net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90 #include <linux/export.h>
  91
  92 /* Set to 3 to get tracing... */
  93 #define ACONF_DEBUG 2
  94
  95 #if ACONF_DEBUG >= 3
  96 #define ADBG(x) printk x
  97 #else
  98 #define ADBG(x)
  99 #endif
 100
 101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 102
 103 static inline u32 cstamp_delta(unsigned long cstamp)
 104 {
 105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 106 }
 107
 108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 111
 112 #ifdef CONFIG_SYSCTL
 113 static void addrconf_sysctl_register(struct inet6_dev *idev);
 114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 115 #else
 116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 117 {
 118 }
 119
 120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 121 {
 122 }
 123 #endif
 124
 125 #ifdef CONFIG_IPV6_PRIVACY
 126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 128 static void ipv6_regen_rndid(unsigned long data);
 129 #endif
 130
 131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 132 static int ipv6_count_addresses(struct inet6_dev *idev);
 133
 134 /*
 135  *      Configured unicast address hash table
 136  */
 137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 138 static DEFINE_SPINLOCK(addrconf_hash_lock);
 139
 140 static void addrconf_verify(unsigned long);
 141
 142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 143 static DEFINE_SPINLOCK(addrconf_verify_lock);
 144
 145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 147
 148 static void addrconf_type_change(struct net_device *dev,
 149                                  unsigned long event);
 150 static int addrconf_ifdown(struct net_device *dev, int how);
 151
 152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 153 static void addrconf_dad_timer(unsigned long data);
 154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 155 static void addrconf_dad_run(struct inet6_dev *idev);
 156 static void addrconf_rs_timer(unsigned long data);
 157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159
 160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 161                                 struct prefix_info *pinfo);
 162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 163                                struct net_device *dev);
 164
 165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 166
 167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 168         .forwarding             = 0,
 169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 170         .mtu6                   = IPV6_MIN_MTU,
 171         .accept_ra              = 1,
 172         .accept_redirects       = 1,
 173         .autoconf               = 1,
 174         .force_mld_version      = 0,
 175         .dad_transmits          = 1,
 176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 179 #ifdef CONFIG_IPV6_PRIVACY
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185 #endif
 186         .max_addresses          = IPV6_MAX_ADDRESSES,
 187         .accept_ra_defrtr       = 1,
 188         .accept_ra_pinfo        = 1,
 189 #ifdef CONFIG_IPV6_ROUTER_PREF
 190         .accept_ra_rtr_pref     = 1,
 191         .rtr_probe_interval     = 60 * HZ,
 192 #ifdef CONFIG_IPV6_ROUTE_INFO
 193         .accept_ra_rt_info_max_plen = 0,
 194 #endif
 195 #endif
 196         .proxy_ndp              = 0,
 197         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 198         .disable_ipv6           = 0,
 199         .accept_dad             = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .dad_transmits          = 1,
 210         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 211         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 212         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 213 #ifdef CONFIG_IPV6_PRIVACY
 214         .use_tempaddr           = 0,
 215         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 216         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 217         .regen_max_retry        = REGEN_MAX_RETRY,
 218         .max_desync_factor      = MAX_DESYNC_FACTOR,
 219 #endif
 220         .max_addresses          = IPV6_MAX_ADDRESSES,
 221         .accept_ra_defrtr       = 1,
 222         .accept_ra_pinfo        = 1,
 223 #ifdef CONFIG_IPV6_ROUTER_PREF
 224         .accept_ra_rtr_pref     = 1,
 225         .rtr_probe_interval     = 60 * HZ,
 226 #ifdef CONFIG_IPV6_ROUTE_INFO
 227         .accept_ra_rt_info_max_plen = 0,
 228 #endif
 229 #endif
 230         .proxy_ndp              = 0,
 231         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 232         .disable_ipv6           = 0,
 233         .accept_dad             = 1,
 234 };
 235
 236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 241
 242 /* Check if a valid qdisc is available */
 243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 244 {
 245         return !qdisc_tx_is_noop(dev);
 246 }
 247
 248 /* Check if a route is valid prefix route */
 249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 250 {
 251         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 252 }
 253
 254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 255 {
 256         if (del_timer(&ifp->timer))
 257                 __in6_ifa_put(ifp);
 258 }
 259
 260 enum addrconf_timer_t {
 261         AC_NONE,
 262         AC_DAD,
 263         AC_RS,
 264 };
 265
 266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 267                                enum addrconf_timer_t what,
 268                                unsigned long when)
 269 {
 270         if (!del_timer(&ifp->timer))
 271                 in6_ifa_hold(ifp);
 272
 273         switch (what) {
 274         case AC_DAD:
 275                 ifp->timer.function = addrconf_dad_timer;
 276                 break;
 277         case AC_RS:
 278                 ifp->timer.function = addrconf_rs_timer;
 279                 break;
 280         default:
 281                 break;
 282         }
 283         ifp->timer.expires = jiffies + when;
 284         add_timer(&ifp->timer);
 285 }
 286
 287 static int snmp6_alloc_dev(struct inet6_dev *idev)
 288 {
 289         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 290                           sizeof(struct ipstats_mib),
 291                           __alignof__(struct ipstats_mib)) < 0)
 292                 goto err_ip;
 293         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 294                                         GFP_KERNEL);
 295         if (!idev->stats.icmpv6dev)
 296                 goto err_icmp;
 297         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 298                                            GFP_KERNEL);
 299         if (!idev->stats.icmpv6msgdev)
 300                 goto err_icmpmsg;
 301
 302         return 0;
 303
 304 err_icmpmsg:
 305         kfree(idev->stats.icmpv6dev);
 306 err_icmp:
 307         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 308 err_ip:
 309         return -ENOMEM;
 310 }
 311
 312 static void snmp6_free_dev(struct inet6_dev *idev)
 313 {
 314         kfree(idev->stats.icmpv6msgdev);
 315         kfree(idev->stats.icmpv6dev);
 316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 317 }
 318
 319 /* Nobody refers to this device, we may destroy it. */
 320
 321 void in6_dev_finish_destroy(struct inet6_dev *idev)
 322 {
 323         struct net_device *dev = idev->dev;
 324
 325         WARN_ON(!list_empty(&idev->addr_list));
 326         WARN_ON(idev->mc_list != NULL);
 327
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 pr_warning("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         kfree_rcu(idev, rcu);
 338 }
 339
 340 EXPORT_SYMBOL(in6_dev_finish_destroy);
 341
 342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 343 {
 344         struct inet6_dev *ndev;
 345
 346         ASSERT_RTNL();
 347
 348         if (dev->mtu < IPV6_MIN_MTU)
 349                 return NULL;
 350
 351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 352
 353         if (ndev == NULL)
 354                 return NULL;
 355
 356         rwlock_init(&ndev->lock);
 357         ndev->dev = dev;
 358         INIT_LIST_HEAD(&ndev->addr_list);
 359
 360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (ndev->nd_parms == NULL) {
 365                 kfree(ndev);
 366                 return NULL;
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 dev_put(dev);
 379                 kfree(ndev);
 380                 return NULL;
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG((KERN_WARNING
 385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name));
 387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 388                 ndev->dead = 1;
 389                 in6_dev_finish_destroy(ndev);
 390                 return NULL;
 391         }
 392
 393         /* One reference from device.  We must do this before
 394          * we invoke __ipv6_regen_rndid().
 395          */
 396         in6_dev_hold(ndev);
 397
 398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 399                 ndev->cnf.accept_dad = -1;
 400
 401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Multicast RS\n",
 405                        dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 ndev->cnf.use_tempaddr = -1;
 419         } else {
 420                 in6_dev_hold(ndev);
 421                 ipv6_regen_rndid((unsigned long) ndev);
 422         }
 423 #endif
 424
 425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 426                 ndev->if_flags |= IF_READY;
 427
 428         ipv6_mc_init_dev(ndev);
 429         ndev->tstamp = jiffies;
 430         addrconf_sysctl_register(ndev);
 431         /* protected by rtnl_lock */
 432         rcu_assign_pointer(dev->ip6_ptr, ndev);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         /* Join all-router multicast group if forwarding is set */
 438         if (ndev->cnf.forwarding && dev && (dev->flags & IFF_MULTICAST))
 439                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 #ifdef CONFIG_SYSCTL
 463 static void dev_forward_change(struct inet6_dev *idev)
 464 {
 465         struct net_device *dev;
 466         struct inet6_ifaddr *ifa;
 467
 468         if (!idev)
 469                 return;
 470         dev = idev->dev;
 471         if (idev->cnf.forwarding)
 472                 dev_disable_lro(dev);
 473         if (dev && (dev->flags & IFF_MULTICAST)) {
 474                 if (idev->cnf.forwarding)
 475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 476                 else
 477                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 478         }
 479
 480         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 481                 if (ifa->flags&IFA_F_TENTATIVE)
 482                         continue;
 483                 if (idev->cnf.forwarding)
 484                         addrconf_join_anycast(ifa);
 485                 else
 486                         addrconf_leave_anycast(ifa);
 487         }
 488 }
 489
 490
 491 static void addrconf_forward_change(struct net *net, __s32 newf)
 492 {
 493         struct net_device *dev;
 494         struct inet6_dev *idev;
 495
 496         for_each_netdev(net, dev) {
 497                 idev = __in6_dev_get(dev);
 498                 if (idev) {
 499                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 500                         idev->cnf.forwarding = newf;
 501                         if (changed)
 502                                 dev_forward_change(idev);
 503                 }
 504         }
 505 }
 506
 507 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 508 {
 509         struct net *net;
 510
 511         net = (struct net *)table->extra2;
 512         if (p == &net->ipv6.devconf_dflt->forwarding)
 513                 return 0;
 514
 515         if (!rtnl_trylock()) {
 516                 /* Restore the original values before restarting */
 517                 *p = old;
 518                 return restart_syscall();
 519         }
 520
 521         if (p == &net->ipv6.devconf_all->forwarding) {
 522                 __s32 newf = net->ipv6.devconf_all->forwarding;
 523                 net->ipv6.devconf_dflt->forwarding = newf;
 524                 addrconf_forward_change(net, newf);
 525         } else if ((!*p) ^ (!old))
 526                 dev_forward_change((struct inet6_dev *)table->extra1);
 527         rtnl_unlock();
 528
 529         if (*p)
 530                 rt6_purge_dflt_routers(net);
 531         return 1;
 532 }
 533 #endif
 534
 535 /* Nobody refers to this ifaddr, destroy it */
 536 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 537 {
 538         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 539
 540 #ifdef NET_REFCNT_DEBUG
 541         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 542 #endif
 543
 544         in6_dev_put(ifp->idev);
 545
 546         if (del_timer(&ifp->timer))
 547                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 548
 549         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 550                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 551                 return;
 552         }
 553         dst_release(&ifp->rt->dst);
 554
 555         kfree_rcu(ifp, rcu);
 556 }
 557
 558 static void
 559 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 560 {
 561         struct list_head *p;
 562         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 563
 564         /*
 565          * Each device address list is sorted in order of scope -
 566          * global before linklocal.
 567          */
 568         list_for_each(p, &idev->addr_list) {
 569                 struct inet6_ifaddr *ifa
 570                         = list_entry(p, struct inet6_ifaddr, if_list);
 571                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 572                         break;
 573         }
 574
 575         list_add_tail(&ifp->if_list, p);
 576 }
 577
 578 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 579 {
 580         /*
 581          * We perform the hash function over the last 64 bits of the address
 582          * This will include the IEEE address token on links that support it.
 583          */
 584         return jhash_2words((__force u32)addr->s6_addr32[2],
 585                             (__force u32)addr->s6_addr32[3], 0)
 586                 & (IN6_ADDR_HSIZE - 1);
 587 }
 588
 589 /* On success it returns ifp with increased reference count */
 590
 591 static struct inet6_ifaddr *
 592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 593               int scope, u32 flags)
 594 {
 595         struct inet6_ifaddr *ifa = NULL;
 596         struct rt6_info *rt;
 597         unsigned int hash;
 598         int err = 0;
 599         int addr_type = ipv6_addr_type(addr);
 600
 601         if (addr_type == IPV6_ADDR_ANY ||
 602             addr_type & IPV6_ADDR_MULTICAST ||
 603             (!(idev->dev->flags & IFF_LOOPBACK) &&
 604              addr_type & IPV6_ADDR_LOOPBACK))
 605                 return ERR_PTR(-EADDRNOTAVAIL);
 606
 607         rcu_read_lock_bh();
 608         if (idev->dead) {
 609                 err = -ENODEV;                  /*XXX*/
 610                 goto out2;
 611         }
 612
 613         if (idev->cnf.disable_ipv6) {
 614                 err = -EACCES;
 615                 goto out2;
 616         }
 617
 618         spin_lock(&addrconf_hash_lock);
 619
 620         /* Ignore adding duplicate addresses on an interface */
 621         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 622                 ADBG(("ipv6_add_addr: already assigned\n"));
 623                 err = -EEXIST;
 624                 goto out;
 625         }
 626
 627         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 628
 629         if (ifa == NULL) {
 630                 ADBG(("ipv6_add_addr: malloc failed\n"));
 631                 err = -ENOBUFS;
 632                 goto out;
 633         }
 634
 635         rt = addrconf_dst_alloc(idev, addr, 0);
 636         if (IS_ERR(rt)) {
 637                 err = PTR_ERR(rt);
 638                 goto out;
 639         }
 640
 641         ipv6_addr_copy(&ifa->addr, addr);
 642
 643         spin_lock_init(&ifa->lock);
 644         spin_lock_init(&ifa->state_lock);
 645         init_timer(&ifa->timer);
 646         INIT_HLIST_NODE(&ifa->addr_lst);
 647         ifa->timer.data = (unsigned long) ifa;
 648         ifa->scope = scope;
 649         ifa->prefix_len = pfxlen;
 650         ifa->flags = flags | IFA_F_TENTATIVE;
 651         ifa->cstamp = ifa->tstamp = jiffies;
 652
 653         ifa->rt = rt;
 654
 655         /*
 656          * part one of RFC 4429, section 3.3
 657          * We should not configure an address as
 658          * optimistic if we do not yet know the link
 659          * layer address of our nexhop router
 660          */
 661
 662         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 663                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 664
 665         ifa->idev = idev;
 666         in6_dev_hold(idev);
 667         /* For caller */
 668         in6_ifa_hold(ifa);
 669
 670         /* Add to big hash table */
 671         hash = ipv6_addr_hash(addr);
 672
 673         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 674         spin_unlock(&addrconf_hash_lock);
 675
 676         write_lock(&idev->lock);
 677         /* Add to inet6_dev unicast addr list. */
 678         ipv6_link_dev_addr(idev, ifa);
 679
 680 #ifdef CONFIG_IPV6_PRIVACY
 681         if (ifa->flags&IFA_F_TEMPORARY) {
 682                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 683                 in6_ifa_hold(ifa);
 684         }
 685 #endif
 686
 687         in6_ifa_hold(ifa);
 688         write_unlock(&idev->lock);
 689 out2:
 690         rcu_read_unlock_bh();
 691
 692         if (likely(err == 0))
 693                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 694         else {
 695                 kfree(ifa);
 696                 ifa = ERR_PTR(err);
 697         }
 698
 699         return ifa;
 700 out:
 701         spin_unlock(&addrconf_hash_lock);
 702         goto out2;
 703 }
 704
 705 /* This function wants to get referenced ifp and releases it before return */
 706
 707 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 708 {
 709         struct inet6_ifaddr *ifa, *ifn;
 710         struct inet6_dev *idev = ifp->idev;
 711         int state;
 712         int deleted = 0, onlink = 0;
 713         unsigned long expires = jiffies;
 714
 715         spin_lock_bh(&ifp->state_lock);
 716         state = ifp->state;
 717         ifp->state = INET6_IFADDR_STATE_DEAD;
 718         spin_unlock_bh(&ifp->state_lock);
 719
 720         if (state == INET6_IFADDR_STATE_DEAD)
 721                 goto out;
 722
 723         spin_lock_bh(&addrconf_hash_lock);
 724         hlist_del_init_rcu(&ifp->addr_lst);
 725         spin_unlock_bh(&addrconf_hash_lock);
 726
 727         write_lock_bh(&idev->lock);
 728 #ifdef CONFIG_IPV6_PRIVACY
 729         if (ifp->flags&IFA_F_TEMPORARY) {
 730                 list_del(&ifp->tmp_list);
 731                 if (ifp->ifpub) {
 732                         in6_ifa_put(ifp->ifpub);
 733                         ifp->ifpub = NULL;
 734                 }
 735                 __in6_ifa_put(ifp);
 736         }
 737 #endif
 738
 739         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 740                 if (ifa == ifp) {
 741                         list_del_init(&ifp->if_list);
 742                         __in6_ifa_put(ifp);
 743
 744                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 745                                 break;
 746                         deleted = 1;
 747                         continue;
 748                 } else if (ifp->flags & IFA_F_PERMANENT) {
 749                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 750                                               ifp->prefix_len)) {
 751                                 if (ifa->flags & IFA_F_PERMANENT) {
 752                                         onlink = 1;
 753                                         if (deleted)
 754                                                 break;
 755                                 } else {
 756                                         unsigned long lifetime;
 757
 758                                         if (!onlink)
 759                                                 onlink = -1;
 760
 761                                         spin_lock(&ifa->lock);
 762
 763                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 764                                         /*
 765                                          * Note: Because this address is
 766                                          * not permanent, lifetime <
 767                                          * LONG_MAX / HZ here.
