SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static int addrconf_ifdown(struct net_device *dev, int how);
 141
 142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 143 static void addrconf_dad_timer(unsigned long data);
 144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 145 static void addrconf_dad_run(struct inet6_dev *idev);
 146 static void addrconf_rs_timer(unsigned long data);
 147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 149
 150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 151                                 struct prefix_info *pinfo);
 152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 153                               struct net_device *dev);
 154
 155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 156
 157 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 158         .forwarding             = 0,
 159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 160         .mtu6                   = IPV6_MIN_MTU,
 161         .accept_ra              = 1,
 162         .accept_redirects       = 1,
 163         .autoconf               = 1,
 164         .force_mld_version      = 0,
 165         .dad_transmits          = 1,
 166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 169 #ifdef CONFIG_IPV6_PRIVACY
 170         .use_tempaddr           = 0,
 171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 173         .regen_max_retry        = REGEN_MAX_RETRY,
 174         .max_desync_factor      = MAX_DESYNC_FACTOR,
 175 #endif
 176         .max_addresses          = IPV6_MAX_ADDRESSES,
 177         .accept_ra_defrtr       = 1,
 178         .accept_ra_pinfo        = 1,
 179 #ifdef CONFIG_IPV6_ROUTER_PREF
 180         .accept_ra_rtr_pref     = 1,
 181         .rtr_probe_interval     = 60 * HZ,
 182 #ifdef CONFIG_IPV6_ROUTE_INFO
 183         .accept_ra_rt_info_max_plen = 0,
 184 #endif
 185 #endif
 186         .proxy_ndp              = 0,
 187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 188         .disable_ipv6           = 0,
 189         .accept_dad             = 1,
 190 };
 191
 192 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 193         .forwarding             = 0,
 194         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 195         .mtu6                   = IPV6_MIN_MTU,
 196         .accept_ra              = 1,
 197         .accept_redirects       = 1,
 198         .autoconf               = 1,
 199         .dad_transmits          = 1,
 200         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 201         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 202         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 203 #ifdef CONFIG_IPV6_PRIVACY
 204         .use_tempaddr           = 0,
 205         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 206         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 207         .regen_max_retry        = REGEN_MAX_RETRY,
 208         .max_desync_factor      = MAX_DESYNC_FACTOR,
 209 #endif
 210         .max_addresses          = IPV6_MAX_ADDRESSES,
 211         .accept_ra_defrtr       = 1,
 212         .accept_ra_pinfo        = 1,
 213 #ifdef CONFIG_IPV6_ROUTER_PREF
 214         .accept_ra_rtr_pref     = 1,
 215         .rtr_probe_interval     = 60 * HZ,
 216 #ifdef CONFIG_IPV6_ROUTE_INFO
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 1,
 224 };
 225
 226 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 227 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 228 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 229 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 230 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 231
 232 /* Check if a valid qdisc is available */
 233 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 234 {
 235         return !qdisc_tx_is_noop(dev);
 236 }
 237
 238 /* Check if a route is valid prefix route */
 239 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 240 {
 241         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 242 }
 243
 244 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 245 {
 246         if (del_timer(&ifp->timer))
 247                 __in6_ifa_put(ifp);
 248 }
 249
 250 enum addrconf_timer_t
 251 {
 252         AC_NONE,
 253         AC_DAD,
 254         AC_RS,
 255 };
 256
 257 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 258                                enum addrconf_timer_t what,
 259                                unsigned long when)
 260 {
 261         if (!del_timer(&ifp->timer))
 262                 in6_ifa_hold(ifp);
 263
 264         switch (what) {
 265         case AC_DAD:
 266                 ifp->timer.function = addrconf_dad_timer;
 267                 break;
 268         case AC_RS:
 269                 ifp->timer.function = addrconf_rs_timer;
 270                 break;
 271         default:;
 272         }
 273         ifp->timer.expires = jiffies + when;
 274         add_timer(&ifp->timer);
 275 }
 276
 277 static int snmp6_alloc_dev(struct inet6_dev *idev)
 278 {
 279         if (snmp_mib_init((void **)idev->stats.ipv6,
 280                           sizeof(struct ipstats_mib)) < 0)
 281                 goto err_ip;
 282         if (snmp_mib_init((void **)idev->stats.icmpv6,
 283                           sizeof(struct icmpv6_mib)) < 0)
 284                 goto err_icmp;
 285         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 286                           sizeof(struct icmpv6msg_mib)) < 0)
 287                 goto err_icmpmsg;
 288
 289         return 0;
 290
 291 err_icmpmsg:
 292         snmp_mib_free((void **)idev->stats.icmpv6);
 293 err_icmp:
 294         snmp_mib_free((void **)idev->stats.ipv6);
 295 err_ip:
 296         return -ENOMEM;
 297 }
 298
 299 static void snmp6_free_dev(struct inet6_dev *idev)
 300 {
 301         snmp_mib_free((void **)idev->stats.icmpv6msg);
 302         snmp_mib_free((void **)idev->stats.icmpv6);
 303         snmp_mib_free((void **)idev->stats.ipv6);
 304 }
 305
 306 /* Nobody refers to this device, we may destroy it. */
 307
 308 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 309 {
 310         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 311         kfree(idev);
 312 }
 313
 314 void in6_dev_finish_destroy(struct inet6_dev *idev)
 315 {
 316         struct net_device *dev = idev->dev;
 317
 318         WARN_ON(idev->addr_list != NULL);
 319         WARN_ON(idev->mc_list != NULL);
 320
 321 #ifdef NET_REFCNT_DEBUG
 322         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 323 #endif
 324         dev_put(dev);
 325         if (!idev->dead) {
 326                 printk("Freeing alive inet6 device %p\n", idev);
 327                 return;
 328         }
 329         snmp6_free_dev(idev);
 330         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 331 }
 332
 333 EXPORT_SYMBOL(in6_dev_finish_destroy);
 334
 335 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 336 {
 337         struct inet6_dev *ndev;
 338
 339         ASSERT_RTNL();
 340
 341         if (dev->mtu < IPV6_MIN_MTU)
 342                 return NULL;
 343
 344         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 345
 346         if (ndev == NULL)
 347                 return NULL;
 348
 349         rwlock_init(&ndev->lock);
 350         ndev->dev = dev;
 351         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 352         ndev->cnf.mtu6 = dev->mtu;
 353         ndev->cnf.sysctl = NULL;
 354         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 355         if (ndev->nd_parms == NULL) {
 356                 kfree(ndev);
 357                 return NULL;
 358         }
 359         if (ndev->cnf.forwarding)
 360                 dev_disable_lro(dev);
 361         /* We refer to the device */
 362         dev_hold(dev);
 363
 364         if (snmp6_alloc_dev(ndev) < 0) {
 365                 ADBG((KERN_WARNING
 366                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 367                         __func__, dev->name));
 368                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 369                 ndev->dead = 1;
 370                 in6_dev_finish_destroy(ndev);
 371                 return NULL;
 372         }
 373
 374         if (snmp6_register_dev(ndev) < 0) {
 375                 ADBG((KERN_WARNING
 376                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 377                         __func__, dev->name));
 378                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 379                 ndev->dead = 1;
 380                 in6_dev_finish_destroy(ndev);
 381                 return NULL;
 382         }
 383
 384         /* One reference from device.  We must do this before
 385          * we invoke __ipv6_regen_rndid().
 386          */
 387         in6_dev_hold(ndev);
 388
 389         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 390                 ndev->cnf.accept_dad = -1;
 391
 392 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 393         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 394                 printk(KERN_INFO
 395                        "%s: Disabled Multicast RS\n",
 396                        dev->name);
 397                 ndev->cnf.rtr_solicits = 0;
 398         }
 399 #endif
 400
 401 #ifdef CONFIG_IPV6_PRIVACY
 402         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 403         if ((dev->flags&IFF_LOOPBACK) ||
 404             dev->type == ARPHRD_TUNNEL ||
 405             dev->type == ARPHRD_TUNNEL6 ||
 406             dev->type == ARPHRD_SIT ||
 407             dev->type == ARPHRD_NONE) {
 408                 printk(KERN_INFO
 409                        "%s: Disabled Privacy Extensions\n",
 410                        dev->name);
 411                 ndev->cnf.use_tempaddr = -1;
 412         } else {
 413                 in6_dev_hold(ndev);
 414                 ipv6_regen_rndid((unsigned long) ndev);
 415         }
 416 #endif
 417
 418         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 419                 ndev->if_flags |= IF_READY;
 420
 421         ipv6_mc_init_dev(ndev);
 422         ndev->tstamp = jiffies;
 423         addrconf_sysctl_register(ndev);
 424         /* protected by rtnl_lock */
 425         rcu_assign_pointer(dev->ip6_ptr, ndev);
 426
 427         /* Join all-node multicast group */
 428         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 429
 430         return ndev;
 431 }
 432
 433 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 434 {
 435         struct inet6_dev *idev;
 436
 437         ASSERT_RTNL();
 438
 439         if ((idev = __in6_dev_get(dev)) == NULL) {
 440                 if ((idev = ipv6_add_dev(dev)) == NULL)
 441                         return NULL;
 442         }
 443
 444         if (dev->flags&IFF_UP)
 445                 ipv6_mc_up(idev);
 446         return idev;
 447 }
 448
 449 #ifdef CONFIG_SYSCTL
 450 static void dev_forward_change(struct inet6_dev *idev)
 451 {
 452         struct net_device *dev;
 453         struct inet6_ifaddr *ifa;
 454
 455         if (!idev)
 456                 return;
 457         dev = idev->dev;
 458         if (idev->cnf.forwarding)
 459                 dev_disable_lro(dev);
 460         if (dev && (dev->flags & IFF_MULTICAST)) {
 461                 if (idev->cnf.forwarding)
 462                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 463                 else
 464                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 465         }
 466         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 467                 if (ifa->flags&IFA_F_TENTATIVE)
 468                         continue;
 469                 if (idev->cnf.forwarding)
 470                         addrconf_join_anycast(ifa);
 471                 else
 472                         addrconf_leave_anycast(ifa);
 473         }
 474 }
 475
 476
 477 static void addrconf_forward_change(struct net *net, __s32 newf)
 478 {
 479         struct net_device *dev;
 480         struct inet6_dev *idev;
 481
 482         read_lock(&dev_base_lock);
 483         for_each_netdev(net, dev) {
 484                 rcu_read_lock();
 485                 idev = __in6_dev_get(dev);
 486                 if (idev) {
 487                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 488                         idev->cnf.forwarding = newf;
 489                         if (changed)
 490                                 dev_forward_change(idev);
 491                 }
 492                 rcu_read_unlock();
 493         }
 494         read_unlock(&dev_base_lock);
 495 }
 496
 497 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 498 {
 499         struct net *net;
 500
 501         net = (struct net *)table->extra2;
 502         if (p == &net->ipv6.devconf_dflt->forwarding)
 503                 return 0;
 504
 505         if (!rtnl_trylock())
 506                 return -ERESTARTSYS;
 507
 508         if (p == &net->ipv6.devconf_all->forwarding) {
 509                 __s32 newf = net->ipv6.devconf_all->forwarding;
 510                 net->ipv6.devconf_dflt->forwarding = newf;
 511                 addrconf_forward_change(net, newf);
 512         } else if ((!*p) ^ (!old))
 513                 dev_forward_change((struct inet6_dev *)table->extra1);
 514         rtnl_unlock();
 515
 516         if (*p)
 517                 rt6_purge_dflt_routers(net);
 518         return 1;
 519 }
 520 #endif
 521
 522 /* Nobody refers to this ifaddr, destroy it */
 523
 524 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 525 {
 526         WARN_ON(ifp->if_next != NULL);
 527         WARN_ON(ifp->lst_next != NULL);
 528
 529 #ifdef NET_REFCNT_DEBUG
 530         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 531 #endif
 532
 533         in6_dev_put(ifp->idev);
 534
 535         if (del_timer(&ifp->timer))
 536                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 537
 538         if (!ifp->dead) {
 539                 printk("Freeing alive inet6 address %p\n", ifp);
 540                 return;
 541         }
 542         dst_release(&ifp->rt->u.dst);
 543
 544         kfree(ifp);
 545 }
 546
 547 static void
 548 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 549 {
 550         struct inet6_ifaddr *ifa, **ifap;
 551         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 552
 553         /*
 554          * Each device address list is sorted in order of scope -
 555          * global before linklocal.
