->mtu < IPV6_MIN_MTU)
 352                 return NULL;
 353
 354         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 355
 356         if (ndev == NULL)
 357                 return NULL;
 358
 359         rwlock_init(&ndev->lock);
 360         ndev->dev = dev;
 361         INIT_LIST_HEAD(&ndev->addr_list);
 362
 363         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 364         ndev->cnf.mtu6 = dev->mtu;
 365         ndev->cnf.sysctl = NULL;
 366         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 367         if (ndev->nd_parms == NULL) {
 368                 kfree(ndev);
 369                 return NULL;
 370         }
 371         if (ndev->cnf.forwarding)
 372                 dev_disable_lro(dev);
 373         /* We refer to the device */
 374         dev_hold(dev);
 375
 376         if (snmp6_alloc_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 dev_put(dev);
 382                 kfree(ndev);
 383                 return NULL;
 384         }
 385
 386         if (snmp6_register_dev(ndev) < 0) {
 387                 ADBG((KERN_WARNING
 388                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 389                         __func__, dev->name));
 390                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 391                 ndev->dead = 1;
 392                 in6_dev_finish_destroy(ndev);
 393                 return NULL;
 394         }
 395
 396         /* One reference from device.  We must do this before
 397          * we invoke __ipv6_regen_rndid().
 398          */
 399         in6_dev_hold(ndev);
 400
 401         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 402                 ndev->cnf.accept_dad = -1;
 403
 404 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 405         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 406                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 407                 ndev->cnf.rtr_solicits = 0;
 408         }
 409 #endif
 410
 411 #ifdef CONFIG_IPV6_PRIVACY
 412         INIT_LIST_HEAD(&ndev->tempaddr_list);
 413         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 414         if ((dev->flags&IFF_LOOPBACK) ||
 415             dev->type == ARPHRD_TUNNEL ||
 416             dev->type == ARPHRD_TUNNEL6 ||
 417             dev->type == ARPHRD_SIT ||
 418             dev->type == ARPHRD_NONE) {
 419                 ndev->cnf.use_tempaddr = -1;
 420         } else {
 421                 in6_dev_hold(ndev);
 422                 ipv6_regen_rndid((unsigned long) ndev);
 423         }
 424 #endif
 425
 426         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 427                 ndev->if_flags |= IF_READY;
 428
 429         ipv6_mc_init_dev(ndev);
 430         ndev->tstamp = jiffies;
 431         addrconf_sysctl_register(ndev);
 432         /* protected by rtnl_lock */
 433         rcu_assign_pointer(dev->ip6_ptr, ndev);
 434
 435         /* Join all-node multicast group */
 436         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 437
 438         /* Join all-router multicast group if forwarding is set */
 439         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 440                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 441
 442         return ndev;
 443 }
 444
 445 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 446 {
 447         struct inet6_dev *idev;
 448
 449         ASSERT_RTNL();
 450
 451         idev = __in6_dev_get(dev);
 452         if (!idev) {
 453                 idev = ipv6_add_dev(dev);
 454                 if (!idev)
 455                         return NULL;
 456         }
 457
 458         if (dev->flags&IFF_UP)
 459                 ipv6_mc_up(idev);
 460         return idev;
 461 }
 462
 463 #ifdef CONFIG_SYSCTL
 464 static void dev_forward_change(struct inet6_dev *idev)
 465 {
 466         struct net_device *dev;
 467         struct inet6_ifaddr *ifa;
 468
 469         if (!idev)
 470                 return;
 471         dev = idev->dev;
 472         if (idev->cnf.forwarding)
 473                 dev_disable_lro(dev);
 474         if (dev && (dev->flags & IFF_MULTICAST)) {
 475                 if (idev->cnf.forwarding)
 476                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 477                 else
 478                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 479         }
 480
 481         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 482                 if (ifa->flags&IFA_F_TENTATIVE)
 483                         continue;
 484                 if (idev->cnf.forwarding)
 485                         addrconf_join_anycast(ifa);
 486                 else
 487                         addrconf_leave_anycast(ifa);
 488         }
 489 }
 490
 491
 492 static void addrconf_forward_change(struct net *net, __s32 newf)
 493 {
 494         struct net_device *dev;
 495         struct inet6_dev *idev;
 496
 497         for_each_netdev(net, dev) {
 498                 idev = __in6_dev_get(dev);
 499                 if (idev) {
 500                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 501                         idev->cnf.forwarding = newf;
 502                         if (changed)
 503                                 dev_forward_change(idev);
 504                 }
 505         }
 506 }
 507
 508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 509 {
 510         struct net *net;
 511         int old;
 512
 513         if (!rtnl_trylock())
 514                 return restart_syscall();
 515
 516         net = (struct net *)table->extra2;
 517         old = *p;
 518         *p = newf;
 519
 520         if (p == &net->ipv6.devconf_dflt->forwarding) {
 521                 rtnl_unlock();
 522                 return 0;
 523         }
 524
 525         if (p == &net->ipv6.devconf_all->forwarding) {
 526                 net->ipv6.devconf_dflt->forwarding = newf;
 527                 addrconf_forward_change(net, newf);
 528         } else if ((!newf) ^ (!old))
 529                 dev_forward_change((struct inet6_dev *)table->extra1);
 530         rtnl_unlock();
 531
 532         if (newf)
 533                 rt6_purge_dflt_routers(net);
 534         return 1;
 535 }
 536 #endif
 537
 538 /* Nobody refers to this ifaddr, destroy it */
 539 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 540 {
 541         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 542
 543 #ifdef NET_REFCNT_DEBUG
 544         pr_debug("%s\n", __func__);
 545 #endif
 546
 547         in6_dev_put(ifp->idev);
 548
 549         if (del_timer(&ifp->timer))
 550                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 551
 552         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 553                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 554                 return;
 555         }
 556         dst_release(&ifp->rt->dst);
 557
 558         kfree_rcu(ifp, rcu);
 559 }
 560
 561 static void
 562 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 563 {
 564         struct list_head *p;
 565         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 566
 567         /*
 568          * Each device address list is sorted in order of scope -
 569          * global before linklocal.
 570          */
 571         list_for_each(p, &idev->addr_list) {
 572                 struct inet6_ifaddr *ifa
 573                         = list_entry(p, struct inet6_ifaddr, if_list);
 574                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 575                         break;
 576         }
 577
 578         list_add_tail(&ifp->if_list, p);
 579 }
 580
 581 static u32 inet6_addr_hash(const struct in6_addr *addr)
 582 {
 583         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, false);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ifa->addr = *addr;
 639
 640         spin_lock_init(&ifa->lock);
 641         spin_lock_init(&ifa->state_lock);
 642         init_timer(&ifa->timer);
 643         INIT_HLIST_NODE(&ifa->addr_lst);
 644         ifa->timer.data = (unsigned long) ifa;
 645         ifa->scope = scope;
 646         ifa->prefix_len = pfxlen;
 647         ifa->flags = flags | IFA_F_TENTATIVE;
 648         ifa->cstamp = ifa->tstamp = jiffies;
 649
 650         ifa->rt = rt;
 651
 652         ifa->idev = idev;
 653         in6_dev_hold(idev);
 654         /* For caller */
 655         in6_ifa_hold(ifa);
 656
 657         /* Add to big hash table */
 658         hash = inet6_addr_hash(addr);
 659
 660         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 661         spin_unlock(&addrconf_hash_lock);
 662
 663         write_lock(&idev->lock);
 664         /* Add to inet6_dev unicast addr list. */
 665         ipv6_link_dev_addr(idev, ifa);
 666
 667 #ifdef CONFIG_IPV6_PRIVACY
 668         if (ifa->flags&IFA_F_TEMPORARY) {
 669                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 670                 in6_ifa_hold(ifa);
 671         }
 672 #endif
 673
 674         in6_ifa_hold(ifa);
 675         write_unlock(&idev->lock);
 676 out2:
 677         rcu_read_unlock_bh();
 678
 679         if (likely(err == 0))
 680                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 681         else {
 682                 kfree(ifa);
 683                 ifa = ERR_PTR(err);
 684         }
 685
 686         return ifa;
 687 out:
 688         spin_unlock(&addrconf_hash_lock);
 689         goto out2;
 690 }
 691
 692 /* This function wants to get referenced ifp and releases it before return */
 693
 694 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 695 {
 696         struct inet6_ifaddr *ifa, *ifn;
 697         struct inet6_dev *idev = ifp->idev;
 698         int state;
 699         int deleted = 0, onlink = 0;
 700         unsigned long expires = jiffies;
 701
 702         spin_lock_bh(&ifp->state_lock);
 703         state = ifp->state;
 704         ifp->state = INET6_IFADDR_STATE_DEAD;
 705         spin_unlock_bh(&ifp->state_lock);
 706
 707         if (state == INET6_IFADDR_STATE_DEAD)
 708                 goto out;
 709
 710         spin_lock_bh(&addrconf_hash_lock);
 711         hlist_del_init_rcu(&ifp->addr_lst);
 712         spin_unlock_bh(&addrconf_hash_lock);
 713
 714         write_lock_bh(&idev->lock);
 715 #ifdef CONFIG_IPV6_PRIVACY
 716         if (ifp->flags&IFA_F_TEMPORARY) {
 717                 list_del(&ifp->tmp_list);
 718                 if (ifp->ifpub) {
 719                         in6_ifa_put(ifp->ifpub);
 720                         ifp->ifpub = NULL;
 721                 }
 722                 __in6_ifa_put(ifp);
 723         }
 724 #endif
 725
 726         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 727                 if (ifa == ifp) {
 728                         list_del_init(&ifp->if_list);
 729                         __in6_ifa_put(ifp);
 730
 731                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 732                                 break;
 733                         deleted = 1;
 734                         continue;
 735                 } else if (ifp->flags & IFA_F_PERMANENT) {
 736                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 737                                               ifp->prefix_len)) {
 738                                 if (ifa->flags & IFA_F_PERMANENT) {
 739                                         onlink = 1;
 740                                         if (deleted)
 741                                                 break;
 742                                 } else {
 743                                         unsigned long lifetime;
 744
 745                                         if (!onlink)
 746                                                 onlink = -1;
 747
 748                                         spin_lock(&ifa->lock);
 749
 750                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 751                                         /*
 752                                          * Note: Because this address is
 753                                          * not permanent, lifetime <
 754                                          * LONG_MAX / HZ here.
 755                                          */
 756                                         if (time_before(expires,
 757                                                         ifa->tstamp + lifetime * HZ))
 758                                                 expires = ifa->tstamp + lifetime * HZ;
 759                                         spin_unlock(&ifa->lock);
 760                                 }
 761                         }
 762                 }
 763         }
 764         write_unlock_bh(&idev->lock);
 765
 766         addrconf_del_timer(ifp);
 767
 768         ipv6_ifa_notify(RTM_DELADDR, ifp);
 769
 770         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 771
 772         /*
 773          * Purge or update corresponding prefix
 774          *
 775          * 1) we don't purge prefix here if address was not permanent.
 776          *    prefix is managed by its own lifetime.
 777          * 2) if there're no addresses, delete prefix.
 778          * 3) if there're still other permanent address(es),
 779          *    corresponding prefix is still permanent.
 780          * 4) otherwise, update prefix lifetime to the
 781          *    longest valid lifetime among the corresponding
 782          *    addresses on the device.
 783          *    Note: subsequent RA will update lifetime.
 784          *
 785          * --yoshfuji
 786          */
 787         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 788                 struct in6_addr prefix;
 789                 struct rt6_info *rt;
 790                 struct net *net = dev_net(ifp->idev->dev);
 791                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 792                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 793
 794                 if (rt && addrconf_is_prefix_route(rt)) {
 795                         if (onlink == 0) {
 796                                 ip6_del_rt(rt);
 797                                 rt = NULL;
 798                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 799                                 rt6_set_expires(rt, expires);
 800                         }
 801                 }
 802                 dst_release(&rt->dst);
 803         }
 804
 805         /* clean up prefsrc entries */
 806         rt6_remove_prefsrc(ifp);
 807 out:
 808         in6_ifa_put(ifp);
 809 }
 810
 811 #ifdef CONFIG_IPV6_PRIVACY
 812 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 813 {
 814         struct inet6_dev *idev = ifp->idev;
 815         struct in6_addr addr, *tmpaddr;
 816         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 817         unsigned long regen_advance;
 818         int tmp_plen;
 819         int ret = 0;
 820         int max_addresses;
 821         u32 addr_flags;
 822         unsigned long now = jiffies;
 823
 824         write_lock(&idev->lock);
 825         if (ift) {
 826                 spin_lock_bh(&ift->lock);
 827                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 828                 spin_unlock_bh(&ift->lock);
 829                 tmpaddr = &addr;
 830         } else {
 831                 tmpaddr = NULL;
 832         }
 833 retry:
 834         in6_dev_hold(idev);
 835         if (idev->cnf.use_tempaddr <= 0) {
 836                 write_unlock(&idev->lock);
 837                 pr_info("%s: use_tempaddr is disabled\n", __func__);
 838                 in6_dev_put(idev);
 839                 ret = -1;
 840                 goto out;
 841         }
 842         spin_lock_bh(&ifp->lock);
 843         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 844                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 845                 spin_unlock_bh(&ifp->lock);
 846                 write_unlock(&idev->lock);
 847                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
 848                         __func__);
 849                 in6_dev_put(idev);
 850                 ret = -1;
 851                 goto out;
 852         }
 853         in6_ifa_hold(ifp);
 854         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 855         __ipv6_try_regen_rndid(idev, tmpaddr);
 856         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 857         age = (now - ifp->tstamp) / HZ;
 858         tmp_valid_lft = min_t(__u32,
 859                               ifp->valid_lft,
 860                               idev->cnf.temp_valid_lft + age);
 861         tmp_prefered_lft = min_t(__u32,
 862                                  ifp->prefered_lft,
 863                                  idev->cnf.temp_prefered_lft + age -
 864                                  idev->cnf.max_desync_factor);
 865         tmp_plen = ifp->prefix_len;
 866         max_addresses = idev->cnf.max_addresses;
 867         tmp_tstamp = ifp->tstamp;
 868         spin_unlock_bh(&ifp->lock);
 869
 870         regen_advance = idev->cnf.regen_max_retry *
 871                         idev->cnf.dad_transmits *
 872                         idev->nd_parms->retrans_time / HZ;
 873         write_unlock(&idev->lock);
 874
 875         /* A temporary address is created only if this calculated Preferred
 876          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 877          * an implementation must not create a temporary address with a zero
 878          * Preferred Lifetime.
 879          */
 880         if (tmp_prefered_lft <= regen_advance) {
 881                 in6_ifa_put(ifp);
 882                 in6_dev_put(idev);
 883                 ret = -1;
 884                 goto out;
 885         }
 886
 887         addr_flags = IFA_F_TEMPORARY;
 888         /* set in addrconf_prefix_rcv() */
 889         if (ifp->flags & IFA_F_OPTIMISTIC)
 890                 addr_flags |= IFA_F_OPTIMISTIC;
 891
 892         ift = !max_addresses ||
 893               ipv6_count_addresses(idev) < max_addresses ?