 768                                          */
 769                                         if (time_before(expires,
 770                                                         ifa->tstamp + lifetime * HZ))
 771                                                 expires = ifa->tstamp + lifetime * HZ;
 772                                         spin_unlock(&ifa->lock);
 773                                 }
 774                         }
 775                 }
 776         }
 777         write_unlock_bh(&idev->lock);
 778
 779         addrconf_del_timer(ifp);
 780
 781         ipv6_ifa_notify(RTM_DELADDR, ifp);
 782
 783         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 784
 785         /*
 786          * Purge or update corresponding prefix
 787          *
 788          * 1) we don't purge prefix here if address was not permanent.
 789          *    prefix is managed by its own lifetime.
 790          * 2) if there're no addresses, delete prefix.
 791          * 3) if there're still other permanent address(es),
 792          *    corresponding prefix is still permanent.
 793          * 4) otherwise, update prefix lifetime to the
 794          *    longest valid lifetime among the corresponding
 795          *    addresses on the device.
 796          *    Note: subsequent RA will update lifetime.
 797          *
 798          * --yoshfuji
 799          */
 800         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 801                 struct in6_addr prefix;
 802                 struct rt6_info *rt;
 803                 struct net *net = dev_net(ifp->idev->dev);
 804                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 805                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 806
 807                 if (rt && addrconf_is_prefix_route(rt)) {
 808                         if (onlink == 0) {
 809                                 ip6_del_rt(rt);
 810                                 rt = NULL;
 811                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 812                                 rt->rt6i_expires = expires;
 813                                 rt->rt6i_flags |= RTF_EXPIRES;
 814                         }
 815                 }
 816                 dst_release(&rt->dst);
 817         }
 818
 819         /* clean up prefsrc entries */
 820         rt6_remove_prefsrc(ifp);
 821 out:
 822         in6_ifa_put(ifp);
 823 }
 824
 825 #ifdef CONFIG_IPV6_PRIVACY
 826 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 827 {
 828         struct inet6_dev *idev = ifp->idev;
 829         struct in6_addr addr, *tmpaddr;
 830         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 831         unsigned long regen_advance;
 832         int tmp_plen;
 833         int ret = 0;
 834         int max_addresses;
 835         u32 addr_flags;
 836         unsigned long now = jiffies;
 837
 838         write_lock(&idev->lock);
 839         if (ift) {
 840                 spin_lock_bh(&ift->lock);
 841                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 842                 spin_unlock_bh(&ift->lock);
 843                 tmpaddr = &addr;
 844         } else {
 845                 tmpaddr = NULL;
 846         }
 847 retry:
 848         in6_dev_hold(idev);
 849         if (idev->cnf.use_tempaddr <= 0) {
 850                 write_unlock(&idev->lock);
 851                 printk(KERN_INFO
 852                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 853                 in6_dev_put(idev);
 854                 ret = -1;
 855                 goto out;
 856         }
 857         spin_lock_bh(&ifp->lock);
 858         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 859                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 860                 spin_unlock_bh(&ifp->lock);
 861                 write_unlock(&idev->lock);
 862                 printk(KERN_WARNING
 863                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 864                 in6_dev_put(idev);
 865                 ret = -1;
 866                 goto out;
 867         }
 868         in6_ifa_hold(ifp);
 869         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 870         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 871                 spin_unlock_bh(&ifp->lock);
 872                 write_unlock(&idev->lock);
 873                 printk(KERN_WARNING
 874                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 875                 in6_ifa_put(ifp);
 876                 in6_dev_put(idev);
 877                 ret = -1;
 878                 goto out;
 879         }
 880         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 881         age = (now - ifp->tstamp) / HZ;
 882         tmp_valid_lft = min_t(__u32,
 883                               ifp->valid_lft,
 884                               idev->cnf.temp_valid_lft + age);
 885         tmp_prefered_lft = min_t(__u32,
 886                                  ifp->prefered_lft,
 887                                  idev->cnf.temp_prefered_lft + age -
 888                                  idev->cnf.max_desync_factor);
 889         tmp_plen = ifp->prefix_len;
 890         max_addresses = idev->cnf.max_addresses;
 891         tmp_tstamp = ifp->tstamp;
 892         spin_unlock_bh(&ifp->lock);
 893
 894         regen_advance = idev->cnf.regen_max_retry *
 895                         idev->cnf.dad_transmits *
 896                         idev->nd_parms->retrans_time / HZ;
 897         write_unlock(&idev->lock);
 898
 899         /* A temporary address is created only if this calculated Preferred
 900          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 901          * an implementation must not create a temporary address with a zero
 902          * Preferred Lifetime.
 903          * Use age calculation as in addrconf_verify to avoid unnecessary
 904          * temporary addresses being generated.
 905          */
 906         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 907         if (tmp_prefered_lft <= regen_advance + age) {
 908                 in6_ifa_put(ifp);
 909                 in6_dev_put(idev);
 910                 ret = -1;
 911                 goto out;
 912         }
 913
 914         addr_flags = IFA_F_TEMPORARY;
 915         /* set in addrconf_prefix_rcv() */
 916         if (ifp->flags & IFA_F_OPTIMISTIC)
 917                 addr_flags |= IFA_F_OPTIMISTIC;
 918
 919         ift = ipv6_add_addr(idev, &addr, tmp_plen,
 920                             ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 921                             addr_flags);
 922         if (IS_ERR(ift)) {
 923                 in6_ifa_put(ifp);
 924                 in6_dev_put(idev);
 925                 printk(KERN_INFO
 926                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 927                 tmpaddr = &addr;
 928                 write_lock(&idev->lock);
 929                 goto retry;
 930         }
 931
 932         spin_lock_bh(&ift->lock);
 933         ift->ifpub = ifp;
 934         ift->valid_lft = tmp_valid_lft;
 935         ift->prefered_lft = tmp_prefered_lft;
 936         ift->cstamp = now;
 937         ift->tstamp = tmp_tstamp;
 938         spin_unlock_bh(&ift->lock);
 939
 940         addrconf_dad_start(ift, 0);
 941         in6_ifa_put(ift);
 942         in6_dev_put(idev);
 943 out:
 944         return ret;
 945 }
 946 #endif
 947
 948 /*
 949  *      Choose an appropriate source address (RFC3484)
 950  */
 951 enum {
 952         IPV6_SADDR_RULE_INIT = 0,
 953         IPV6_SADDR_RULE_LOCAL,
 954         IPV6_SADDR_RULE_SCOPE,
 955         IPV6_SADDR_RULE_PREFERRED,
 956 #ifdef CONFIG_IPV6_MIP6
 957         IPV6_SADDR_RULE_HOA,
 958 #endif
 959         IPV6_SADDR_RULE_OIF,
 960         IPV6_SADDR_RULE_LABEL,
 961 #ifdef CONFIG_IPV6_PRIVACY
 962         IPV6_SADDR_RULE_PRIVACY,
 963 #endif
 964         IPV6_SADDR_RULE_ORCHID,
 965         IPV6_SADDR_RULE_PREFIX,
 966         IPV6_SADDR_RULE_MAX
 967 };
 968
 969 struct ipv6_saddr_score {
 970         int                     rule;
 971         int                     addr_type;
 972         struct inet6_ifaddr     *ifa;
 973         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 974         int                     scopedist;
 975         int                     matchlen;
 976 };
 977
 978 struct ipv6_saddr_dst {
 979         const struct in6_addr *addr;
 980         int ifindex;
 981         int scope;
 982         int label;
 983         unsigned int prefs;
 984 };
 985
 986 static inline int ipv6_saddr_preferred(int type)
 987 {
 988         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 989                 return 1;
 990         return 0;
 991 }
 992
 993 static int ipv6_get_saddr_eval(struct net *net,
 994                                struct ipv6_saddr_score *score,
 995                                struct ipv6_saddr_dst *dst,
 996                                int i)
 997 {
 998         int ret;
 999
1000         if (i <= score->rule) {
1001                 switch (i) {
1002                 case IPV6_SADDR_RULE_SCOPE:
1003                         ret = score->scopedist;
1004                         break;
1005                 case IPV6_SADDR_RULE_PREFIX:
1006                         ret = score->matchlen;
1007                         break;
1008                 default:
1009                         ret = !!test_bit(i, score->scorebits);
1010                 }
1011                 goto out;
1012         }
1013
1014         switch (i) {
1015         case IPV6_SADDR_RULE_INIT:
1016                 /* Rule 0: remember if hiscore is not ready yet */
1017                 ret = !!score->ifa;
1018                 break;
1019         case IPV6_SADDR_RULE_LOCAL:
1020                 /* Rule 1: Prefer same address */
1021                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1022                 break;
1023         case IPV6_SADDR_RULE_SCOPE:
1024                 /* Rule 2: Prefer appropriate scope
1025                  *
1026                  *      ret
1027                  *       ^
1028                  *    -1 |  d 15
1029                  *    ---+--+-+---> scope
1030                  *       |
1031                  *       |             d is scope of the destination.
1032                  *  B-d  |  \
1033                  *       |   \      <- smaller scope is better if
1034                  *  B-15 |    \        if scope is enough for destinaion.
1035                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1036                  * d-C-1 | /
1037                  *       |/         <- greater is better
1038                  *   -C  /             if scope is not enough for destination.
1039                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1040                  *
1041                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1042                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1043                  * Assume B = 0 and we get C > 29.
1044                  */
1045                 ret = __ipv6_addr_src_scope(score->addr_type);
1046                 if (ret >= dst->scope)
1047                         ret = -ret;
1048                 else
1049                         ret -= 128;     /* 30 is enough */
1050                 score->scopedist = ret;
1051                 break;
1052         case IPV6_SADDR_RULE_PREFERRED:
1053                 /* Rule 3: Avoid deprecated and optimistic addresses */
1054                 ret = ipv6_saddr_preferred(score->addr_type) ||
1055                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1056                 break;
1057 #ifdef CONFIG_IPV6_MIP6
1058         case IPV6_SADDR_RULE_HOA:
1059             {
1060                 /* Rule 4: Prefer home address */
1061                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1062                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1063                 break;
1064             }
1065 #endif
1066         case IPV6_SADDR_RULE_OIF:
1067                 /* Rule 5: Prefer outgoing interface */
1068                 ret = (!dst->ifindex ||
1069                        dst->ifindex == score->ifa->idev->dev->ifindex);
1070                 break;
1071         case IPV6_SADDR_RULE_LABEL:
1072                 /* Rule 6: Prefer matching label */
1073                 ret = ipv6_addr_label(net,
1074                                       &score->ifa->addr, score->addr_type,
1075                                       score->ifa->idev->dev->ifindex) == dst->label;
1076                 break;
1077 #ifdef CONFIG_IPV6_PRIVACY
1078         case IPV6_SADDR_RULE_PRIVACY:
1079             {
1080                 /* Rule 7: Prefer public address
1081                  * Note: prefer temporary address if use_tempaddr >= 2
1082                  */
1083                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1084                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1085                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1086                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1087                 break;
1088             }
1089 #endif
1090         case IPV6_SADDR_RULE_ORCHID:
1091                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1092                  *          non-ORCHID vs non-ORCHID
1093                  */
1094                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1095                         ipv6_addr_orchid(dst->addr));
1096                 break;
1097         case IPV6_SADDR_RULE_PREFIX:
1098                 /* Rule 8: Use longest matching prefix */
1099                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1100                                                        dst->addr);
1101                 break;
1102         default:
1103                 ret = 0;
1104         }
1105
1106         if (ret)
1107                 __set_bit(i, score->scorebits);
1108         score->rule = i;
1109 out:
1110         return ret;
1111 }
1112
1113 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1114                        const struct in6_addr *daddr, unsigned int prefs,
1115                        struct in6_addr *saddr)
1116 {
1117         struct ipv6_saddr_score scores[2],
1118                                 *score = &scores[0], *hiscore = &scores[1];
1119         struct ipv6_saddr_dst dst;
1120         struct net_device *dev;
1121         int dst_type;
1122
1123         dst_type = __ipv6_addr_type(daddr);
1124         dst.addr = daddr;
1125         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1126         dst.scope = __ipv6_addr_src_scope(dst_type);
1127         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1128         dst.prefs = prefs;
1129
1130         hiscore->rule = -1;
1131         hiscore->ifa = NULL;
1132
1133         rcu_read_lock();
1134
1135         for_each_netdev_rcu(net, dev) {
1136                 struct inet6_dev *idev;
1137
1138                 /* Candidate Source Address (section 4)
1139                  *  - multicast and link-local destination address,
1140                  *    the set of candidate source address MUST only
1141                  *    include addresses assigned to interfaces
1142                  *    belonging to the same link as the outgoing
1143                  *    interface.
1144                  * (- For site-local destination addresses, the
1145                  *    set of candidate source addresses MUST only
1146                  *    include addresses assigned to interfaces
1147                  *    belonging to the same site as the outgoing
1148                  *    interface.)
1149                  */
1150                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1151                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1152                     dst.ifindex && dev->ifindex != dst.ifindex)
1153                         continue;
1154
1155                 idev = __in6_dev_get(dev);
1156                 if (!idev)
1157                         continue;
1158
1159                 read_lock_bh(&idev->lock);
1160                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1161                         int i;
1162
1163                         /*
1164                          * - Tentative Address (RFC2462 section 5.4)
1165                          *  - A tentative address is not considered
1166                          *    "assigned to an interface" in the traditional
1167                          *    sense, unless it is also flagged as optimistic.
1168                          * - Candidate Source Address (section 4)
1169                          *  - In any case, anycast addresses, multicast
1170                          *    addresses, and the unspecified address MUST
1171                          *    NOT be included in a candidate set.
1172                          */
1173                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1174                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1175                                 continue;
1176
1177                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1178
1179                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1180                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1181                                 LIMIT_NETDEBUG(KERN_DEBUG
1182                                                "ADDRCONF: unspecified / multicast address "
1183                                                "assigned as unicast address on %s",
1184                                                dev->name);
1185                                 continue;
1186                         }
1187
1188                         score->rule = -1;
1189                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1190
1191                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1192                                 int minihiscore, miniscore;
1193
1194                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1195                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1196
1197                                 if (minihiscore > miniscore) {
1198                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1199                                             score->scopedist > 0) {
1200                                                 /*
1201                                                  * special case:
1202                                                  * each remaining entry
1203                                                  * has too small (not enough)
1204                                                  * scope, because ifa entries
1205                                                  * are sorted by their scope
1206                                                  * values.