 556          */
 557         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 558              ifap = &ifa->if_next) {
 559                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 560                         break;
 561         }
 562
 563         ifp->if_next = *ifap;
 564         *ifap = ifp;
 565 }
 566
 567 /*
 568  *      Hash function taken from net_alias.c
 569  */
 570 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 571 {
 572         __u32 word;
 573
 574         /*
 575          * We perform the hash function over the last 64 bits of the address
 576          * This will include the IEEE address token on links that support it.
 577          */
 578
 579         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 580         word ^= (word >> 16);
 581         word ^= (word >> 8);
 582
 583         return ((word ^ (word >> 4)) & 0x0f);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         struct net *net = dev_net(idev->dev);
 595         int hash;
 596         int err = 0;
 597         int addr_type = ipv6_addr_type(addr);
 598
 599         if (addr_type == IPV6_ADDR_ANY ||
 600             addr_type & IPV6_ADDR_MULTICAST ||
 601             (!(idev->dev->flags & IFF_LOOPBACK) &&
 602              addr_type & IPV6_ADDR_LOOPBACK))
 603                 return ERR_PTR(-EADDRNOTAVAIL);
 604
 605         rcu_read_lock_bh();
 606         if (idev->dead) {
 607                 err = -ENODEV;                  /*XXX*/
 608                 goto out2;
 609         }
 610
 611         if (idev->cnf.disable_ipv6 || net->ipv6.devconf_all->disable_ipv6) {
 612                 err = -EACCES;
 613                 goto out2;
 614         }
 615
 616         write_lock(&addrconf_hash_lock);
 617
 618         /* Ignore adding duplicate addresses on an interface */
 619         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 620                 ADBG(("ipv6_add_addr: already assigned\n"));
 621                 err = -EEXIST;
 622                 goto out;
 623         }
 624
 625         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 626
 627         if (ifa == NULL) {
 628                 ADBG(("ipv6_add_addr: malloc failed\n"));
 629                 err = -ENOBUFS;
 630                 goto out;
 631         }
 632
 633         rt = addrconf_dst_alloc(idev, addr, 0);
 634         if (IS_ERR(rt)) {
 635                 err = PTR_ERR(rt);
 636                 goto out;
 637         }
 638
 639         ipv6_addr_copy(&ifa->addr, addr);
 640
 641         spin_lock_init(&ifa->lock);
 642         init_timer(&ifa->timer);
 643         ifa->timer.data = (unsigned long) ifa;
 644         ifa->scope = scope;
 645         ifa->prefix_len = pfxlen;
 646         ifa->flags = flags | IFA_F_TENTATIVE;
 647         ifa->cstamp = ifa->tstamp = jiffies;
 648
 649         ifa->rt = rt;
 650
 651         /*
 652          * part one of RFC 4429, section 3.3
 653          * We should not configure an address as
 654          * optimistic if we do not yet know the link
 655          * layer address of our nexhop router
 656          */
 657
 658         if (rt->rt6i_nexthop == NULL)
 659                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 660
 661         ifa->idev = idev;
 662         in6_dev_hold(idev);
 663         /* For caller */
 664         in6_ifa_hold(ifa);
 665
 666         /* Add to big hash table */
 667         hash = ipv6_addr_hash(addr);
 668
 669         ifa->lst_next = inet6_addr_lst[hash];
 670         inet6_addr_lst[hash] = ifa;
 671         in6_ifa_hold(ifa);
 672         write_unlock(&addrconf_hash_lock);
 673
 674         write_lock(&idev->lock);
 675         /* Add to inet6_dev unicast addr list. */
 676         ipv6_link_dev_addr(idev, ifa);
 677
 678 #ifdef CONFIG_IPV6_PRIVACY
 679         if (ifa->flags&IFA_F_TEMPORARY) {
 680                 ifa->tmp_next = idev->tempaddr_list;
 681                 idev->tempaddr_list = ifa;
 682                 in6_ifa_hold(ifa);
 683         }
 684 #endif
 685
 686         in6_ifa_hold(ifa);
 687         write_unlock(&idev->lock);
 688 out2:
 689         rcu_read_unlock_bh();
 690
 691         if (likely(err == 0))
 692                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 693         else {
 694                 kfree(ifa);
 695                 ifa = ERR_PTR(err);
 696         }
 697
 698         return ifa;
 699 out:
 700         write_unlock(&addrconf_hash_lock);
 701         goto out2;
 702 }
 703
 704 /* This function wants to get referenced ifp and releases it before return */
 705
 706 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 707 {
 708         struct inet6_ifaddr *ifa, **ifap;
 709         struct inet6_dev *idev = ifp->idev;
 710         int hash;
 711         int deleted = 0, onlink = 0;
 712         unsigned long expires = jiffies;
 713
 714         hash = ipv6_addr_hash(&ifp->addr);
 715
 716         ifp->dead = 1;
 717
 718         write_lock_bh(&addrconf_hash_lock);
 719         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 720              ifap = &ifa->lst_next) {
 721                 if (ifa == ifp) {
 722                         *ifap = ifa->lst_next;
 723                         __in6_ifa_put(ifp);
 724                         ifa->lst_next = NULL;
 725                         break;
 726                 }
 727         }
 728         write_unlock_bh(&addrconf_hash_lock);
 729
 730         write_lock_bh(&idev->lock);
 731 #ifdef CONFIG_IPV6_PRIVACY
 732         if (ifp->flags&IFA_F_TEMPORARY) {
 733                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 734                      ifap = &ifa->tmp_next) {
 735                         if (ifa == ifp) {
 736                                 *ifap = ifa->tmp_next;
 737                                 if (ifp->ifpub) {
 738                                         in6_ifa_put(ifp->ifpub);
 739                                         ifp->ifpub = NULL;
 740                                 }
 741                                 __in6_ifa_put(ifp);
 742                                 ifa->tmp_next = NULL;
 743                                 break;
 744                         }
 745                 }
 746         }
 747 #endif
 748
 749         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 750                 if (ifa == ifp) {
 751                         *ifap = ifa->if_next;
 752                         __in6_ifa_put(ifp);
 753                         ifa->if_next = NULL;
 754                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 755                                 break;
 756                         deleted = 1;
 757                         continue;
 758                 } else if (ifp->flags & IFA_F_PERMANENT) {
 759                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 760                                               ifp->prefix_len)) {
 761                                 if (ifa->flags & IFA_F_PERMANENT) {
 762                                         onlink = 1;
 763                                         if (deleted)
 764                                                 break;
 765                                 } else {
 766                                         unsigned long lifetime;
 767
 768                                         if (!onlink)
 769                                                 onlink = -1;
 770
 771                                         spin_lock(&ifa->lock);
 772
 773                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 774                                         /*
 775                                          * Note: Because this address is
 776                                          * not permanent, lifetime <
 777                                          * LONG_MAX / HZ here.
 778                                          */
 779                                         if (time_before(expires,
 780                                                         ifa->tstamp + lifetime * HZ))
 781                                                 expires = ifa->tstamp + lifetime * HZ;
 782                                         spin_unlock(&ifa->lock);
 783                                 }
 784                         }
 785                 }
 786                 ifap = &ifa->if_next;
 787         }
 788         write_unlock_bh(&idev->lock);
 789
 790         addrconf_del_timer(ifp);
 791
 792         ipv6_ifa_notify(RTM_DELADDR, ifp);
 793
 794         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 795
 796         /*
 797          * Purge or update corresponding prefix
 798          *
 799          * 1) we don't purge prefix here if address was not permanent.
 800          *    prefix is managed by its own lifetime.
 801          * 2) if there're no addresses, delete prefix.
 802          * 3) if there're still other permanent address(es),
 803          *    corresponding prefix is still permanent.
 804          * 4) otherwise, update prefix lifetime to the
 805          *    longest valid lifetime among the corresponding
 806          *    addresses on the device.
 807          *    Note: subsequent RA will update lifetime.