 894                 ipv6_add_addr(idev, &addr, tmp_plen,
 895                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 896                               addr_flags) : NULL;
 897         if (!ift || IS_ERR(ift)) {
 898                 in6_ifa_put(ifp);
 899                 in6_dev_put(idev);
 900                 pr_info("%s: retry temporary address regeneration\n", __func__);
 901                 tmpaddr = &addr;
 902                 write_lock(&idev->lock);
 903                 goto retry;
 904         }
 905
 906         spin_lock_bh(&ift->lock);
 907         ift->ifpub = ifp;
 908         ift->valid_lft = tmp_valid_lft;
 909         ift->prefered_lft = tmp_prefered_lft;
 910         ift->cstamp = now;
 911         ift->tstamp = tmp_tstamp;
 912         spin_unlock_bh(&ift->lock);
 913
 914         addrconf_dad_start(ift);
 915         in6_ifa_put(ift);
 916         in6_dev_put(idev);
 917 out:
 918         return ret;
 919 }
 920 #endif
 921
 922 /*
 923  *      Choose an appropriate source address (RFC3484)
 924  */
 925 enum {
 926         IPV6_SADDR_RULE_INIT = 0,
 927         IPV6_SADDR_RULE_LOCAL,
 928         IPV6_SADDR_RULE_SCOPE,
 929         IPV6_SADDR_RULE_PREFERRED,
 930 #ifdef CONFIG_IPV6_MIP6
 931         IPV6_SADDR_RULE_HOA,
 932 #endif
 933         IPV6_SADDR_RULE_OIF,
 934         IPV6_SADDR_RULE_LABEL,
 935 #ifdef CONFIG_IPV6_PRIVACY
 936         IPV6_SADDR_RULE_PRIVACY,
 937 #endif
 938         IPV6_SADDR_RULE_ORCHID,
 939         IPV6_SADDR_RULE_PREFIX,
 940         IPV6_SADDR_RULE_MAX
 941 };
 942
 943 struct ipv6_saddr_score {
 944         int                     rule;
 945         int                     addr_type;
 946         struct inet6_ifaddr     *ifa;
 947         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 948         int                     scopedist;
 949         int                     matchlen;
 950 };
 951
 952 struct ipv6_saddr_dst {
 953         const struct in6_addr *addr;
 954         int ifindex;
 955         int scope;
 956         int label;
 957         unsigned int prefs;
 958 };
 959
 960 static inline int ipv6_saddr_preferred(int type)
 961 {
 962         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 963                 return 1;
 964         return 0;
 965 }
 966
 967 static int ipv6_get_saddr_eval(struct net *net,
 968                                struct ipv6_saddr_score *score,
 969                                struct ipv6_saddr_dst *dst,
 970                                int i)
 971 {
 972         int ret;
 973
 974         if (i <= score->rule) {
 975                 switch (i) {
 976                 case IPV6_SADDR_RULE_SCOPE:
 977                         ret = score->scopedist;
 978                         break;
 979                 case IPV6_SADDR_RULE_PREFIX:
 980                         ret = score->matchlen;
 981                         break;
 982                 default:
 983                         ret = !!test_bit(i, score->scorebits);
 984                 }
 985                 goto out;
 986         }
 987
 988         switch (i) {
 989         case IPV6_SADDR_RULE_INIT:
 990                 /* Rule 0: remember if hiscore is not ready yet */
 991                 ret = !!score->ifa;
 992                 break;
 993         case IPV6_SADDR_RULE_LOCAL:
 994                 /* Rule 1: Prefer same address */
 995                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
 996                 break;
 997         case IPV6_SADDR_RULE_SCOPE:
 998                 /* Rule 2: Prefer appropriate scope
 999                  *
1000                  *      ret
1001                  *       ^
1002                  *    -1 |  d 15
1003                  *    ---+--+-+---> scope
1004                  *       |
1005                  *       |             d is scope of the destination.
1006                  *  B-d  |  \
1007                  *       |   \      <- smaller scope is better if
1008                  *  B-15 |    \        if scope is enough for destinaion.
1009                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1010                  * d-C-1 | /
1011                  *       |/         <- greater is better
1012                  *   -C  /             if scope is not enough for destination.
1013                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1014                  *
1015                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1016                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1017                  * Assume B = 0 and we get C > 29.
1018                  */
1019                 ret = __ipv6_addr_src_scope(score->addr_type);
1020                 if (ret >= dst->scope)
1021                         ret = -ret;
1022                 else
1023                         ret -= 128;     /* 30 is enough */
1024                 score->scopedist = ret;
1025                 break;
1026         case IPV6_SADDR_RULE_PREFERRED:
1027                 /* Rule 3: Avoid deprecated and optimistic addresses */
1028                 ret = ipv6_saddr_preferred(score->addr_type) ||
1029                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1030                 break;
1031 #ifdef CONFIG_IPV6_MIP6
1032         case IPV6_SADDR_RULE_HOA:
1033             {
1034                 /* Rule 4: Prefer home address */
1035                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1036                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1037                 break;
1038             }
1039 #endif
1040         case IPV6_SADDR_RULE_OIF:
1041                 /* Rule 5: Prefer outgoing interface */
1042                 ret = (!dst->ifindex ||
1043                        dst->ifindex == score->ifa->idev->dev->ifindex);
1044                 break;
1045         case IPV6_SADDR_RULE_LABEL:
1046                 /* Rule 6: Prefer matching label */
1047                 ret = ipv6_addr_label(net,
1048                                       &score->ifa->addr, score->addr_type,
1049                                       score->ifa->idev->dev->ifindex) == dst->label;
1050                 break;
1051 #ifdef CONFIG_IPV6_PRIVACY
1052         case IPV6_SADDR_RULE_PRIVACY:
1053             {
1054                 /* Rule 7: Prefer public address
1055                  * Note: prefer temporary address if use_tempaddr >= 2
1056                  */
1057                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1058                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1059                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1060                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1061                 break;
1062             }
1063 #endif
1064         case IPV6_SADDR_RULE_ORCHID:
1065                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1066                  *          non-ORCHID vs non-ORCHID
1067                  */
1068                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1069                         ipv6_addr_orchid(dst->addr));
1070                 break;
1071         case IPV6_SADDR_RULE_PREFIX:
1072                 /* Rule 8: Use longest matching prefix */
1073                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1074                 if (ret > score->ifa->prefix_len)
1075                         ret = score->ifa->prefix_len;
1076                 score->matchlen = ret;
1077                 break;
1078         default:
1079                 ret = 0;
1080         }
1081
1082         if (ret)
1083                 __set_bit(i, score->scorebits);
1084         score->rule = i;
1085 out:
1086         return ret;
1087 }
1088
1089 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1090                        const struct in6_addr *daddr, unsigned int prefs,
1091                        struct in6_addr *saddr)
1092 {
1093         struct ipv6_saddr_score scores[2],
1094                                 *score = &scores[0], *hiscore = &scores[1];
1095         struct ipv6_saddr_dst dst;
1096         struct net_device *dev;
1097         int dst_type;
1098
1099         dst_type = __ipv6_addr_type(daddr);
1100         dst.addr = daddr;
1101         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1102         dst.scope = __ipv6_addr_src_scope(dst_type);
1103         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1104         dst.prefs = prefs;
1105
1106         hiscore->rule = -1;
1107         hiscore->ifa = NULL;
1108
1109         rcu_read_lock();
1110
1111         for_each_netdev_rcu(net, dev) {
1112                 struct inet6_dev *idev;
1113
1114                 /* Candidate Source Address (section 4)
1115                  *  - multicast and link-local destination address,
1116                  *    the set of candidate source address MUST only
1117                  *    include addresses assigned to interfaces
1118                  *    belonging to the same link as the outgoing
1119                  *    interface.
1120                  * (- For site-local destination addresses, the
1121                  *    set of candidate source addresses MUST only
1122                  *    include addresses assigned to interfaces
1123                  *    belonging to the same site as the outgoing
1124                  *    interface.)
1125                  */
1126                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1127                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1128                     dst.ifindex && dev->ifindex != dst.ifindex)
1129                         continue;
1130
1131                 idev = __in6_dev_get(dev);
1132                 if (!idev)
1133                         continue;
1134
1135                 read_lock_bh(&idev->lock);
1136                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1137                         int i;
1138
1139                         /*
1140                          * - Tentative Address (RFC2462 section 5.4)
1141                          *  - A tentative address is not considered
1142                          *    "assigned to an interface" in the traditional
1143                          *    sense, unless it is also flagged as optimistic.
1144                          * - Candidate Source Address (section 4)
1145                          *  - In any case, anycast addresses, multicast
1146                          *    addresses, and the unspecified address MUST
1147                          *    NOT be included in a candidate set.
1148                          */
1149                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1150                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1151                                 continue;
1152
1153                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1154
1155                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1156                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1157                                 LIMIT_NETDEBUG(KERN_DEBUG
1158                                                "ADDRCONF: unspecified / multicast address "
1159                                                "assigned as unicast address on %s",
1160                                                dev->name);
1161                                 continue;
1162                         }
1163
1164                         score->rule = -1;
1165                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1166
1167                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1168                                 int minihiscore, miniscore;
1169
1170                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1171                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1172
1173                                 if (minihiscore > miniscore) {
1174                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1175                                             score->scopedist > 0) {
1176                                                 /*
1177                                                  * special case:
1178                                                  * each remaining entry
1179                                                  * has too small (not enough)
1180                                                  * scope, because ifa entries
1181                                                  * are sorted by their scope
1182                                                  * values.
1183                                                  */
1184                                                 goto try_nextdev;
1185                                         }
1186                                         break;
1187                                 } else if (minihiscore < miniscore) {
1188                                         if (hiscore->ifa)
1189                                                 in6_ifa_put(hiscore->ifa);
1190
1191                                         in6_ifa_hold(score->ifa);
1192
1193                                         swap(hiscore, score);
1194
1195                                         /* restore our iterator */
1196                                         score->ifa = hiscore->ifa;
1197
1198                                         break;
1199                                 }
1200                         }
1201                 }
1202 try_nextdev:
1203                 read_unlock_bh(&idev->lock);
1204         }
1205         rcu_read_unlock();
1206
1207         if (!hiscore->ifa)
1208                 return -EADDRNOTAVAIL;
1209
1210         *saddr = hiscore->ifa->addr;
1211         in6_ifa_put(hiscore->ifa);
1212         return 0;
1213 }
1214 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1215
1216 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1217                     unsigned char banned_flags)
1218 {
1219         struct inet6_dev *idev;
1220         int err = -EADDRNOTAVAIL;
1221
1222         rcu_read_lock();
1223         idev = __in6_dev_get(dev);
1224         if (idev) {
1225                 struct inet6_ifaddr *ifp;
1226
1227                 read_lock_bh(&idev->lock);
1228                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1229                         if (ifp->scope == IFA_LINK &&
1230                             !(ifp->flags & banned_flags)) {
1231                                 *addr = ifp->addr;
1232                                 err = 0;
1233                                 break;
1234                         }
1235                 }
1236                 read_unlock_bh(&idev->lock);
1237         }
1238         rcu_read_unlock();
1239         return err;
1240 }
1241
1242 static int ipv6_count_addresses(struct inet6_dev *idev)
1243 {
1244         int cnt = 0;
1245         struct inet6_ifaddr *ifp;
1246
1247         read_lock_bh(&idev->lock);
1248         list_for_each_entry(ifp, &idev->addr_list, if_list)
1249                 cnt++;
1250         read_unlock_bh(&idev->lock);
1251         return cnt;
1252 }
1253
1254 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1255                   struct net_device *dev, int strict)
1256 {
1257         struct inet6_ifaddr *ifp;
1258         struct hlist_node *node;
1259         unsigned int hash = inet6_addr_hash(addr);
1260
1261         rcu_read_lock_bh();
1262         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1263                 if (!net_eq(dev_net(ifp->idev->dev), net))
1264                         continue;
1265                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1266                     !(ifp->flags&IFA_F_TENTATIVE) &&
1267                     (dev == NULL || ifp->idev->dev == dev ||
1268                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1269                         rcu_read_unlock_bh();
1270                         return 1;
1271                 }
1272         }
1273
1274         rcu_read_unlock_bh();
1275         return 0;
1276 }
1277 EXPORT_SYMBOL(ipv6_chk_addr);
1278
1279 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1280                                struct net_device *dev)
1281 {
1282         unsigned int hash = inet6_addr_hash(addr);
1283         struct inet6_ifaddr *ifp;
1284         struct hlist_node *node;
1285
1286         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1287                 if (!net_eq(dev_net(ifp->idev->dev), net))
1288                         continue;
1289                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1290                         if (dev == NULL || ifp->idev->dev == dev)
1291                                 return true;
1292                 }
1293         }
1294         return false;
1295 }
1296
1297 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1298 {
1299         struct inet6_dev *idev;
1300         struct inet6_ifaddr *ifa;
1301         int     onlink;
1302
1303         onlink = 0;
1304         rcu_read_lock();
1305         idev = __in6_dev_get(dev);
1306         if (idev) {
1307                 read_lock_bh(&idev->lock);
1308                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1309                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1310                                                    ifa->prefix_len);
1311                         if (onlink)
1312                                 break;
1313                 }
1314                 read_unlock_bh(&idev->lock);
1315         }
1316         rcu_read_unlock();
1317         return onlink;
1318 }
1319 EXPORT_SYMBOL(ipv6_chk_prefix);
1320
1321 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1322                                      struct net_device *dev, int strict)
1323 {
1324         struct inet6_ifaddr *ifp, *result = NULL;
1325         unsigned int hash = inet6_addr_hash(addr);
1326         struct hlist_node *node;
1327
1328         rcu_read_lock_bh();
1329         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1330                 if (!net_eq(dev_net(ifp->idev->dev), net))
1331                         continue;
1332                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1333                         if (dev == NULL || ifp->idev->dev == dev ||
1334                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1335                                 result = ifp;
1336                                 in6_ifa_hold(ifp);
1337                                 break;
1338                         }
1339                 }
1340         }
1341         rcu_read_unlock_bh();
1342
1343         return result;
1344 }
1345
1346 /* Gets referenced address, destroys ifaddr */
1347
1348 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1349 {
1350         if (ifp->flags&IFA_F_PERMANENT) {
1351                 spin_lock_bh(&ifp->lock);
1352                 addrconf_del_timer(ifp);
1353                 ifp->flags |= IFA_F_TENTATIVE;
1354                 if (dad_failed)
1355                         ifp->flags |= IFA_F_DADFAILED;
1356                 spin_unlock_bh(&ifp->lock);
1357                 if (dad_failed)
1358                         ipv6_ifa_notify(0, ifp);
1359                 in6_ifa_put(ifp);
1360 #ifdef CONFIG_IPV6_PRIVACY
1361         } else if (ifp->flags&IFA_F_TEMPORARY) {
1362                 struct inet6_ifaddr *ifpub;
1363                 spin_lock_bh(&ifp->lock);
1364                 ifpub = ifp->ifpub;
1365                 if (ifpub) {
1366                         in6_ifa_hold(ifpub);
1367                         spin_unlock_bh(&ifp->lock);
1368                         ipv6_create_tempaddr(ifpub, ifp);
1369                         in6_ifa_put(ifpub);
1370                 } else {
1371                         spin_unlock_bh(&ifp->lock);
1372                 }
1373                 ipv6_del_addr(ifp);
1374 #endif
1375         } else
1376                 ipv6_del_addr(ifp);
1377 }
1378
1379 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1380 {
1381         int err = -ENOENT;
1382
1383         spin_lock(&ifp->state_lock);
1384         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1385                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1386                 err = 0;
1387         }
1388         spin_unlock(&ifp->state_lock);
1389
1390         return err;
1391 }
1392
1393 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1394 {
1395         struct inet6_dev *idev = ifp->idev;
1396
1397         if (addrconf_dad_end(ifp)) {
1398                 in6_ifa_put(ifp);
1399                 return;
1400         }
1401
1402         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1403                              ifp->idev->dev->name, &ifp->addr);
1404
1405         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1406                 struct in6_addr addr;
1407
1408                 addr.s6_addr32[0] = htonl(0xfe800000);
1409                 addr.s6_addr32[1] = 0;
1410
1411                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1412                     ipv6_addr_equal(&ifp->addr, &addr)) {
1413                         /* DAD failed for link-local based on MAC address */
1414                         idev->cnf.disable_ipv6 = 1;
1415
1416                         pr_info("%s: IPv6 being disabled!\n",
1417                                 ifp->idev->dev->name);
1418                 }
1419         }
1420
1421         addrconf_dad_stop(ifp, 1);
1422 }
1423
1424 /* Join to solicited addr multicast group. */
1425
1426 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1427 {
1428         struct in6_addr maddr;
1429
1430         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1431                 return;
1432
1433         addrconf_addr_solict_mult(addr, &maddr);
1434         ipv6_dev_mc_inc(dev, &maddr);
1435 }
1436
1437 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1438 {
1439         struct in6_addr maddr;
1440
1441         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1442                 return;
1443
1444         addrconf_addr_solict_mult(addr, &maddr);
1445         __ipv6_dev_mc_dec(idev, &maddr);
1446 }
1447
1448 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1449 {
1450         struct in6_addr addr;
1451         if (ifp->prefix_len == 127) /* RFC 6164 */
1452                 return;
1453         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1454         if (ipv6_addr_any(&addr))
1455                 return;
1456         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1457 }
1458
1459 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1460 {
1461         struct in6_addr addr;
1462         if (ifp->prefix_len == 127) /* RFC 6164 */
1463                 return;
1464         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1465         if (ipv6_addr_any(&addr))
1466                 return;
1467         __ipv6_dev_ac_dec(ifp->idev, &addr);
1468 }
1469
1470 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1471 {
1472         if (dev->addr_len != ETH_ALEN)
1473                 return -1;
1474         memcpy(eui, dev->dev_addr, 3);
1475         memcpy(eui + 5, dev->dev_addr + 3, 3);
1476
1477         /*
1478          * The zSeries OSA network cards can be shared among various
1479          * OS instances, but the OSA cards have only one MAC address.