1207                                                  */
1208                                                 goto try_nextdev;
1209                                         }
1210                                         break;
1211                                 } else if (minihiscore < miniscore) {
1212                                         if (hiscore->ifa)
1213                                                 in6_ifa_put(hiscore->ifa);
1214
1215                                         in6_ifa_hold(score->ifa);
1216
1217                                         swap(hiscore, score);
1218
1219                                         /* restore our iterator */
1220                                         score->ifa = hiscore->ifa;
1221
1222                                         break;
1223                                 }
1224                         }
1225                 }
1226 try_nextdev:
1227                 read_unlock_bh(&idev->lock);
1228         }
1229         rcu_read_unlock();
1230
1231         if (!hiscore->ifa)
1232                 return -EADDRNOTAVAIL;
1233
1234         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1235         in6_ifa_put(hiscore->ifa);
1236         return 0;
1237 }
1238 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1239
1240 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1241                       unsigned char banned_flags)
1242 {
1243         struct inet6_ifaddr *ifp;
1244         int err = -EADDRNOTAVAIL;
1245
1246         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1247                 if (ifp->scope == IFA_LINK &&
1248                     !(ifp->flags & banned_flags)) {
1249                         ipv6_addr_copy(addr, &ifp->addr);
1250                         err = 0;
1251                         break;
1252                 }
1253         }
1254         return err;
1255 }
1256
1257 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1258                     unsigned char banned_flags)
1259 {
1260         struct inet6_dev *idev;
1261         int err = -EADDRNOTAVAIL;
1262
1263         rcu_read_lock();
1264         idev = __in6_dev_get(dev);
1265         if (idev) {
1266                 read_lock_bh(&idev->lock);
1267                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1268                 read_unlock_bh(&idev->lock);
1269         }
1270         rcu_read_unlock();
1271         return err;
1272 }
1273
1274 static int ipv6_count_addresses(struct inet6_dev *idev)
1275 {
1276         int cnt = 0;
1277         struct inet6_ifaddr *ifp;
1278
1279         read_lock_bh(&idev->lock);
1280         list_for_each_entry(ifp, &idev->addr_list, if_list)
1281                 cnt++;
1282         read_unlock_bh(&idev->lock);
1283         return cnt;
1284 }
1285
1286 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1287                   struct net_device *dev, int strict)
1288 {
1289         struct inet6_ifaddr *ifp;
1290         struct hlist_node *node;
1291         unsigned int hash = ipv6_addr_hash(addr);
1292
1293         rcu_read_lock_bh();
1294         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1295                 if (!net_eq(dev_net(ifp->idev->dev), net))
1296                         continue;
1297                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1298                     !(ifp->flags&IFA_F_TENTATIVE) &&
1299                     (dev == NULL || ifp->idev->dev == dev ||
1300                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1301                         rcu_read_unlock_bh();
1302                         return 1;
1303                 }
1304         }
1305
1306         rcu_read_unlock_bh();
1307         return 0;
1308 }
1309 EXPORT_SYMBOL(ipv6_chk_addr);
1310
1311 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1312                                struct net_device *dev)
1313 {
1314         unsigned int hash = ipv6_addr_hash(addr);
1315         struct inet6_ifaddr *ifp;
1316         struct hlist_node *node;
1317
1318         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1319                 if (!net_eq(dev_net(ifp->idev->dev), net))
1320                         continue;
1321                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1322                         if (dev == NULL || ifp->idev->dev == dev)
1323                                 return true;
1324                 }
1325         }
1326         return false;
1327 }
1328
1329 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1330 {
1331         struct inet6_dev *idev;
1332         struct inet6_ifaddr *ifa;
1333         int     onlink;
1334
1335         onlink = 0;
1336         rcu_read_lock();
1337         idev = __in6_dev_get(dev);
1338         if (idev) {
1339                 read_lock_bh(&idev->lock);
1340                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1341                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1342                                                    ifa->prefix_len);
1343                         if (onlink)
1344                                 break;
1345                 }
1346                 read_unlock_bh(&idev->lock);
1347         }
1348         rcu_read_unlock();
1349         return onlink;
1350 }
1351
1352 EXPORT_SYMBOL(ipv6_chk_prefix);
1353
1354 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1355                                      struct net_device *dev, int strict)
1356 {
1357         struct inet6_ifaddr *ifp, *result = NULL;
1358         unsigned int hash = ipv6_addr_hash(addr);
1359         struct hlist_node *node;
1360
1361         rcu_read_lock_bh();
1362         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1363                 if (!net_eq(dev_net(ifp->idev->dev), net))
1364                         continue;
1365                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1366                         if (dev == NULL || ifp->idev->dev == dev ||
1367                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1368                                 result = ifp;
1369                                 in6_ifa_hold(ifp);
1370                                 break;
1371                         }
1372                 }
1373         }
1374         rcu_read_unlock_bh();
1375
1376         return result;
1377 }
1378
1379 /* Gets referenced address, destroys ifaddr */
1380
1381 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1382 {
1383         if (ifp->flags&IFA_F_PERMANENT) {
1384                 spin_lock_bh(&ifp->lock);
1385                 addrconf_del_timer(ifp);
1386                 ifp->flags |= IFA_F_TENTATIVE;
1387                 if (dad_failed)
1388                         ifp->flags |= IFA_F_DADFAILED;
1389                 spin_unlock_bh(&ifp->lock);
1390                 if (dad_failed)
1391                         ipv6_ifa_notify(0, ifp);
1392                 in6_ifa_put(ifp);
1393 #ifdef CONFIG_IPV6_PRIVACY
1394         } else if (ifp->flags&IFA_F_TEMPORARY) {
1395                 struct inet6_ifaddr *ifpub;
1396                 spin_lock_bh(&ifp->lock);
1397                 ifpub = ifp->ifpub;
1398                 if (ifpub) {
1399                         in6_ifa_hold(ifpub);
1400                         spin_unlock_bh(&ifp->lock);
1401                         ipv6_create_tempaddr(ifpub, ifp);
1402                         in6_ifa_put(ifpub);
1403                 } else {
1404                         spin_unlock_bh(&ifp->lock);
1405                 }
1406                 ipv6_del_addr(ifp);
1407 #endif
1408         } else
1409                 ipv6_del_addr(ifp);
1410 }
1411
1412 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1413 {
1414         int err = -ENOENT;
1415
1416         spin_lock(&ifp->state_lock);
1417         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1418                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1419                 err = 0;
1420         }
1421         spin_unlock(&ifp->state_lock);
1422
1423         return err;
1424 }
1425
1426 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1427 {
1428         struct inet6_dev *idev = ifp->idev;
1429
1430         if (addrconf_dad_end(ifp)) {
1431                 in6_ifa_put(ifp);
1432                 return;
1433         }
1434
1435         if (net_ratelimit())
1436                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1437                         ifp->idev->dev->name, &ifp->addr);
1438
1439         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1440                 struct in6_addr addr;
1441
1442                 addr.s6_addr32[0] = htonl(0xfe800000);
1443                 addr.s6_addr32[1] = 0;
1444
1445                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1446                     ipv6_addr_equal(&ifp->addr, &addr)) {
1447                         /* DAD failed for link-local based on MAC address */
1448                         idev->cnf.disable_ipv6 = 1;
1449
1450                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1451                                 ifp->idev->dev->name);
1452                 }
1453         }
1454
1455         addrconf_dad_stop(ifp, 1);
1456 }
1457
1458 /* Join to solicited addr multicast group. */
1459
1460 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1461 {
1462         struct in6_addr maddr;
1463
1464         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1465                 return;
1466
1467         addrconf_addr_solict_mult(addr, &maddr);
1468         ipv6_dev_mc_inc(dev, &maddr);
1469 }
1470
1471 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1472 {
1473         struct in6_addr maddr;
1474
1475         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1476                 return;
1477
1478         addrconf_addr_solict_mult(addr, &maddr);
1479         __ipv6_dev_mc_dec(idev, &maddr);
1480 }
1481
1482 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1483 {
1484         struct in6_addr addr;
1485         if (ifp->prefix_len == 127) /* RFC 6164 */
1486                 return;
1487         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1488         if (ipv6_addr_any(&addr))
1489                 return;
1490         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1491 }
1492
1493 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1494 {
1495         struct in6_addr addr;
1496         if (ifp->prefix_len == 127) /* RFC 6164 */
1497                 return;
1498         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1499         if (ipv6_addr_any(&addr))
1500                 return;
1501         __ipv6_dev_ac_dec(ifp->idev, &addr);
1502 }
1503
1504 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1505 {
1506         if (dev->addr_len != ETH_ALEN)
1507                 return -1;
1508         memcpy(eui, dev->dev_addr, 3);
1509         memcpy(eui + 5, dev->dev_addr + 3, 3);
1510
1511         /*
1512          * The zSeries OSA network cards can be shared among various
1513          * OS instances, but the OSA cards have only one MAC address.
1514          * This leads to duplicate address conflicts in conjunction
1515          * with IPv6 if more than one instance uses the same card.
1516          *
1517          * The driver for these cards can deliver a unique 16-bit
1518          * identifier for each instance sharing the same card.  It is
1519          * placed instead of 0xFFFE in the interface identifier.  The
1520          * "u" bit of the interface identifier is not inverted in this
1521          * case.  Hence the resulting interface identifier has local
1522          * scope according to RFC2373.
1523          */
1524         if (dev->dev_id) {
1525                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1526                 eui[4] = dev->dev_id & 0xFF;
1527         } else {
1528                 eui[3] = 0xFF;
1529                 eui[4] = 0xFE;
1530                 eui[0] ^= 2;
1531         }
1532         return 0;
1533 }
1534
1535 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1536 {
1537         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1538         if (dev->addr_len != ARCNET_ALEN)
1539                 return -1;
1540         memset(eui, 0, 7);
1541         eui[7] = *(u8*)dev->dev_addr;
1542         return 0;
1543 }
1544
1545 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1546 {
1547         if (dev->addr_len != INFINIBAND_ALEN)
1548                 return -1;
1549         memcpy(eui, dev->dev_addr + 12, 8);
1550         eui[0] |= 2;
1551         return 0;
1552 }
1553
1554 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1555 {
1556         if (addr == 0)
1557                 return -1;
1558         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1559                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1560                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1561                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1562                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1563                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1564         eui[1] = 0;
1565         eui[2] = 0x5E;
1566         eui[3] = 0xFE;
1567         memcpy(eui + 4, &addr, 4);
1568         return 0;
1569 }
1570
1571 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1572 {
1573         if (dev->priv_flags & IFF_ISATAP)
1574                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1575         return -1;
1576 }
1577
1578 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1579 {
1580         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1581 }
1582
1583 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1584 {
1585         switch (dev->type) {
1586         case ARPHRD_ETHER:
1587         case ARPHRD_FDDI:
1588         case ARPHRD_IEEE802_TR:
1589                 return addrconf_ifid_eui48(eui, dev);
1590         case ARPHRD_ARCNET:
1591                 return addrconf_ifid_arcnet(eui, dev);
1592         case ARPHRD_INFINIBAND:
1593                 return addrconf_ifid_infiniband(eui, dev);
1594         case ARPHRD_SIT:
1595                 return addrconf_ifid_sit(eui, dev);
1596         case ARPHRD_IPGRE:
1597                 return addrconf_ifid_gre(eui, dev);
1598         }
1599         return -1;
1600 }
1601
1602 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1603 {
1604         int err = -1;
1605         struct inet6_ifaddr *ifp;
1606
1607         read_lock_bh(&idev->lock);
1608         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1609                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1610                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1611                         err = 0;
1612                         break;
1613                 }
1614         }
1615         read_unlock_bh(&idev->lock);
1616         return err;
1617 }
1618
1619 #ifdef CONFIG_IPV6_PRIVACY
1620 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1621 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1622 {
1623 regen:
1624         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1625         idev->rndid[0] &= ~0x02;
1626
1627         /*
1628          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1629          * check if generated address is not inappropriate
1630          *
1631          *  - Reserved subnet anycast (RFC 2526)
1632          *      11111101 11....11 1xxxxxxx
1633          *  - ISATAP (RFC4214) 6.1
1634          *      00-00-5E-FE-xx-xx-xx-xx
1635          *  - value 0
1636          *  - XXX: already assigned to an address on the device
1637          */
1638         if (idev->rndid[0] == 0xfd &&
1639             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1640             (idev->rndid[7]&0x80))
1641                 goto regen;
1642         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1643                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1644                         goto regen;
1645                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1646                         goto regen;
1647         }
1648
1649         return 0;
1650 }
1651
1652 static void ipv6_regen_rndid(unsigned long data)
1653 {
1654         struct inet6_dev *idev = (struct inet6_dev *) data;
1655         unsigned long expires;
1656
1657         rcu_read_lock_bh();
1658         write_lock_bh(&idev->lock);
1659
1660         if (idev->dead)
1661                 goto out;
1662
1663         if (__ipv6_regen_rndid(idev) < 0)
1664                 goto out;
1665
1666         expires = jiffies +
1667                 idev->cnf.temp_prefered_lft * HZ -
1668                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1669                 idev->cnf.max_desync_factor * HZ;
1670         if (time_before(expires, jiffies)) {
1671                 printk(KERN_WARNING
1672                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1673                         idev->dev->name);
1674                 goto out;
1675         }
1676
1677         if (!mod_timer(&idev->regen_timer, expires))
1678                 in6_dev_hold(idev);
1679
1680 out:
1681         write_unlock_bh(&idev->lock);
1682         rcu_read_unlock_bh();
1683         in6_dev_put(idev);
1684 }
1685
1686 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1687         int ret = 0;
1688
1689         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1690                 ret = __ipv6_regen_rndid(idev);
1691         return ret;
1692 }
1693 #endif
1694
1695 /*
1696  *      Add prefix route.
1697  */
1698
1699 static void
1700 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1701                       unsigned long expires, u32 flags)
1702 {
1703         struct fib6_config cfg = {
1704                 .fc_table = RT6_TABLE_PREFIX,
1705                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1706                 .fc_ifindex = dev->ifindex,
1707                 .fc_expires = expires,
1708                 .fc_dst_len = plen,
1709                 .fc_flags = RTF_UP | flags,
1710                 .fc_nlinfo.nl_net = dev_net(dev),
1711                 .fc_protocol = RTPROT_KERNEL,
1712         };
1713
1714         ipv6_addr_copy(&cfg.fc_dst, pfx);
1715
1716         /* Prevent useless cloning on PtP SIT.
1717            This thing is done here expecting that the whole
1718            class of non-broadcast devices need not cloning.
1719          */
1720 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1721         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1722                 cfg.fc_flags |= RTF_NONEXTHOP;
1723 #endif
1724
1725         ip6_route_add(&cfg);
1726 }
1727
1728
1729 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1730                                                   int plen,
1731                                                   const struct net_device *dev,
1732                                                   u32 flags, u32 noflags)
1733 {
1734         struct fib6_node *fn;
1735         struct rt6_info *rt = NULL;
1736         struct fib6_table *table;
1737
1738         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1739         if (table == NULL)
1740                 return NULL;
1741
1742         write_lock_bh(&table->tb6_lock);
1743         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1744         if (!fn)
1745                 goto out;
1746         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1747                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1748                         continue;
1749                 if ((rt->rt6i_flags & flags) != flags)
1750                         continue;
1751                 if ((rt->rt6i_flags & noflags) != 0)
1752                         continue;
1753                 dst_hold(&rt->dst);
1754                 break;
1755         }
1756 out:
1757         write_unlock_bh(&table->tb6_lock);
1758         return rt;
1759 }
1760
1761
1762 /* Create "default" multicast route to the interface */
1763
1764 static void addrconf_add_mroute(struct net_device *dev)
1765 {
1766         struct fib6_config cfg = {
1767                 .fc_table = RT6_TABLE_LOCAL,
1768                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1769                 .fc_ifindex = dev->ifindex,
1770                 .fc_dst_len = 8,
1771                 .fc_flags = RTF_UP,
1772                 .fc_nlinfo.nl_net = dev_net(dev),
1773         };
1774
1775         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1776
1777         ip6_route_add(&cfg);
1778 }
1779
1780 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1781 static void sit_route_add(struct net_device *dev)
1782 {
1783         struct fib6_config cfg = {
1784                 .fc_table = RT6_TABLE_MAIN,
1785                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1786                 .fc_ifindex = dev->ifindex,
1787                 .fc_dst_len = 96,
1788                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1789                 .fc_nlinfo.nl_net = dev_net(dev),
1790         };
1791
1792         /* prefix length - 96 bits "::d.d.d.d" */
1793         ip6_route_add(&cfg);
1794 }
1795 #endif
1796
1797 static void addrconf_add_lroute(struct net_device *dev)
1798 {
1799         struct in6_addr addr;
1800
1801         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1802         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1803 }
1804
1805 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1806 {
1807         struct inet6_dev *idev;
1808
1809         ASSERT_RTNL();
1810
1811         idev = ipv6_find_idev(dev);
1812         if (!idev)
1813                 return ERR_PTR(-ENOBUFS);
1814
1815         if (idev->cnf.disable_ipv6)
1816                 return ERR_PTR(-EACCES);
1817
1818         /* Add default multicast route */
1819         if (!(dev->flags & IFF_LOOPBACK))
1820                 addrconf_add_mroute(dev);
1821
1822         /* Add link local route */
1823         addrconf_add_lroute(dev);
1824         return idev;
1825 }
1826
1827 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1828 {
1829         struct prefix_info *pinfo;
1830         __u32 valid_lft;
1831         __u32 prefered_lft;
1832         int addr_type;
1833         struct inet6_dev *in6_dev;
1834         struct net *net = dev_net(dev);
1835
1836         pinfo = (struct prefix_info *) opt;
1837
1838         if (len < sizeof(struct prefix_info)) {
1839                 ADBG(("addrconf: prefix option too short\n"));
1840                 return;
1841         }
1842
1843         /*
1844          *      Validation checks ([ADDRCONF], page 19)
1845          */
1846
1847         addr_type = ipv6_addr_type(&pinfo->prefix);
1848
1849         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1850                 return;
1851
1852         valid_lft = ntohl(pinfo->valid);
1853         prefered_lft = ntohl(pinfo->prefered);
1854
1855         if (prefered_lft > valid_lft) {
1856                 if (net_ratelimit())
1857                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1858                 return;
1859         }
1860
1861         in6_dev = in6_dev_get(dev);
1862
1863         if (in6_dev == NULL) {
1864                 if (net_ratelimit())
1865                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1866                 return;
1867         }
1868
1869         /*
1870          *      Two things going on here:
1871          *      1) Add routes for on-link prefixes
1872          *      2) Configure prefixes with the auto flag set
1873          */
1874
1875         if (pinfo->onlink) {
1876                 struct rt6_info *rt;
1877                 unsigned long rt_expires;
1878
1879                 /* Avoid arithmetic overflow. Really, we could
1880                  * save rt_expires in seconds, likely valid_lft,
1881                  * but it would require division in fib gc, that it
1882                  * not good.
1883                  */
1884                 if (HZ > USER_HZ)
1885                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1886                 else
1887                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1888
1889                 if (addrconf_finite_timeout(rt_expires))
1890                         rt_expires *= HZ;
1891
1892                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1893                                                pinfo->prefix_len,
1894                                                dev,
1895                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1896                                                RTF_GATEWAY | RTF_DEFAULT);
1897
1898                 if (rt) {
1899                         /* Autoconf prefix route */
1900                         if (valid_lft == 0) {
1901                                 ip6_del_rt(rt);
1902                                 rt = NULL;
1903                         } else if (addrconf_finite_timeout(rt_expires)) {
1904                                 /* not infinity */
1905                                 rt->rt6i_expires = jiffies + rt_expires;
1906                                 rt->rt6i_flags |= RTF_EXPIRES;
1907                         } else {
1908                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1909                                 rt->rt6i_expires = 0;
1910                         }
1911                 } else if (valid_lft) {
1912                         clock_t expires = 0;
1913                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1914                         if (addrconf_finite_timeout(rt_expires)) {
1915                                 /* not infinity */
1916                                 flags |= RTF_EXPIRES;
1917                                 expires = jiffies_to_clock_t(rt_expires);
1918                         }
1919                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1920                                               dev, expires, flags);
1921                 }
1922                 if (rt)
1923                         dst_release(&rt->dst);
1924         }
1925
1926         /* Try to figure out our local address for this prefix */
1927
1928         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1929                 struct inet6_ifaddr * ifp;
1930                 struct in6_addr addr;
1931                 int create = 0, update_lft = 0;
1932
1933                 if (pinfo->prefix_len == 64) {
1934                         memcpy(&addr, &pinfo->prefix, 8);
1935                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1936                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1937                                 in6_dev_put(in6_dev);
1938                                 return;
1939                         }
1940                         goto ok;
1941                 }
1942                 if (net_ratelimit())
1943                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1944                                pinfo->prefix_len);
1945                 in6_dev_put(in6_dev);
1946                 return;
1947
1948 ok:
1949
1950                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1951
1952                 if (ifp == NULL && valid_lft) {
1953                         int max_addresses = in6_dev->cnf.max_addresses;
1954                         u32 addr_flags = 0;
1955
1956 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1957                         if (in6_dev->cnf.optimistic_dad &&
1958                             !net->ipv6.devconf_all->forwarding)
1959                                 addr_flags = IFA_F_OPTIMISTIC;
1960 #endif
1961
1962                         /* Do not allow to create too much of autoconfigured
1963                          * addresses; this would be too easy way to crash kernel.