 808          *
 809          * --yoshfuji
 810          */
 811         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 812                 struct in6_addr prefix;
 813                 struct rt6_info *rt;
 814                 struct net *net = dev_net(ifp->idev->dev);
 815                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 816                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 817
 818                 if (rt && addrconf_is_prefix_route(rt)) {
 819                         if (onlink == 0) {
 820                                 ip6_del_rt(rt);
 821                                 rt = NULL;
 822                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 823                                 rt->rt6i_expires = expires;
 824                                 rt->rt6i_flags |= RTF_EXPIRES;
 825                         }
 826                 }
 827                 dst_release(&rt->u.dst);
 828         }
 829
 830         in6_ifa_put(ifp);
 831 }
 832
 833 #ifdef CONFIG_IPV6_PRIVACY
 834 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 835 {
 836         struct inet6_dev *idev = ifp->idev;
 837         struct in6_addr addr, *tmpaddr;
 838         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 839         unsigned long regen_advance;
 840         int tmp_plen;
 841         int ret = 0;
 842         int max_addresses;
 843         u32 addr_flags;
 844
 845         write_lock(&idev->lock);
 846         if (ift) {
 847                 spin_lock_bh(&ift->lock);
 848                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 849                 spin_unlock_bh(&ift->lock);
 850                 tmpaddr = &addr;
 851         } else {
 852                 tmpaddr = NULL;
 853         }
 854 retry:
 855         in6_dev_hold(idev);
 856         if (idev->cnf.use_tempaddr <= 0) {
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_INFO
 859                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         spin_lock_bh(&ifp->lock);
 865         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 866                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 871                 in6_dev_put(idev);
 872                 ret = -1;
 873                 goto out;
 874         }
 875         in6_ifa_hold(ifp);
 876         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 877         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 878                 spin_unlock_bh(&ifp->lock);
 879                 write_unlock(&idev->lock);
 880                 printk(KERN_WARNING
 881                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 882                 in6_ifa_put(ifp);
 883                 in6_dev_put(idev);
 884                 ret = -1;
 885                 goto out;
 886         }
 887         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 888         tmp_valid_lft = min_t(__u32,
 889                               ifp->valid_lft,
 890                               idev->cnf.temp_valid_lft);
 891         tmp_prefered_lft = min_t(__u32,
 892                                  ifp->prefered_lft,
 893                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 894         tmp_plen = ifp->prefix_len;
 895         max_addresses = idev->cnf.max_addresses;
 896         tmp_cstamp = ifp->cstamp;
 897         tmp_tstamp = ifp->tstamp;
 898         spin_unlock_bh(&ifp->lock);
 899
 900         regen_advance = idev->cnf.regen_max_retry *
 901                         idev->cnf.dad_transmits *
 902                         idev->nd_parms->retrans_time / HZ;
 903         write_unlock(&idev->lock);
 904
 905         /* A temporary address is created only if this calculated Preferred
 906          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 907          * an implementation must not create a temporary address with a zero
 908          * Preferred Lifetime.
 909          */
 910         if (tmp_prefered_lft <= regen_advance) {
 911                 in6_ifa_put(ifp);
 912                 in6_dev_put(idev);
 913                 ret = -1;
 914                 goto out;
 915         }
 916
 917         addr_flags = IFA_F_TEMPORARY;
 918         /* set in addrconf_prefix_rcv() */
 919         if (ifp->flags & IFA_F_OPTIMISTIC)
 920                 addr_flags |= IFA_F_OPTIMISTIC;
 921
 922         ift = !max_addresses ||
 923               ipv6_count_addresses(idev) < max_addresses ?
 924                 ipv6_add_addr(idev, &addr, tmp_plen,
 925                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 926                               addr_flags) : NULL;
 927         if (!ift || IS_ERR(ift)) {
 928                 in6_ifa_put(ifp);
 929                 in6_dev_put(idev);
 930                 printk(KERN_INFO
 931                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 932                 tmpaddr = &addr;
 933                 write_lock(&idev->lock);
 934                 goto retry;
 935         }
 936
 937         spin_lock_bh(&ift->lock);
 938         ift->ifpub = ifp;
 939         ift->valid_lft = tmp_valid_lft;
 940         ift->prefered_lft = tmp_prefered_lft;
 941         ift->cstamp = tmp_cstamp;
 942         ift->tstamp = tmp_tstamp;
 943         spin_unlock_bh(&ift->lock);
 944
 945         addrconf_dad_start(ift, 0);
 946         in6_ifa_put(ift);
 947         in6_dev_put(idev);
 948 out:
 949         return ret;
 950 }
 951 #endif
 952
 953 /*
 954  *      Choose an appropriate source address (RFC3484)
 955  */
 956 enum {
 957         IPV6_SADDR_RULE_INIT = 0,
 958         IPV6_SADDR_RULE_LOCAL,
 959         IPV6_SADDR_RULE_SCOPE,
 960         IPV6_SADDR_RULE_PREFERRED,
 961 #ifdef CONFIG_IPV6_MIP6
 962         IPV6_SADDR_RULE_HOA,
 963 #endif
 964         IPV6_SADDR_RULE_OIF,
 965         IPV6_SADDR_RULE_LABEL,
 966 #ifdef CONFIG_IPV6_PRIVACY
 967         IPV6_SADDR_RULE_PRIVACY,
 968 #endif
 969         IPV6_SADDR_RULE_ORCHID,
 970         IPV6_SADDR_RULE_PREFIX,
 971         IPV6_SADDR_RULE_MAX
 972 };
 973
 974 struct ipv6_saddr_score {
 975         int                     rule;
 976         int                     addr_type;
 977         struct inet6_ifaddr     *ifa;
 978         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 979         int                     scopedist;
 980         int                     matchlen;
 981 };
 982
 983 struct ipv6_saddr_dst {
 984         const struct in6_addr *addr;
 985         int ifindex;
 986         int scope;
 987         int label;
 988         unsigned int prefs;
 989 };
 990
 991 static inline int ipv6_saddr_preferred(int type)
 992 {
 993         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 994                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 995                 return 1;
 996         return 0;
 997 }
 998
 999 static int ipv6_get_saddr_eval(struct net *net,
1000                                struct ipv6_saddr_score *score,
1001                                struct ipv6_saddr_dst *dst,
1002                                int i)
1003 {
1004         int ret;
1005
1006         if (i <= score->rule) {
1007                 switch (i) {
1008                 case IPV6_SADDR_RULE_SCOPE:
1009                         ret = score->scopedist;
1010                         break;
1011                 case IPV6_SADDR_RULE_PREFIX:
1012                         ret = score->matchlen;
1013                         break;
1014                 default:
1015                         ret = !!test_bit(i, score->scorebits);
1016                 }
1017                 goto out;
1018         }
1019
1020         switch (i) {
1021         case IPV6_SADDR_RULE_INIT:
1022                 /* Rule 0: remember if hiscore is not ready yet */
1023                 ret = !!score->ifa;
1024                 break;
1025         case IPV6_SADDR_RULE_LOCAL:
1026                 /* Rule 1: Prefer same address */
1027                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1028                 break;
1029         case IPV6_SADDR_RULE_SCOPE:
1030                 /* Rule 2: Prefer appropriate scope
1031                  *
1032                  *      ret
1033                  *       ^
1034                  *    -1 |  d 15
1035                  *    ---+--+-+---> scope
1036                  *       |
1037                  *       |             d is scope of the destination.
1038                  *  B-d  |  \
1039                  *       |   \      <- smaller scope is better if
1040                  *  B-15 |    \        if scope is enough for destinaion.
1041                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1042                  * d-C-1 | /
1043                  *       |/         <- greater is better
1044                  *   -C  /             if scope is not enough for destination.
1045                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1046                  *
1047                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1048                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1049                  * Assume B = 0 and we get C > 29.
1050                  */
1051                 ret = __ipv6_addr_src_scope(score->addr_type);
1052                 if (ret >= dst->scope)
1053                         ret = -ret;
1054                 else
1055                         ret -= 128;     /* 30 is enough */
1056                 score->scopedist = ret;
1057                 break;
1058         case IPV6_SADDR_RULE_PREFERRED:
1059                 /* Rule 3: Avoid deprecated and optimistic addresses */
1060                 ret = ipv6_saddr_preferred(score->addr_type) ||
1061                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1062                 break;
1063 #ifdef CONFIG_IPV6_MIP6
1064         case IPV6_SADDR_RULE_HOA:
1065             {
1066                 /* Rule 4: Prefer home address */
1067                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1068                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1069                 break;
1070             }
1071 #endif
1072         case IPV6_SADDR_RULE_OIF:
1073                 /* Rule 5: Prefer outgoing interface */
1074                 ret = (!dst->ifindex ||
1075                        dst->ifindex == score->ifa->idev->dev->ifindex);
1076                 break;
1077         case IPV6_SADDR_RULE_LABEL:
1078                 /* Rule 6: Prefer matching label */
1079                 ret = ipv6_addr_label(net,
1080                                       &score->ifa->addr, score->addr_type,
1081                                       score->ifa->idev->dev->ifindex) == dst->label;
1082                 break;
1083 #ifdef CONFIG_IPV6_PRIVACY
1084         case IPV6_SADDR_RULE_PRIVACY:
1085             {
1086                 /* Rule 7: Prefer public address
1087                  * Note: prefer temprary address if use_tempaddr >= 2
1088                  */
1089                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1090                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1091                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1092                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1093                 break;
1094             }
1095 #endif
1096         case IPV6_SADDR_RULE_ORCHID:
1097                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1098                  *          non-ORCHID vs non-ORCHID
1099                  */
1100                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1101                         ipv6_addr_orchid(dst->addr));
1102                 break;
1103         case IPV6_SADDR_RULE_PREFIX:
1104                 /* Rule 8: Use longest matching prefix */
1105                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1106                                                        dst->addr);
1107                 break;
1108         default:
1109                 ret = 0;
1110         }
1111
1112         if (ret)
1113                 __set_bit(i, score->scorebits);
1114         score->rule = i;
1115 out:
1116         return ret;
1117 }
1118
1119 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1120                        const struct in6_addr *daddr, unsigned int prefs,
1121                        struct in6_addr *saddr)
1122 {
1123         struct ipv6_saddr_score scores[2],
1124                                 *score = &scores[0], *hiscore = &scores[1];
1125         struct ipv6_saddr_dst dst;
1126         struct net_device *dev;
1127         int dst_type;
1128
1129         dst_type = __ipv6_addr_type(daddr);
1130         dst.addr = daddr;
1131         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1132         dst.scope = __ipv6_addr_src_scope(dst_type);
1133         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1134         dst.prefs = prefs;
1135
1136         hiscore->rule = -1;
1137         hiscore->ifa = NULL;
1138
1139         read_lock(&dev_base_lock);
1140         rcu_read_lock();
1141
1142         for_each_netdev(net, dev) {
1143                 struct inet6_dev *idev;
1144
1145                 /* Candidate Source Address (section 4)
1146                  *  - multicast and link-local destination address,
1147                  *    the set of candidate source address MUST only
1148                  *    include addresses assigned to interfaces
1149                  *    belonging to the same link as the outgoing
1150                  *    interface.