1480          * This leads to duplicate address conflicts in conjunction
1481          * with IPv6 if more than one instance uses the same card.
1482          *
1483          * The driver for these cards can deliver a unique 16-bit
1484          * identifier for each instance sharing the same card.  It is
1485          * placed instead of 0xFFFE in the interface identifier.  The
1486          * "u" bit of the interface identifier is not inverted in this
1487          * case.  Hence the resulting interface identifier has local
1488          * scope according to RFC2373.
1489          */
1490         if (dev->dev_id) {
1491                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1492                 eui[4] = dev->dev_id & 0xFF;
1493         } else {
1494                 eui[3] = 0xFF;
1495                 eui[4] = 0xFE;
1496                 eui[0] ^= 2;
1497         }
1498         return 0;
1499 }
1500
1501 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1502 {
1503         if (dev->addr_len != IEEE802154_ADDR_LEN)
1504                 return -1;
1505         memcpy(eui, dev->dev_addr, 8);
1506         return 0;
1507 }
1508
1509 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1510 {
1511         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1512         if (dev->addr_len != ARCNET_ALEN)
1513                 return -1;
1514         memset(eui, 0, 7);
1515         eui[7] = *(u8 *)dev->dev_addr;
1516         return 0;
1517 }
1518
1519 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1520 {
1521         if (dev->addr_len != INFINIBAND_ALEN)
1522                 return -1;
1523         memcpy(eui, dev->dev_addr + 12, 8);
1524         eui[0] |= 2;
1525         return 0;
1526 }
1527
1528 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1529 {
1530         if (addr == 0)
1531                 return -1;
1532         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1533                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1534                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1535                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1536                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1537                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1538         eui[1] = 0;
1539         eui[2] = 0x5E;
1540         eui[3] = 0xFE;
1541         memcpy(eui + 4, &addr, 4);
1542         return 0;
1543 }
1544
1545 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1546 {
1547         if (dev->priv_flags & IFF_ISATAP)
1548                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1549         return -1;
1550 }
1551
1552 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1553 {
1554         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1555 }
1556
1557 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1558 {
1559         switch (dev->type) {
1560         case ARPHRD_ETHER:
1561         case ARPHRD_FDDI:
1562                 return addrconf_ifid_eui48(eui, dev);
1563         case ARPHRD_ARCNET:
1564                 return addrconf_ifid_arcnet(eui, dev);
1565         case ARPHRD_INFINIBAND:
1566                 return addrconf_ifid_infiniband(eui, dev);
1567         case ARPHRD_SIT:
1568                 return addrconf_ifid_sit(eui, dev);
1569         case ARPHRD_IPGRE:
1570                 return addrconf_ifid_gre(eui, dev);
1571         case ARPHRD_IEEE802154:
1572                 return addrconf_ifid_eui64(eui, dev);
1573         }
1574         return -1;
1575 }
1576
1577 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1578 {
1579         int err = -1;
1580         struct inet6_ifaddr *ifp;
1581
1582         read_lock_bh(&idev->lock);
1583         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1584                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1585                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1586                         err = 0;
1587                         break;
1588                 }
1589         }
1590         read_unlock_bh(&idev->lock);
1591         return err;
1592 }
1593
1594 #ifdef CONFIG_IPV6_PRIVACY
1595 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1596 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1597 {
1598 regen:
1599         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1600         idev->rndid[0] &= ~0x02;
1601
1602         /*
1603          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1604          * check if generated address is not inappropriate
1605          *
1606          *  - Reserved subnet anycast (RFC 2526)
1607          *      11111101 11....11 1xxxxxxx
1608          *  - ISATAP (RFC4214) 6.1
1609          *      00-00-5E-FE-xx-xx-xx-xx
1610          *  - value 0
1611          *  - XXX: already assigned to an address on the device
1612          */
1613         if (idev->rndid[0] == 0xfd &&
1614             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1615             (idev->rndid[7]&0x80))
1616                 goto regen;
1617         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1618                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1619                         goto regen;
1620                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1621                         goto regen;
1622         }
1623 }
1624
1625 static void ipv6_regen_rndid(unsigned long data)
1626 {
1627         struct inet6_dev *idev = (struct inet6_dev *) data;
1628         unsigned long expires;
1629
1630         rcu_read_lock_bh();
1631         write_lock_bh(&idev->lock);
1632
1633         if (idev->dead)
1634                 goto out;
1635
1636         __ipv6_regen_rndid(idev);
1637
1638         expires = jiffies +
1639                 idev->cnf.temp_prefered_lft * HZ -
1640                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1641                 idev->cnf.max_desync_factor * HZ;
1642         if (time_before(expires, jiffies)) {
1643                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1644                         __func__, idev->dev->name);
1645                 goto out;
1646         }
1647
1648         if (!mod_timer(&idev->regen_timer, expires))
1649                 in6_dev_hold(idev);
1650
1651 out:
1652         write_unlock_bh(&idev->lock);
1653         rcu_read_unlock_bh();
1654         in6_dev_put(idev);
1655 }
1656
1657 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1658 {
1659         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1660                 __ipv6_regen_rndid(idev);
1661 }
1662 #endif
1663
1664 /*
1665  *      Add prefix route.
1666  */
1667
1668 static void
1669 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1670                       unsigned long expires, u32 flags)
1671 {
1672         struct fib6_config cfg = {
1673                 .fc_table = RT6_TABLE_PREFIX,
1674                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1675                 .fc_ifindex = dev->ifindex,
1676                 .fc_expires = expires,
1677                 .fc_dst_len = plen,
1678                 .fc_flags = RTF_UP | flags,
1679                 .fc_nlinfo.nl_net = dev_net(dev),
1680                 .fc_protocol = RTPROT_KERNEL,
1681         };
1682
1683         cfg.fc_dst = *pfx;
1684
1685         /* Prevent useless cloning on PtP SIT.
1686            This thing is done here expecting that the whole
1687            class of non-broadcast devices need not cloning.
1688          */
1689 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1690         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1691                 cfg.fc_flags |= RTF_NONEXTHOP;
1692 #endif
1693
1694         ip6_route_add(&cfg);
1695 }
1696
1697
1698 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1699                                                   int plen,
1700                                                   const struct net_device *dev,
1701                                                   u32 flags, u32 noflags)
1702 {
1703         struct fib6_node *fn;
1704         struct rt6_info *rt = NULL;
1705         struct fib6_table *table;
1706
1707         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1708         if (table == NULL)
1709                 return NULL;
1710
1711         read_lock_bh(&table->tb6_lock);
1712         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1713         if (!fn)
1714                 goto out;
1715         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1716                 if (rt->dst.dev->ifindex != dev->ifindex)
1717                         continue;
1718                 if ((rt->rt6i_flags & flags) != flags)
1719                         continue;
1720                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1721                         continue;
1722                 dst_hold(&rt->dst);
1723                 break;
1724         }
1725 out:
1726         read_unlock_bh(&table->tb6_lock);
1727         return rt;
1728 }
1729
1730
1731 /* Create "default" multicast route to the interface */
1732
1733 static void addrconf_add_mroute(struct net_device *dev)
1734 {
1735         struct fib6_config cfg = {
1736                 .fc_table = RT6_TABLE_LOCAL,
1737                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1738                 .fc_ifindex = dev->ifindex,
1739                 .fc_dst_len = 8,
1740                 .fc_flags = RTF_UP,
1741                 .fc_nlinfo.nl_net = dev_net(dev),
1742         };
1743
1744         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1745
1746         ip6_route_add(&cfg);
1747 }
1748
1749 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1750 static void sit_route_add(struct net_device *dev)
1751 {
1752         struct fib6_config cfg = {
1753                 .fc_table = RT6_TABLE_MAIN,
1754                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1755                 .fc_ifindex = dev->ifindex,
1756                 .fc_dst_len = 96,
1757                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1758                 .fc_nlinfo.nl_net = dev_net(dev),
1759         };
1760
1761         /* prefix length - 96 bits "::d.d.d.d" */
1762         ip6_route_add(&cfg);
1763 }
1764 #endif
1765
1766 static void addrconf_add_lroute(struct net_device *dev)
1767 {
1768         struct in6_addr addr;
1769
1770         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1771         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1772 }
1773
1774 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1775 {
1776         struct inet6_dev *idev;
1777
1778         ASSERT_RTNL();
1779
1780         idev = ipv6_find_idev(dev);
1781         if (!idev)
1782                 return ERR_PTR(-ENOBUFS);
1783
1784         if (idev->cnf.disable_ipv6)
1785                 return ERR_PTR(-EACCES);
1786
1787         /* Add default multicast route */
1788         if (!(dev->flags & IFF_LOOPBACK))
1789                 addrconf_add_mroute(dev);
1790
1791         /* Add link local route */
1792         addrconf_add_lroute(dev);
1793         return idev;
1794 }
1795
1796 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1797 {
1798         struct prefix_info *pinfo;
1799         __u32 valid_lft;
1800         __u32 prefered_lft;
1801         int addr_type;
1802         struct inet6_dev *in6_dev;
1803         struct net *net = dev_net(dev);
1804
1805         pinfo = (struct prefix_info *) opt;
1806
1807         if (len < sizeof(struct prefix_info)) {
1808                 ADBG(("addrconf: prefix option too short\n"));
1809                 return;
1810         }
1811
1812         /*
1813          *      Validation checks ([ADDRCONF], page 19)
1814          */
1815
1816         addr_type = ipv6_addr_type(&pinfo->prefix);
1817
1818         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1819                 return;
1820
1821         valid_lft = ntohl(pinfo->valid);
1822         prefered_lft = ntohl(pinfo->prefered);
1823
1824         if (prefered_lft > valid_lft) {
1825                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1826                 return;
1827         }
1828
1829         in6_dev = in6_dev_get(dev);
1830
1831         if (in6_dev == NULL) {
1832                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1833                                     dev->name);
1834                 return;
1835         }
1836
1837         /*
1838          *      Two things going on here:
1839          *      1) Add routes for on-link prefixes
1840          *      2) Configure prefixes with the auto flag set
1841          */
1842
1843         if (pinfo->onlink) {
1844                 struct rt6_info *rt;
1845                 unsigned long rt_expires;
1846
1847                 /* Avoid arithmetic overflow. Really, we could
1848                  * save rt_expires in seconds, likely valid_lft,
1849                  * but it would require division in fib gc, that it
1850                  * not good.
1851                  */
1852                 if (HZ > USER_HZ)
1853                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1854                 else
1855                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1856
1857                 if (addrconf_finite_timeout(rt_expires))
1858                         rt_expires *= HZ;
1859
1860                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1861                                                pinfo->prefix_len,
1862                                                dev,
1863                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1864                                                RTF_GATEWAY | RTF_DEFAULT);
1865
1866                 if (rt) {
1867                         /* Autoconf prefix route */
1868                         if (valid_lft == 0) {
1869                                 ip6_del_rt(rt);
1870                                 rt = NULL;
1871                         } else if (addrconf_finite_timeout(rt_expires)) {
1872                                 /* not infinity */
1873                                 rt6_set_expires(rt, jiffies + rt_expires);
1874                         } else {
1875                                 rt6_clean_expires(rt);
1876                         }
1877                 } else if (valid_lft) {
1878                         clock_t expires = 0;
1879                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1880                         if (addrconf_finite_timeout(rt_expires)) {
1881                                 /* not infinity */
1882                                 flags |= RTF_EXPIRES;
1883                                 expires = jiffies_to_clock_t(rt_expires);
1884                         }
1885                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1886                                               dev, expires, flags);
1887                 }
1888                 if (rt)
1889                         dst_release(&rt->dst);
1890         }
1891
1892         /* Try to figure out our local address for this prefix */
1893
1894         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1895                 struct inet6_ifaddr *ifp;
1896                 struct in6_addr addr;
1897                 int create = 0, update_lft = 0;
1898
1899                 if (pinfo->prefix_len == 64) {
1900                         memcpy(&addr, &pinfo->prefix, 8);
1901                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1902                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1903                                 in6_dev_put(in6_dev);
1904                                 return;
1905                         }
1906                         goto ok;
1907                 }
1908                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
1909                                     pinfo->prefix_len);
1910                 in6_dev_put(in6_dev);
1911                 return;
1912
1913 ok:
1914
1915                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1916
1917                 if (ifp == NULL && valid_lft) {
1918                         int max_addresses = in6_dev->cnf.max_addresses;
1919                         u32 addr_flags = 0;
1920
1921 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1922                         if (in6_dev->cnf.optimistic_dad &&
1923                             !net->ipv6.devconf_all->forwarding && sllao)
1924                                 addr_flags = IFA_F_OPTIMISTIC;
1925 #endif
1926
1927                         /* Do not allow to create too much of autoconfigured
1928                          * addresses; this would be too easy way to crash kernel.
1929                          */
1930                         if (!max_addresses ||
1931                             ipv6_count_addresses(in6_dev) < max_addresses)
1932                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1933                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1934                                                     addr_flags);
1935
1936                         if (!ifp || IS_ERR(ifp)) {
1937                                 in6_dev_put(in6_dev);
1938                                 return;
1939                         }
1940
1941                         update_lft = create = 1;
1942                         ifp->cstamp = jiffies;
1943                         addrconf_dad_start(ifp);
1944                 }
1945
1946                 if (ifp) {
1947                         int flags;
1948                         unsigned long now;
1949 #ifdef CONFIG_IPV6_PRIVACY
1950                         struct inet6_ifaddr *ift;
1951 #endif
1952                         u32 stored_lft;
1953
1954                         /* update lifetime (RFC2462 5.5.3 e) */
1955                         spin_lock(&ifp->lock);
1956                         now = jiffies;
1957                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1958                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1959                         else
1960                                 stored_lft = 0;
1961                         if (!update_lft && stored_lft) {
1962                                 if (valid_lft > MIN_VALID_LIFETIME ||
1963                                     valid_lft > stored_lft)
1964                                         update_lft = 1;
1965                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1966                                         /* valid_lft <= stored_lft is always true */
1967                                         /*
1968                                          * RFC 4862 Section 5.5.3e:
1969                                          * "Note that the preferred lifetime of
1970                                          *  the corresponding address is always
1971                                          *  reset to the Preferred Lifetime in
1972                                          *  the received Prefix Information
1973                                          *  option, regardless of whether the
1974                                          *  valid lifetime is also reset or
1975                                          *  ignored."
1976                                          *
1977                                          *  So if the preferred lifetime in
1978                                          *  this advertisement is different
1979                                          *  than what we have stored, but the
1980                                          *  valid lifetime is invalid, just
1981                                          *  reset prefered_lft.
1982                                          *
1983                                          *  We must set the valid lifetime
1984                                          *  to the stored lifetime since we'll
1985                                          *  be updating the timestamp below,
1986                                          *  else we'll set it back to the
1987                                          *  minimum.
1988                                          */
1989                                         if (prefered_lft != ifp->prefered_lft) {
1990                                                 valid_lft = stored_lft;
1991                                                 update_lft = 1;
1992                                         }
1993                                 } else {
1994                                         valid_lft = MIN_VALID_LIFETIME;
1995                                         if (valid_lft < prefered_lft)
1996                                                 prefered_lft = valid_lft;
1997                                         update_lft = 1;
1998                                 }
1999                         }
2000
2001                         if (update_lft) {
2002                                 ifp->valid_lft = valid_lft;
2003                                 ifp->prefered_lft = prefered_lft;
2004                                 ifp->tstamp = now;
2005                                 flags = ifp->flags;
2006                                 ifp->flags &= ~IFA_F_DEPRECATED;
2007                                 spin_unlock(&ifp->lock);
2008
2009                                 if (!(flags&IFA_F_TENTATIVE))
2010                                         ipv6_ifa_notify(0, ifp);
2011                         } else
2012                                 spin_unlock(&ifp->lock);
2013
2014 #ifdef CONFIG_IPV6_PRIVACY
2015                         read_lock_bh(&in6_dev->lock);
2016                         /* update all temporary addresses in the list */
2017                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2018                                             tmp_list) {
2019                                 int age, max_valid, max_prefered;
2020
2021                                 if (ifp != ift->ifpub)
2022                                         continue;
2023
2024                                 /*
2025                                  * RFC 4941 section 3.3:
2026                                  * If a received option will extend the lifetime
2027                                  * of a public address, the lifetimes of
2028                                  * temporary addresses should be extended,
2029                                  * subject to the overall constraint that no
2030                                  * temporary addresses should ever remain
2031                                  * "valid" or "preferred" for a time longer than
2032                                  * (TEMP_VALID_LIFETIME) or
2033                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2034                                  * respectively.