1964                          */
1965                         if (!max_addresses ||
1966                             ipv6_count_addresses(in6_dev) < max_addresses)
1967                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1968                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1969                                                     addr_flags);
1970
1971                         if (!ifp || IS_ERR(ifp)) {
1972                                 in6_dev_put(in6_dev);
1973                                 return;
1974                         }
1975
1976                         update_lft = create = 1;
1977                         ifp->cstamp = jiffies;
1978                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1979                 }
1980
1981                 if (ifp) {
1982                         int flags;
1983                         unsigned long now;
1984 #ifdef CONFIG_IPV6_PRIVACY
1985                         struct inet6_ifaddr *ift;
1986 #endif
1987                         u32 stored_lft;
1988
1989                         /* update lifetime (RFC2462 5.5.3 e) */
1990                         spin_lock(&ifp->lock);
1991                         now = jiffies;
1992                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1993                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1994                         else
1995                                 stored_lft = 0;
1996                         if (!update_lft && stored_lft) {
1997                                 if (valid_lft > MIN_VALID_LIFETIME ||
1998                                     valid_lft > stored_lft)
1999                                         update_lft = 1;
2000                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2001                                         /* valid_lft <= stored_lft is always true */
2002                                         /*
2003                                          * RFC 4862 Section 5.5.3e:
2004                                          * "Note that the preferred lifetime of
2005                                          *  the corresponding address is always
2006                                          *  reset to the Preferred Lifetime in
2007                                          *  the received Prefix Information
2008                                          *  option, regardless of whether the
2009                                          *  valid lifetime is also reset or
2010                                          *  ignored."
2011                                          *
2012                                          *  So if the preferred lifetime in
2013                                          *  this advertisement is different
2014                                          *  than what we have stored, but the
2015                                          *  valid lifetime is invalid, just
2016                                          *  reset prefered_lft.
2017                                          *
2018                                          *  We must set the valid lifetime
2019                                          *  to the stored lifetime since we'll
2020                                          *  be updating the timestamp below,
2021                                          *  else we'll set it back to the
2022                                          *  minimum.
2023                                          */
2024                                         if (prefered_lft != ifp->prefered_lft) {
2025                                                 valid_lft = stored_lft;
2026                                                 update_lft = 1;
2027                                         }
2028                                 } else {
2029                                         valid_lft = MIN_VALID_LIFETIME;
2030                                         if (valid_lft < prefered_lft)
2031                                                 prefered_lft = valid_lft;
2032                                         update_lft = 1;
2033                                 }
2034                         }
2035
2036                         if (update_lft) {
2037                                 ifp->valid_lft = valid_lft;
2038                                 ifp->prefered_lft = prefered_lft;
2039                                 ifp->tstamp = now;
2040                                 flags = ifp->flags;
2041                                 ifp->flags &= ~IFA_F_DEPRECATED;
2042                                 spin_unlock(&ifp->lock);
2043
2044                                 if (!(flags&IFA_F_TENTATIVE))
2045                                         ipv6_ifa_notify(0, ifp);
2046                         } else
2047                                 spin_unlock(&ifp->lock);
2048
2049 #ifdef CONFIG_IPV6_PRIVACY
2050                         read_lock_bh(&in6_dev->lock);
2051                         /* update all temporary addresses in the list */
2052                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2053                                             tmp_list) {
2054                                 int age, max_valid, max_prefered;
2055
2056                                 if (ifp != ift->ifpub)
2057                                         continue;
2058
2059                                 /*
2060                                  * RFC 4941 section 3.3:
2061                                  * If a received option will extend the lifetime
2062                                  * of a public address, the lifetimes of
2063                                  * temporary addresses should be extended,
2064                                  * subject to the overall constraint that no
2065                                  * temporary addresses should ever remain
2066                                  * "valid" or "preferred" for a time longer than
2067                                  * (TEMP_VALID_LIFETIME) or
2068                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2069                                  * respectively.
2070                                  */
2071                                 age = (now - ift->cstamp) / HZ;
2072                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2073                                 if (max_valid < 0)
2074                                         max_valid = 0;
2075
2076                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2077                                                in6_dev->cnf.max_desync_factor -
2078                                                age;
2079                                 if (max_prefered < 0)
2080                                         max_prefered = 0;
2081
2082                                 if (valid_lft > max_valid)
2083                                         valid_lft = max_valid;
2084
2085                                 if (prefered_lft > max_prefered)
2086                                         prefered_lft = max_prefered;
2087
2088                                 spin_lock(&ift->lock);
2089                                 flags = ift->flags;
2090                                 ift->valid_lft = valid_lft;
2091                                 ift->prefered_lft = prefered_lft;
2092                                 ift->tstamp = now;
2093                                 if (prefered_lft > 0)
2094                                         ift->flags &= ~IFA_F_DEPRECATED;
2095
2096                                 spin_unlock(&ift->lock);
2097                                 if (!(flags&IFA_F_TENTATIVE))
2098                                         ipv6_ifa_notify(0, ift);
2099                         }
2100
2101                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2102                                 /*
2103                                  * When a new public address is created as
2104                                  * described in [ADDRCONF], also create a new
2105                                  * temporary address. Also create a temporary
2106                                  * address if it's enabled but no temporary
2107                                  * address currently exists.
2108                                  */
2109                                 read_unlock_bh(&in6_dev->lock);
2110                                 ipv6_create_tempaddr(ifp, NULL);
2111                         } else {
2112                                 read_unlock_bh(&in6_dev->lock);
2113                         }
2114 #endif
2115                         in6_ifa_put(ifp);
2116                         addrconf_verify(0);
2117                 }
2118         }
2119         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2120         in6_dev_put(in6_dev);
2121 }
2122
2123 /*
2124  *      Set destination address.
2125  *      Special case for SIT interfaces where we create a new "virtual"
2126  *      device.
2127  */
2128 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2129 {
2130         struct in6_ifreq ireq;
2131         struct net_device *dev;
2132         int err = -EINVAL;
2133
2134         rtnl_lock();
2135
2136         err = -EFAULT;
2137         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2138                 goto err_exit;
2139
2140         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2141
2142         err = -ENODEV;
2143         if (dev == NULL)
2144                 goto err_exit;
2145
2146 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2147         if (dev->type == ARPHRD_SIT) {
2148                 const struct net_device_ops *ops = dev->netdev_ops;
2149                 struct ifreq ifr;
2150                 struct ip_tunnel_parm p;
2151
2152                 err = -EADDRNOTAVAIL;
2153                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2154                         goto err_exit;
2155
2156                 memset(&p, 0, sizeof(p));
2157                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2158                 p.iph.saddr = 0;
2159                 p.iph.version = 4;
2160                 p.iph.ihl = 5;
2161                 p.iph.protocol = IPPROTO_IPV6;
2162                 p.iph.ttl = 64;
2163                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2164
2165                 if (ops->ndo_do_ioctl) {
2166                         mm_segment_t oldfs = get_fs();
2167
2168                         set_fs(KERNEL_DS);
2169                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2170                         set_fs(oldfs);
2171                 } else
2172                         err = -EOPNOTSUPP;
2173
2174                 if (err == 0) {
2175                         err = -ENOBUFS;
2176                         dev = __dev_get_by_name(net, p.name);
2177                         if (!dev)
2178                                 goto err_exit;
2179                         err = dev_open(dev);
2180                 }
2181         }
2182 #endif
2183
2184 err_exit:
2185         rtnl_unlock();
2186         return err;
2187 }
2188
2189 /*
2190  *      Manual configuration of address on an interface
2191  */
2192 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2193                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2194                           __u32 valid_lft)
2195 {
2196         struct inet6_ifaddr *ifp;
2197         struct inet6_dev *idev;
2198         struct net_device *dev;
2199         int scope;
2200         u32 flags;
2201         clock_t expires;
2202         unsigned long timeout;
2203
2204         ASSERT_RTNL();
2205
2206         if (plen > 128)
2207                 return -EINVAL;
2208
2209         /* check the lifetime */
2210         if (!valid_lft || prefered_lft > valid_lft)
2211                 return -EINVAL;
2212
2213         dev = __dev_get_by_index(net, ifindex);
2214         if (!dev)
2215                 return -ENODEV;
2216
2217         idev = addrconf_add_dev(dev);
2218         if (IS_ERR(idev))
2219                 return PTR_ERR(idev);
2220
2221         scope = ipv6_addr_scope(pfx);
2222
2223         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2224         if (addrconf_finite_timeout(timeout)) {
2225                 expires = jiffies_to_clock_t(timeout * HZ);
2226                 valid_lft = timeout;
2227                 flags = RTF_EXPIRES;
2228         } else {
2229                 expires = 0;
2230                 flags = 0;
2231                 ifa_flags |= IFA_F_PERMANENT;
2232         }
2233
2234         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2235         if (addrconf_finite_timeout(timeout)) {
2236                 if (timeout == 0)
2237                         ifa_flags |= IFA_F_DEPRECATED;
2238                 prefered_lft = timeout;
2239         }
2240
2241         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2242
2243         if (!IS_ERR(ifp)) {
2244                 spin_lock_bh(&ifp->lock);
2245                 ifp->valid_lft = valid_lft;
2246                 ifp->prefered_lft = prefered_lft;
2247                 ifp->tstamp = jiffies;
2248                 spin_unlock_bh(&ifp->lock);
2249
2250                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2251                                       expires, flags);
2252                 /*
2253                  * Note that section 3.1 of RFC 4429 indicates
2254                  * that the Optimistic flag should not be set for
2255                  * manually configured addresses
2256                  */
2257                 addrconf_dad_start(ifp, 0);
2258                 in6_ifa_put(ifp);
2259                 addrconf_verify(0);
2260                 return 0;
2261         }
2262
2263         return PTR_ERR(ifp);
2264 }
2265
2266 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2267                           unsigned int plen)
2268 {
2269         struct inet6_ifaddr *ifp;
2270         struct inet6_dev *idev;
2271         struct net_device *dev;
2272
2273         if (plen > 128)
2274                 return -EINVAL;
2275
2276         dev = __dev_get_by_index(net, ifindex);
2277         if (!dev)
2278                 return -ENODEV;
2279
2280         if ((idev = __in6_dev_get(dev)) == NULL)
2281                 return -ENXIO;
2282
2283         read_lock_bh(&idev->lock);
2284         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2285                 if (ifp->prefix_len == plen &&
2286                     ipv6_addr_equal(pfx, &ifp->addr)) {
2287                         in6_ifa_hold(ifp);
2288                         read_unlock_bh(&idev->lock);
2289
2290                         ipv6_del_addr(ifp);
2291
2292                         /* If the last address is deleted administratively,
2293                            disable IPv6 on this interface.
2294                          */
2295                         if (list_empty(&idev->addr_list))
2296                                 addrconf_ifdown(idev->dev, 1);
2297                         return 0;
2298                 }
2299         }
2300         read_unlock_bh(&idev->lock);
2301         return -EADDRNOTAVAIL;
2302 }
2303
2304
2305 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2306 {
2307         struct in6_ifreq ireq;
2308         int err;
2309
2310         if (!capable(CAP_NET_ADMIN))
2311                 return -EPERM;
2312
2313         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2314                 return -EFAULT;
2315
2316         rtnl_lock();
2317         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2318                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2319                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2320         rtnl_unlock();
2321         return err;
2322 }
2323
2324 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2325 {
2326         struct in6_ifreq ireq;
2327         int err;
2328
2329         if (!capable(CAP_NET_ADMIN))
2330                 return -EPERM;
2331
2332         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2333                 return -EFAULT;
2334
2335         rtnl_lock();
2336         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2337                              ireq.ifr6_prefixlen);
2338         rtnl_unlock();
2339         return err;
2340 }
2341
2342 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2343                      int plen, int scope)
2344 {
2345         struct inet6_ifaddr *ifp;
2346
2347         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2348         if (!IS_ERR(ifp)) {
2349                 spin_lock_bh(&ifp->lock);
2350                 ifp->flags &= ~IFA_F_TENTATIVE;
2351                 spin_unlock_bh(&ifp->lock);
2352                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2353                 in6_ifa_put(ifp);
2354         }
2355 }
2356
2357 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2358 static void sit_add_v4_addrs(struct inet6_dev *idev)
2359 {
2360         struct in6_addr addr;
2361         struct net_device *dev;
2362         struct net *net = dev_net(idev->dev);
2363         int scope;
2364
2365         ASSERT_RTNL();
2366
2367         memset(&addr, 0, sizeof(struct in6_addr));
2368         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2369
2370         if (idev->dev->flags&IFF_POINTOPOINT) {
2371                 addr.s6_addr32[0] = htonl(0xfe800000);
2372                 scope = IFA_LINK;
2373         } else {
2374                 scope = IPV6_ADDR_COMPATv4;
2375         }
2376
2377         if (addr.s6_addr32[3]) {
2378                 add_addr(idev, &addr, 128, scope);
2379                 return;
2380         }
2381
2382         for_each_netdev(net, dev) {
2383                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2384                 if (in_dev && (dev->flags & IFF_UP)) {
2385                         struct in_ifaddr * ifa;
2386
2387                         int flag = scope;
2388
2389                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2390                                 int plen;
2391
2392                                 addr.s6_addr32[3] = ifa->ifa_local;
2393
2394                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2395                                         continue;
2396                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2397                                         if (idev->dev->flags&IFF_POINTOPOINT)
2398                                                 continue;
2399                                         flag |= IFA_HOST;
2400                                 }
2401                                 if (idev->dev->flags&IFF_POINTOPOINT)
2402                                         plen = 64;
2403                                 else
2404                                         plen = 96;
2405
2406                                 add_addr(idev, &addr, plen, flag);
2407                         }
2408                 }
2409         }
2410 }
2411 #endif
2412
2413 static void init_loopback(struct net_device *dev)
2414 {
2415         struct inet6_dev  *idev;
2416         struct net_device *sp_dev;
2417         struct inet6_ifaddr *sp_ifa;
2418         struct rt6_info *sp_rt;
2419
2420         /* ::1 */
2421
2422         ASSERT_RTNL();
2423
2424         if ((idev = ipv6_find_idev(dev)) == NULL) {
2425                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2426                 return;
2427         }
2428
2429         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2430
2431         /* Add routes to other interface's IPv6 addresses */
2432         for_each_netdev(dev_net(dev), sp_dev) {
2433                 if (!strcmp(sp_dev->name, dev->name))
2434                         continue;
2435
2436                 idev = __in6_dev_get(sp_dev);
2437                 if (!idev)
2438                         continue;
2439
2440                 read_lock_bh(&idev->lock);
2441                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2442
2443                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2444                                 continue;
2445
2446                         if (sp_ifa->rt)
2447                                 continue;
2448
2449                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2450
2451                         /* Failure cases are ignored */
2452                         if (!IS_ERR(sp_rt)) {
2453                                 sp_ifa->rt = sp_rt;
2454                                 ip6_ins_rt(sp_rt);
2455                         }
2456                 }
2457                 read_unlock_bh(&idev->lock);
2458         }
2459 }
2460
2461 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2462 {
2463         struct inet6_ifaddr * ifp;
2464         u32 addr_flags = IFA_F_PERMANENT;
2465
2466 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2467         if (idev->cnf.optimistic_dad &&
2468             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2469                 addr_flags |= IFA_F_OPTIMISTIC;
2470 #endif
2471
2472
2473         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2474         if (!IS_ERR(ifp)) {
2475                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2476                 addrconf_dad_start(ifp, 0);
2477                 in6_ifa_put(ifp);
2478         }
2479 }
2480
2481 static void addrconf_dev_config(struct net_device *dev)
2482 {
2483         struct in6_addr addr;
2484         struct inet6_dev    * idev;
2485
2486         ASSERT_RTNL();
2487