1151                  * (- For site-local destination addresses, the
1152                  *    set of candidate source addresses MUST only
1153                  *    include addresses assigned to interfaces
1154                  *    belonging to the same site as the outgoing
1155                  *    interface.)
1156                  */
1157                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1158                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1159                     dst.ifindex && dev->ifindex != dst.ifindex)
1160                         continue;
1161
1162                 idev = __in6_dev_get(dev);
1163                 if (!idev)
1164                         continue;
1165
1166                 read_lock_bh(&idev->lock);
1167                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1168                         int i;
1169
1170                         /*
1171                          * - Tentative Address (RFC2462 section 5.4)
1172                          *  - A tentative address is not considered
1173                          *    "assigned to an interface" in the traditional
1174                          *    sense, unless it is also flagged as optimistic.
1175                          * - Candidate Source Address (section 4)
1176                          *  - In any case, anycast addresses, multicast
1177                          *    addresses, and the unspecified address MUST
1178                          *    NOT be included in a candidate set.
1179                          */
1180                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1181                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1182                                 continue;
1183
1184                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1185
1186                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1187                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1188                                 LIMIT_NETDEBUG(KERN_DEBUG
1189                                                "ADDRCONF: unspecified / multicast address "
1190                                                "assigned as unicast address on %s",
1191                                                dev->name);
1192                                 continue;
1193                         }
1194
1195                         score->rule = -1;
1196                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1197
1198                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1199                                 int minihiscore, miniscore;
1200
1201                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1202                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1203
1204                                 if (minihiscore > miniscore) {
1205                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1206                                             score->scopedist > 0) {
1207                                                 /*
1208                                                  * special case:
1209                                                  * each remaining entry
1210                                                  * has too small (not enough)
1211                                                  * scope, because ifa entries
1212                                                  * are sorted by their scope
1213                                                  * values.
1214                                                  */
1215                                                 goto try_nextdev;
1216                                         }
1217                                         break;
1218                                 } else if (minihiscore < miniscore) {
1219                                         if (hiscore->ifa)
1220                                                 in6_ifa_put(hiscore->ifa);
1221
1222                                         in6_ifa_hold(score->ifa);
1223
1224                                         swap(hiscore, score);
1225
1226                                         /* restore our iterator */
1227                                         score->ifa = hiscore->ifa;
1228
1229                                         break;
1230                                 }
1231                         }
1232                 }
1233 try_nextdev:
1234                 read_unlock_bh(&idev->lock);
1235         }
1236         rcu_read_unlock();
1237         read_unlock(&dev_base_lock);
1238
1239         if (!hiscore->ifa)
1240                 return -EADDRNOTAVAIL;
1241
1242         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243         in6_ifa_put(hiscore->ifa);
1244         return 0;
1245 }
1246
1247 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1248
1249 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1250                     unsigned char banned_flags)
1251 {
1252         struct inet6_dev *idev;
1253         int err = -EADDRNOTAVAIL;
1254
1255         rcu_read_lock();
1256         if ((idev = __in6_dev_get(dev)) != NULL) {
1257                 struct inet6_ifaddr *ifp;
1258
1259                 read_lock_bh(&idev->lock);
1260                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1261                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1262                                 ipv6_addr_copy(addr, &ifp->addr);
1263                                 err = 0;
1264                                 break;
1265                         }
1266                 }
1267                 read_unlock_bh(&idev->lock);
1268         }
1269         rcu_read_unlock();
1270         return err;
1271 }
1272
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 {
1275         int cnt = 0;
1276         struct inet6_ifaddr *ifp;
1277
1278         read_lock_bh(&idev->lock);
1279         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1280                 cnt++;
1281         read_unlock_bh(&idev->lock);
1282         return cnt;
1283 }
1284
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286                   struct net_device *dev, int strict)
1287 {
1288         struct inet6_ifaddr * ifp;
1289         u8 hash = ipv6_addr_hash(addr);
1290
1291         read_lock_bh(&addrconf_hash_lock);
1292         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1293                 if (!net_eq(dev_net(ifp->idev->dev), net))
1294                         continue;
1295                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1296                     !(ifp->flags&IFA_F_TENTATIVE)) {
1297                         if (dev == NULL || ifp->idev->dev == dev ||
1298                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1299                                 break;
1300                 }
1301         }
1302         read_unlock_bh(&addrconf_hash_lock);
1303         return ifp != NULL;
1304 }
1305 EXPORT_SYMBOL(ipv6_chk_addr);
1306
1307 static
1308 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309                        struct net_device *dev)
1310 {
1311         struct inet6_ifaddr * ifp;
1312         u8 hash = ipv6_addr_hash(addr);
1313
1314         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1315                 if (!net_eq(dev_net(ifp->idev->dev), net))
1316                         continue;
1317                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1318                         if (dev == NULL || ifp->idev->dev == dev)
1319                                 break;
1320                 }
1321         }
1322         return ifp != NULL;
1323 }
1324
1325 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1326 {
1327         struct inet6_dev *idev;
1328         struct inet6_ifaddr *ifa;
1329         int     onlink;
1330
1331         onlink = 0;
1332         rcu_read_lock();
1333         idev = __in6_dev_get(dev);
1334         if (idev) {
1335                 read_lock_bh(&idev->lock);
1336                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1337                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1338                                                    ifa->prefix_len);
1339                         if (onlink)
1340                                 break;
1341                 }
1342                 read_unlock_bh(&idev->lock);
1343         }
1344         rcu_read_unlock();
1345         return onlink;
1346 }
1347
1348 EXPORT_SYMBOL(ipv6_chk_prefix);
1349
1350 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1351                                      struct net_device *dev, int strict)
1352 {
1353         struct inet6_ifaddr * ifp;
1354         u8 hash = ipv6_addr_hash(addr);
1355
1356         read_lock_bh(&addrconf_hash_lock);
1357         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1358                 if (!net_eq(dev_net(ifp->idev->dev), net))
1359                         continue;
1360                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1361                         if (dev == NULL || ifp->idev->dev == dev ||
1362                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1363                                 in6_ifa_hold(ifp);
1364                                 break;
1365                         }
1366                 }
1367         }
1368         read_unlock_bh(&addrconf_hash_lock);
1369
1370         return ifp;
1371 }
1372
1373 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1374 {
1375         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1376         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1377         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1378         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1379         int sk_ipv6only = ipv6_only_sock(sk);
1380         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1381         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1382         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1383
1384         if (!sk2_rcv_saddr && !sk_ipv6only)
1385                 return 1;
1386
1387         if (addr_type2 == IPV6_ADDR_ANY &&
1388             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1389                 return 1;
1390
1391         if (addr_type == IPV6_ADDR_ANY &&
1392             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1393                 return 1;
1394
1395         if (sk2_rcv_saddr6 &&
1396             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1397                 return 1;
1398
1399         if (addr_type == IPV6_ADDR_MAPPED &&
1400             !sk2_ipv6only &&
1401             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1402                 return 1;
1403
1404         return 0;
1405 }
1406
1407 /* Gets referenced address, destroys ifaddr */
1408
1409 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1410 {
1411         if (ifp->flags&IFA_F_PERMANENT) {
1412                 spin_lock_bh(&ifp->lock);
1413                 addrconf_del_timer(ifp);
1414                 ifp->flags |= IFA_F_TENTATIVE;
1415                 spin_unlock_bh(&ifp->lock);
1416                 in6_ifa_put(ifp);
1417 #ifdef CONFIG_IPV6_PRIVACY
1418         } else if (ifp->flags&IFA_F_TEMPORARY) {
1419                 struct inet6_ifaddr *ifpub;
1420                 spin_lock_bh(&ifp->lock);
1421                 ifpub = ifp->ifpub;
1422                 if (ifpub) {
1423                         in6_ifa_hold(ifpub);
1424                         spin_unlock_bh(&ifp->lock);
1425                         ipv6_create_tempaddr(ifpub, ifp);
1426                         in6_ifa_put(ifpub);
1427                 } else {
1428                         spin_unlock_bh(&ifp->lock);
1429                 }
1430                 ipv6_del_addr(ifp);
1431 #endif
1432         } else
1433                 ipv6_del_addr(ifp);
1434 }
1435
1436 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1437 {
1438         struct inet6_dev *idev = ifp->idev;
1439
1440         if (net_ratelimit())
1441                 printk(KERN_INFO "%s: IPv6 duplicate address detected!\n",
1442                         ifp->idev->dev->name);
1443
1444         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1445                 struct in6_addr addr;
1446
1447                 addr.s6_addr32[0] = htonl(0xfe800000);
1448                 addr.s6_addr32[1] = 0;
1449
1450                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1451                     ipv6_addr_equal(&ifp->addr, &addr)) {
1452                         /* DAD failed for link-local based on MAC address */
1453                         idev->cnf.disable_ipv6 = 1;
1454
1455                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1456                                 ifp->idev->dev->name);
1457                 }
1458         }
1459
1460         addrconf_dad_stop(ifp);
1461 }
1462
1463 /* Join to solicited addr multicast group. */
1464
1465 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1466 {
1467         struct in6_addr maddr;
1468
1469         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1470                 return;
1471
1472         addrconf_addr_solict_mult(addr, &maddr);
1473         ipv6_dev_mc_inc(dev, &maddr);
1474 }
1475
1476 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1477 {
1478         struct in6_addr maddr;
1479
1480         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1481                 return;
1482
1483         addrconf_addr_solict_mult(addr, &maddr);
1484         __ipv6_dev_mc_dec(idev, &maddr);
1485 }
1486
1487 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1488 {
1489         struct in6_addr addr;
1490         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1491         if (ipv6_addr_any(&addr))
1492                 return;
1493         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1494 }
1495
1496 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1497 {
1498         struct in6_addr addr;
1499         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1500         if (ipv6_addr_any(&addr))
1501                 return;
1502         __ipv6_dev_ac_dec(ifp->idev, &addr);
1503 }
1504
1505 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1506 {
1507         if (dev->addr_len != ETH_ALEN)
1508                 return -1;
1509         memcpy(eui, dev->dev_addr, 3);
1510         memcpy(eui + 5, dev->dev_addr + 3, 3);
1511
1512         /*
1513          * The zSeries OSA network cards can be shared among various
1514          * OS instances, but the OSA cards have only one MAC address.
1515          * This leads to duplicate address conflicts in conjunction
1516          * with IPv6 if more than one instance uses the same card.
1517          *
1518          * The driver for these cards can deliver a unique 16-bit
1519          * identifier for each instance sharing the same card.  It is
1520          * placed instead of 0xFFFE in the interface identifier.  The
1521          * "u" bit of the interface identifier is not inverted in this
1522          * case.  Hence the resulting interface identifier has local
1523          * scope according to RFC2373.