2035                                  */
2036                                 age = (now - ift->cstamp) / HZ;
2037                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2038                                 if (max_valid < 0)
2039                                         max_valid = 0;
2040
2041                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2042                                                in6_dev->cnf.max_desync_factor -
2043                                                age;
2044                                 if (max_prefered < 0)
2045                                         max_prefered = 0;
2046
2047                                 if (valid_lft > max_valid)
2048                                         valid_lft = max_valid;
2049
2050                                 if (prefered_lft > max_prefered)
2051                                         prefered_lft = max_prefered;
2052
2053                                 spin_lock(&ift->lock);
2054                                 flags = ift->flags;
2055                                 ift->valid_lft = valid_lft;
2056                                 ift->prefered_lft = prefered_lft;
2057                                 ift->tstamp = now;
2058                                 if (prefered_lft > 0)
2059                                         ift->flags &= ~IFA_F_DEPRECATED;
2060
2061                                 spin_unlock(&ift->lock);
2062                                 if (!(flags&IFA_F_TENTATIVE))
2063                                         ipv6_ifa_notify(0, ift);
2064                         }
2065
2066                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2067                                 /*
2068                                  * When a new public address is created as
2069                                  * described in [ADDRCONF], also create a new
2070                                  * temporary address. Also create a temporary
2071                                  * address if it's enabled but no temporary
2072                                  * address currently exists.
2073                                  */
2074                                 read_unlock_bh(&in6_dev->lock);
2075                                 ipv6_create_tempaddr(ifp, NULL);
2076                         } else {
2077                                 read_unlock_bh(&in6_dev->lock);
2078                         }
2079 #endif
2080                         in6_ifa_put(ifp);
2081                         addrconf_verify(0);
2082                 }
2083         }
2084         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2085         in6_dev_put(in6_dev);
2086 }
2087
2088 /*
2089  *      Set destination address.
2090  *      Special case for SIT interfaces where we create a new "virtual"
2091  *      device.
2092  */
2093 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2094 {
2095         struct in6_ifreq ireq;
2096         struct net_device *dev;
2097         int err = -EINVAL;
2098
2099         rtnl_lock();
2100
2101         err = -EFAULT;
2102         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2103                 goto err_exit;
2104
2105         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2106
2107         err = -ENODEV;
2108         if (dev == NULL)
2109                 goto err_exit;
2110
2111 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2112         if (dev->type == ARPHRD_SIT) {
2113                 const struct net_device_ops *ops = dev->netdev_ops;
2114                 struct ifreq ifr;
2115                 struct ip_tunnel_parm p;
2116
2117                 err = -EADDRNOTAVAIL;
2118                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2119                         goto err_exit;
2120
2121                 memset(&p, 0, sizeof(p));
2122                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2123                 p.iph.saddr = 0;
2124                 p.iph.version = 4;
2125                 p.iph.ihl = 5;
2126                 p.iph.protocol = IPPROTO_IPV6;
2127                 p.iph.ttl = 64;
2128                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2129
2130                 if (ops->ndo_do_ioctl) {
2131                         mm_segment_t oldfs = get_fs();
2132
2133                         set_fs(KERNEL_DS);
2134                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2135                         set_fs(oldfs);
2136                 } else
2137                         err = -EOPNOTSUPP;
2138
2139                 if (err == 0) {
2140                         err = -ENOBUFS;
2141                         dev = __dev_get_by_name(net, p.name);
2142                         if (!dev)
2143                                 goto err_exit;
2144                         err = dev_open(dev);
2145                 }
2146         }
2147 #endif
2148
2149 err_exit:
2150         rtnl_unlock();
2151         return err;
2152 }
2153
2154 /*
2155  *      Manual configuration of address on an interface
2156  */
2157 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2158                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2159                           __u32 valid_lft)
2160 {
2161         struct inet6_ifaddr *ifp;
2162         struct inet6_dev *idev;
2163         struct net_device *dev;
2164         int scope;
2165         u32 flags;
2166         clock_t expires;
2167         unsigned long timeout;
2168
2169         ASSERT_RTNL();
2170
2171         if (plen > 128)
2172                 return -EINVAL;
2173
2174         /* check the lifetime */
2175         if (!valid_lft || prefered_lft > valid_lft)
2176                 return -EINVAL;
2177
2178         dev = __dev_get_by_index(net, ifindex);
2179         if (!dev)
2180                 return -ENODEV;
2181
2182         idev = addrconf_add_dev(dev);
2183         if (IS_ERR(idev))
2184                 return PTR_ERR(idev);
2185
2186         scope = ipv6_addr_scope(pfx);
2187
2188         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2189         if (addrconf_finite_timeout(timeout)) {
2190                 expires = jiffies_to_clock_t(timeout * HZ);
2191                 valid_lft = timeout;
2192                 flags = RTF_EXPIRES;
2193         } else {
2194                 expires = 0;
2195                 flags = 0;
2196                 ifa_flags |= IFA_F_PERMANENT;
2197         }
2198
2199         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2200         if (addrconf_finite_timeout(timeout)) {
2201                 if (timeout == 0)
2202                         ifa_flags |= IFA_F_DEPRECATED;
2203                 prefered_lft = timeout;
2204         }
2205
2206         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2207
2208         if (!IS_ERR(ifp)) {
2209                 spin_lock_bh(&ifp->lock);
2210                 ifp->valid_lft = valid_lft;
2211                 ifp->prefered_lft = prefered_lft;
2212                 ifp->tstamp = jiffies;
2213                 spin_unlock_bh(&ifp->lock);
2214
2215                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2216                                       expires, flags);
2217                 /*
2218                  * Note that section 3.1 of RFC 4429 indicates
2219                  * that the Optimistic flag should not be set for
2220                  * manually configured addresses
2221                  */
2222                 addrconf_dad_start(ifp);
2223                 in6_ifa_put(ifp);
2224                 addrconf_verify(0);
2225                 return 0;
2226         }
2227
2228         return PTR_ERR(ifp);
2229 }
2230
2231 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2232                           unsigned int plen)
2233 {
2234         struct inet6_ifaddr *ifp;
2235         struct inet6_dev *idev;
2236         struct net_device *dev;
2237
2238         if (plen > 128)
2239                 return -EINVAL;
2240
2241         dev = __dev_get_by_index(net, ifindex);
2242         if (!dev)
2243                 return -ENODEV;
2244
2245         if ((idev = __in6_dev_get(dev)) == NULL)
2246                 return -ENXIO;
2247
2248         read_lock_bh(&idev->lock);
2249         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2250                 if (ifp->prefix_len == plen &&
2251                     ipv6_addr_equal(pfx, &ifp->addr)) {
2252                         in6_ifa_hold(ifp);
2253                         read_unlock_bh(&idev->lock);
2254
2255                         ipv6_del_addr(ifp);
2256
2257                         /* If the last address is deleted administratively,
2258                            disable IPv6 on this interface.
2259                          */
2260                         if (list_empty(&idev->addr_list))
2261                                 addrconf_ifdown(idev->dev, 1);
2262                         return 0;
2263                 }
2264         }
2265         read_unlock_bh(&idev->lock);
2266         return -EADDRNOTAVAIL;
2267 }
2268
2269
2270 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2271 {
2272         struct in6_ifreq ireq;
2273         int err;
2274
2275         if (!capable(CAP_NET_ADMIN))
2276                 return -EPERM;
2277
2278         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2279                 return -EFAULT;
2280
2281         rtnl_lock();
2282         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2283                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2284                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2285         rtnl_unlock();
2286         return err;
2287 }
2288
2289 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2290 {
2291         struct in6_ifreq ireq;
2292         int err;
2293
2294         if (!capable(CAP_NET_ADMIN))
2295                 return -EPERM;
2296
2297         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2298                 return -EFAULT;
2299
2300         rtnl_lock();
2301         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2302                              ireq.ifr6_prefixlen);
2303         rtnl_unlock();
2304         return err;
2305 }
2306
2307 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2308                      int plen, int scope)
2309 {
2310         struct inet6_ifaddr *ifp;
2311
2312         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2313         if (!IS_ERR(ifp)) {
2314                 spin_lock_bh(&ifp->lock);
2315                 ifp->flags &= ~IFA_F_TENTATIVE;
2316                 spin_unlock_bh(&ifp->lock);
2317                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2318                 in6_ifa_put(ifp);
2319         }
2320 }
2321
2322 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2323 static void sit_add_v4_addrs(struct inet6_dev *idev)
2324 {
2325         struct in6_addr addr;
2326         struct net_device *dev;
2327         struct net *net = dev_net(idev->dev);
2328         int scope;
2329
2330         ASSERT_RTNL();
2331
2332         memset(&addr, 0, sizeof(struct in6_addr));
2333         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2334
2335         if (idev->dev->flags&IFF_POINTOPOINT) {
2336                 addr.s6_addr32[0] = htonl(0xfe800000);
2337                 scope = IFA_LINK;
2338         } else {
2339                 scope = IPV6_ADDR_COMPATv4;
2340         }
2341
2342         if (addr.s6_addr32[3]) {
2343                 add_addr(idev, &addr, 128, scope);
2344                 return;
2345         }
2346
2347         for_each_netdev(net, dev) {
2348                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2349                 if (in_dev && (dev->flags & IFF_UP)) {
2350                         struct in_ifaddr *ifa;
2351
2352                         int flag = scope;
2353
2354                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2355                                 int plen;
2356
2357                                 addr.s6_addr32[3] = ifa->ifa_local;
2358
2359                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2360                                         continue;
2361                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2362                                         if (idev->dev->flags&IFF_POINTOPOINT)
2363                                                 continue;
2364                                         flag |= IFA_HOST;
2365                                 }
2366                                 if (idev->dev->flags&IFF_POINTOPOINT)
2367                                         plen = 64;
2368                                 else
2369                                         plen = 96;
2370
2371                                 add_addr(idev, &addr, plen, flag);
2372                         }
2373                 }
2374         }
2375 }
2376 #endif
2377
2378 static void init_loopback(struct net_device *dev)
2379 {
2380         struct inet6_dev  *idev;
2381
2382         /* ::1 */
2383
2384         ASSERT_RTNL();
2385
2386         if ((idev = ipv6_find_idev(dev)) == NULL) {
2387                 pr_debug("%s: add_dev failed\n", __func__);
2388                 return;
2389         }
2390
2391         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2392 }
2393
2394 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2395 {
2396         struct inet6_ifaddr *ifp;
2397         u32 addr_flags = IFA_F_PERMANENT;
2398
2399 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2400         if (idev->cnf.optimistic_dad &&
2401             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2402                 addr_flags |= IFA_F_OPTIMISTIC;
2403 #endif
2404
2405
2406         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2407         if (!IS_ERR(ifp)) {
2408                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2409                 addrconf_dad_start(ifp);
2410                 in6_ifa_put(ifp);
2411         }
2412 }
2413
2414 static void addrconf_dev_config(struct net_device *dev)
2415 {
2416         struct in6_addr addr;
2417         struct inet6_dev *idev;
2418
2419         ASSERT_RTNL();
2420
2421         if ((dev->type != ARPHRD_ETHER) &&
2422             (dev->type != ARPHRD_FDDI) &&
2423             (dev->type != ARPHRD_ARCNET) &&
2424             (dev->type != ARPHRD_INFINIBAND) &&
2425             (dev->type != ARPHRD_IEEE802154)) {
2426                 /* Alas, we support only Ethernet autoconfiguration. */
2427                 return;
2428         }
2429
2430         idev = addrconf_add_dev(dev);
2431         if (IS_ERR(idev))
2432                 return;
2433
2434         memset(&addr, 0, sizeof(struct in6_addr));
2435         addr.s6_addr32[0] = htonl(0xFE800000);
2436
2437         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2438                 addrconf_add_linklocal(idev, &addr);
2439 }
2440
2441 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2442 static void addrconf_sit_config(struct net_device *dev)
2443 {
2444         struct inet6_dev *idev;
2445
2446         ASSERT_RTNL();
2447
2448         /*
2449          * Configure the tunnel with one of our IPv4
2450          * addresses... we should configure all of
2451          * our v4 addrs in the tunnel
2452          */
2453
2454         if ((idev = ipv6_find_idev(dev)) == NULL) {
2455                 pr_debug("%s: add_dev failed\n", __func__);
2456                 return;
2457         }
2458
2459         if (dev->priv_flags & IFF_ISATAP) {
2460                 struct in6_addr addr;
2461
2462                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2463                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2464                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2465                         addrconf_add_linklocal(idev, &addr);
2466                 return;
2467         }
2468
2469         sit_add_v4_addrs(idev);
2470
2471         if (dev->flags&IFF_POINTOPOINT) {
2472                 addrconf_add_mroute(dev);
2473                 addrconf_add_lroute(dev);
2474         } else
2475                 sit_route_add(dev);
2476 }
2477 #endif
2478
2479 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2480 static void addrconf_gre_config(struct net_device *dev)
2481 {
2482         struct inet6_dev *idev;
2483         struct in6_addr addr;
2484
2485         pr_info("%s(%s)\n", __func__, dev->name);
2486
2487         ASSERT_RTNL();
2488
2489         if ((idev = ipv6_find_idev(dev)) == NULL) {
2490                 pr_debug("%s: add_dev failed\n", __func__);
2491                 return;
2492         }
2493
2494         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2495         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2496
2497         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2498                 addrconf_add_linklocal(idev, &addr);
2499 }
2500 #endif
2501
2502 static inline int
2503 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2504 {
2505         struct in6_addr lladdr;
2506
2507         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2508                 addrconf_add_linklocal(idev, &lladdr);
2509                 return 0;
2510         }
2511         return -1;
2512 }
2513
2514 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2515 {
2516         struct net_device *link_dev;
2517         struct net *net = dev_net(idev->dev);
2518
2519         /* first try to inherit the link-local address from the link device */
2520         if (idev->dev->iflink &&
2521             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2522                 if (!ipv6_inherit_linklocal(idev, link_dev))
2523                         return;
2524         }
2525         /* then try to inherit it from any device */
2526         for_each_netdev(net, link_dev) {
2527                 if (!ipv6_inherit_linklocal(idev, link_dev))
2528                         return;
2529         }
2530         pr_debug("init ip6-ip6: add_linklocal failed\n");
2531 }
2532
2533 /*
2534  * Autoconfigure tunnel with a link-local address so routing protocols,
2535  * DHCPv6, MLD etc. can be run over the virtual link
2536  */
2537
2538 static void addrconf_ip6_tnl_config(struct net_device *dev)
2539 {
2540         struct inet6_dev *idev;
2541
2542         ASSERT_RTNL();
2543
2544         idev = addrconf_add_dev(dev);
2545         if (IS_ERR(idev)) {
2546                 pr_debug("init ip6-ip6: add_dev failed\n");
2547                 return;
2548         }
2549         ip6_tnl_add_linklocal(idev);
2550 }
2551
2552 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2553                            void *data)
2554 {
2555         struct net_device *dev = (struct net_device *) data;
2556         struct inet6_dev *idev = __in6_dev_get(dev);
2557         int run_pending = 0;
2558         int err;
2559
2560         switch (event) {
2561         case NETDEV_REGISTER:
2562                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2563                         idev = ipv6_add_dev(dev);
2564                         if (!idev)
2565                                 return notifier_from_errno(-ENOMEM);
2566                 }
2567                 break;
2568
2569         case NETDEV_UP:
2570         case NETDEV_CHANGE:
2571                 if (dev->flags & IFF_SLAVE)
2572                         break;
2573
2574                 if (event == NETDEV_UP) {
2575                         if (!addrconf_qdisc_ok(dev)) {
2576                                 /* device is not ready yet. */
2577                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2578                                         dev->name);
2579                                 break;
2580                         }
2581
2582                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2583                                 idev = ipv6_add_dev(dev);
2584
2585                         if (idev) {
2586                                 idev->if_flags |= IF_READY;
2587                                 run_pending = 1;
2588                         }
2589                 } else {
2590                         if (!addrconf_qdisc_ok(dev)) {
2591                                 /* device is still not ready. */
2592                                 break;
2593                         }
2594
2595                         if (idev) {
2596                                 if (idev->if_flags & IF_READY)
2597                                         /* device is already configured. */
2598                                         break;
2599                                 idev->if_flags |= IF_READY;
2600                         }
2601
2602                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2603                                 dev->name);
2604
2605                         run_pending = 1;
2606                 }
2607
2608                 switch (dev->type) {
2609 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2610                 case ARPHRD_SIT:
2611                         addrconf_sit_config(dev);
2612                         break;
2613 #endif
2614 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2615                 case ARPHRD_IPGRE:
2616                         addrconf_gre_config(dev);
2617                         break;
2618 #endif
2619                 case ARPHRD_TUNNEL6:
2620                         addrconf_ip6_tnl_config(dev);
2621                         break;
2622                 case ARPHRD_LOOPBACK:
2623                         init_loopback(dev);
2624                         break;
2625
2626                 default:
2627                         addrconf_dev_config(dev);
2628                         break;
2629                 }
2630
2631                 if (idev) {
2632                         if (run_pending)
2633                                 addrconf_dad_run(idev);
2634
2635                         /*
2636                          * If the MTU changed during the interface down,
2637                          * when the interface up, the changed MTU must be
2638                          * reflected in the idev as well as routers.