2488         if ((dev->type != ARPHRD_ETHER) &&
2489             (dev->type != ARPHRD_FDDI) &&
2490             (dev->type != ARPHRD_IEEE802_TR) &&
2491             (dev->type != ARPHRD_ARCNET) &&
2492             (dev->type != ARPHRD_INFINIBAND)) {
2493                 /* Alas, we support only Ethernet autoconfiguration. */
2494                 return;
2495         }
2496
2497         idev = addrconf_add_dev(dev);
2498         if (IS_ERR(idev))
2499                 return;
2500
2501         memset(&addr, 0, sizeof(struct in6_addr));
2502         addr.s6_addr32[0] = htonl(0xFE800000);
2503
2504         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2505                 addrconf_add_linklocal(idev, &addr);
2506 }
2507
2508 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2509 static void addrconf_sit_config(struct net_device *dev)
2510 {
2511         struct inet6_dev *idev;
2512
2513         ASSERT_RTNL();
2514
2515         /*
2516          * Configure the tunnel with one of our IPv4
2517          * addresses... we should configure all of
2518          * our v4 addrs in the tunnel
2519          */
2520
2521         if ((idev = ipv6_find_idev(dev)) == NULL) {
2522                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2523                 return;
2524         }
2525
2526         if (dev->priv_flags & IFF_ISATAP) {
2527                 struct in6_addr addr;
2528
2529                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2530                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2531                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2532                         addrconf_add_linklocal(idev, &addr);
2533                 return;
2534         }
2535
2536         sit_add_v4_addrs(idev);
2537
2538         if (dev->flags&IFF_POINTOPOINT) {
2539                 addrconf_add_mroute(dev);
2540                 addrconf_add_lroute(dev);
2541         } else
2542                 sit_route_add(dev);
2543 }
2544 #endif
2545
2546 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2547 static void addrconf_gre_config(struct net_device *dev)
2548 {
2549         struct inet6_dev *idev;
2550         struct in6_addr addr;
2551
2552         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2553
2554         ASSERT_RTNL();
2555
2556         if ((idev = ipv6_find_idev(dev)) == NULL) {
2557                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2558                 return;
2559         }
2560
2561         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2562         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2563
2564         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2565                 addrconf_add_linklocal(idev, &addr);
2566 }
2567 #endif
2568
2569 static inline int
2570 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2571 {
2572         struct in6_addr lladdr;
2573
2574         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2575                 addrconf_add_linklocal(idev, &lladdr);
2576                 return 0;
2577         }
2578         return -1;
2579 }
2580
2581 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2582 {
2583         struct net_device *link_dev;
2584         struct net *net = dev_net(idev->dev);
2585
2586         /* first try to inherit the link-local address from the link device */
2587         if (idev->dev->iflink &&
2588             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2589                 if (!ipv6_inherit_linklocal(idev, link_dev))
2590                         return;
2591         }
2592         /* then try to inherit it from any device */
2593         for_each_netdev(net, link_dev) {
2594                 if (!ipv6_inherit_linklocal(idev, link_dev))
2595                         return;
2596         }
2597         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2598 }
2599
2600 /*
2601  * Autoconfigure tunnel with a link-local address so routing protocols,
2602  * DHCPv6, MLD etc. can be run over the virtual link
2603  */
2604
2605 static void addrconf_ip6_tnl_config(struct net_device *dev)
2606 {
2607         struct inet6_dev *idev;
2608
2609         ASSERT_RTNL();
2610
2611         idev = addrconf_add_dev(dev);
2612         if (IS_ERR(idev)) {
2613                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2614                 return;
2615         }
2616         ip6_tnl_add_linklocal(idev);
2617 }
2618
2619 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2620                            void * data)
2621 {
2622         struct net_device *dev = (struct net_device *) data;
2623         struct inet6_dev *idev = __in6_dev_get(dev);
2624         int run_pending = 0;
2625         int err;
2626
2627         switch (event) {
2628         case NETDEV_REGISTER:
2629                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2630                         idev = ipv6_add_dev(dev);
2631                         if (!idev)
2632                                 return notifier_from_errno(-ENOMEM);
2633                 }
2634                 break;
2635
2636         case NETDEV_UP:
2637         case NETDEV_CHANGE:
2638                 if (dev->flags & IFF_SLAVE)
2639                         break;
2640
2641                 if (event == NETDEV_UP) {
2642                         if (!addrconf_qdisc_ok(dev)) {
2643                                 /* device is not ready yet. */
2644                                 printk(KERN_INFO
2645                                         "ADDRCONF(NETDEV_UP): %s: "
2646                                         "link is not ready\n",
2647                                         dev->name);
2648                                 break;
2649                         }
2650
2651                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2652                                 idev = ipv6_add_dev(dev);
2653
2654                         if (idev) {
2655                                 idev->if_flags |= IF_READY;
2656                                 run_pending = 1;
2657                         }
2658                 } else {
2659                         if (!addrconf_qdisc_ok(dev)) {
2660                                 /* device is still not ready. */
2661                                 break;
2662                         }
2663
2664                         if (idev) {
2665                                 if (idev->if_flags & IF_READY)
2666                                         /* device is already configured. */
2667                                         break;
2668                                 idev->if_flags |= IF_READY;
2669                         }
2670
2671                         printk(KERN_INFO
2672                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2673                                         "link becomes ready\n",
2674                                         dev->name);
2675
2676                         run_pending = 1;
2677                 }
2678
2679                 switch (dev->type) {
2680 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2681                 case ARPHRD_SIT:
2682                         addrconf_sit_config(dev);
2683                         break;
2684 #endif
2685 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2686                 case ARPHRD_IPGRE:
2687                         addrconf_gre_config(dev);
2688                         break;
2689 #endif
2690                 case ARPHRD_TUNNEL6:
2691                         addrconf_ip6_tnl_config(dev);
2692                         break;
2693                 case ARPHRD_LOOPBACK:
2694                         init_loopback(dev);
2695                         break;
2696
2697                 default:
2698                         addrconf_dev_config(dev);
2699                         break;
2700                 }
2701
2702                 if (idev) {
2703                         if (run_pending)
2704                                 addrconf_dad_run(idev);
2705
2706                         /*
2707                          * If the MTU changed during the interface down,
2708                          * when the interface up, the changed MTU must be
2709                          * reflected in the idev as well as routers.
2710                          */
2711                         if (idev->cnf.mtu6 != dev->mtu &&
2712                             dev->mtu >= IPV6_MIN_MTU) {
2713                                 rt6_mtu_change(dev, dev->mtu);
2714                                 idev->cnf.mtu6 = dev->mtu;
2715                         }
2716                         idev->tstamp = jiffies;
2717                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2718
2719                         /*
2720                          * If the changed mtu during down is lower than
2721                          * IPV6_MIN_MTU stop IPv6 on this interface.
2722                          */
2723                         if (dev->mtu < IPV6_MIN_MTU)
2724                                 addrconf_ifdown(dev, 1);
2725                 }
2726                 break;
2727
2728         case NETDEV_CHANGEMTU:
2729                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2730                         rt6_mtu_change(dev, dev->mtu);
2731                         idev->cnf.mtu6 = dev->mtu;
2732                         break;
2733                 }
2734
2735                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2736                         idev = ipv6_add_dev(dev);
2737                         if (idev)
2738                                 break;
2739                 }
2740
2741                 /*
2742                  * MTU falled under IPV6_MIN_MTU.
2743                  * Stop IPv6 on this interface.
2744                  */
2745
2746         case NETDEV_DOWN:
2747         case NETDEV_UNREGISTER:
2748                 /*
2749                  *      Remove all addresses from this interface.
2750                  */
2751                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2752                 break;
2753
2754         case NETDEV_CHANGENAME:
2755                 if (idev) {
2756                         snmp6_unregister_dev(idev);
2757                         addrconf_sysctl_unregister(idev);
2758                         addrconf_sysctl_register(idev);
2759                         err = snmp6_register_dev(idev);
2760                         if (err)
2761                                 return notifier_from_errno(err);
2762                 }
2763                 break;
2764
2765         case NETDEV_PRE_TYPE_CHANGE:
2766         case NETDEV_POST_TYPE_CHANGE:
2767                 addrconf_type_change(dev, event);
2768                 break;
2769         }
2770
2771         return NOTIFY_OK;
2772 }
2773
2774 /*
2775  *      addrconf module should be notified of a device going up
2776  */
2777 static struct notifier_block ipv6_dev_notf = {
2778         .notifier_call = addrconf_notify,
2779 };
2780
2781 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2782 {
2783         struct inet6_dev *idev;
2784         ASSERT_RTNL();
2785
2786         idev = __in6_dev_get(dev);
2787
2788         if (event == NETDEV_POST_TYPE_CHANGE)
2789                 ipv6_mc_remap(idev);
2790         else if (event == NETDEV_PRE_TYPE_CHANGE)
2791                 ipv6_mc_unmap(idev);
2792 }
2793
2794 static int addrconf_ifdown(struct net_device *dev, int how)
2795 {
2796         struct net *net = dev_net(dev);
2797         struct inet6_dev *idev;
2798         struct inet6_ifaddr *ifa;
2799         int state, i;
2800
2801         ASSERT_RTNL();
2802
2803         rt6_ifdown(net, dev);
2804         neigh_ifdown(&nd_tbl, dev);
2805
2806         idev = __in6_dev_get(dev);
2807         if (idev == NULL)
2808                 return -ENODEV;
2809
2810         /*
2811          * Step 1: remove reference to ipv6 device from parent device.
2812          *         Do not dev_put!
2813          */
2814         if (how) {
2815                 idev->dead = 1;
2816
2817                 /* protected by rtnl_lock */
2818                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2819
2820                 /* Step 1.5: remove snmp6 entry */
2821                 snmp6_unregister_dev(idev);
2822
2823         }
2824
2825         /* Step 2: clear hash table */
2826         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2827                 struct hlist_head *h = &inet6_addr_lst[i];
2828                 struct hlist_node *n;
2829
2830                 spin_lock_bh(&addrconf_hash_lock);
2831         restart:
2832                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2833                         if (ifa->idev == idev) {
2834                                 hlist_del_init_rcu(&ifa->addr_lst);
2835                                 addrconf_del_timer(ifa);
2836                                 goto restart;
2837                         }
2838                 }
2839                 spin_unlock_bh(&addrconf_hash_lock);
2840         }
2841
2842         write_lock_bh(&idev->lock);
2843
2844         /* Step 2: clear flags for stateless addrconf */
2845         if (!how)
2846                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2847
2848 #ifdef CONFIG_IPV6_PRIVACY
2849         if (how && del_timer(&idev->regen_timer))
2850                 in6_dev_put(idev);
2851
2852         /* Step 3: clear tempaddr list */
2853         while (!list_empty(&idev->tempaddr_list)) {
2854                 ifa = list_first_entry(&idev->tempaddr_list,
2855                                        struct inet6_ifaddr, tmp_list);
2856                 list_del(&ifa->tmp_list);
2857                 write_unlock_bh(&idev->lock);
2858                 spin_lock_bh(&ifa->lock);
2859
2860                 if (ifa->ifpub) {
2861                         in6_ifa_put(ifa->ifpub);
2862                         ifa->ifpub = NULL;
2863                 }
2864                 spin_unlock_bh(&ifa->lock);
2865                 in6_ifa_put(ifa);
2866                 write_lock_bh(&idev->lock);
2867         }
2868 #endif
2869
2870         while (!list_empty(&idev->addr_list)) {
2871                 ifa = list_first_entry(&idev->addr_list,
2872                                        struct inet6_ifaddr, if_list);
2873                 addrconf_del_timer(ifa);
2874
2875                 list_del(&ifa->if_list);
2876
2877                 write_unlock_bh(&idev->lock);
2878
2879                 spin_lock_bh(&ifa->state_lock);
2880                 state = ifa->state;
2881                 ifa->state = INET6_IFADDR_STATE_DEAD;
2882                 spin_unlock_bh(&ifa->state_lock);
2883
2884                 if (state != INET6_IFADDR_STATE_DEAD) {
2885                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2886                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2887                 }
2888                 in6_ifa_put(ifa);
2889
2890                 write_lock_bh(&idev->lock);
2891         }
2892
2893         write_unlock_bh(&idev->lock);
2894
2895         /* Step 5: Discard multicast list */
2896         if (how)
2897                 ipv6_mc_destroy_dev(idev);
2898         else
2899                 ipv6_mc_down(idev);
2900
2901         idev->tstamp = jiffies;
2902
2903         /* Last: Shot the device (if unregistered) */
2904         if (how) {
2905                 addrconf_sysctl_unregister(idev);
2906                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2907                 neigh_ifdown(&nd_tbl, dev);
2908                 in6_dev_put(idev);
2909         }
2910         return 0;
2911 }
2912
2913 static void addrconf_rs_timer(unsigned long data)
2914 {
2915         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2916         struct inet6_dev *idev = ifp->idev;
2917
2918         read_lock(&idev->lock);
2919         if (idev->dead || !(idev->if_flags & IF_READY))
2920                 goto out;
2921
2922         if (idev->cnf.forwarding)
2923                 goto out;
2924
2925         /* Announcement received after solicitation was sent */
2926         if (idev->if_flags & IF_RA_RCVD)
2927                 goto out;
2928
2929         spin_lock(&ifp->lock);
2930         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2931                 /* The wait after the last probe can be shorter */
2932                 addrconf_mod_timer(ifp, AC_RS,
2933                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2934                                    idev->cnf.rtr_solicit_delay :
2935                                    idev->cnf.rtr_solicit_interval);
2936                 spin_unlock(&ifp->lock);
2937
2938                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2939         } else {
2940                 spin_unlock(&ifp->lock);
2941                 /*
2942                  * Note: we do not support deprecated "all on-link"
2943                  * assumption any longer.
2944                  */
2945                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2946                        idev->dev->name);
2947         }
2948
2949 out:
2950         read_unlock(&idev->lock);
2951         in6_ifa_put(ifp);
2952 }
2953
2954 /*
2955  *      Duplicate Address Detection
2956  */
2957 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2958 {
2959         unsigned long rand_num;
2960         struct inet6_dev *idev = ifp->idev;
2961
2962         if (ifp->flags & IFA_F_OPTIMISTIC)
2963                 rand_num = 0;
2964         else
2965                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2966
2967         ifp->probes = idev->cnf.dad_transmits;
2968         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2969 }
2970
2971 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2972 {
2973         struct inet6_dev *idev = ifp->idev;
2974         struct net_device *dev = idev->dev;
2975
2976         addrconf_join_solict(dev, &ifp->addr);
2977
2978         net_srandom(ifp->addr.s6_addr32[3]);
2979
2980         read_lock_bh(&idev->lock);
2981         spin_lock(&ifp->lock);
2982         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2983                 goto out;
2984
2985         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2986             idev->cnf.accept_dad < 1 ||
2987             !(ifp->flags&IFA_F_TENTATIVE) ||
2988             ifp->flags & IFA_F_NODAD) {
2989                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2990                 spin_unlock(&ifp->lock);
2991                 read_unlock_bh(&idev->lock);
2992
2993                 addrconf_dad_completed(ifp);
2994                 return;
2995         }
2996
2997         if (!(idev->if_flags & IF_READY)) {
2998                 spin_unlock(&ifp->lock);
2999                 read_unlock_bh(&idev->lock);
3000                 /*
3001                  * If the device is not ready:
3002                  * - keep it tentative if it is a permanent address.
3003                  * - otherwise, kill it.
3004                  */
3005                 in6_ifa_hold(ifp);
3006                 addrconf_dad_stop(ifp, 0);
3007                 return;
3008         }
3009
3010         /*
3011          * Optimistic nodes can start receiving
3012          * Frames right away
3013          */
3014         if (ifp->flags & IFA_F_OPTIMISTIC)
3015                 ip6_ins_rt(ifp->rt);
3016
3017         addrconf_dad_kick(ifp);
3018 out:
3019         spin_unlock(&ifp->lock);
3020         read_unlock_bh(&idev->lock);
3021 }
3022
3023 static void addrconf_dad_timer(unsigned long data)
3024 {
3025         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3026         struct inet6_dev *idev = ifp->idev;
3027         struct in6_addr mcaddr;
3028
3029         if (!ifp->probes && addrconf_dad_end(ifp))
3030                 goto out;
3031
3032         read_lock(&idev->lock);
3033         if (idev->dead || !(idev->if_flags & IF_READY)) {
3034                 read_unlock(&idev->lock);
3035                 goto out;
3036         }
3037
3038         spin_lock(&ifp->lock);
3039         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3040                 spin_unlock(&ifp->lock);
3041                 read_unlock(&idev->lock);
3042                 goto out;
3043         }
3044
3045         if (ifp->probes == 0) {
3046                 /*
3047                  * DAD was successful
3048                  */
3049
3050                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3051                 spin_unlock(&ifp->lock);
3052                 read_unlock(&idev->lock);
3053
3054                 addrconf_dad_completed(ifp);
3055
3056                 goto out;
3057         }
3058
3059         ifp->probes--;
3060         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3061         spin_unlock(&ifp->lock);
3062         read_unlock(&idev->lock);
3063
3064         /* send a neighbour solicitation for our addr */
3065         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3066         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3067 out:
3068         in6_ifa_put(ifp);
3069 }
3070
3071 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3072 {
3073         struct net_device *dev = ifp->idev->dev;
3074
3075         /*
3076          *      Configure the address for reception. Now it is valid.
3077          */
3078
3079         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3080
3081         /* If added prefix is link local and we are prepared to process
3082            router advertisements, start sending router solicitations.