1524          */
1525         if (dev->dev_id) {
1526                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1527                 eui[4] = dev->dev_id & 0xFF;
1528         } else {
1529                 eui[3] = 0xFF;
1530                 eui[4] = 0xFE;
1531                 eui[0] ^= 2;
1532         }
1533         return 0;
1534 }
1535
1536 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1537 {
1538         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1539         if (dev->addr_len != ARCNET_ALEN)
1540                 return -1;
1541         memset(eui, 0, 7);
1542         eui[7] = *(u8*)dev->dev_addr;
1543         return 0;
1544 }
1545
1546 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1547 {
1548         if (dev->addr_len != INFINIBAND_ALEN)
1549                 return -1;
1550         memcpy(eui, dev->dev_addr + 12, 8);
1551         eui[0] |= 2;
1552         return 0;
1553 }
1554
1555 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1556 {
1557         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1558                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1559                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1560                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1561                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1562                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1563         eui[1] = 0;
1564         eui[2] = 0x5E;
1565         eui[3] = 0xFE;
1566         memcpy(eui + 4, &addr, 4);
1567         return 0;
1568 }
1569 EXPORT_SYMBOL(__ipv6_isatap_ifid);
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 ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1579 {
1580         switch (dev->type) {
1581         case ARPHRD_ETHER:
1582         case ARPHRD_FDDI:
1583         case ARPHRD_IEEE802_TR:
1584                 return addrconf_ifid_eui48(eui, dev);
1585         case ARPHRD_ARCNET:
1586                 return addrconf_ifid_arcnet(eui, dev);
1587         case ARPHRD_INFINIBAND:
1588                 return addrconf_ifid_infiniband(eui, dev);
1589         case ARPHRD_SIT:
1590                 return addrconf_ifid_sit(eui, dev);
1591         }
1592         return -1;
1593 }
1594
1595 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1596 {
1597         int err = -1;
1598         struct inet6_ifaddr *ifp;
1599
1600         read_lock_bh(&idev->lock);
1601         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1602                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1603                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1604                         err = 0;
1605                         break;
1606                 }
1607         }
1608         read_unlock_bh(&idev->lock);
1609         return err;
1610 }
1611
1612 #ifdef CONFIG_IPV6_PRIVACY
1613 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1614 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1615 {
1616 regen:
1617         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1618         idev->rndid[0] &= ~0x02;
1619
1620         /*
1621          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1622          * check if generated address is not inappropriate
1623          *
1624          *  - Reserved subnet anycast (RFC 2526)
1625          *      11111101 11....11 1xxxxxxx
1626          *  - ISATAP (RFC4214) 6.1
1627          *      00-00-5E-FE-xx-xx-xx-xx
1628          *  - value 0
1629          *  - XXX: already assigned to an address on the device
1630          */
1631         if (idev->rndid[0] == 0xfd &&
1632             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1633             (idev->rndid[7]&0x80))
1634                 goto regen;
1635         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1636                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1637                         goto regen;
1638                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1639                         goto regen;
1640         }
1641
1642         return 0;
1643 }
1644
1645 static void ipv6_regen_rndid(unsigned long data)
1646 {
1647         struct inet6_dev *idev = (struct inet6_dev *) data;
1648         unsigned long expires;
1649
1650         rcu_read_lock_bh();
1651         write_lock_bh(&idev->lock);
1652
1653         if (idev->dead)
1654                 goto out;
1655
1656         if (__ipv6_regen_rndid(idev) < 0)
1657                 goto out;
1658
1659         expires = jiffies +
1660                 idev->cnf.temp_prefered_lft * HZ -
1661                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1662         if (time_before(expires, jiffies)) {
1663                 printk(KERN_WARNING
1664                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1665                         idev->dev->name);
1666                 goto out;
1667         }
1668
1669         if (!mod_timer(&idev->regen_timer, expires))
1670                 in6_dev_hold(idev);
1671
1672 out:
1673         write_unlock_bh(&idev->lock);
1674         rcu_read_unlock_bh();
1675         in6_dev_put(idev);
1676 }
1677
1678 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1679         int ret = 0;
1680
1681         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1682                 ret = __ipv6_regen_rndid(idev);
1683         return ret;
1684 }
1685 #endif
1686
1687 /*
1688  *      Add prefix route.
1689  */
1690
1691 static void
1692 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1693                       unsigned long expires, u32 flags)
1694 {
1695         struct fib6_config cfg = {
1696                 .fc_table = RT6_TABLE_PREFIX,
1697                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1698                 .fc_ifindex = dev->ifindex,
1699                 .fc_expires = expires,
1700                 .fc_dst_len = plen,
1701                 .fc_flags = RTF_UP | flags,
1702                 .fc_nlinfo.nl_net = dev_net(dev),
1703                 .fc_protocol = RTPROT_KERNEL,
1704         };
1705
1706         ipv6_addr_copy(&cfg.fc_dst, pfx);
1707
1708         /* Prevent useless cloning on PtP SIT.
1709            This thing is done here expecting that the whole
1710            class of non-broadcast devices need not cloning.
1711          */
1712 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1713         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1714                 cfg.fc_flags |= RTF_NONEXTHOP;
1715 #endif
1716
1717         ip6_route_add(&cfg);
1718 }
1719
1720 /* Create "default" multicast route to the interface */
1721
1722 static void addrconf_add_mroute(struct net_device *dev)
1723 {
1724         struct fib6_config cfg = {
1725                 .fc_table = RT6_TABLE_LOCAL,
1726                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1727                 .fc_ifindex = dev->ifindex,
1728                 .fc_dst_len = 8,
1729                 .fc_flags = RTF_UP,
1730                 .fc_nlinfo.nl_net = dev_net(dev),
1731         };
1732
1733         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1734
1735         ip6_route_add(&cfg);
1736 }
1737
1738 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1739 static void sit_route_add(struct net_device *dev)
1740 {
1741         struct fib6_config cfg = {
1742                 .fc_table = RT6_TABLE_MAIN,
1743                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1744                 .fc_ifindex = dev->ifindex,
1745                 .fc_dst_len = 96,
1746                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1747                 .fc_nlinfo.nl_net = dev_net(dev),
1748         };
1749
1750         /* prefix length - 96 bits "::d.d.d.d" */
1751         ip6_route_add(&cfg);
1752 }
1753 #endif
1754
1755 static void addrconf_add_lroute(struct net_device *dev)
1756 {
1757         struct in6_addr addr;
1758
1759         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1760         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1761 }
1762
1763 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1764 {
1765         struct inet6_dev *idev;
1766
1767         ASSERT_RTNL();
1768
1769         if ((idev = ipv6_find_idev(dev)) == NULL)
1770                 return NULL;
1771
1772         /* Add default multicast route */
1773         addrconf_add_mroute(dev);
1774
1775         /* Add link local route */
1776         addrconf_add_lroute(dev);
1777         return idev;
1778 }
1779
1780 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1781 {
1782         struct prefix_info *pinfo;
1783         __u32 valid_lft;
1784         __u32 prefered_lft;
1785         int addr_type;
1786         struct inet6_dev *in6_dev;
1787
1788         pinfo = (struct prefix_info *) opt;
1789
1790         if (len < sizeof(struct prefix_info)) {
1791                 ADBG(("addrconf: prefix option too short\n"));
1792                 return;
1793         }
1794
1795         /*
1796          *      Validation checks ([ADDRCONF], page 19)
1797          */
1798
1799         addr_type = ipv6_addr_type(&pinfo->prefix);
1800
1801         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1802                 return;
1803
1804         valid_lft = ntohl(pinfo->valid);
1805         prefered_lft = ntohl(pinfo->prefered);
1806
1807         if (prefered_lft > valid_lft) {
1808                 if (net_ratelimit())
1809                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1810                 return;
1811         }
1812
1813         in6_dev = in6_dev_get(dev);
1814
1815         if (in6_dev == NULL) {
1816                 if (net_ratelimit())
1817                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1818                 return;
1819         }
1820
1821         /*
1822          *      Two things going on here:
1823          *      1) Add routes for on-link prefixes
1824          *      2) Configure prefixes with the auto flag set
1825          */
1826
1827         if (pinfo->onlink) {
1828                 struct rt6_info *rt;
1829                 unsigned long rt_expires;
1830
1831                 /* Avoid arithmetic overflow. Really, we could
1832                  * save rt_expires in seconds, likely valid_lft,
1833                  * but it would require division in fib gc, that it
1834                  * not good.
1835                  */
1836                 if (HZ > USER_HZ)
1837                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1838                 else
1839                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1840
1841                 if (addrconf_finite_timeout(rt_expires))
1842                         rt_expires *= HZ;
1843
1844                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1845                                 dev->ifindex, 1);
1846
1847                 if (rt && addrconf_is_prefix_route(rt)) {
1848                         /* Autoconf prefix route */
1849                         if (valid_lft == 0) {
1850                                 ip6_del_rt(rt);
1851                                 rt = NULL;
1852                         } else if (addrconf_finite_timeout(rt_expires)) {
1853                                 /* not infinity */
1854                                 rt->rt6i_expires = jiffies + rt_expires;
1855                                 rt->rt6i_flags |= RTF_EXPIRES;
1856                         } else {
1857                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1858                                 rt->rt6i_expires = 0;
1859                         }
1860                 } else if (valid_lft) {
1861                         clock_t expires = 0;
1862                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1863                         if (addrconf_finite_timeout(rt_expires)) {
1864                                 /* not infinity */
1865                                 flags |= RTF_EXPIRES;
1866                                 expires = jiffies_to_clock_t(rt_expires);
1867                         }
1868                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1869                                               dev, expires, flags);
1870                 }
1871                 if (rt)
1872                         dst_release(&rt->u.dst);
1873         }
1874
1875         /* Try to figure out our local address for this prefix */
1876
1877         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1878                 struct inet6_ifaddr * ifp;
1879                 struct in6_addr addr;
1880                 int create = 0, update_lft = 0;
1881                 struct net *net = dev_net(dev);
1882
1883                 if (pinfo->prefix_len == 64) {
1884                         memcpy(&addr, &pinfo->prefix, 8);
1885                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1886                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1887                                 in6_dev_put(in6_dev);
1888                                 return;
1889                         }
1890                         goto ok;
1891                 }
1892                 if (net_ratelimit())
1893                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1894                                pinfo->prefix_len);
1895                 in6_dev_put(in6_dev);
1896                 return;
1897
1898 ok:
1899
1900                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1901
1902                 if (ifp == NULL && valid_lft) {
1903                         int max_addresses = in6_dev->cnf.max_addresses;
1904                         u32 addr_flags = 0;
1905
1906 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1907                         if (in6_dev->cnf.optimistic_dad &&
1908                             !net->ipv6.devconf_all->forwarding)
1909                                 addr_flags = IFA_F_OPTIMISTIC;
1910 #endif
1911
1912                         /* Do not allow to create too much of autoconfigured
1913                          * addresses; this would be too easy way to crash kernel.