2639                          */
2640                         if (idev->cnf.mtu6 != dev->mtu &&
2641                             dev->mtu >= IPV6_MIN_MTU) {
2642                                 rt6_mtu_change(dev, dev->mtu);
2643                                 idev->cnf.mtu6 = dev->mtu;
2644                         }
2645                         idev->tstamp = jiffies;
2646                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2647
2648                         /*
2649                          * If the changed mtu during down is lower than
2650                          * IPV6_MIN_MTU stop IPv6 on this interface.
2651                          */
2652                         if (dev->mtu < IPV6_MIN_MTU)
2653                                 addrconf_ifdown(dev, 1);
2654                 }
2655                 break;
2656
2657         case NETDEV_CHANGEMTU:
2658                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2659                         rt6_mtu_change(dev, dev->mtu);
2660                         idev->cnf.mtu6 = dev->mtu;
2661                         break;
2662                 }
2663
2664                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2665                         idev = ipv6_add_dev(dev);
2666                         if (idev)
2667                                 break;
2668                 }
2669
2670                 /*
2671                  * MTU falled under IPV6_MIN_MTU.
2672                  * Stop IPv6 on this interface.
2673                  */
2674
2675         case NETDEV_DOWN:
2676         case NETDEV_UNREGISTER:
2677                 /*
2678                  *      Remove all addresses from this interface.
2679                  */
2680                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2681                 break;
2682
2683         case NETDEV_CHANGENAME:
2684                 if (idev) {
2685                         snmp6_unregister_dev(idev);
2686                         addrconf_sysctl_unregister(idev);
2687                         addrconf_sysctl_register(idev);
2688                         err = snmp6_register_dev(idev);
2689                         if (err)
2690                                 return notifier_from_errno(err);
2691                 }
2692                 break;
2693
2694         case NETDEV_PRE_TYPE_CHANGE:
2695         case NETDEV_POST_TYPE_CHANGE:
2696                 addrconf_type_change(dev, event);
2697                 break;
2698         }
2699
2700         return NOTIFY_OK;
2701 }
2702
2703 /*
2704  *      addrconf module should be notified of a device going up
2705  */
2706 static struct notifier_block ipv6_dev_notf = {
2707         .notifier_call = addrconf_notify,
2708 };
2709
2710 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2711 {
2712         struct inet6_dev *idev;
2713         ASSERT_RTNL();
2714
2715         idev = __in6_dev_get(dev);
2716
2717         if (event == NETDEV_POST_TYPE_CHANGE)
2718                 ipv6_mc_remap(idev);
2719         else if (event == NETDEV_PRE_TYPE_CHANGE)
2720                 ipv6_mc_unmap(idev);
2721 }
2722
2723 static int addrconf_ifdown(struct net_device *dev, int how)
2724 {
2725         struct net *net = dev_net(dev);
2726         struct inet6_dev *idev;
2727         struct inet6_ifaddr *ifa;
2728         int state, i;
2729
2730         ASSERT_RTNL();
2731
2732         rt6_ifdown(net, dev);
2733         neigh_ifdown(&nd_tbl, dev);
2734
2735         idev = __in6_dev_get(dev);
2736         if (idev == NULL)
2737                 return -ENODEV;
2738
2739         /*
2740          * Step 1: remove reference to ipv6 device from parent device.
2741          *         Do not dev_put!
2742          */
2743         if (how) {
2744                 idev->dead = 1;
2745
2746                 /* protected by rtnl_lock */
2747                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2748
2749                 /* Step 1.5: remove snmp6 entry */
2750                 snmp6_unregister_dev(idev);
2751
2752         }
2753
2754         /* Step 2: clear hash table */
2755         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2756                 struct hlist_head *h = &inet6_addr_lst[i];
2757                 struct hlist_node *n;
2758
2759                 spin_lock_bh(&addrconf_hash_lock);
2760         restart:
2761                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2762                         if (ifa->idev == idev) {
2763                                 hlist_del_init_rcu(&ifa->addr_lst);
2764                                 addrconf_del_timer(ifa);
2765                                 goto restart;
2766                         }
2767                 }
2768                 spin_unlock_bh(&addrconf_hash_lock);
2769         }
2770
2771         write_lock_bh(&idev->lock);
2772
2773         /* Step 2: clear flags for stateless addrconf */
2774         if (!how)
2775                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2776
2777 #ifdef CONFIG_IPV6_PRIVACY
2778         if (how && del_timer(&idev->regen_timer))
2779                 in6_dev_put(idev);
2780
2781         /* Step 3: clear tempaddr list */
2782         while (!list_empty(&idev->tempaddr_list)) {
2783                 ifa = list_first_entry(&idev->tempaddr_list,
2784                                        struct inet6_ifaddr, tmp_list);
2785                 list_del(&ifa->tmp_list);
2786                 write_unlock_bh(&idev->lock);
2787                 spin_lock_bh(&ifa->lock);
2788
2789                 if (ifa->ifpub) {
2790                         in6_ifa_put(ifa->ifpub);
2791                         ifa->ifpub = NULL;
2792                 }
2793                 spin_unlock_bh(&ifa->lock);
2794                 in6_ifa_put(ifa);
2795                 write_lock_bh(&idev->lock);
2796         }
2797 #endif
2798
2799         while (!list_empty(&idev->addr_list)) {
2800                 ifa = list_first_entry(&idev->addr_list,
2801                                        struct inet6_ifaddr, if_list);
2802                 addrconf_del_timer(ifa);
2803
2804                 list_del(&ifa->if_list);
2805
2806                 write_unlock_bh(&idev->lock);
2807
2808                 spin_lock_bh(&ifa->state_lock);
2809                 state = ifa->state;
2810                 ifa->state = INET6_IFADDR_STATE_DEAD;
2811                 spin_unlock_bh(&ifa->state_lock);
2812
2813                 if (state != INET6_IFADDR_STATE_DEAD) {
2814                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2815                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2816                 }
2817                 in6_ifa_put(ifa);
2818
2819                 write_lock_bh(&idev->lock);
2820         }
2821
2822         write_unlock_bh(&idev->lock);
2823
2824         /* Step 5: Discard multicast list */
2825         if (how)
2826                 ipv6_mc_destroy_dev(idev);
2827         else
2828                 ipv6_mc_down(idev);
2829
2830         idev->tstamp = jiffies;
2831
2832         /* Last: Shot the device (if unregistered) */
2833         if (how) {
2834                 addrconf_sysctl_unregister(idev);
2835                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2836                 neigh_ifdown(&nd_tbl, dev);
2837                 in6_dev_put(idev);
2838         }
2839         return 0;
2840 }
2841
2842 static void addrconf_rs_timer(unsigned long data)
2843 {
2844         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2845         struct inet6_dev *idev = ifp->idev;
2846
2847         read_lock(&idev->lock);
2848         if (idev->dead || !(idev->if_flags & IF_READY))
2849                 goto out;
2850
2851         if (idev->cnf.forwarding)
2852                 goto out;
2853
2854         /* Announcement received after solicitation was sent */
2855         if (idev->if_flags & IF_RA_RCVD)
2856                 goto out;
2857
2858         spin_lock(&ifp->lock);
2859         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2860                 /* The wait after the last probe can be shorter */
2861                 addrconf_mod_timer(ifp, AC_RS,
2862                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2863                                    idev->cnf.rtr_solicit_delay :
2864                                    idev->cnf.rtr_solicit_interval);
2865                 spin_unlock(&ifp->lock);
2866
2867                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2868         } else {
2869                 spin_unlock(&ifp->lock);
2870                 /*
2871                  * Note: we do not support deprecated "all on-link"
2872                  * assumption any longer.
2873                  */
2874                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
2875         }
2876
2877 out:
2878         read_unlock(&idev->lock);
2879         in6_ifa_put(ifp);
2880 }
2881
2882 /*
2883  *      Duplicate Address Detection
2884  */
2885 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2886 {
2887         unsigned long rand_num;
2888         struct inet6_dev *idev = ifp->idev;
2889
2890         if (ifp->flags & IFA_F_OPTIMISTIC)
2891                 rand_num = 0;
2892         else
2893                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2894
2895         ifp->probes = idev->cnf.dad_transmits;
2896         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2897 }
2898
2899 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
2900 {
2901         struct inet6_dev *idev = ifp->idev;
2902         struct net_device *dev = idev->dev;
2903
2904         addrconf_join_solict(dev, &ifp->addr);
2905
2906         net_srandom(ifp->addr.s6_addr32[3]);
2907
2908         read_lock_bh(&idev->lock);
2909         spin_lock(&ifp->lock);
2910         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2911                 goto out;
2912
2913         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2914             idev->cnf.accept_dad < 1 ||
2915             !(ifp->flags&IFA_F_TENTATIVE) ||
2916             ifp->flags & IFA_F_NODAD) {
2917                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2918                 spin_unlock(&ifp->lock);
2919                 read_unlock_bh(&idev->lock);
2920
2921                 addrconf_dad_completed(ifp);
2922                 return;
2923         }
2924
2925         if (!(idev->if_flags & IF_READY)) {
2926                 spin_unlock(&ifp->lock);
2927                 read_unlock_bh(&idev->lock);
2928                 /*
2929                  * If the device is not ready:
2930                  * - keep it tentative if it is a permanent address.
2931                  * - otherwise, kill it.
2932                  */
2933                 in6_ifa_hold(ifp);
2934                 addrconf_dad_stop(ifp, 0);
2935                 return;
2936         }
2937
2938         /*
2939          * Optimistic nodes can start receiving
2940          * Frames right away
2941          */
2942         if (ifp->flags & IFA_F_OPTIMISTIC)
2943                 ip6_ins_rt(ifp->rt);
2944
2945         addrconf_dad_kick(ifp);
2946 out:
2947         spin_unlock(&ifp->lock);
2948         read_unlock_bh(&idev->lock);
2949 }
2950
2951 static void addrconf_dad_timer(unsigned long data)
2952 {
2953         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2954         struct inet6_dev *idev = ifp->idev;
2955         struct in6_addr mcaddr;
2956
2957         if (!ifp->probes && addrconf_dad_end(ifp))
2958                 goto out;
2959
2960         read_lock(&idev->lock);
2961         if (idev->dead || !(idev->if_flags & IF_READY)) {
2962                 read_unlock(&idev->lock);
2963                 goto out;
2964         }
2965
2966         spin_lock(&ifp->lock);
2967         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2968                 spin_unlock(&ifp->lock);
2969                 read_unlock(&idev->lock);
2970                 goto out;
2971         }
2972
2973         if (ifp->probes == 0) {
2974                 /*
2975                  * DAD was successful
2976                  */
2977
2978                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2979                 spin_unlock(&ifp->lock);
2980                 read_unlock(&idev->lock);
2981
2982                 addrconf_dad_completed(ifp);
2983
2984                 goto out;
2985         }
2986
2987         ifp->probes--;
2988         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2989         spin_unlock(&ifp->lock);
2990         read_unlock(&idev->lock);
2991
2992         /* send a neighbour solicitation for our addr */
2993         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2994         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2995 out:
2996         in6_ifa_put(ifp);
2997 }
2998
2999 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3000 {
3001         struct net_device *dev = ifp->idev->dev;
3002
3003         /*
3004          *      Configure the address for reception. Now it is valid.
3005          */
3006
3007         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3008
3009         /* If added prefix is link local and we are prepared to process
3010            router advertisements, start sending router solicitations.