3083          */
3084
3085         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3086              ifp->idev->cnf.accept_ra == 2) &&
3087             ifp->idev->cnf.rtr_solicits > 0 &&
3088             (dev->flags&IFF_LOOPBACK) == 0 &&
3089             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3090                 /*
3091                  *      If a host as already performed a random delay
3092                  *      [...] as part of DAD [...] there is no need
3093                  *      to delay again before sending the first RS
3094                  */
3095                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3096
3097                 spin_lock_bh(&ifp->lock);
3098                 ifp->probes = 1;
3099                 ifp->idev->if_flags |= IF_RS_SENT;
3100                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3101                 spin_unlock_bh(&ifp->lock);
3102         }
3103 }
3104
3105 static void addrconf_dad_run(struct inet6_dev *idev)
3106 {
3107         struct inet6_ifaddr *ifp;
3108
3109         read_lock_bh(&idev->lock);
3110         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3111                 spin_lock(&ifp->lock);
3112                 if (ifp->flags & IFA_F_TENTATIVE &&
3113                     ifp->state == INET6_IFADDR_STATE_DAD)
3114                         addrconf_dad_kick(ifp);
3115                 spin_unlock(&ifp->lock);
3116         }
3117         read_unlock_bh(&idev->lock);
3118 }
3119
3120 #ifdef CONFIG_PROC_FS
3121 struct if6_iter_state {
3122         struct seq_net_private p;
3123         int bucket;
3124 };
3125
3126 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3127 {
3128         struct inet6_ifaddr *ifa = NULL;
3129         struct if6_iter_state *state = seq->private;
3130         struct net *net = seq_file_net(seq);
3131
3132         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3133                 struct hlist_node *n;
3134                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3135                                          addr_lst)
3136                         if (net_eq(dev_net(ifa->idev->dev), net))
3137                                 return ifa;
3138         }
3139         return NULL;
3140 }
3141
3142 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3143                                          struct inet6_ifaddr *ifa)
3144 {
3145         struct if6_iter_state *state = seq->private;
3146         struct net *net = seq_file_net(seq);
3147         struct hlist_node *n = &ifa->addr_lst;
3148
3149         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3150                 if (net_eq(dev_net(ifa->idev->dev), net))
3151                         return ifa;
3152
3153         while (++state->bucket < IN6_ADDR_HSIZE) {
3154                 hlist_for_each_entry_rcu_bh(ifa, n,
3155                                      &inet6_addr_lst[state->bucket], addr_lst) {
3156                         if (net_eq(dev_net(ifa->idev->dev), net))
3157                                 return ifa;
3158                 }
3159         }
3160
3161         return NULL;
3162 }
3163
3164 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3165 {
3166         struct inet6_ifaddr *ifa = if6_get_first(seq);
3167
3168         if (ifa)
3169                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3170                         --pos;
3171         return pos ? NULL : ifa;
3172 }
3173
3174 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3175         __acquires(rcu_bh)
3176 {
3177         rcu_read_lock_bh();
3178         return if6_get_idx(seq, *pos);
3179 }
3180
3181 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3182 {
3183         struct inet6_ifaddr *ifa;
3184
3185         ifa = if6_get_next(seq, v);
3186         ++*pos;
3187         return ifa;
3188 }
3189
3190 static void if6_seq_stop(struct seq_file *seq, void *v)
3191         __releases(rcu_bh)
3192 {
3193         rcu_read_unlock_bh();
3194 }
3195
3196 static int if6_seq_show(struct seq_file *seq, void *v)
3197 {
3198         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3199         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3200                    &ifp->addr,
3201                    ifp->idev->dev->ifindex,
3202                    ifp->prefix_len,
3203                    ifp->scope,
3204                    ifp->flags,
3205                    ifp->idev->dev->name);
3206         return 0;
3207 }
3208
3209 static const struct seq_operations if6_seq_ops = {
3210         .start  = if6_seq_start,
3211         .next   = if6_seq_next,
3212         .show   = if6_seq_show,
3213         .stop   = if6_seq_stop,
3214 };
3215
3216 static int if6_seq_open(struct inode *inode, struct file *file)
3217 {
3218         return seq_open_net(inode, file, &if6_seq_ops,
3219                             sizeof(struct if6_iter_state));
3220 }
3221
3222 static const struct file_operations if6_fops = {
3223         .owner          = THIS_MODULE,
3224         .open           = if6_seq_open,
3225         .read           = seq_read,
3226         .llseek         = seq_lseek,
3227         .release        = seq_release_net,
3228 };
3229
3230 static int __net_init if6_proc_net_init(struct net *net)
3231 {
3232         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3233                 return -ENOMEM;
3234         return 0;
3235 }
3236
3237 static void __net_exit if6_proc_net_exit(struct net *net)
3238 {
3239        proc_net_remove(net, "if_inet6");
3240 }
3241
3242 static struct pernet_operations if6_proc_net_ops = {
3243        .init = if6_proc_net_init,
3244        .exit = if6_proc_net_exit,
3245 };
3246
3247 int __init if6_proc_init(void)
3248 {
3249         return register_pernet_subsys(&if6_proc_net_ops);
3250 }
3251
3252 void if6_proc_exit(void)
3253 {
3254         unregister_pernet_subsys(&if6_proc_net_ops);
3255 }
3256 #endif  /* CONFIG_PROC_FS */
3257
3258 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3259 /* Check if address is a home address configured on any interface. */
3260 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3261 {
3262         int ret = 0;
3263         struct inet6_ifaddr *ifp = NULL;
3264         struct hlist_node *n;
3265         unsigned int hash = ipv6_addr_hash(addr);
3266
3267         rcu_read_lock_bh();
3268         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3269                 if (!net_eq(dev_net(ifp->idev->dev), net))
3270                         continue;
3271                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3272                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3273                         ret = 1;
3274                         break;
3275                 }
3276         }
3277         rcu_read_unlock_bh();
3278         return ret;
3279 }
3280 #endif
3281
3282 /*
3283  *      Periodic address status verification
3284  */
3285
3286 static void addrconf_verify(unsigned long foo)
3287 {
3288         unsigned long now, next, next_sec, next_sched;
3289         struct inet6_ifaddr *ifp;
3290         struct hlist_node *node;
3291         int i;
3292
3293         rcu_read_lock_bh();
3294         spin_lock(&addrconf_verify_lock);
3295         now = jiffies;
3296         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3297
3298         del_timer(&addr_chk_timer);
3299
3300         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3301 restart:
3302                 hlist_for_each_entry_rcu_bh(ifp, node,
3303                                          &inet6_addr_lst[i], addr_lst) {
3304                         unsigned long age;
3305
3306                         if (ifp->flags & IFA_F_PERMANENT)
3307                                 continue;
3308
3309                         spin_lock(&ifp->lock);
3310                         /* We try to batch several events at once. */
3311                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3312
3313                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3314                             age >= ifp->valid_lft) {
3315                                 spin_unlock(&ifp->lock);
3316                                 in6_ifa_hold(ifp);
3317                                 ipv6_del_addr(ifp);
3318                                 goto restart;
3319                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3320                                 spin_unlock(&ifp->lock);
3321                                 continue;
3322                         } else if (age >= ifp->prefered_lft) {
3323                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3324                                 int deprecate = 0;
3325
3326                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3327                                         deprecate = 1;
3328                                         ifp->flags |= IFA_F_DEPRECATED;
3329                                 }
3330
3331                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3332                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3333
3334                                 spin_unlock(&ifp->lock);
3335
3336                                 if (deprecate) {
3337                                         in6_ifa_hold(ifp);
3338
3339                                         ipv6_ifa_notify(0, ifp);
3340                                         in6_ifa_put(ifp);
3341                                         goto restart;
3342                                 }
3343 #ifdef CONFIG_IPV6_PRIVACY
3344                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3345                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3346                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3347                                         ifp->idev->cnf.dad_transmits *
3348                                         ifp->idev->nd_parms->retrans_time / HZ;
3349
3350                                 if (age >= ifp->prefered_lft - regen_advance) {
3351                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3352                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3353                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3354                                         if (!ifp->regen_count && ifpub) {
3355                                                 ifp->regen_count++;
3356                                                 in6_ifa_hold(ifp);
3357                                                 in6_ifa_hold(ifpub);
3358                                                 spin_unlock(&ifp->lock);
3359
3360                                                 spin_lock(&ifpub->lock);
3361                                                 ifpub->regen_count = 0;
3362                                                 spin_unlock(&ifpub->lock);
3363                                                 ipv6_create_tempaddr(ifpub, ifp);
3364                                                 in6_ifa_put(ifpub);
3365                                                 in6_ifa_put(ifp);
3366                                                 goto restart;
3367                                         }
3368                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3369                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3370                                 spin_unlock(&ifp->lock);
3371 #endif
3372                         } else {
3373                                 /* ifp->prefered_lft <= ifp->valid_lft */
3374                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3375                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3376                                 spin_unlock(&ifp->lock);
3377                         }
3378                 }
3379         }
3380
3381         next_sec = round_jiffies_up(next);
3382         next_sched = next;
3383
3384         /* If rounded timeout is accurate enough, accept it. */
3385         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3386                 next_sched = next_sec;
3387
3388         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3389         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3390                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3391
3392         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3393               now, next, next_sec, next_sched));
3394
3395         addr_chk_timer.expires = next_sched;
3396         add_timer(&addr_chk_timer);
3397         spin_unlock(&addrconf_verify_lock);
3398         rcu_read_unlock_bh();
3399 }
3400
3401 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3402 {
3403         struct in6_addr *pfx = NULL;
3404
3405         if (addr)
3406                 pfx = nla_data(addr);
3407
3408         if (local) {
3409                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3410                         pfx = NULL;
3411                 else
3412                         pfx = nla_data(local);
3413         }
3414
3415         return pfx;
3416 }
3417
3418 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3419         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3420         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3421         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3422 };
3423
3424 static int
3425 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3426 {
3427         struct net *net = sock_net(skb->sk);
3428         struct ifaddrmsg *ifm;
3429         struct nlattr *tb[IFA_MAX+1];
3430         struct in6_addr *pfx;
3431         int err;
3432
3433         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3434         if (err < 0)
3435                 return err;
3436
3437         ifm = nlmsg_data(nlh);
3438         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3439         if (pfx == NULL)
3440                 return -EINVAL;
3441
3442         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3443 }
3444
3445 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3446                              u32 prefered_lft, u32 valid_lft)
3447 {
3448         u32 flags;
3449         clock_t expires;
3450         unsigned long timeout;
3451
3452         if (!valid_lft || (prefered_lft > valid_lft))
3453                 return -EINVAL;
3454
3455         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3456         if (addrconf_finite_timeout(timeout)) {
3457                 expires = jiffies_to_clock_t(timeout * HZ);
3458                 valid_lft = timeout;
3459                 flags = RTF_EXPIRES;
3460         } else {
3461                 expires = 0;
3462                 flags = 0;
3463                 ifa_flags |= IFA_F_PERMANENT;
3464         }
3465
3466         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3467         if (addrconf_finite_timeout(timeout)) {
3468                 if (timeout == 0)
3469                         ifa_flags |= IFA_F_DEPRECATED;
3470                 prefered_lft = timeout;
3471         }
3472
3473         spin_lock_bh(&ifp->lock);
3474         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3475         ifp->tstamp = jiffies;
3476         ifp->valid_lft = valid_lft;
3477         ifp->prefered_lft = prefered_lft;
3478
3479         spin_unlock_bh(&ifp->lock);
3480         if (!(ifp->flags&IFA_F_TENTATIVE))
3481                 ipv6_ifa_notify(0, ifp);
3482
3483         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3484                               expires, flags);
3485         addrconf_verify(0);
3486
3487         return 0;
3488 }
3489
3490 static int
3491 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3492 {
3493         struct net *net = sock_net(skb->sk);
3494         struct ifaddrmsg *ifm;
3495         struct nlattr *tb[IFA_MAX+1];
3496         struct in6_addr *pfx;
3497         struct inet6_ifaddr *ifa;
3498         struct net_device *dev;
3499         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3500         u8 ifa_flags;
3501         int err;
3502
3503         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3504         if (err < 0)
3505                 return err;
3506
3507         ifm = nlmsg_data(nlh);
3508         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3509         if (pfx == NULL)
3510                 return -EINVAL;
3511
3512         if (tb[IFA_CACHEINFO]) {
3513                 struct ifa_cacheinfo *ci;
3514
3515                 ci = nla_data(tb[IFA_CACHEINFO]);
3516                 valid_lft = ci->ifa_valid;
3517                 preferred_lft = ci->ifa_prefered;
3518         } else {
3519                 preferred_lft = INFINITY_LIFE_TIME;
3520                 valid_lft = INFINITY_LIFE_TIME;
3521         }
3522
3523         dev =  __dev_get_by_index(net, ifm->ifa_index);
3524         if (dev == NULL)
3525                 return -ENODEV;
3526
3527         /* We ignore other flags so far. */
3528         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3529
3530         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3531         if (ifa == NULL) {
3532                 /*
3533                  * It would be best to check for !NLM_F_CREATE here but
3534                  * userspace alreay relies on not having to provide this.