1914                          */
1915                         if (!max_addresses ||
1916                             ipv6_count_addresses(in6_dev) < max_addresses)
1917                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1918                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1919                                                     addr_flags);
1920
1921                         if (!ifp || IS_ERR(ifp)) {
1922                                 in6_dev_put(in6_dev);
1923                                 return;
1924                         }
1925
1926                         update_lft = create = 1;
1927                         ifp->cstamp = jiffies;
1928                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1929                 }
1930
1931                 if (ifp) {
1932                         int flags;
1933                         unsigned long now;
1934 #ifdef CONFIG_IPV6_PRIVACY
1935                         struct inet6_ifaddr *ift;
1936 #endif
1937                         u32 stored_lft;
1938
1939                         /* update lifetime (RFC2462 5.5.3 e) */
1940                         spin_lock(&ifp->lock);
1941                         now = jiffies;
1942                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1943                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1944                         else
1945                                 stored_lft = 0;
1946                         if (!update_lft && stored_lft) {
1947                                 if (valid_lft > MIN_VALID_LIFETIME ||
1948                                     valid_lft > stored_lft)
1949                                         update_lft = 1;
1950                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1951                                         /* valid_lft <= stored_lft is always true */
1952                                         /* XXX: IPsec */
1953                                         update_lft = 0;
1954                                 } else {
1955                                         valid_lft = MIN_VALID_LIFETIME;
1956                                         if (valid_lft < prefered_lft)
1957                                                 prefered_lft = valid_lft;
1958                                         update_lft = 1;
1959                                 }
1960                         }
1961
1962                         if (update_lft) {
1963                                 ifp->valid_lft = valid_lft;
1964                                 ifp->prefered_lft = prefered_lft;
1965                                 ifp->tstamp = now;
1966                                 flags = ifp->flags;
1967                                 ifp->flags &= ~IFA_F_DEPRECATED;
1968                                 spin_unlock(&ifp->lock);
1969
1970                                 if (!(flags&IFA_F_TENTATIVE))
1971                                         ipv6_ifa_notify(0, ifp);
1972                         } else
1973                                 spin_unlock(&ifp->lock);
1974
1975 #ifdef CONFIG_IPV6_PRIVACY
1976                         read_lock_bh(&in6_dev->lock);
1977                         /* update all temporary addresses in the list */
1978                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1979                                 /*
1980                                  * When adjusting the lifetimes of an existing
1981                                  * temporary address, only lower the lifetimes.
1982                                  * Implementations must not increase the
1983                                  * lifetimes of an existing temporary address
1984                                  * when processing a Prefix Information Option.
1985                                  */
1986                                 if (ifp != ift->ifpub)
1987                                         continue;
1988
1989                                 spin_lock(&ift->lock);
1990                                 flags = ift->flags;
1991                                 if (ift->valid_lft > valid_lft &&
1992                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1993                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1994                                 if (ift->prefered_lft > prefered_lft &&
1995                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1996                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1997                                 spin_unlock(&ift->lock);
1998                                 if (!(flags&IFA_F_TENTATIVE))
1999                                         ipv6_ifa_notify(0, ift);
2000                         }
2001
2002                         if (create && in6_dev->cnf.use_tempaddr > 0) {
2003                                 /*
2004                                  * When a new public address is created as described in [ADDRCONF],
2005                                  * also create a new temporary address.
2006                                  */
2007                                 read_unlock_bh(&in6_dev->lock);
2008                                 ipv6_create_tempaddr(ifp, NULL);
2009                         } else {
2010                                 read_unlock_bh(&in6_dev->lock);
2011                         }
2012 #endif
2013                         in6_ifa_put(ifp);
2014                         addrconf_verify(0);
2015                 }
2016         }
2017         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2018         in6_dev_put(in6_dev);
2019 }
2020
2021 /*
2022  *      Set destination address.
2023  *      Special case for SIT interfaces where we create a new "virtual"
2024  *      device.
2025  */
2026 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2027 {
2028         struct in6_ifreq ireq;
2029         struct net_device *dev;
2030         int err = -EINVAL;
2031
2032         rtnl_lock();
2033
2034         err = -EFAULT;
2035         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2036                 goto err_exit;
2037
2038         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2039
2040         err = -ENODEV;
2041         if (dev == NULL)
2042                 goto err_exit;
2043
2044 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2045         if (dev->type == ARPHRD_SIT) {
2046                 const struct net_device_ops *ops = dev->netdev_ops;
2047                 struct ifreq ifr;
2048                 struct ip_tunnel_parm p;
2049
2050                 err = -EADDRNOTAVAIL;
2051                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2052                         goto err_exit;
2053
2054                 memset(&p, 0, sizeof(p));
2055                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2056                 p.iph.saddr = 0;
2057                 p.iph.version = 4;
2058                 p.iph.ihl = 5;
2059                 p.iph.protocol = IPPROTO_IPV6;
2060                 p.iph.ttl = 64;
2061                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2062
2063                 if (ops->ndo_do_ioctl) {
2064                         mm_segment_t oldfs = get_fs();
2065
2066                         set_fs(KERNEL_DS);
2067                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2068                         set_fs(oldfs);
2069                 } else
2070                         err = -EOPNOTSUPP;
2071
2072                 if (err == 0) {
2073                         err = -ENOBUFS;
2074                         dev = __dev_get_by_name(net, p.name);
2075                         if (!dev)
2076                                 goto err_exit;
2077                         err = dev_open(dev);
2078                 }
2079         }
2080 #endif
2081
2082 err_exit:
2083         rtnl_unlock();
2084         return err;
2085 }
2086
2087 /*
2088  *      Manual configuration of address on an interface
2089  */
2090 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2091                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2092                           __u32 valid_lft)
2093 {
2094         struct inet6_ifaddr *ifp;
2095         struct inet6_dev *idev;
2096         struct net_device *dev;
2097         int scope;
2098         u32 flags;
2099         clock_t expires;
2100         unsigned long timeout;
2101
2102         ASSERT_RTNL();
2103
2104         if (plen > 128)
2105                 return -EINVAL;
2106
2107         /* check the lifetime */
2108         if (!valid_lft || prefered_lft > valid_lft)
2109                 return -EINVAL;
2110
2111         dev = __dev_get_by_index(net, ifindex);
2112         if (!dev)
2113                 return -ENODEV;
2114
2115         if ((idev = addrconf_add_dev(dev)) == NULL)
2116                 return -ENOBUFS;
2117
2118         scope = ipv6_addr_scope(pfx);
2119
2120         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2121         if (addrconf_finite_timeout(timeout)) {
2122                 expires = jiffies_to_clock_t(timeout * HZ);
2123                 valid_lft = timeout;
2124                 flags = RTF_EXPIRES;
2125         } else {
2126                 expires = 0;
2127                 flags = 0;
2128                 ifa_flags |= IFA_F_PERMANENT;
2129         }
2130
2131         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2132         if (addrconf_finite_timeout(timeout)) {
2133                 if (timeout == 0)
2134                         ifa_flags |= IFA_F_DEPRECATED;
2135                 prefered_lft = timeout;
2136         }
2137
2138         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2139
2140         if (!IS_ERR(ifp)) {
2141                 spin_lock_bh(&ifp->lock);
2142                 ifp->valid_lft = valid_lft;
2143                 ifp->prefered_lft = prefered_lft;
2144                 ifp->tstamp = jiffies;
2145                 spin_unlock_bh(&ifp->lock);
2146
2147                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2148                                       expires, flags);
2149                 /*
2150                  * Note that section 3.1 of RFC 4429 indicates
2151                  * that the Optimistic flag should not be set for
2152                  * manually configured addresses
2153                  */
2154                 addrconf_dad_start(ifp, 0);
2155                 in6_ifa_put(ifp);
2156                 addrconf_verify(0);
2157                 return 0;
2158         }
2159
2160         return PTR_ERR(ifp);
2161 }
2162
2163 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2164                           unsigned int plen)
2165 {
2166         struct inet6_ifaddr *ifp;
2167         struct inet6_dev *idev;
2168         struct net_device *dev;
2169
2170         if (plen > 128)
2171                 return -EINVAL;
2172
2173         dev = __dev_get_by_index(net, ifindex);
2174         if (!dev)
2175                 return -ENODEV;
2176
2177         if ((idev = __in6_dev_get(dev)) == NULL)
2178                 return -ENXIO;
2179
2180         read_lock_bh(&idev->lock);
2181         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2182                 if (ifp->prefix_len == plen &&
2183                     ipv6_addr_equal(pfx, &ifp->addr)) {
2184                         in6_ifa_hold(ifp);
2185                         read_unlock_bh(&idev->lock);
2186
2187                         ipv6_del_addr(ifp);
2188
2189                         /* If the last address is deleted administratively,
2190                            disable IPv6 on this interface.