3011          */
3012
3013         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3014              ifp->idev->cnf.accept_ra == 2) &&
3015             ifp->idev->cnf.rtr_solicits > 0 &&
3016             (dev->flags&IFF_LOOPBACK) == 0 &&
3017             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3018                 /*
3019                  *      If a host as already performed a random delay
3020                  *      [...] as part of DAD [...] there is no need
3021                  *      to delay again before sending the first RS
3022                  */
3023                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3024
3025                 spin_lock_bh(&ifp->lock);
3026                 ifp->probes = 1;
3027                 ifp->idev->if_flags |= IF_RS_SENT;
3028                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3029                 spin_unlock_bh(&ifp->lock);
3030         }
3031 }
3032
3033 static void addrconf_dad_run(struct inet6_dev *idev)
3034 {
3035         struct inet6_ifaddr *ifp;
3036
3037         read_lock_bh(&idev->lock);
3038         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3039                 spin_lock(&ifp->lock);
3040                 if (ifp->flags & IFA_F_TENTATIVE &&
3041                     ifp->state == INET6_IFADDR_STATE_DAD)
3042                         addrconf_dad_kick(ifp);
3043                 spin_unlock(&ifp->lock);
3044         }
3045         read_unlock_bh(&idev->lock);
3046 }
3047
3048 #ifdef CONFIG_PROC_FS
3049 struct if6_iter_state {
3050         struct seq_net_private p;
3051         int bucket;
3052         int offset;
3053 };
3054
3055 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3056 {
3057         struct inet6_ifaddr *ifa = NULL;
3058         struct if6_iter_state *state = seq->private;
3059         struct net *net = seq_file_net(seq);
3060         int p = 0;
3061
3062         /* initial bucket if pos is 0 */
3063         if (pos == 0) {
3064                 state->bucket = 0;
3065                 state->offset = 0;
3066         }
3067
3068         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3069                 struct hlist_node *n;
3070                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3071                                          addr_lst) {
3072                         /* sync with offset */
3073                         if (p < state->offset) {
3074                                 p++;
3075                                 continue;
3076                         }
3077                         state->offset++;
3078                         if (net_eq(dev_net(ifa->idev->dev), net))
3079                                 return ifa;
3080                 }
3081
3082                 /* prepare for next bucket */
3083                 state->offset = 0;
3084                 p = 0;
3085         }
3086         return NULL;
3087 }
3088
3089 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3090                                          struct inet6_ifaddr *ifa)
3091 {
3092         struct if6_iter_state *state = seq->private;
3093         struct net *net = seq_file_net(seq);
3094         struct hlist_node *n = &ifa->addr_lst;
3095
3096         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3097                 state->offset++;
3098                 if (net_eq(dev_net(ifa->idev->dev), net))
3099                         return ifa;
3100         }
3101
3102         while (++state->bucket < IN6_ADDR_HSIZE) {
3103                 state->offset = 0;
3104                 hlist_for_each_entry_rcu_bh(ifa, n,
3105                                      &inet6_addr_lst[state->bucket], addr_lst) {
3106                         state->offset++;
3107                         if (net_eq(dev_net(ifa->idev->dev), net))
3108                                 return ifa;
3109                 }
3110         }
3111
3112         return NULL;
3113 }
3114
3115 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3116         __acquires(rcu_bh)
3117 {
3118         rcu_read_lock_bh();
3119         return if6_get_first(seq, *pos);
3120 }
3121
3122 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3123 {
3124         struct inet6_ifaddr *ifa;
3125
3126         ifa = if6_get_next(seq, v);
3127         ++*pos;
3128         return ifa;
3129 }
3130
3131 static void if6_seq_stop(struct seq_file *seq, void *v)
3132         __releases(rcu_bh)
3133 {
3134         rcu_read_unlock_bh();
3135 }
3136
3137 static int if6_seq_show(struct seq_file *seq, void *v)
3138 {
3139         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3140         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3141                    &ifp->addr,
3142                    ifp->idev->dev->ifindex,
3143                    ifp->prefix_len,
3144                    ifp->scope,
3145                    ifp->flags,
3146                    ifp->idev->dev->name);
3147         return 0;
3148 }
3149
3150 static const struct seq_operations if6_seq_ops = {
3151         .start  = if6_seq_start,
3152         .next   = if6_seq_next,
3153         .show   = if6_seq_show,
3154         .stop   = if6_seq_stop,
3155 };
3156
3157 static int if6_seq_open(struct inode *inode, struct file *file)
3158 {
3159         return seq_open_net(inode, file, &if6_seq_ops,
3160                             sizeof(struct if6_iter_state));
3161 }
3162
3163 static const struct file_operations if6_fops = {
3164         .owner          = THIS_MODULE,
3165         .open           = if6_seq_open,
3166         .read           = seq_read,
3167         .llseek         = seq_lseek,
3168         .release        = seq_release_net,
3169 };
3170
3171 static int __net_init if6_proc_net_init(struct net *net)
3172 {
3173         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3174                 return -ENOMEM;
3175         return 0;
3176 }
3177
3178 static void __net_exit if6_proc_net_exit(struct net *net)
3179 {
3180        proc_net_remove(net, "if_inet6");
3181 }
3182
3183 static struct pernet_operations if6_proc_net_ops = {
3184        .init = if6_proc_net_init,
3185        .exit = if6_proc_net_exit,
3186 };
3187
3188 int __init if6_proc_init(void)
3189 {
3190         return register_pernet_subsys(&if6_proc_net_ops);
3191 }
3192
3193 void if6_proc_exit(void)
3194 {
3195         unregister_pernet_subsys(&if6_proc_net_ops);
3196 }
3197 #endif  /* CONFIG_PROC_FS */
3198
3199 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3200 /* Check if address is a home address configured on any interface. */
3201 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3202 {
3203         int ret = 0;
3204         struct inet6_ifaddr *ifp = NULL;
3205         struct hlist_node *n;
3206         unsigned int hash = inet6_addr_hash(addr);
3207
3208         rcu_read_lock_bh();
3209         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3210                 if (!net_eq(dev_net(ifp->idev->dev), net))
3211                         continue;
3212                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3213                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3214                         ret = 1;
3215                         break;
3216                 }
3217         }
3218         rcu_read_unlock_bh();
3219         return ret;
3220 }
3221 #endif
3222
3223 /*
3224  *      Periodic address status verification
3225  */
3226
3227 static void addrconf_verify(unsigned long foo)
3228 {
3229         unsigned long now, next, next_sec, next_sched;
3230         struct inet6_ifaddr *ifp;
3231         struct hlist_node *node;
3232         int i;
3233
3234         rcu_read_lock_bh();
3235         spin_lock(&addrconf_verify_lock);
3236         now = jiffies;
3237         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3238
3239         del_timer(&addr_chk_timer);
3240
3241         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3242 restart:
3243                 hlist_for_each_entry_rcu_bh(ifp, node,
3244                                          &inet6_addr_lst[i], addr_lst) {
3245                         unsigned long age;
3246
3247                         if (ifp->flags & IFA_F_PERMANENT)
3248                                 continue;
3249
3250                         spin_lock(&ifp->lock);
3251                         /* We try to batch several events at once. */
3252                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3253
3254                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3255                             age >= ifp->valid_lft) {
3256                                 spin_unlock(&ifp->lock);
3257                                 in6_ifa_hold(ifp);
3258                                 ipv6_del_addr(ifp);
3259                                 goto restart;
3260                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3261                                 spin_unlock(&ifp->lock);
3262                                 continue;
3263                         } else if (age >= ifp->prefered_lft) {
3264                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3265                                 int deprecate = 0;
3266
3267                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3268                                         deprecate = 1;
3269                                         ifp->flags |= IFA_F_DEPRECATED;
3270                                 }
3271
3272                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3273                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3274
3275                                 spin_unlock(&ifp->lock);
3276
3277                                 if (deprecate) {
3278                                         in6_ifa_hold(ifp);
3279
3280                                         ipv6_ifa_notify(0, ifp);
3281                                         in6_ifa_put(ifp);
3282                                         goto restart;
3283                                 }
3284 #ifdef CONFIG_IPV6_PRIVACY
3285                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3286                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3287                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3288                                         ifp->idev->cnf.dad_transmits *
3289                                         ifp->idev->nd_parms->retrans_time / HZ;
3290
3291                                 if (age >= ifp->prefered_lft - regen_advance) {
3292                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3293                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3294                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3295                                         if (!ifp->regen_count && ifpub) {
3296                                                 ifp->regen_count++;
3297                                                 in6_ifa_hold(ifp);
3298                                                 in6_ifa_hold(ifpub);
3299                                                 spin_unlock(&ifp->lock);
3300
3301                                                 spin_lock(&ifpub->lock);
3302                                                 ifpub->regen_count = 0;
3303                                                 spin_unlock(&ifpub->lock);
3304                                                 ipv6_create_tempaddr(ifpub, ifp);
3305                                                 in6_ifa_put(ifpub);
3306                                                 in6_ifa_put(ifp);
3307                                                 goto restart;
3308                                         }
3309                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3310                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3311                                 spin_unlock(&ifp->lock);
3312 #endif
3313                         } else {
3314                                 /* ifp->prefered_lft <= ifp->valid_lft */
3315                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3316                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3317                                 spin_unlock(&ifp->lock);
3318                         }
3319                 }
3320         }
3321
3322         next_sec = round_jiffies_up(next);
3323         next_sched = next;
3324
3325         /* If rounded timeout is accurate enough, accept it. */
3326         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3327                 next_sched = next_sec;
3328
3329         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3330         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3331                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3332
3333         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3334               now, next, next_sec, next_sched));
3335
3336         addr_chk_timer.expires = next_sched;
3337         add_timer(&addr_chk_timer);
3338         spin_unlock(&addrconf_verify_lock);
3339         rcu_read_unlock_bh();
3340 }
3341
3342 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3343 {
3344         struct in6_addr *pfx = NULL;
3345
3346         if (addr)
3347                 pfx = nla_data(addr);
3348
3349         if (local) {
3350                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3351                         pfx = NULL;
3352                 else
3353                         pfx = nla_data(local);
3354         }
3355
3356         return pfx;
3357 }
3358
3359 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3360         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3361         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3362         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3363 };
3364
3365 static int
3366 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3367 {
3368         struct net *net = sock_net(skb->sk);
3369         struct ifaddrmsg *ifm;
3370         struct nlattr *tb[IFA_MAX+1];
3371         struct in6_addr *pfx;
3372         int err;
3373
3374         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3375         if (err < 0)
3376                 return err;
3377
3378         ifm = nlmsg_data(nlh);
3379         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3380         if (pfx == NULL)
3381                 return -EINVAL;
3382
3383         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3384 }
3385
3386 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3387                              u32 prefered_lft, u32 valid_lft)
3388 {
3389         u32 flags;
3390         clock_t expires;
3391         unsigned long timeout;
3392
3393         if (!valid_lft || (prefered_lft > valid_lft))
3394                 return -EINVAL;
3395
3396         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3397         if (addrconf_finite_timeout(timeout)) {
3398                 expires = jiffies_to_clock_t(timeout * HZ);
3399                 valid_lft = timeout;
3400                 flags = RTF_EXPIRES;
3401         } else {
3402                 expires = 0;
3403                 flags = 0;
3404                 ifa_flags |= IFA_F_PERMANENT;
3405         }
3406
3407         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3408         if (addrconf_finite_timeout(timeout)) {
3409                 if (timeout == 0)
3410                         ifa_flags |= IFA_F_DEPRECATED;
3411                 prefered_lft = timeout;
3412         }
3413
3414         spin_lock_bh(&ifp->lock);
3415         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3416         ifp->tstamp = jiffies;
3417         ifp->valid_lft = valid_lft;
3418         ifp->prefered_lft = prefered_lft;
3419
3420         spin_unlock_bh(&ifp->lock);
3421         if (!(ifp->flags&IFA_F_TENTATIVE))
3422                 ipv6_ifa_notify(0, ifp);
3423
3424         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3425                               expires, flags);
3426         addrconf_verify(0);
3427
3428         return 0;
3429 }
3430
3431 static int
3432 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3433 {
3434         struct net *net = sock_net(skb->sk);
3435         struct ifaddrmsg *ifm;
3436         struct nlattr *tb[IFA_MAX+1];
3437         struct in6_addr *pfx;
3438         struct inet6_ifaddr *ifa;
3439         struct net_device *dev;
3440         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3441         u8 ifa_flags;
3442         int err;
3443
3444         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3445         if (err < 0)
3446                 return err;
3447
3448         ifm = nlmsg_data(nlh);
3449         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3450         if (pfx == NULL)
3451                 return -EINVAL;
3452
3453         if (tb[IFA_CACHEINFO]) {
3454                 struct ifa_cacheinfo *ci;
3455
3456                 ci = nla_data(tb[IFA_CACHEINFO]);
3457                 valid_lft = ci->ifa_valid;
3458                 preferred_lft = ci->ifa_prefered;
3459         } else {
3460                 preferred_lft = INFINITY_LIFE_TIME;
3461                 valid_lft = INFINITY_LIFE_TIME;
3462         }
3463
3464         dev =  __dev_get_by_index(net, ifm->ifa_index);
3465         if (dev == NULL)
3466                 return -ENODEV;
3467
3468         /* We ignore other flags so far. */
3469         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3470
3471         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3472         if (ifa == NULL) {
3473                 /*
3474                  * It would be best to check for !NLM_F_CREATE here but
3475                  * userspace alreay relies on not having to provide this.
3476                  */
3477                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3478                                       ifm->ifa_prefixlen, ifa_flags,
3479                                       preferred_lft, valid_lft);
3480         }
3481
3482         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3483             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3484                 err = -EEXIST;
3485         else
3486                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3487
3488         in6_ifa_put(ifa);
3489
3490         return err;
3491 }
3492
3493 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3494                           u8 scope, int ifindex)
3495 {
3496         struct ifaddrmsg *ifm;
3497
3498         ifm = nlmsg_data(nlh);
3499         ifm->ifa_family = AF_INET6;
3500         ifm->ifa_prefixlen = prefixlen;
3501         ifm->ifa_flags = flags;
3502         ifm->ifa_scope = scope;
3503         ifm->ifa_index = ifindex;
3504 }
3505
3506 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3507                          unsigned long tstamp, u32 preferred, u32 valid)
3508 {
3509         struct ifa_cacheinfo ci;
3510
3511         ci.cstamp = cstamp_delta(cstamp);
3512         ci.tstamp = cstamp_delta(tstamp);
3513         ci.ifa_prefered = preferred;
3514         ci.ifa_valid = valid;
3515
3516         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3517 }
3518
3519 static inline int rt_scope(int ifa_scope)
3520 {
3521         if (ifa_scope & IFA_HOST)
3522                 return RT_SCOPE_HOST;
3523         else if (ifa_scope & IFA_LINK)
3524                 return RT_SCOPE_LINK;
3525         else if (ifa_scope & IFA_SITE)
3526                 return RT_SCOPE_SITE;
3527         else
3528                 return RT_SCOPE_UNIVERSE;
3529 }
3530
3531 static inline int inet6_ifaddr_msgsize(void)
3532 {
3533         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3534                + nla_total_size(16) /* IFA_ADDRESS */
3535                + nla_total_size(sizeof(struct ifa_cacheinfo));
3536 }
3537
3538 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3539                              u32 portid, u32 seq, int event, unsigned int flags)
3540 {
3541         struct nlmsghdr  *nlh;
3542         u32 preferred, valid;
3543
3544         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3545         if (nlh == NULL)
3546                 return -EMSGSIZE;
3547
3548         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3549                       ifa->idev->dev->ifindex);
3550
3551         if (!(ifa->flags&IFA_F_PERMANENT)) {
3552                 preferred = ifa->prefered_lft;
3553                 valid = ifa->valid_lft;
3554                 if (preferred != INFINITY_LIFE_TIME) {
3555                         long tval = (jiffies - ifa->tstamp)/HZ;
3556                         if (preferred > tval)
3557                                 preferred -= tval;
3558                         else
3559                                 preferred = 0;
3560                         if (valid != INFINITY_LIFE_TIME) {
3561                                 if (valid > tval)
3562                                         valid -= tval;
3563                                 else
3564                                         valid = 0;
3565                         }
3566                 }
3567         } else {
3568                 preferred = INFINITY_LIFE_TIME;
3569                 valid = INFINITY_LIFE_TIME;
3570         }
3571
3572         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3573             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3574                 nlmsg_cancel(skb, nlh);
3575                 return -EMSGSIZE;
3576         }
3577
3578         return nlmsg_end(skb, nlh);
3579 }
3580
3581 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3582                                 u32 portid, u32 seq, int event, u16 flags)
3583 {
3584         struct nlmsghdr  *nlh;
3585         u8 scope = RT_SCOPE_UNIVERSE;
3586         int ifindex = ifmca->idev->dev->ifindex;
3587
3588         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3589                 scope = RT_SCOPE_SITE;
3590
3591         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3592         if (nlh == NULL)
3593                 return -EMSGSIZE;
3594
3595         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3596         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3597             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3598                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3599                 nlmsg_cancel(skb, nlh);
3600                 return -EMSGSIZE;
3601         }
3602
3603         return nlmsg_end(skb, nlh);
3604 }
3605
3606 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3607                                 u32 portid, u32 seq, int event, unsigned int flags)
3608 {
3609         struct nlmsghdr  *nlh;
3610         u8 scope = RT_SCOPE_UNIVERSE;
3611         int ifindex = ifaca->aca_idev->dev->ifindex;
3612
3613         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3614                 scope = RT_SCOPE_SITE;
3615
3616         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3617         if (nlh == NULL)
3618                 return -EMSGSIZE;
3619
3620         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3621         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3622             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3623                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3624                 nlmsg_cancel(skb, nlh);
3625                 return -EMSGSIZE;
3626         }
3627
3628         return nlmsg_end(skb, nlh);
3629 }
3630
3631 enum addr_type_t {
3632         UNICAST_ADDR,
3633         MULTICAST_ADDR,
3634         ANYCAST_ADDR,
3635 };
3636
3637 /* called with rcu_read_lock() */
3638 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3639                           struct netlink_callback *cb, enum addr_type_t type,
3640                           int s_ip_idx, int *p_ip_idx)
3641 {
3642         struct ifmcaddr6 *ifmca;
3643         struct ifacaddr6 *ifaca;
3644         int err = 1;
3645         int ip_idx = *p_ip_idx;
3646
3647         read_lock_bh(&idev->lock);
3648         switch (type) {
3649         case UNICAST_ADDR: {
3650                 struct inet6_ifaddr *ifa;
3651
3652                 /* unicast address incl. temp addr */
3653                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3654                         if (++ip_idx < s_ip_idx)
3655                                 continue;
3656                         err = inet6_fill_ifaddr(skb, ifa,
3657                                                 NETLINK_CB(cb->skb).portid,
3658                                                 cb->nlh->nlmsg_seq,