3535                  */
3536                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3537                                       ifm->ifa_prefixlen, ifa_flags,
3538                                       preferred_lft, valid_lft);
3539         }
3540
3541         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3542             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3543                 err = -EEXIST;
3544         else
3545                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3546
3547         in6_ifa_put(ifa);
3548
3549         return err;
3550 }
3551
3552 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3553                           u8 scope, int ifindex)
3554 {
3555         struct ifaddrmsg *ifm;
3556
3557         ifm = nlmsg_data(nlh);
3558         ifm->ifa_family = AF_INET6;
3559         ifm->ifa_prefixlen = prefixlen;
3560         ifm->ifa_flags = flags;
3561         ifm->ifa_scope = scope;
3562         ifm->ifa_index = ifindex;
3563 }
3564
3565 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3566                          unsigned long tstamp, u32 preferred, u32 valid)
3567 {
3568         struct ifa_cacheinfo ci;
3569
3570         ci.cstamp = cstamp_delta(cstamp);
3571         ci.tstamp = cstamp_delta(tstamp);
3572         ci.ifa_prefered = preferred;
3573         ci.ifa_valid = valid;
3574
3575         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3576 }
3577
3578 static inline int rt_scope(int ifa_scope)
3579 {
3580         if (ifa_scope & IFA_HOST)
3581                 return RT_SCOPE_HOST;
3582         else if (ifa_scope & IFA_LINK)
3583                 return RT_SCOPE_LINK;
3584         else if (ifa_scope & IFA_SITE)
3585                 return RT_SCOPE_SITE;
3586         else
3587                 return RT_SCOPE_UNIVERSE;
3588 }
3589
3590 static inline int inet6_ifaddr_msgsize(void)
3591 {
3592         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3593                + nla_total_size(16) /* IFA_ADDRESS */
3594                + nla_total_size(sizeof(struct ifa_cacheinfo));
3595 }
3596
3597 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3598                              u32 pid, u32 seq, int event, unsigned int flags)
3599 {
3600         struct nlmsghdr  *nlh;
3601         u32 preferred, valid;
3602
3603         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3604         if (nlh == NULL)
3605                 return -EMSGSIZE;
3606
3607         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3608                       ifa->idev->dev->ifindex);
3609
3610         if (!(ifa->flags&IFA_F_PERMANENT)) {
3611                 preferred = ifa->prefered_lft;
3612                 valid = ifa->valid_lft;
3613                 if (preferred != INFINITY_LIFE_TIME) {
3614                         long tval = (jiffies - ifa->tstamp)/HZ;
3615                         if (preferred > tval)
3616                                 preferred -= tval;
3617                         else
3618                                 preferred = 0;
3619                         if (valid != INFINITY_LIFE_TIME) {
3620                                 if (valid > tval)
3621                                         valid -= tval;
3622                                 else
3623                                         valid = 0;
3624                         }
3625                 }
3626         } else {
3627                 preferred = INFINITY_LIFE_TIME;
3628                 valid = INFINITY_LIFE_TIME;
3629         }
3630
3631         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3632             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3633                 nlmsg_cancel(skb, nlh);
3634                 return -EMSGSIZE;
3635         }
3636
3637         return nlmsg_end(skb, nlh);
3638 }
3639
3640 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3641                                 u32 pid, u32 seq, int event, u16 flags)
3642 {
3643         struct nlmsghdr  *nlh;
3644         u8 scope = RT_SCOPE_UNIVERSE;
3645         int ifindex = ifmca->idev->dev->ifindex;
3646
3647         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3648                 scope = RT_SCOPE_SITE;
3649
3650         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3651         if (nlh == NULL)
3652                 return -EMSGSIZE;
3653
3654         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3655         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3656             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3657                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3658                 nlmsg_cancel(skb, nlh);
3659                 return -EMSGSIZE;
3660         }
3661
3662         return nlmsg_end(skb, nlh);
3663 }
3664
3665 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3666                                 u32 pid, u32 seq, int event, unsigned int flags)
3667 {
3668         struct nlmsghdr  *nlh;
3669         u8 scope = RT_SCOPE_UNIVERSE;
3670         int ifindex = ifaca->aca_idev->dev->ifindex;
3671
3672         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3673                 scope = RT_SCOPE_SITE;
3674
3675         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3676         if (nlh == NULL)
3677                 return -EMSGSIZE;
3678
3679         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3680         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3681             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3682                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3683                 nlmsg_cancel(skb, nlh);
3684                 return -EMSGSIZE;
3685         }
3686
3687         return nlmsg_end(skb, nlh);
3688 }
3689
3690 enum addr_type_t {
3691         UNICAST_ADDR,
3692         MULTICAST_ADDR,
3693         ANYCAST_ADDR,
3694 };
3695
3696 /* called with rcu_read_lock() */
3697 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3698                           struct netlink_callback *cb, enum addr_type_t type,
3699                           int s_ip_idx, int *p_ip_idx)
3700 {
3701         struct ifmcaddr6 *ifmca;
3702         struct ifacaddr6 *ifaca;
3703         int err = 1;
3704         int ip_idx = *p_ip_idx;
3705
3706         read_lock_bh(&idev->lock);
3707         switch (type) {
3708         case UNICAST_ADDR: {
3709                 struct inet6_ifaddr *ifa;
3710
3711                 /* unicast address incl. temp addr */
3712                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3713                         if (++ip_idx < s_ip_idx)
3714                                 continue;
3715                         err = inet6_fill_ifaddr(skb, ifa,
3716                                                 NETLINK_CB(cb->skb).pid,
3717                                                 cb->nlh->nlmsg_seq,
3718                                                 RTM_NEWADDR,
3719                                                 NLM_F_MULTI);
3720                         if (err <= 0)
3721                                 break;
3722                 }
3723                 break;
3724         }
3725         case MULTICAST_ADDR:
3726                 /* multicast address */
3727                 for (ifmca = idev->mc_list; ifmca;
3728                      ifmca = ifmca->next, ip_idx++) {
3729                         if (ip_idx < s_ip_idx)
3730                                 continue;
3731                         err = inet6_fill_ifmcaddr(skb, ifmca,
3732                                                   NETLINK_CB(cb->skb).pid,
3733                                                   cb->nlh->nlmsg_seq,
3734                                                   RTM_GETMULTICAST,
3735                                                   NLM_F_MULTI);
3736                         if (err <= 0)
3737                                 break;
3738                 }
3739                 break;
3740         case ANYCAST_ADDR:
3741                 /* anycast address */
3742                 for (ifaca = idev->ac_list; ifaca;
3743                      ifaca = ifaca->aca_next, ip_idx++) {
3744                         if (ip_idx < s_ip_idx)
3745                                 continue;
3746                         err = inet6_fill_ifacaddr(skb, ifaca,
3747                                                   NETLINK_CB(cb->skb).pid,
3748                                                   cb->nlh->nlmsg_seq,
3749                                                   RTM_GETANYCAST,
3750                                                   NLM_F_MULTI);
3751                         if (err <= 0)
3752                                 break;
3753                 }
3754                 break;
3755         default:
3756                 break;
3757         }
3758         read_unlock_bh(&idev->lock);
3759         *p_ip_idx = ip_idx;
3760         return err;
3761 }
3762
3763 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3764                            enum addr_type_t type)
3765 {
3766         struct net *net = sock_net(skb->sk);
3767         int h, s_h;
3768         int idx, ip_idx;
3769         int s_idx, s_ip_idx;
3770         struct net_device *dev;
3771         struct inet6_dev *idev;
3772         struct hlist_head *head;
3773         struct hlist_node *node;
3774
3775         s_h = cb->args[0];
3776         s_idx = idx = cb->args[1];
3777         s_ip_idx = ip_idx = cb->args[2];
3778
3779         rcu_read_lock();
3780         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3781                 idx = 0;
3782                 head = &net->dev_index_head[h];
3783                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3784                         if (idx < s_idx)
3785                                 goto cont;
3786                         if (h > s_h || idx > s_idx)
3787                                 s_ip_idx = 0;
3788                         ip_idx = 0;
3789                         idev = __in6_dev_get(dev);
3790                         if (!idev)
3791                                 goto cont;
3792
3793                         if (in6_dump_addrs(idev, skb, cb, type,
3794                                            s_ip_idx, &ip_idx) <= 0)
3795                                 goto done;
3796 cont:
3797                         idx++;
3798                 }
3799         }
3800 done:
3801         rcu_read_unlock();
3802         cb->args[0] = h;
3803         cb->args[1] = idx;
3804         cb->args[2] = ip_idx;
3805
3806         return skb->len;
3807 }
3808
3809 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3810 {
3811         enum addr_type_t type = UNICAST_ADDR;
3812
3813         return inet6_dump_addr(skb, cb, type);
3814 }
3815
3816 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3817 {
3818         enum addr_type_t type = MULTICAST_ADDR;
3819
3820         return inet6_dump_addr(skb, cb, type);
3821 }
3822
3823
3824 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3825 {
3826         enum addr_type_t type = ANYCAST_ADDR;
3827
3828         return inet6_dump_addr(skb, cb, type);
3829 }
3830
3831 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3832                              void *arg)
3833 {
3834         struct net *net = sock_net(in_skb->sk);
3835         struct ifaddrmsg *ifm;
3836         struct nlattr *tb[IFA_MAX+1];
3837         struct in6_addr *addr = NULL;
3838         struct net_device *dev = NULL;
3839         struct inet6_ifaddr *ifa;
3840         struct sk_buff *skb;
3841         int err;
3842
3843         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3844         if (err < 0)
3845                 goto errout;
3846
3847         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3848         if (addr == NULL) {
3849                 err = -EINVAL;
3850                 goto errout;
3851         }
3852
3853         ifm = nlmsg_data(nlh);
3854         if (ifm->ifa_index)
3855                 dev = __dev_get_by_index(net, ifm->ifa_index);
3856
3857         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3858         if (!ifa) {
3859                 err = -EADDRNOTAVAIL;
3860                 goto errout;
3861         }
3862
3863         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3864         if (!skb) {
3865                 err = -ENOBUFS;
3866                 goto errout_ifa;
3867         }
3868
3869         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3870                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3871         if (err < 0) {
3872                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3873                 WARN_ON(err == -EMSGSIZE);
3874                 kfree_skb(skb);
3875                 goto errout_ifa;
3876         }
3877         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3878 errout_ifa:
3879         in6_ifa_put(ifa);
3880 errout:
3881         return err;
3882 }
3883
3884 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3885 {
3886         struct sk_buff *skb;
3887         struct net *net = dev_net(ifa->idev->dev);
3888         int err = -ENOBUFS;
3889
3890         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3891         if (skb == NULL)
3892                 goto errout;
3893
3894         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3895         if (err < 0) {
3896                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3897                 WARN_ON(err == -EMSGSIZE);
3898                 kfree_skb(skb);
3899                 goto errout;
3900         }
3901         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3902         return;
3903 errout:
3904         if (err < 0)
3905                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3906 }
3907
3908 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3909                                 __s32 *array, int bytes)
3910 {
3911         BUG_ON(bytes < (DEVCONF_MAX * 4));
3912
3913         memset(array, 0, bytes);
3914         array[DEVCONF_FORWARDING] = cnf->forwarding;
3915         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3916         array[DEVCONF_MTU6] = cnf->mtu6;
3917         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3918         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3919         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3920         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3921         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3922         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3923                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3924         array[DEVCONF_RTR_SOLICIT_DELAY] =
3925                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3926         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3927 #ifdef CONFIG_IPV6_PRIVACY
3928         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3929         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3930         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3931         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3932         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3933 #endif
3934         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3935         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3936         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3937 #ifdef CONFIG_IPV6_ROUTER_PREF
3938         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3939         array[DEVCONF_RTR_PROBE_INTERVAL] =
3940                 jiffies_to_msecs(cnf->rtr_probe_interval);
3941 #ifdef CONFIG_IPV6_ROUTE_INFO
3942         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3943 #endif
3944 #endif
3945         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3946         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3947 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3948         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3949 #endif
3950 #ifdef CONFIG_IPV6_MROUTE
3951         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3952 #endif
3953         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3954         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3955         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3956 }
3957
3958 static inline size_t inet6_ifla6_size(void)
3959 {
3960         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3961              + nla_total_size(sizeof(struct ifla_cacheinfo))
3962              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3963              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3964              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3965 }
3966
3967 static inline size_t inet6_if_nlmsg_size(void)
3968 {
3969         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3970                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3971                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3972                + nla_total_size(4) /* IFLA_MTU */
3973                + nla_total_size(4) /* IFLA_LINK */
3974                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3975 }
3976
3977 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3978                                       int items, int bytes)
3979 {
3980         int i;
3981         int pad = bytes - sizeof(u64) * items;
3982         BUG_ON(pad < 0);
3983
3984         /* Use put_unaligned() because stats may not be aligned for u64. */
3985         put_unaligned(items, &stats[0]);
3986         for (i = 1; i < items; i++)
3987                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3988
3989         memset(&stats[items], 0, pad);
3990 }
3991
3992 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3993                                       int items, int bytes, size_t syncpoff)
3994 {
3995         int i;
3996         int pad = bytes - sizeof(u64) * items;
3997         BUG_ON(pad < 0);
3998
3999         /* Use put_unaligned() because stats may not be aligned for u64. */
4000         put_unaligned(items, &stats[0]);
4001         for (i = 1; i < items; i++)
4002                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4003
4004         memset(&stats[items], 0, pad);
4005 }
4006
4007 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4008                              int bytes)
4009 {
4010         switch (attrtype) {
4011         case IFLA_INET6_STATS:
4012                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4013                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4014                 break;
4015         case IFLA_INET6_ICMP6STATS:
4016                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4017                 break;
4018         }
4019 }
4020
4021 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4022 {
4023         struct nlattr *nla;
4024         struct ifla_cacheinfo ci;
4025
4026         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
4027
4028         ci.max_reasm_len = IPV6_MAXPLEN;
4029         ci.tstamp = cstamp_delta(idev->tstamp);
4030         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4031         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4032         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
4033
4034         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4035         if (nla == NULL)
4036                 goto nla_put_failure;
4037         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4038
4039         /* XXX - MC not implemented */
4040
4041         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4042         if (nla == NULL)
4043                 goto nla_put_failure;
4044         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4045
4046         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4047         if (nla == NULL)
4048                 goto nla_put_failure;
4049         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4050
4051         return 0;
4052
4053 nla_put_failure:
4054         return -EMSGSIZE;
4055 }
4056
4057 static size_t inet6_get_link_af_size(const struct net_device *dev)
4058 {
4059         if (!__in6_dev_get(dev))
4060                 return 0;
4061
4062         return inet6_ifla6_size();
4063 }
4064
4065 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4066 {
4067         struct inet6_dev *idev = __in6_dev_get(dev);
4068
4069         if (!idev)
4070                 return -ENODATA;
4071
4072         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4073                 return -EMSGSIZE;
4074
4075         return 0;
4076 }
4077
4078 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4079                              u32 pid, u32 seq, int event, unsigned int flags)
4080 {
4081         struct net_device *dev = idev->dev;
4082         struct ifinfomsg *hdr;
4083         struct nlmsghdr *nlh;
4084         void *protoinfo;
4085
4086         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4087         if (nlh == NULL)
4088                 return -EMSGSIZE;
4089
4090         hdr = nlmsg_data(nlh);
4091         hdr->ifi_family = AF_INET6;
4092         hdr->__ifi_pad = 0;
4093         hdr->ifi_type = dev->type;
4094         hdr->ifi_index = dev->ifindex;
4095         hdr->ifi_flags = dev_get_flags(dev);
4096         hdr->ifi_change = 0;
4097
4098         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4099
4100         if (dev->addr_len)
4101                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4102
4103         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4104         if (dev->ifindex != dev->iflink)
4105                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4106
4107         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4108         if (protoinfo == NULL)
4109                 goto nla_put_failure;
4110
4111         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4112                 goto nla_put_failure;
4113
4114         nla_nest_end(skb, protoinfo);
4115         return nlmsg_end(skb, nlh);
4116
4117 nla_put_failure:
4118         nlmsg_cancel(skb, nlh);
4119         return -EMSGSIZE;
4120 }
4121
4122 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4123 {
4124         struct net *net = sock_net(skb->sk);
4125         int h, s_h;
4126         int idx = 0, s_idx;
4127         struct net_device *dev;
4128         struct inet6_dev *idev;
4129         struct hlist_head *head;
4130         struct hlist_node *node;
4131
4132         s_h = cb->args[0];
4133         s_idx = cb->args[1];
4134
4135         rcu_read_lock();
4136         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4137                 idx = 0;
4138                 head = &net->dev_index_head[h];
4139                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4140                         if (idx < s_idx)
4141                                 goto cont;
4142                         idev = __in6_dev_get(dev);
4143                         if (!idev)
4144                                 goto cont;
4145                         if (inet6_fill_ifinfo(skb, idev,
4146                                               NETLINK_CB(cb->skb).pid,
4147                                               cb->nlh->nlmsg_seq,
4148                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4149                                 goto out;
4150 cont:
4151                         idx++;
4152                 }
4153         }
4154 out:
4155         rcu_read_unlock();
4156         cb->args[1] = idx;
4157         cb->args[0] = h;
4158
4159         return skb->len;
4160 }
4161
4162 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4163 {
4164         struct sk_buff *skb;
4165         struct net *net = dev_net(idev->dev);
4166         int err = -ENOBUFS;
4167
4168         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4169         if (skb == NULL)
4170                 goto errout;
4171
4172         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4173         if (err < 0) {
4174                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4175                 WARN_ON(err == -EMSGSIZE);
4176                 kfree_skb(skb);
4177                 goto errout;
4178         }
4179         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4180         return;
4181 errout:
4182         if (err < 0)
4183                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4184 }
4185
4186 static inline size_t inet6_prefix_nlmsg_size(void)
4187 {
4188         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4189                + nla_total_size(sizeof(struct in6_addr))
4190                + nla_total_size(sizeof(struct prefix_cacheinfo));
4191 }
4192
4193 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4194                              struct prefix_info *pinfo, u32 pid, u32 seq,
4195                              int event, unsigned int flags)
4196 {
4197         struct prefixmsg *pmsg;
4198         struct nlmsghdr *nlh;
4199         struct prefix_cacheinfo ci;
4200
4201         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4202         if (nlh == NULL)
4203                 return -EMSGSIZE;
4204
4205         pmsg = nlmsg_data(nlh);
4206         pmsg->prefix_family = AF_INET6;
4207         pmsg->prefix_pad1 = 0;
4208         pmsg->prefix_pad2 = 0;
4209         pmsg->prefix_ifindex = idev->dev->ifindex;
4210         pmsg->prefix_len = pinfo->prefix_len;
4211         pmsg->prefix_type = pinfo->type;
4212         pmsg->prefix_pad3 = 0;
4213         pmsg->prefix_flags = 0;
4214         if (pinfo->onlink)
4215                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4216         if (pinfo->autoconf)
4217                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4218
4219         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4220
4221         ci.preferred_time = ntohl(pinfo->prefered);
4222         ci.valid_time = ntohl(pinfo->valid);
4223         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4224
4225         return nlmsg_end(skb, nlh);
4226
4227 nla_put_failure:
4228         nlmsg_cancel(skb, nlh);
4229         return -EMSGSIZE;
4230 }
4231
4232 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4233                          struct prefix_info *pinfo)
4234 {
4235         struct sk_buff *skb;
4236         struct net *net = dev_net(idev->dev);
4237         int err = -ENOBUFS;
4238
4239         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4240         if (skb == NULL)
4241                 goto errout;
4242
4243         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4244         if (err < 0) {
4245                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4246                 WARN_ON(err == -EMSGSIZE);
4247                 kfree_skb(skb);
4248                 goto errout;
4249         }
4250         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4251         return;
4252 errout:
4253         if (err < 0)
4254                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4255 }
4256
4257 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4258 {
4259         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4260
4261         switch (event) {
4262         case RTM_NEWADDR:
4263                 /*
4264                  * If the address was optimistic
4265                  * we inserted the route at the start of
4266                  * our DAD process, so we don't need
4267                  * to do it again
4268                  */
4269                 if (!(ifp->rt->rt6i_node))
4270                         ip6_ins_rt(ifp->rt);
4271                 if (ifp->idev->cnf.forwarding)
4272                         addrconf_join_anycast(ifp);
4273                 break;
4274         case RTM_DELADDR:
4275                 if (ifp->idev->cnf.forwarding)
4276                         addrconf_leave_anycast(ifp);
4277                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4278                 dst_hold(&ifp->rt->dst);
4279
4280                 if (ip6_del_rt(ifp->rt))
4281                         dst_free(&ifp->rt->dst);
4282                 break;
4283         }
4284 }
4285
4286 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4287 {
4288         rcu_read_lock_bh();
4289         if (likely(ifp->idev->dead == 0))
4290                 __ipv6_ifa_notify(event, ifp);
4291         rcu_read_unlock_bh();
4292 }
4293
4294 #ifdef CONFIG_SYSCTL
4295
4296 static
4297 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4298                            void __user *buffer, size_t *lenp, loff_t *ppos)
4299 {
4300         int *valp = ctl->data;
4301         int val = *valp;
4302         loff_t pos = *ppos;
4303         int ret;
4304
4305         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4306
4307         if (write)
4308                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4309         if (ret)
4310                 *ppos = pos;
4311         return ret;
4312 }
4313
4314 static void dev_disable_change(struct inet6_dev *idev)
4315 {
4316         if (!idev || !idev->dev)
4317                 return;
4318
4319         if (idev->cnf.disable_ipv6)
4320                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4321         else
4322                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4323 }
4324
4325 static void addrconf_disable_change(struct net *net, __s32 newf)
4326 {
4327         struct net_device *dev;
4328         struct inet6_dev *idev;
4329
4330         rcu_read_lock();
4331         for_each_netdev_rcu(net, dev) {
4332                 idev = __in6_dev_get(dev);
4333                 if (idev) {
4334                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4335                         idev->cnf.disable_ipv6 = newf;
4336                         if (changed)
4337                                 dev_disable_change(idev);
4338                 }
4339         }
4340         rcu_read_unlock();
4341 }
4342
4343 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4344 {
4345         struct net *net;
4346
4347         net = (struct net *)table->extra2;
4348
4349         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4350                 return 0;
4351
4352         if (!rtnl_trylock()) {
4353                 /* Restore the original values before restarting */
4354                 *p = old;
4355                 return restart_syscall();
4356         }
4357
4358         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4359                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4360                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4361                 addrconf_disable_change(net, newf);
4362         } else if ((!*p) ^ (!old))
4363                 dev_disable_change((struct inet6_dev *)table->extra1);
4364
4365         rtnl_unlock();
4366         return 0;
4367 }
4368
4369 static
4370 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4371                             void __user *buffer, size_t *lenp, loff_t *ppos)
4372 {
4373         int *valp = ctl->data;
4374         int val = *valp;
4375         loff_t pos = *ppos;
4376         int ret;
4377
4378         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4379
4380         if (write)
4381                 ret = addrconf_disable_ipv6(ctl, valp, val);
4382         if (ret)
4383                 *ppos = pos;
4384         return ret;
4385 }
4386
4387 static struct addrconf_sysctl_table
4388 {
4389         struct ctl_table_header *sysctl_header;
4390         ctl_table addrconf_vars[DEVCONF_MAX+1];
4391         char *dev_name;
4392 } addrconf_sysctl __read_mostly = {
4393         .sysctl_header = NULL,
4394         .addrconf_vars = {
4395                 {
4396                         .procname       = "forwarding",
4397                         .data           = &ipv6_devconf.forwarding,
4398                         .maxlen         = sizeof(int),
4399                         .mode           = 0644,
4400                         .proc_handler   = addrconf_sysctl_forward,
4401                 },
4402                 {
4403                         .procname       = "hop_limit",
4404                         .data           = &ipv6_devconf.hop_limit,
4405                         .maxlen         = sizeof(int),
4406                         .mode           = 0644,
4407                         .proc_handler   = proc_dointvec,
4408                 },
4409                 {
4410                         .procname       = "mtu",
4411                         .data           = &ipv6_devconf.mtu6,
4412                         .maxlen         = sizeof(int),
4413                         .mode           = 0644,
4414                         .proc_handler   = proc_dointvec,
4415                 },
4416                 {
4417                         .procname       = "accept_ra",
4418                         .data           = &ipv6_devconf.accept_ra,
4419                         .maxlen         = sizeof(int),
4420                         .mode           = 0644,
4421                         .proc_handler   = proc_dointvec,
4422                 },
4423                 {
4424                         .procname       = "accept_redirects",
4425                         .data           = &ipv6_devconf.accept_redirects,
4426                         .maxlen         = sizeof(int),
4427                         .mode           = 0644,
4428                         .proc_handler   = proc_dointvec,
4429                 },
4430                 {
4431                         .procname       = "autoconf",
4432                         .data           = &ipv6_devconf.autoconf,
4433                         .maxlen         = sizeof(int),
4434                         .mode           = 0644,
4435                         .proc_handler   = proc_dointvec,
4436                 },
4437                 {
4438                         .procname       = "dad_transmits",
4439                         .data           = &ipv6_devconf.dad_transmits,
4440                         .maxlen         = sizeof(int),
4441                         .mode           = 0644,
4442                         .proc_handler   = proc_dointvec,
4443                 },
4444                 {
4445                         .procname       = "router_solicitations",
4446                         .data           = &ipv6_devconf.rtr_solicits,
4447                         .maxlen         = sizeof(int),
4448                         .mode           = 0644,
4449                         .proc_handler   = proc_dointvec,
4450                 },
4451                 {
4452                         .procname       = "router_solicitation_interval",
4453                         .data           = &ipv6_devconf.rtr_solicit_interval,
4454                         .maxlen         = sizeof(int),
4455                         .mode           = 0644,
4456                         .proc_handler   = proc_dointvec_jiffies,
4457                 },
4458                 {
4459                         .procname       = "router_solicitation_delay",
4460                         .data           = &ipv6_devconf.rtr_solicit_delay,
4461                         .maxlen         = sizeof(int),
4462                         .mode           = 0644,
4463                         .proc_handler   = proc_dointvec_jiffies,
4464                 },
4465                 {
4466                         .procname       = "force_mld_version",
4467                         .data           = &ipv6_devconf.force_mld_version,
4468                         .maxlen         = sizeof(int),
4469                         .mode           = 0644,
4470                         .proc_handler   = proc_dointvec,
4471                 },
4472 #ifdef CONFIG_IPV6_PRIVACY
4473                 {
4474                         .procname       = "use_tempaddr",
4475                         .data           = &ipv6_devconf.use_tempaddr,
4476                         .maxlen         = sizeof(int),
4477                         .mode           = 0644,
4478                         .proc_handler   = proc_dointvec,
4479                 },
4480                 {
4481                         .procname       = "temp_valid_lft",
4482                         .data           = &ipv6_devconf.temp_valid_lft,
4483                         .maxlen         = sizeof(int),
4484                         .mode           = 0644,
4485                         .proc_handler   = proc_dointvec,
4486                 },
4487                 {
4488                         .procname       = "temp_prefered_lft",
4489                         .data           = &ipv6_devconf.temp_prefered_lft,
4490                         .maxlen         = sizeof(int),
4491                         .mode           = 0644,
4492                         .proc_handler   = proc_dointvec,
4493                 },
4494                 {
4495                         .procname       = "regen_max_retry",
4496                         .data           = &ipv6_devconf.regen_max_retry,
4497                         .maxlen         = sizeof(int),
4498                         .mode           = 0644,
4499                         .proc_handler   = proc_dointvec,
4500                 },
4501                 {
4502                         .procname       = "max_desync_factor",
4503                         .data           = &ipv6_devconf.max_desync_factor,
4504                         .maxlen         = sizeof(int),
4505                         .mode           = 0644,
4506                         .proc_handler   = proc_dointvec,
4507                 },
4508 #endif
4509                 {
4510                         .procname       = "max_addresses",
4511                         .data           = &ipv6_devconf.max_addresses,
4512                         .maxlen         = sizeof(int),
4513                         .mode           = 0644,
4514                         .proc_handler   = proc_dointvec,
4515                 },
4516                 {
4517                         .procname       = "accept_ra_defrtr",
4518                         .data           = &ipv6_devconf.accept_ra_defrtr,
4519                         .maxlen         = sizeof(int),
4520                         .mode           = 0644,
4521                         .proc_handler   = proc_dointvec,
4522                 },
4523                 {
4524                         .procname       = "accept_ra_pinfo",
4525                         .data           = &ipv6_devconf.accept_ra_pinfo,
4526                         .maxlen         = sizeof(int),
4527                         .mode           = 0644,
4528                         .proc_handler   = proc_dointvec,
4529                 },
4530 #ifdef CONFIG_IPV6_ROUTER_PREF
4531                 {
4532                         .procname       = "accept_ra_rtr_pref",
4533                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4534                         .maxlen         = sizeof(int),
4535                         .mode           = 0644,
4536                         .proc_handler   = proc_dointvec,
4537                 },
4538                 {
4539                         .procname       = "router_probe_interval",
4540                         .data           = &ipv6_devconf.rtr_probe_interval,
4541                         .maxlen         = sizeof(int),
4542                         .mode           = 0644,
4543                         .proc_handler   = proc_dointvec_jiffies,
4544                 },
4545 #ifdef CONFIG_IPV6_ROUTE_INFO
4546                 {
4547                         .procname       = "accept_ra_rt_info_max_plen",
4548                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4549                         .maxlen         = sizeof(int),
4550                         .mode           = 0644,
4551                         .proc_handler   = proc_dointvec,
4552                 },
4553 #endif
4554 #endif
4555                 {
4556                         .procname       = "proxy_ndp",
4557                         .data           = &ipv6_devconf.proxy_ndp,
4558                         .maxlen         = sizeof(int),
4559                         .mode           = 0644,
4560                         .proc_handler   = proc_dointvec,
4561                 },
4562                 {
4563                         .procname       = "accept_source_route",
4564                         .data           = &ipv6_devconf.accept_source_route,
4565                         .maxlen         = sizeof(int),
4566                         .mode           = 0644,
4567                         .proc_handler   = proc_dointvec,
4568                 },
4569 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4570                 {
4571                         .procname       = "optimistic_dad",
4572                         .data           = &ipv6_devconf.optimistic_dad,
4573                         .maxlen         = sizeof(int),
4574                         .mode           = 0644,
4575                         .proc_handler   = proc_dointvec,
4576
4577                 },
4578 #endif
4579 #ifdef CONFIG_IPV6_MROUTE
4580                 {
4581                         .procname       = "mc_forwarding",
4582                         .data           = &ipv6_devconf.mc_forwarding,
4583                         .maxlen         = sizeof(int),
4584                         .mode           = 0444,
4585                         .proc_handler   = proc_dointvec,
4586                 },
4587 #endif
4588                 {
4589                         .procname       = "disable_ipv6",
4590                         .data           = &ipv6_devconf.disable_ipv6,
4591                         .maxlen         = sizeof(int),
4592                         .mode           = 0644,
4593                         .proc_handler   = addrconf_sysctl_disable,
4594                 },
4595                 {
4596                         .procname       = "accept_dad",
4597                         .data           = &ipv6_devconf.accept_dad,
4598                         .maxlen         = sizeof(int),
4599                         .mode           = 0644,
4600                         .proc_handler   = proc_dointvec,
4601                 },
4602                 {
4603                         .procname       = "force_tllao",
4604                         .data           = &ipv6_devconf.force_tllao,
4605                         .maxlen         = sizeof(int),
4606                         .mode           = 0644,
4607                         .proc_handler   = proc_dointvec
4608                 },
4609                 {
4610                         /* sentinel */
4611                 }
4612         },
4613 };
4614
4615 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4616                 struct inet6_dev *idev, struct ipv6_devconf *p)
4617 {
4618         int i;
4619         struct addrconf_sysctl_table *t;
4620
4621 #define ADDRCONF_CTL_PATH_DEV   3
4622
4623         struct ctl_path addrconf_ctl_path[] = {
4624                 { .procname = "net", },
4625                 { .procname = "ipv6", },
4626                 { .procname = "conf", },
4627                 { /* to be set */ },
4628                 { },
4629         };
4630
4631
4632         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4633         if (t == NULL)
4634                 goto out;
4635
4636         for (i = 0; t->addrconf_vars[i].data; i++) {
4637                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4638                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4639                 t->addrconf_vars[i].extra2 = net;
4640         }
4641
4642         /*
4643          * Make a copy of dev_name, because '.procname' is regarded as const
4644          * by sysctl and we wouldn't want anyone to change it under our feet
4645          * (see SIOCSIFNAME).
4646          */
4647         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4648         if (!t->dev_name)
4649                 goto free;
4650
4651         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4652
4653         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4654                         t->addrconf_vars);
4655         if (t->sysctl_header == NULL)
4656                 goto free_procname;
4657
4658         p->sysctl = t;
4659         return 0;
4660
4661 free_procname:
4662         kfree(t->dev_name);
4663 free:
4664         kfree(t);
4665 out:
4666         return -ENOBUFS;
4667 }
4668
4669 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4670 {
4671         struct addrconf_sysctl_table *t;
4672
4673         if (p->sysctl == NULL)
4674                 return;
4675
4676         t = p->sysctl;
4677         p->sysctl = NULL;
4678         unregister_net_sysctl_table(t->sysctl_header);
4679         kfree(t->dev_name);
4680         kfree(t);
4681 }
4682
4683 static void addrconf_sysctl_register(struct inet6_dev *idev)
4684 {
4685         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4686                               &ndisc_ifinfo_sysctl_change);
4687         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4688                                         idev, &idev->cnf);
4689 }
4690
4691 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4692 {
4693         __addrconf_sysctl_unregister(&idev->cnf);
4694         neigh_sysctl_unregister(idev->nd_parms);
4695 }
4696
4697
4698 #endif
4699
4700 static int __net_init addrconf_init_net(struct net *net)
4701 {
4702         int err = -ENOMEM;
4703         struct ipv6_devconf *all, *dflt;
4704
4705         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
4706         if (all == NULL)
4707                 goto err_alloc_all;
4708
4709         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4710         if (dflt == NULL)
4711                 goto err_alloc_dflt;
4712
4713         /* these will be inherited by all namespaces */
4714         dflt->autoconf = ipv6_defaults.autoconf;
4715         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4716
4717         net->ipv6.devconf_all = all;
4718         net->ipv6.devconf_dflt = dflt;
4719
4720 #ifdef CONFIG_SYSCTL
4721         err = __addrconf_sysctl_register(net, "all", NULL, all);
4722         if (err < 0)
4723                 goto err_reg_all;
4724
4725         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4726         if (err < 0)
4727                 goto err_reg_dflt;
4728 #endif
4729         return 0;
4730
4731 #ifdef CONFIG_SYSCTL
4732 err_reg_dflt:
4733         __addrconf_sysctl_unregister(all);
4734 err_reg_all:
4735         kfree(dflt);
4736 #endif
4737 err_alloc_dflt:
4738         kfree(all);
4739 err_alloc_all:
4740         return err;
4741 }
4742
4743 static void __net_exit addrconf_exit_net(struct net *net)
4744 {
4745 #ifdef CONFIG_SYSCTL
4746         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4747         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4748 #endif
4749         if (!net_eq(net, &init_net)) {
4750                 kfree(net->ipv6.devconf_dflt);
4751                 kfree(net->ipv6.devconf_all);
4752         }
4753 }
4754
4755 static struct pernet_operations addrconf_ops = {
4756         .init = addrconf_init_net,
4757         .exit = addrconf_exit_net,
4758 };
4759
4760 /*
4761  *      Device notifier
4762  */
4763
4764 int register_inet6addr_notifier(struct notifier_block *nb)
4765 {
4766         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4767 }
4768 EXPORT_SYMBOL(register_inet6addr_notifier);
4769
4770 int unregister_inet6addr_notifier(struct notifier_block *nb)
4771 {
4772         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4773 }
4774 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4775
4776 static struct rtnl_af_ops inet6_ops = {
4777         .family           = AF_INET6,
4778         .fill_link_af     = inet6_fill_link_af,
4779         .get_link_af_size = inet6_get_link_af_size,
4780 };
4781
4782 /*
4783  *      Init / cleanup code
4784  */
4785
4786 int __init addrconf_init(void)
4787 {
4788         int i, err;
4789
4790         err = ipv6_addr_label_init();
4791         if (err < 0) {
4792                 printk(KERN_CRIT "IPv6 Addrconf:"
4793                        " cannot initialize default policy table: %d.\n", err);
4794                 goto out;
4795         }
4796
4797         err = register_pernet_subsys(&addrconf_ops);
4798         if (err < 0)
4799                 goto out_addrlabel;
4800
4801         /* The addrconf netdev notifier requires that loopback_dev
4802          * has it's ipv6 private information allocated and setup
4803          * before it can bring up and give link-local addresses
4804          * to other devices which are up.
4805          *
4806          * Unfortunately, loopback_dev is not necessarily the first
4807          * entry in the global dev_base list of net devices.  In fact,
4808          * it is likely to be the very last entry on that list.
4809          * So this causes the notifier registry below to try and
4810          * give link-local addresses to all devices besides loopback_dev
4811          * first, then loopback_dev, which cases all the non-loopback_dev
4812          * devices to fail to get a link-local address.
4813          *
4814          * So, as a temporary fix, allocate the ipv6 structure for
4815          * loopback_dev first by hand.
4816          * Longer term, all of the dependencies ipv6 has upon the loopback
4817          * device and it being up should be removed.
4818          */
4819         rtnl_lock();
4820         if (!ipv6_add_dev(init_net.loopback_dev))
4821                 err = -ENOMEM;
4822         rtnl_unlock();
4823         if (err)
4824                 goto errlo;
4825
4826         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4827                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4828
4829         register_netdevice_notifier(&ipv6_dev_notf);
4830
4831         addrconf_verify(0);
4832
4833         err = rtnl_af_register(&inet6_ops);
4834         if (err < 0)
4835                 goto errout_af;
4836
4837         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4838                               NULL);
4839         if (err < 0)
4840                 goto errout;
4841
4842         /* Only the first call to __rtnl_register can fail */
4843         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4844         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4845         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4846                         inet6_dump_ifaddr, NULL);
4847         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4848                         inet6_dump_ifmcaddr, NULL);
4849         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4850                         inet6_dump_ifacaddr, NULL);
4851
4852         ipv6_addr_label_rtnl_register();
4853
4854         return 0;
4855 errout:
4856         rtnl_af_unregister(&inet6_ops);
4857 errout_af:
4858         unregister_netdevice_notifier(&ipv6_dev_notf);
4859 errlo:
4860         unregister_pernet_subsys(&addrconf_ops);
4861 out_addrlabel:
4862         ipv6_addr_label_cleanup();
4863 out:
4864         return err;
4865 }
4866
4867 void addrconf_cleanup(void)
4868 {
4869         struct net_device *dev;
4870         int i;
4871
4872         unregister_netdevice_notifier(&ipv6_dev_notf);
4873         unregister_pernet_subsys(&addrconf_ops);
4874         ipv6_addr_label_cleanup();
4875
4876         rtnl_lock();
4877
4878         __rtnl_af_unregister(&inet6_ops);
4879
4880         /* clean dev list */
4881         for_each_netdev(&init_net, dev) {
4882                 if (__in6_dev_get(dev) == NULL)
4883                         continue;
4884                 addrconf_ifdown(dev, 1);
4885         }
4886         addrconf_ifdown(init_net.loopback_dev, 2);
4887
4888         /*
4889          *      Check hash table.
4890          */
4891         spin_lock_bh(&addrconf_hash_lock);
4892         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4893                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4894         spin_unlock_bh(&addrconf_hash_lock);
4895
4896         del_timer(&addr_chk_timer);
4897         rtnl_unlock();
4898 }