2191                          */
2192                         if (idev->addr_list == NULL)
2193                                 addrconf_ifdown(idev->dev, 1);
2194                         return 0;
2195                 }
2196         }
2197         read_unlock_bh(&idev->lock);
2198         return -EADDRNOTAVAIL;
2199 }
2200
2201
2202 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2203 {
2204         struct in6_ifreq ireq;
2205         int err;
2206
2207         if (!capable(CAP_NET_ADMIN))
2208                 return -EPERM;
2209
2210         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2211                 return -EFAULT;
2212
2213         rtnl_lock();
2214         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2215                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2216                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2217         rtnl_unlock();
2218         return err;
2219 }
2220
2221 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2222 {
2223         struct in6_ifreq ireq;
2224         int err;
2225
2226         if (!capable(CAP_NET_ADMIN))
2227                 return -EPERM;
2228
2229         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2230                 return -EFAULT;
2231
2232         rtnl_lock();
2233         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2234                              ireq.ifr6_prefixlen);
2235         rtnl_unlock();
2236         return err;
2237 }
2238
2239 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2240                      int plen, int scope)
2241 {
2242         struct inet6_ifaddr *ifp;
2243
2244         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2245         if (!IS_ERR(ifp)) {
2246                 spin_lock_bh(&ifp->lock);
2247                 ifp->flags &= ~IFA_F_TENTATIVE;
2248                 spin_unlock_bh(&ifp->lock);
2249                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2250                 in6_ifa_put(ifp);
2251         }
2252 }
2253
2254 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2255 static void sit_add_v4_addrs(struct inet6_dev *idev)
2256 {
2257         struct in6_addr addr;
2258         struct net_device *dev;
2259         struct net *net = dev_net(idev->dev);
2260         int scope;
2261
2262         ASSERT_RTNL();
2263
2264         memset(&addr, 0, sizeof(struct in6_addr));
2265         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2266
2267         if (idev->dev->flags&IFF_POINTOPOINT) {
2268                 addr.s6_addr32[0] = htonl(0xfe800000);
2269                 scope = IFA_LINK;
2270         } else {
2271                 scope = IPV6_ADDR_COMPATv4;
2272         }
2273
2274         if (addr.s6_addr32[3]) {
2275                 add_addr(idev, &addr, 128, scope);
2276                 return;
2277         }
2278
2279         for_each_netdev(net, dev) {
2280                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2281                 if (in_dev && (dev->flags & IFF_UP)) {
2282                         struct in_ifaddr * ifa;
2283
2284                         int flag = scope;
2285
2286                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2287                                 int plen;
2288
2289                                 addr.s6_addr32[3] = ifa->ifa_local;
2290
2291                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2292                                         continue;
2293                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2294                                         if (idev->dev->flags&IFF_POINTOPOINT)
2295                                                 continue;
2296                                         flag |= IFA_HOST;
2297                                 }
2298                                 if (idev->dev->flags&IFF_POINTOPOINT)
2299                                         plen = 64;
2300                                 else
2301                                         plen = 96;
2302
2303                                 add_addr(idev, &addr, plen, flag);
2304                         }
2305                 }
2306         }
2307 }
2308 #endif
2309
2310 static void init_loopback(struct net_device *dev)
2311 {
2312         struct inet6_dev  *idev;
2313
2314         /* ::1 */
2315
2316         ASSERT_RTNL();
2317
2318         if ((idev = ipv6_find_idev(dev)) == NULL) {
2319                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2320                 return;
2321         }
2322
2323         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2324 }
2325
2326 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2327 {
2328         struct inet6_ifaddr * ifp;
2329         u32 addr_flags = IFA_F_PERMANENT;
2330
2331 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2332         if (idev->cnf.optimistic_dad &&
2333             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2334                 addr_flags |= IFA_F_OPTIMISTIC;
2335 #endif
2336
2337
2338         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2339         if (!IS_ERR(ifp)) {
2340                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2341                 addrconf_dad_start(ifp, 0);
2342                 in6_ifa_put(ifp);
2343         }
2344 }
2345
2346 static void addrconf_dev_config(struct net_device *dev)
2347 {
2348         struct in6_addr addr;
2349         struct inet6_dev    * idev;
2350
2351         ASSERT_RTNL();
2352
2353         if ((dev->type != ARPHRD_ETHER) &&
2354             (dev->type != ARPHRD_FDDI) &&
2355             (dev->type != ARPHRD_IEEE802_TR) &&
2356             (dev->type != ARPHRD_ARCNET) &&
2357             (dev->type != ARPHRD_INFINIBAND)) {
2358                 /* Alas, we support only Ethernet autoconfiguration. */
2359                 return;
2360         }
2361
2362         idev = addrconf_add_dev(dev);
2363         if (idev == NULL)
2364                 return;
2365
2366         memset(&addr, 0, sizeof(struct in6_addr));
2367         addr.s6_addr32[0] = htonl(0xFE800000);
2368
2369         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2370                 addrconf_add_linklocal(idev, &addr);
2371 }
2372
2373 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2374 static void addrconf_sit_config(struct net_device *dev)
2375 {
2376         struct inet6_dev *idev;
2377
2378         ASSERT_RTNL();
2379
2380         /*
2381          * Configure the tunnel with one of our IPv4
2382          * addresses... we should configure all of
2383          * our v4 addrs in the tunnel
2384          */
2385
2386         if ((idev = ipv6_find_idev(dev)) == NULL) {
2387                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2388                 return;
2389         }
2390
2391         if (dev->priv_flags & IFF_ISATAP) {
2392                 struct in6_addr addr;
2393
2394                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2395                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2396                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2397                         addrconf_add_linklocal(idev, &addr);
2398                 return;
2399         }
2400
2401         sit_add_v4_addrs(idev);
2402
2403         if (dev->flags&IFF_POINTOPOINT) {
2404                 addrconf_add_mroute(dev);
2405                 addrconf_add_lroute(dev);
2406         } else
2407                 sit_route_add(dev);
2408 }
2409 #endif
2410
2411 static inline int
2412 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2413 {
2414         struct in6_addr lladdr;
2415
2416         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2417                 addrconf_add_linklocal(idev, &lladdr);
2418                 return 0;
2419         }
2420         return -1;
2421 }
2422
2423 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2424 {
2425         struct net_device *link_dev;
2426         struct net *net = dev_net(idev->dev);
2427
2428         /* first try to inherit the link-local address from the link device */
2429         if (idev->dev->iflink &&
2430             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2431                 if (!ipv6_inherit_linklocal(idev, link_dev))
2432                         return;
2433         }
2434         /* then try to inherit it from any device */
2435         for_each_netdev(net, link_dev) {
2436                 if (!ipv6_inherit_linklocal(idev, link_dev))
2437                         return;
2438         }
2439         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2440 }
2441
2442 /*
2443  * Autoconfigure tunnel with a link-local address so routing protocols,
2444  * DHCPv6, MLD etc. can be run over the virtual link
2445  */
2446
2447 static void addrconf_ip6_tnl_config(struct net_device *dev)
2448 {
2449         struct inet6_dev *idev;
2450
2451         ASSERT_RTNL();
2452
2453         if ((idev = addrconf_add_dev(dev)) == NULL) {
2454                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2455                 return;
2456         }
2457         ip6_tnl_add_linklocal(idev);
2458 }
2459
2460 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2461                            void * data)
2462 {
2463         struct net_device *dev = (struct net_device *) data;
2464         struct inet6_dev *idev = __in6_dev_get(dev);
2465         int run_pending = 0;
2466         int err;
2467
2468         switch(event) {
2469         case NETDEV_REGISTER:
2470                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2471                         idev = ipv6_add_dev(dev);
2472                         if (!idev)
2473                                 return notifier_from_errno(-ENOMEM);
2474                 }
2475                 break;
2476         case NETDEV_UP:
2477         case NETDEV_CHANGE:
2478                 if (dev->flags & IFF_SLAVE)
2479                         break;
2480
2481                 if (event == NETDEV_UP) {
2482                         if (!addrconf_qdisc_ok(dev)) {
2483                                 /* device is not ready yet. */
2484                                 printk(KERN_INFO
2485                                         "ADDRCONF(NETDEV_UP): %s: "
2486                                         "link is not ready\n",
2487                                         dev->name);
2488                                 break;
2489                         }
2490
2491                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2492                                 idev = ipv6_add_dev(dev);
2493
2494                         if (idev) {
2495                                 idev->if_flags |= IF_READY;
2496                                 run_pending = 1;
2497                         }
2498                 } else {
2499                         if (!addrconf_qdisc_ok(dev)) {
2500                                 /* device is still not ready. */
2501                                 break;
2502                         }
2503
2504                         if (idev) {
2505                                 if (idev->if_flags & IF_READY) {
2506                                         /* device is already configured. */
2507                                         break;
2508                                 }
2509                                 idev->if_flags |= IF_READY;
2510                         }
2511
2512                         printk(KERN_INFO
2513                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2514                                         "link becomes ready\n",
2515                                         dev->name);
2516
2517                         run_pending = 1;
2518                 }
2519
2520                 switch(dev->type) {
2521 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2522                 case ARPHRD_SIT:
2523                         addrconf_sit_config(dev);
2524                         break;
2525 #endif
2526                 case ARPHRD_TUNNEL6:
2527                         addrconf_ip6_tnl_config(dev);
2528                         break;
2529                 case ARPHRD_LOOPBACK:
2530                         init_loopback(dev);
2531                         break;
2532
2533                 default:
2534                         addrconf_dev_config(dev);
2535                         break;
2536                 }
2537                 if (idev) {
2538                         if (run_pending)
2539                                 addrconf_dad_run(idev);
2540
2541                         /* If the MTU changed during the interface down, when the
2542                            interface up, the changed MTU must be reflected in the
2543                            idev as well as routers.
2544                          */
2545                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2546                                 rt6_mtu_change(dev, dev->mtu);
2547                                 idev->cnf.mtu6 = dev->mtu;
2548                         }
2549                         idev->tstamp = jiffies;
2550                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2551                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2552                            stop IPv6 on this interface.
2553                          */
2554                         if (dev->mtu < IPV6_MIN_MTU)
2555                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2556                 }
2557                 break;
2558
2559         case NETDEV_CHANGEMTU:
2560                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2561                         rt6_mtu_change(dev, dev->mtu);
2562                         idev->cnf.mtu6 = dev->mtu;
2563                         break;
2564                 }
2565
2566                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2567                         idev = ipv6_add_dev(dev);
2568                         if (idev)
2569                                 break;
2570                 }
2571
2572                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2573
2574         case NETDEV_DOWN:
2575         case NETDEV_UNREGISTER:
2576                 /*
2577                  *      Remove all addresses from this interface.
2578                  */
2579                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2580                 break;
2581
2582         case NETDEV_CHANGENAME:
2583                 if (idev) {
2584                         snmp6_unregister_dev(idev);
2585                         addrconf_sysctl_unregister(idev);
2586                         addrconf_sysctl_register(idev);
2587                         err = snmp6_register_dev(idev);
2588                         if (err)
2589                                 return notifier_from_errno(err);
2590                 }
2591                 break;
2592         }
2593
2594         return NOTIFY_OK;
2595 }
2596
2597 /*
2598  *      addrconf module should be notified of a device going up
2599  */
2600 static struct notifier_block ipv6_dev_notf = {
2601         .notifier_call = addrconf_notify,
2602         .priority = 0
2603 };
2604
2605 static int addrconf_ifdown(struct net_device *dev, int how)
2606 {
2607         struct inet6_dev *idev;
2608         struct inet6_ifaddr *ifa, **bifa;
2609         struct net *net = dev_net(dev);
2610         int i;
2611
2612         ASSERT_RTNL();
2613
2614         rt6_ifdown(net, dev);
2615         neigh_ifdown(&nd_tbl, dev);
2616
2617         idev = __in6_dev_get(dev);
2618         if (idev == NULL)
2619                 return -ENODEV;
2620
2621         /* Step 1: remove reference to ipv6 device from parent device.