3659                                                 RTM_NEWADDR,
3660                                                 NLM_F_MULTI);
3661                         if (err <= 0)
3662                                 break;
3663                 }
3664                 break;
3665         }
3666         case MULTICAST_ADDR:
3667                 /* multicast address */
3668                 for (ifmca = idev->mc_list; ifmca;
3669                      ifmca = ifmca->next, ip_idx++) {
3670                         if (ip_idx < s_ip_idx)
3671                                 continue;
3672                         err = inet6_fill_ifmcaddr(skb, ifmca,
3673                                                   NETLINK_CB(cb->skb).portid,
3674                                                   cb->nlh->nlmsg_seq,
3675                                                   RTM_GETMULTICAST,
3676                                                   NLM_F_MULTI);
3677                         if (err <= 0)
3678                                 break;
3679                 }
3680                 break;
3681         case ANYCAST_ADDR:
3682                 /* anycast address */
3683                 for (ifaca = idev->ac_list; ifaca;
3684                      ifaca = ifaca->aca_next, ip_idx++) {
3685                         if (ip_idx < s_ip_idx)
3686                                 continue;
3687                         err = inet6_fill_ifacaddr(skb, ifaca,
3688                                                   NETLINK_CB(cb->skb).portid,
3689                                                   cb->nlh->nlmsg_seq,
3690                                                   RTM_GETANYCAST,
3691                                                   NLM_F_MULTI);
3692                         if (err <= 0)
3693                                 break;
3694                 }
3695                 break;
3696         default:
3697                 break;
3698         }
3699         read_unlock_bh(&idev->lock);
3700         *p_ip_idx = ip_idx;
3701         return err;
3702 }
3703
3704 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3705                            enum addr_type_t type)
3706 {
3707         struct net *net = sock_net(skb->sk);
3708         int h, s_h;
3709         int idx, ip_idx;
3710         int s_idx, s_ip_idx;
3711         struct net_device *dev;
3712         struct inet6_dev *idev;
3713         struct hlist_head *head;
3714         struct hlist_node *node;
3715
3716         s_h = cb->args[0];
3717         s_idx = idx = cb->args[1];
3718         s_ip_idx = ip_idx = cb->args[2];
3719
3720         rcu_read_lock();
3721         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3722                 idx = 0;
3723                 head = &net->dev_index_head[h];
3724                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3725                         if (idx < s_idx)
3726                                 goto cont;
3727                         if (h > s_h || idx > s_idx)
3728                                 s_ip_idx = 0;
3729                         ip_idx = 0;
3730                         idev = __in6_dev_get(dev);
3731                         if (!idev)
3732                                 goto cont;
3733
3734                         if (in6_dump_addrs(idev, skb, cb, type,
3735                                            s_ip_idx, &ip_idx) <= 0)
3736                                 goto done;
3737 cont:
3738                         idx++;
3739                 }
3740         }
3741 done:
3742         rcu_read_unlock();
3743         cb->args[0] = h;
3744         cb->args[1] = idx;
3745         cb->args[2] = ip_idx;
3746
3747         return skb->len;
3748 }
3749
3750 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3751 {
3752         enum addr_type_t type = UNICAST_ADDR;
3753
3754         return inet6_dump_addr(skb, cb, type);
3755 }
3756
3757 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3758 {
3759         enum addr_type_t type = MULTICAST_ADDR;
3760
3761         return inet6_dump_addr(skb, cb, type);
3762 }
3763
3764
3765 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3766 {
3767         enum addr_type_t type = ANYCAST_ADDR;
3768
3769         return inet6_dump_addr(skb, cb, type);
3770 }
3771
3772 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3773                              void *arg)
3774 {
3775         struct net *net = sock_net(in_skb->sk);
3776         struct ifaddrmsg *ifm;
3777         struct nlattr *tb[IFA_MAX+1];
3778         struct in6_addr *addr = NULL;
3779         struct net_device *dev = NULL;
3780         struct inet6_ifaddr *ifa;
3781         struct sk_buff *skb;
3782         int err;
3783
3784         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3785         if (err < 0)
3786                 goto errout;
3787
3788         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3789         if (addr == NULL) {
3790                 err = -EINVAL;
3791                 goto errout;
3792         }
3793
3794         ifm = nlmsg_data(nlh);
3795         if (ifm->ifa_index)
3796                 dev = __dev_get_by_index(net, ifm->ifa_index);
3797
3798         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3799         if (!ifa) {
3800                 err = -EADDRNOTAVAIL;
3801                 goto errout;
3802         }
3803
3804         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3805         if (!skb) {
3806                 err = -ENOBUFS;
3807                 goto errout_ifa;
3808         }
3809
3810         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3811                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3812         if (err < 0) {
3813                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3814                 WARN_ON(err == -EMSGSIZE);
3815                 kfree_skb(skb);
3816                 goto errout_ifa;
3817         }
3818         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3819 errout_ifa:
3820         in6_ifa_put(ifa);
3821 errout:
3822         return err;
3823 }
3824
3825 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3826 {
3827         struct sk_buff *skb;
3828         struct net *net = dev_net(ifa->idev->dev);
3829         int err = -ENOBUFS;
3830
3831         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3832         if (skb == NULL)
3833                 goto errout;
3834
3835         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3836         if (err < 0) {
3837                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3838                 WARN_ON(err == -EMSGSIZE);
3839                 kfree_skb(skb);
3840                 goto errout;
3841         }
3842         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3843         return;
3844 errout:
3845         if (err < 0)
3846                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3847 }
3848
3849 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3850                                 __s32 *array, int bytes)
3851 {
3852         BUG_ON(bytes < (DEVCONF_MAX * 4));
3853
3854         memset(array, 0, bytes);
3855         array[DEVCONF_FORWARDING] = cnf->forwarding;
3856         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3857         array[DEVCONF_MTU6] = cnf->mtu6;
3858         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3859         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3860         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3861         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3862         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3863         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3864                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3865         array[DEVCONF_RTR_SOLICIT_DELAY] =
3866                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3867         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3868 #ifdef CONFIG_IPV6_PRIVACY
3869         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3870         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3871         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3872         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3873         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3874 #endif
3875         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3876         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3877         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3878 #ifdef CONFIG_IPV6_ROUTER_PREF
3879         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3880         array[DEVCONF_RTR_PROBE_INTERVAL] =
3881                 jiffies_to_msecs(cnf->rtr_probe_interval);
3882 #ifdef CONFIG_IPV6_ROUTE_INFO
3883         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3884 #endif
3885 #endif
3886         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3887         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3888 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3889         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3890 #endif
3891 #ifdef CONFIG_IPV6_MROUTE
3892         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3893 #endif
3894         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3895         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3896         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3897 }
3898
3899 static inline size_t inet6_ifla6_size(void)
3900 {
3901         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3902              + nla_total_size(sizeof(struct ifla_cacheinfo))
3903              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3904              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3905              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3906 }
3907
3908 static inline size_t inet6_if_nlmsg_size(void)
3909 {
3910         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3911                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3912                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3913                + nla_total_size(4) /* IFLA_MTU */
3914                + nla_total_size(4) /* IFLA_LINK */
3915                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3916 }
3917
3918 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3919                                       int items, int bytes)
3920 {
3921         int i;
3922         int pad = bytes - sizeof(u64) * items;
3923         BUG_ON(pad < 0);
3924
3925         /* Use put_unaligned() because stats may not be aligned for u64. */
3926         put_unaligned(items, &stats[0]);
3927         for (i = 1; i < items; i++)
3928                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3929
3930         memset(&stats[items], 0, pad);
3931 }
3932
3933 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3934                                       int items, int bytes, size_t syncpoff)
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(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3944
3945         memset(&stats[items], 0, pad);
3946 }
3947
3948 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3949                              int bytes)
3950 {
3951         switch (attrtype) {
3952         case IFLA_INET6_STATS:
3953                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3954                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3955                 break;
3956         case IFLA_INET6_ICMP6STATS:
3957                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3958                 break;
3959         }
3960 }
3961
3962 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3963 {
3964         struct nlattr *nla;
3965         struct ifla_cacheinfo ci;
3966
3967         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
3968                 goto nla_put_failure;
3969         ci.max_reasm_len = IPV6_MAXPLEN;
3970         ci.tstamp = cstamp_delta(idev->tstamp);
3971         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3972         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3973         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
3974                 goto nla_put_failure;
3975         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3976         if (nla == NULL)
3977                 goto nla_put_failure;
3978         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3979
3980         /* XXX - MC not implemented */
3981
3982         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3983         if (nla == NULL)
3984                 goto nla_put_failure;
3985         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3986
3987         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3988         if (nla == NULL)
3989                 goto nla_put_failure;
3990         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3991
3992         return 0;
3993
3994 nla_put_failure:
3995         return -EMSGSIZE;
3996 }
3997
3998 static size_t inet6_get_link_af_size(const struct net_device *dev)
3999 {
4000         if (!__in6_dev_get(dev))
4001                 return 0;
4002
4003         return inet6_ifla6_size();
4004 }
4005
4006 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4007 {
4008         struct inet6_dev *idev = __in6_dev_get(dev);
4009
4010         if (!idev)
4011                 return -ENODATA;
4012
4013         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4014                 return -EMSGSIZE;
4015
4016         return 0;
4017 }
4018
4019 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4020                              u32 portid, u32 seq, int event, unsigned int flags)
4021 {
4022         struct net_device *dev = idev->dev;
4023         struct ifinfomsg *hdr;
4024         struct nlmsghdr *nlh;
4025         void *protoinfo;
4026
4027         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4028         if (nlh == NULL)
4029                 return -EMSGSIZE;
4030
4031         hdr = nlmsg_data(nlh);
4032         hdr->ifi_family = AF_INET6;
4033         hdr->__ifi_pad = 0;
4034         hdr->ifi_type = dev->type;
4035         hdr->ifi_index = dev->ifindex;
4036         hdr->ifi_flags = dev_get_flags(dev);
4037         hdr->ifi_change = 0;
4038
4039         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4040             (dev->addr_len &&
4041              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4042             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4043             (dev->ifindex != dev->iflink &&
4044              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4045                 goto nla_put_failure;
4046         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4047         if (protoinfo == NULL)
4048                 goto nla_put_failure;
4049
4050         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4051                 goto nla_put_failure;
4052
4053         nla_nest_end(skb, protoinfo);
4054         return nlmsg_end(skb, nlh);
4055
4056 nla_put_failure:
4057         nlmsg_cancel(skb, nlh);
4058         return -EMSGSIZE;
4059 }
4060
4061 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4062 {
4063         struct net *net = sock_net(skb->sk);
4064         int h, s_h;
4065         int idx = 0, s_idx;
4066         struct net_device *dev;
4067         struct inet6_dev *idev;
4068         struct hlist_head *head;
4069         struct hlist_node *node;
4070
4071         s_h = cb->args[0];
4072         s_idx = cb->args[1];
4073
4074         rcu_read_lock();
4075         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4076                 idx = 0;
4077                 head = &net->dev_index_head[h];
4078                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4079                         if (idx < s_idx)
4080                                 goto cont;
4081                         idev = __in6_dev_get(dev);
4082                         if (!idev)
4083                                 goto cont;
4084                         if (inet6_fill_ifinfo(skb, idev,
4085                                               NETLINK_CB(cb->skb).portid,
4086                                               cb->nlh->nlmsg_seq,
4087                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4088                                 goto out;
4089 cont:
4090                         idx++;
4091                 }
4092         }
4093 out:
4094         rcu_read_unlock();
4095         cb->args[1] = idx;
4096         cb->args[0] = h;
4097
4098         return skb->len;
4099 }
4100
4101 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4102 {
4103         struct sk_buff *skb;
4104         struct net *net = dev_net(idev->dev);
4105         int err = -ENOBUFS;
4106
4107         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4108         if (skb == NULL)
4109                 goto errout;
4110
4111         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4112         if (err < 0) {
4113                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4114                 WARN_ON(err == -EMSGSIZE);
4115                 kfree_skb(skb);
4116                 goto errout;
4117         }
4118         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4119         return;
4120 errout:
4121         if (err < 0)
4122                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4123 }
4124
4125 static inline size_t inet6_prefix_nlmsg_size(void)
4126 {
4127         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4128                + nla_total_size(sizeof(struct in6_addr))
4129                + nla_total_size(sizeof(struct prefix_cacheinfo));
4130 }
4131
4132 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4133                              struct prefix_info *pinfo, u32 portid, u32 seq,
4134                              int event, unsigned int flags)
4135 {
4136         struct prefixmsg *pmsg;
4137         struct nlmsghdr *nlh;
4138         struct prefix_cacheinfo ci;
4139
4140         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4141         if (nlh == NULL)
4142                 return -EMSGSIZE;
4143
4144         pmsg = nlmsg_data(nlh);
4145         pmsg->prefix_family = AF_INET6;
4146         pmsg->prefix_pad1 = 0;
4147         pmsg->prefix_pad2 = 0;
4148         pmsg->prefix_ifindex = idev->dev->ifindex;
4149         pmsg->prefix_len = pinfo->prefix_len;
4150         pmsg->prefix_type = pinfo->type;
4151         pmsg->prefix_pad3 = 0;
4152         pmsg->prefix_flags = 0;
4153         if (pinfo->onlink)
4154                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4155         if (pinfo->autoconf)
4156                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4157
4158         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4159                 goto nla_put_failure;
4160         ci.preferred_time = ntohl(pinfo->prefered);
4161         ci.valid_time = ntohl(pinfo->valid);
4162         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4163                 goto nla_put_failure;
4164         return nlmsg_end(skb, nlh);
4165
4166 nla_put_failure:
4167         nlmsg_cancel(skb, nlh);
4168         return -EMSGSIZE;
4169 }
4170
4171 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4172                          struct prefix_info *pinfo)
4173 {
4174         struct sk_buff *skb;
4175         struct net *net = dev_net(idev->dev);
4176         int err = -ENOBUFS;
4177
4178         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4179         if (skb == NULL)
4180                 goto errout;
4181
4182         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4183         if (err < 0) {
4184                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4185                 WARN_ON(err == -EMSGSIZE);
4186                 kfree_skb(skb);
4187                 goto errout;
4188         }
4189         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4190         return;
4191 errout:
4192         if (err < 0)
4193                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4194 }
4195
4196 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4197 {
4198         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4199
4200         switch (event) {
4201         case RTM_NEWADDR:
4202                 /*
4203                  * If the address was optimistic
4204                  * we inserted the route at the start of
4205                  * our DAD process, so we don't need
4206                  * to do it again
4207                  */
4208                 if (!(ifp->rt->rt6i_node))
4209                         ip6_ins_rt(ifp->rt);
4210                 if (ifp->idev->cnf.forwarding)
4211                         addrconf_join_anycast(ifp);
4212                 break;
4213         case RTM_DELADDR:
4214                 if (ifp->idev->cnf.forwarding)
4215                         addrconf_leave_anycast(ifp);
4216                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4217                 dst_hold(&ifp->rt->dst);
4218
4219                 if (ip6_del_rt(ifp->rt))
4220                         dst_free(&ifp->rt->dst);
4221                 break;
4222         }
4223 }
4224
4225 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4226 {
4227         rcu_read_lock_bh();
4228         if (likely(ifp->idev->dead == 0))
4229                 __ipv6_ifa_notify(event, ifp);
4230         rcu_read_unlock_bh();
4231 }
4232
4233 #ifdef CONFIG_SYSCTL
4234
4235 static
4236 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4237                            void __user *buffer, size_t *lenp, loff_t *ppos)
4238 {
4239         int *valp = ctl->data;
4240         int val = *valp;
4241         loff_t pos = *ppos;
4242         ctl_table lctl;
4243         int ret;
4244
4245         /*
4246          * ctl->data points to idev->cnf.forwarding, we should
4247          * not modify it until we get the rtnl lock.
4248          */
4249         lctl = *ctl;
4250         lctl.data = &val;
4251
4252         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4253
4254         if (write)
4255                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4256         if (ret)
4257                 *ppos = pos;
4258         return ret;
4259 }
4260
4261 static void dev_disable_change(struct inet6_dev *idev)
4262 {
4263         if (!idev || !idev->dev)
4264                 return;
4265
4266         if (idev->cnf.disable_ipv6)
4267                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4268         else
4269                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4270 }
4271
4272 static void addrconf_disable_change(struct net *net, __s32 newf)
4273 {
4274         struct net_device *dev;
4275         struct inet6_dev *idev;
4276
4277         rcu_read_lock();
4278         for_each_netdev_rcu(net, dev) {
4279                 idev = __in6_dev_get(dev);
4280                 if (idev) {
4281                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4282                         idev->cnf.disable_ipv6 = newf;
4283                         if (changed)
4284                                 dev_disable_change(idev);
4285                 }
4286         }
4287         rcu_read_unlock();
4288 }
4289
4290 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4291 {
4292         struct net *net;
4293         int old;
4294
4295         if (!rtnl_trylock())
4296                 return restart_syscall();
4297
4298         net = (struct net *)table->extra2;
4299         old = *p;
4300         *p = newf;
4301
4302         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4303                 rtnl_unlock();
4304                 return 0;
4305         }
4306
4307         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4308                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4309                 addrconf_disable_change(net, newf);
4310         } else if ((!newf) ^ (!old))
4311                 dev_disable_change((struct inet6_dev *)table->extra1);
4312
4313         rtnl_unlock();
4314         return 0;
4315 }
4316
4317 static
4318 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4319                             void __user *buffer, size_t *lenp, loff_t *ppos)
4320 {
4321         int *valp = ctl->data;
4322         int val = *valp;
4323         loff_t pos = *ppos;
4324         ctl_table lctl;
4325         int ret;
4326
4327         /*
4328          * ctl->data points to idev->cnf.disable_ipv6, we should
4329          * not modify it until we get the rtnl lock.
4330          */
4331         lctl = *ctl;
4332         lctl.data = &val;
4333
4334         ret = proc_dointvec(&lctl, 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 } addrconf_sysctl __read_mostly = {
4348         .sysctl_header = NULL,
4349         .addrconf_vars = {
4350                 {
4351                         .procname       = "forwarding",
4352                         .data           = &ipv6_devconf.forwarding,
4353                         .maxlen         = sizeof(int),
4354                         .mode           = 0644,
4355                         .proc_handler   = addrconf_sysctl_forward,
4356                 },
4357                 {
4358                         .procname       = "hop_limit",
4359                         .data           = &ipv6_devconf.hop_limit,
4360                         .maxlen         = sizeof(int),
4361                         .mode           = 0644,
4362                         .proc_handler   = proc_dointvec,
4363                 },
4364                 {
4365                         .procname       = "mtu",
4366                         .data           = &ipv6_devconf.mtu6,
4367                         .maxlen         = sizeof(int),
4368                         .mode           = 0644,
4369                         .proc_handler   = proc_dointvec,
4370                 },
4371                 {
4372                         .procname       = "accept_ra",
4373                         .data           = &ipv6_devconf.accept_ra,
4374                         .maxlen         = sizeof(int),
4375                         .mode           = 0644,
4376                         .proc_handler   = proc_dointvec,
4377                 },
4378                 {
4379                         .procname       = "accept_redirects",
4380                         .data           = &ipv6_devconf.accept_redirects,
4381                         .maxlen         = sizeof(int),
4382                         .mode           = 0644,
4383                         .proc_handler   = proc_dointvec,
4384                 },
4385                 {
4386                         .procname       = "autoconf",
4387                         .data           = &ipv6_devconf.autoconf,
4388                         .maxlen         = sizeof(int),
4389                         .mode           = 0644,
4390                         .proc_handler   = proc_dointvec,
4391                 },
4392                 {
4393                         .procname       = "dad_transmits",
4394                         .data           = &ipv6_devconf.dad_transmits,
4395                         .maxlen         = sizeof(int),
4396                         .mode           = 0644,
4397                         .proc_handler   = proc_dointvec,
4398                 },
4399                 {
4400                         .procname       = "router_solicitations",
4401                         .data           = &ipv6_devconf.rtr_solicits,
4402                         .maxlen         = sizeof(int),
4403                         .mode           = 0644,
4404                         .proc_handler   = proc_dointvec,
4405                 },
4406                 {
4407                         .procname       = "router_solicitation_interval",
4408                         .data           = &ipv6_devconf.rtr_solicit_interval,
4409                         .maxlen         = sizeof(int),
4410                         .mode           = 0644,
4411                         .proc_handler   = proc_dointvec_jiffies,
4412                 },
4413                 {
4414                         .procname       = "router_solicitation_delay",
4415                         .data           = &ipv6_devconf.rtr_solicit_delay,
4416                         .maxlen         = sizeof(int),
4417                         .mode           = 0644,
4418                         .proc_handler   = proc_dointvec_jiffies,
4419                 },
4420                 {
4421                         .procname       = "force_mld_version",
4422                         .data           = &ipv6_devconf.force_mld_version,
4423                         .maxlen         = sizeof(int),
4424                         .mode           = 0644,
4425                         .proc_handler   = proc_dointvec,
4426                 },
4427 #ifdef CONFIG_IPV6_PRIVACY
4428                 {
4429                         .procname       = "use_tempaddr",
4430                         .data           = &ipv6_devconf.use_tempaddr,
4431                         .maxlen         = sizeof(int),
4432                         .mode           = 0644,
4433                         .proc_handler   = proc_dointvec,
4434                 },
4435                 {
4436                         .procname       = "temp_valid_lft",
4437                         .data           = &ipv6_devconf.temp_valid_lft,
4438                         .maxlen         = sizeof(int),
4439                         .mode           = 0644,
4440                         .proc_handler   = proc_dointvec,
4441                 },
4442                 {
4443                         .procname       = "temp_prefered_lft",
4444                         .data           = &ipv6_devconf.temp_prefered_lft,
4445                         .maxlen         = sizeof(int),
4446                         .mode           = 0644,
4447                         .proc_handler   = proc_dointvec,
4448                 },
4449                 {
4450                         .procname       = "regen_max_retry",
4451                         .data           = &ipv6_devconf.regen_max_retry,
4452                         .maxlen         = sizeof(int),
4453                         .mode           = 0644,
4454                         .proc_handler   = proc_dointvec,
4455                 },
4456                 {
4457                         .procname       = "max_desync_factor",
4458                         .data           = &ipv6_devconf.max_desync_factor,
4459                         .maxlen         = sizeof(int),
4460                         .mode           = 0644,
4461                         .proc_handler   = proc_dointvec,
4462                 },
4463 #endif
4464                 {
4465                         .procname       = "max_addresses",
4466                         .data           = &ipv6_devconf.max_addresses,
4467                         .maxlen         = sizeof(int),
4468                         .mode           = 0644,
4469                         .proc_handler   = proc_dointvec,
4470                 },
4471                 {
4472                         .procname       = "accept_ra_defrtr",
4473                         .data           = &ipv6_devconf.accept_ra_defrtr,
4474                         .maxlen         = sizeof(int),
4475                         .mode           = 0644,
4476                         .proc_handler   = proc_dointvec,
4477                 },
4478                 {
4479                         .procname       = "accept_ra_pinfo",
4480                         .data           = &ipv6_devconf.accept_ra_pinfo,
4481                         .maxlen         = sizeof(int),
4482                         .mode           = 0644,
4483                         .proc_handler   = proc_dointvec,
4484                 },
4485 #ifdef CONFIG_IPV6_ROUTER_PREF
4486                 {
4487                         .procname       = "accept_ra_rtr_pref",
4488                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4489                         .maxlen         = sizeof(int),
4490                         .mode           = 0644,
4491                         .proc_handler   = proc_dointvec,
4492                 },
4493                 {
4494                         .procname       = "router_probe_interval",
4495                         .data           = &ipv6_devconf.rtr_probe_interval,
4496                         .maxlen         = sizeof(int),
4497                         .mode           = 0644,
4498                         .proc_handler   = proc_dointvec_jiffies,
4499                 },
4500 #ifdef CONFIG_IPV6_ROUTE_INFO
4501                 {
4502                         .procname       = "accept_ra_rt_info_max_plen",
4503                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4504                         .maxlen         = sizeof(int),
4505                         .mode           = 0644,
4506                         .proc_handler   = proc_dointvec,
4507                 },
4508 #endif
4509 #endif
4510                 {
4511                         .procname       = "proxy_ndp",
4512                         .data           = &ipv6_devconf.proxy_ndp,
4513                         .maxlen         = sizeof(int),
4514                         .mode           = 0644,
4515                         .proc_handler   = proc_dointvec,
4516                 },
4517                 {
4518                         .procname       = "accept_source_route",
4519                         .data           = &ipv6_devconf.accept_source_route,
4520                         .maxlen         = sizeof(int),
4521                         .mode           = 0644,
4522                         .proc_handler   = proc_dointvec,
4523                 },
4524 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4525                 {
4526                         .procname       = "optimistic_dad",
4527                         .data           = &ipv6_devconf.optimistic_dad,
4528                         .maxlen         = sizeof(int),
4529                         .mode           = 0644,
4530                         .proc_handler   = proc_dointvec,
4531
4532                 },
4533 #endif
4534 #ifdef CONFIG_IPV6_MROUTE
4535                 {
4536                         .procname       = "mc_forwarding",
4537                         .data           = &ipv6_devconf.mc_forwarding,
4538                         .maxlen         = sizeof(int),
4539                         .mode           = 0444,
4540                         .proc_handler   = proc_dointvec,
4541                 },
4542 #endif
4543                 {
4544                         .procname       = "disable_ipv6",
4545                         .data           = &ipv6_devconf.disable_ipv6,
4546                         .maxlen         = sizeof(int),
4547                         .mode           = 0644,
4548                         .proc_handler   = addrconf_sysctl_disable,
4549                 },
4550                 {
4551                         .procname       = "accept_dad",
4552                         .data           = &ipv6_devconf.accept_dad,
4553                         .maxlen         = sizeof(int),
4554                         .mode           = 0644,
4555                         .proc_handler   = proc_dointvec,
4556                 },
4557                 {
4558                         .procname       = "force_tllao",
4559                         .data           = &ipv6_devconf.force_tllao,
4560                         .maxlen         = sizeof(int),
4561                         .mode           = 0644,
4562                         .proc_handler   = proc_dointvec
4563                 },
4564                 {
4565                         /* sentinel */
4566                 }
4567         },
4568 };
4569
4570 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4571                 struct inet6_dev *idev, struct ipv6_devconf *p)
4572 {
4573         int i;
4574         struct addrconf_sysctl_table *t;
4575         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4576
4577         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4578         if (t == NULL)
4579                 goto out;
4580
4581         for (i = 0; t->addrconf_vars[i].data; i++) {
4582                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4583                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4584                 t->addrconf_vars[i].extra2 = net;
4585         }
4586
4587         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4588
4589         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4590         if (t->sysctl_header == NULL)
4591                 goto free;
4592
4593         p->sysctl = t;
4594         return 0;
4595
4596 free:
4597         kfree(t);
4598 out:
4599         return -ENOBUFS;
4600 }
4601
4602 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4603 {
4604         struct addrconf_sysctl_table *t;
4605
4606         if (p->sysctl == NULL)
4607                 return;
4608
4609         t = p->sysctl;
4610         p->sysctl = NULL;
4611         unregister_net_sysctl_table(t->sysctl_header);
4612         kfree(t);
4613 }
4614
4615 static void addrconf_sysctl_register(struct inet6_dev *idev)
4616 {
4617         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4618                               &ndisc_ifinfo_sysctl_change);
4619         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4620                                         idev, &idev->cnf);
4621 }
4622
4623 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4624 {
4625         __addrconf_sysctl_unregister(&idev->cnf);
4626         neigh_sysctl_unregister(idev->nd_parms);
4627 }
4628
4629
4630 #endif
4631
4632 static int __net_init addrconf_init_net(struct net *net)
4633 {
4634         int err;
4635         struct ipv6_devconf *all, *dflt;
4636
4637         err = -ENOMEM;
4638         all = &ipv6_devconf;
4639         dflt = &ipv6_devconf_dflt;
4640
4641         if (!net_eq(net, &init_net)) {
4642                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4643                 if (all == NULL)
4644                         goto err_alloc_all;
4645
4646                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4647                 if (dflt == NULL)
4648                         goto err_alloc_dflt;
4649         } else {
4650                 /* these will be inherited by all namespaces */
4651                 dflt->autoconf = ipv6_defaults.autoconf;
4652                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4653         }
4654
4655         net->ipv6.devconf_all = all;
4656         net->ipv6.devconf_dflt = dflt;
4657
4658 #ifdef CONFIG_SYSCTL
4659         err = __addrconf_sysctl_register(net, "all", NULL, all);
4660         if (err < 0)
4661                 goto err_reg_all;
4662
4663         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4664         if (err < 0)
4665                 goto err_reg_dflt;
4666 #endif
4667         return 0;
4668
4669 #ifdef CONFIG_SYSCTL
4670 err_reg_dflt:
4671         __addrconf_sysctl_unregister(all);
4672 err_reg_all:
4673         kfree(dflt);
4674 #endif
4675 err_alloc_dflt:
4676         kfree(all);
4677 err_alloc_all:
4678         return err;
4679 }
4680
4681 static void __net_exit addrconf_exit_net(struct net *net)
4682 {
4683 #ifdef CONFIG_SYSCTL
4684         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4685         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4686 #endif
4687         if (!net_eq(net, &init_net)) {
4688                 kfree(net->ipv6.devconf_dflt);
4689                 kfree(net->ipv6.devconf_all);
4690         }
4691 }
4692
4693 static struct pernet_operations addrconf_ops = {
4694         .init = addrconf_init_net,
4695         .exit = addrconf_exit_net,
4696 };
4697
4698 /*
4699  *      Device notifier
4700  */
4701
4702 int register_inet6addr_notifier(struct notifier_block *nb)
4703 {
4704         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4705 }
4706 EXPORT_SYMBOL(register_inet6addr_notifier);
4707
4708 int unregister_inet6addr_notifier(struct notifier_block *nb)
4709 {
4710         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4711 }
4712 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4713
4714 static struct rtnl_af_ops inet6_ops = {
4715         .family           = AF_INET6,
4716         .fill_link_af     = inet6_fill_link_af,
4717         .get_link_af_size = inet6_get_link_af_size,
4718 };
4719
4720 /*
4721  *      Init / cleanup code
4722  */
4723
4724 int __init addrconf_init(void)
4725 {
4726         int i, err;
4727
4728         err = ipv6_addr_label_init();
4729         if (err < 0) {
4730                 pr_crit("%s: cannot initialize default policy table: %d\n",
4731                         __func__, err);
4732                 goto out;
4733         }
4734
4735         err = register_pernet_subsys(&addrconf_ops);
4736         if (err < 0)
4737                 goto out_addrlabel;
4738
4739         /* The addrconf netdev notifier requires that loopback_dev
4740          * has it's ipv6 private information allocated and setup
4741          * before it can bring up and give link-local addresses
4742          * to other devices which are up.
4743          *
4744          * Unfortunately, loopback_dev is not necessarily the first
4745          * entry in the global dev_base list of net devices.  In fact,
4746          * it is likely to be the very last entry on that list.
4747          * So this causes the notifier registry below to try and
4748          * give link-local addresses to all devices besides loopback_dev
4749          * first, then loopback_dev, which cases all the non-loopback_dev
4750          * devices to fail to get a link-local address.
4751          *
4752          * So, as a temporary fix, allocate the ipv6 structure for
4753          * loopback_dev first by hand.
4754          * Longer term, all of the dependencies ipv6 has upon the loopback
4755          * device and it being up should be removed.
4756          */
4757         rtnl_lock();
4758         if (!ipv6_add_dev(init_net.loopback_dev))
4759                 err = -ENOMEM;
4760         rtnl_unlock();
4761         if (err)
4762                 goto errlo;
4763
4764         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4765                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4766
4767         register_netdevice_notifier(&ipv6_dev_notf);
4768
4769         addrconf_verify(0);
4770
4771         err = rtnl_af_register(&inet6_ops);
4772         if (err < 0)
4773                 goto errout_af;
4774
4775         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4776                               NULL);
4777         if (err < 0)
4778                 goto errout;
4779
4780         /* Only the first call to __rtnl_register can fail */
4781         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4782         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4783         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4784                         inet6_dump_ifaddr, NULL);
4785         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4786                         inet6_dump_ifmcaddr, NULL);
4787         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4788                         inet6_dump_ifacaddr, NULL);
4789
4790         ipv6_addr_label_rtnl_register();
4791
4792         return 0;
4793 errout:
4794         rtnl_af_unregister(&inet6_ops);
4795 errout_af:
4796         unregister_netdevice_notifier(&ipv6_dev_notf);
4797 errlo:
4798         unregister_pernet_subsys(&addrconf_ops);
4799 out_addrlabel:
4800         ipv6_addr_label_cleanup();
4801 out:
4802         return err;
4803 }
4804
4805 void addrconf_cleanup(void)
4806 {
4807         struct net_device *dev;
4808         int i;
4809
4810         unregister_netdevice_notifier(&ipv6_dev_notf);
4811         unregister_pernet_subsys(&addrconf_ops);
4812         ipv6_addr_label_cleanup();
4813
4814         rtnl_lock();
4815
4816         __rtnl_af_unregister(&inet6_ops);
4817
4818         /* clean dev list */
4819         for_each_netdev(&init_net, dev) {
4820                 if (__in6_dev_get(dev) == NULL)
4821                         continue;
4822                 addrconf_ifdown(dev, 1);
4823         }
4824         addrconf_ifdown(init_net.loopback_dev, 2);
4825
4826         /*
4827          *      Check hash table.
4828          */
4829         spin_lock_bh(&addrconf_hash_lock);
4830         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4831                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4832         spin_unlock_bh(&addrconf_hash_lock);
4833
4834         del_timer(&addr_chk_timer);
4835         rtnl_unlock();
4836 }