2622                    Do not dev_put!
2623          */
2624         if (how) {
2625                 idev->dead = 1;
2626
2627                 /* protected by rtnl_lock */
2628                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2629
2630                 /* Step 1.5: remove snmp6 entry */
2631                 snmp6_unregister_dev(idev);
2632
2633         }
2634
2635         /* Step 2: clear hash table */
2636         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2637                 bifa = &inet6_addr_lst[i];
2638
2639                 write_lock_bh(&addrconf_hash_lock);
2640                 while ((ifa = *bifa) != NULL) {
2641                         if (ifa->idev == idev) {
2642                                 *bifa = ifa->lst_next;
2643                                 ifa->lst_next = NULL;
2644                                 addrconf_del_timer(ifa);
2645                                 in6_ifa_put(ifa);
2646                                 continue;
2647                         }
2648                         bifa = &ifa->lst_next;
2649                 }
2650                 write_unlock_bh(&addrconf_hash_lock);
2651         }
2652
2653         write_lock_bh(&idev->lock);
2654
2655         /* Step 3: clear flags for stateless addrconf */
2656         if (!how)
2657                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2658
2659         /* Step 4: clear address list */
2660 #ifdef CONFIG_IPV6_PRIVACY
2661         if (how && del_timer(&idev->regen_timer))
2662                 in6_dev_put(idev);
2663
2664         /* clear tempaddr list */
2665         while ((ifa = idev->tempaddr_list) != NULL) {
2666                 idev->tempaddr_list = ifa->tmp_next;
2667                 ifa->tmp_next = NULL;
2668                 ifa->dead = 1;
2669                 write_unlock_bh(&idev->lock);
2670                 spin_lock_bh(&ifa->lock);
2671
2672                 if (ifa->ifpub) {
2673                         in6_ifa_put(ifa->ifpub);
2674                         ifa->ifpub = NULL;
2675                 }
2676                 spin_unlock_bh(&ifa->lock);
2677                 in6_ifa_put(ifa);
2678                 write_lock_bh(&idev->lock);
2679         }
2680 #endif
2681         while ((ifa = idev->addr_list) != NULL) {
2682                 idev->addr_list = ifa->if_next;
2683                 ifa->if_next = NULL;
2684                 ifa->dead = 1;
2685                 addrconf_del_timer(ifa);
2686                 write_unlock_bh(&idev->lock);
2687
2688                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2689                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2690                 in6_ifa_put(ifa);
2691
2692                 write_lock_bh(&idev->lock);
2693         }
2694         write_unlock_bh(&idev->lock);
2695
2696         /* Step 5: Discard multicast list */
2697
2698         if (how)
2699                 ipv6_mc_destroy_dev(idev);
2700         else
2701                 ipv6_mc_down(idev);
2702
2703         idev->tstamp = jiffies;
2704
2705         /* Shot the device (if unregistered) */
2706
2707         if (how) {
2708                 addrconf_sysctl_unregister(idev);
2709                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2710                 neigh_ifdown(&nd_tbl, dev);
2711                 in6_dev_put(idev);
2712         }
2713         return 0;
2714 }
2715
2716 static void addrconf_rs_timer(unsigned long data)
2717 {
2718         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2719
2720         if (ifp->idev->cnf.forwarding)
2721                 goto out;
2722
2723         if (ifp->idev->if_flags & IF_RA_RCVD) {
2724                 /*
2725                  *      Announcement received after solicitation
2726                  *      was sent
2727                  */
2728                 goto out;
2729         }
2730
2731         spin_lock(&ifp->lock);
2732         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2733                 /* The wait after the last probe can be shorter */
2734                 addrconf_mod_timer(ifp, AC_RS,
2735                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2736                                    ifp->idev->cnf.rtr_solicit_delay :
2737                                    ifp->idev->cnf.rtr_solicit_interval);
2738                 spin_unlock(&ifp->lock);
2739
2740                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2741         } else {
2742                 spin_unlock(&ifp->lock);
2743                 /*
2744                  * Note: we do not support deprecated "all on-link"
2745                  * assumption any longer.
2746                  */
2747                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2748                        ifp->idev->dev->name);
2749         }
2750
2751 out:
2752         in6_ifa_put(ifp);
2753 }
2754
2755 /*
2756  *      Duplicate Address Detection
2757  */
2758 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2759 {
2760         unsigned long rand_num;
2761         struct inet6_dev *idev = ifp->idev;
2762
2763         if (ifp->flags & IFA_F_OPTIMISTIC)
2764                 rand_num = 0;
2765         else
2766                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2767
2768         ifp->probes = idev->cnf.dad_transmits;
2769         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2770 }
2771
2772 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2773 {
2774         struct inet6_dev *idev = ifp->idev;
2775         struct net_device *dev = idev->dev;
2776
2777         addrconf_join_solict(dev, &ifp->addr);
2778
2779         net_srandom(ifp->addr.s6_addr32[3]);
2780
2781         read_lock_bh(&idev->lock);
2782         if (ifp->dead)
2783                 goto out;
2784         spin_lock_bh(&ifp->lock);
2785
2786         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2787             idev->cnf.accept_dad < 1 ||
2788             !(ifp->flags&IFA_F_TENTATIVE) ||
2789             ifp->flags & IFA_F_NODAD) {
2790                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2791                 spin_unlock_bh(&ifp->lock);
2792                 read_unlock_bh(&idev->lock);
2793
2794                 addrconf_dad_completed(ifp);
2795                 return;
2796         }
2797
2798         if (!(idev->if_flags & IF_READY)) {
2799                 spin_unlock_bh(&ifp->lock);
2800                 read_unlock_bh(&idev->lock);
2801                 /*
2802                  * If the defice is not ready:
2803                  * - keep it tentative if it is a permanent address.
2804                  * - otherwise, kill it.
2805                  */
2806                 in6_ifa_hold(ifp);
2807                 addrconf_dad_stop(ifp);
2808                 return;
2809         }
2810
2811         /*
2812          * Optimistic nodes can start receiving
2813          * Frames right away
2814          */
2815         if(ifp->flags & IFA_F_OPTIMISTIC)
2816                 ip6_ins_rt(ifp->rt);
2817
2818         addrconf_dad_kick(ifp);
2819         spin_unlock_bh(&ifp->lock);
2820 out:
2821         read_unlock_bh(&idev->lock);
2822 }
2823
2824 static void addrconf_dad_timer(unsigned long data)
2825 {
2826         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2827         struct inet6_dev *idev = ifp->idev;
2828         struct in6_addr mcaddr;
2829
2830         read_lock_bh(&idev->lock);
2831         if (idev->dead) {
2832                 read_unlock_bh(&idev->lock);
2833                 goto out;
2834         }
2835         spin_lock_bh(&ifp->lock);
2836         if (ifp->probes == 0) {
2837                 /*
2838                  * DAD was successful
2839                  */
2840
2841                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2842                 spin_unlock_bh(&ifp->lock);
2843                 read_unlock_bh(&idev->lock);
2844
2845                 addrconf_dad_completed(ifp);
2846
2847                 goto out;
2848         }
2849
2850         ifp->probes--;
2851         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2852         spin_unlock_bh(&ifp->lock);
2853         read_unlock_bh(&idev->lock);
2854
2855         /* send a neighbour solicitation for our addr */
2856         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2857         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2858 out:
2859         in6_ifa_put(ifp);
2860 }
2861
2862 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2863 {
2864         struct net_device *     dev = ifp->idev->dev;
2865
2866         /*
2867          *      Configure the address for reception. Now it is valid.
2868          */
2869
2870         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2871
2872         /* If added prefix is link local and forwarding is off,
2873            start sending router solicitations.
2874          */
2875
2876         if (ifp->idev->cnf.forwarding == 0 &&
2877             ifp->idev->cnf.rtr_solicits > 0 &&
2878             (dev->flags&IFF_LOOPBACK) == 0 &&
2879             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2880                 /*
2881                  *      If a host as already performed a random delay
2882                  *      [...] as part of DAD [...] there is no need
2883                  *      to delay again before sending the first RS
2884                  */
2885                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2886
2887                 spin_lock_bh(&ifp->lock);
2888                 ifp->probes = 1;
2889                 ifp->idev->if_flags |= IF_RS_SENT;
2890                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2891                 spin_unlock_bh(&ifp->lock);
2892         }
2893 }
2894
2895 static void addrconf_dad_run(struct inet6_dev *idev) {
2896         struct inet6_ifaddr *ifp;
2897
2898         read_lock_bh(&idev->lock);
2899         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2900                 spin_lock_bh(&ifp->lock);
2901                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2902                         spin_unlock_bh(&ifp->lock);
2903                         continue;
2904                 }
2905                 spin_unlock_bh(&ifp->lock);
2906                 addrconf_dad_kick(ifp);
2907         }
2908         read_unlock_bh(&idev->lock);
2909 }
2910
2911 #ifdef CONFIG_PROC_FS
2912 struct if6_iter_state {
2913         struct seq_net_private p;
2914         int bucket;
2915 };
2916
2917 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2918 {
2919         struct inet6_ifaddr *ifa = NULL;
2920         struct if6_iter_state *state = seq->private;
2921         struct net *net = seq_file_net(seq);
2922
2923         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2924                 ifa = inet6_addr_lst[state->bucket];
2925
2926                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2927                         ifa = ifa->lst_next;
2928                 if (ifa)
2929                         break;
2930         }
2931         return ifa;
2932 }
2933
2934 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2935 {
2936         struct if6_iter_state *state = seq->private;
2937         struct net *net = seq_file_net(seq);
2938
2939         ifa = ifa->lst_next;
2940 try_again:
2941         if (ifa) {
2942                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2943                         ifa = ifa->lst_next;
2944                         goto try_again;
2945                 }
2946         }
2947
2948         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2949                 ifa = inet6_addr_lst[state->bucket];
2950                 goto try_again;
2951         }
2952
2953         return ifa;
2954 }
2955
2956 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2957 {
2958         struct inet6_ifaddr *ifa = if6_get_first(seq);
2959
2960         if (ifa)
2961                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2962                         --pos;
2963         return pos ? NULL : ifa;
2964 }
2965
2966 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2967         __acquires(addrconf_hash_lock)
2968 {
2969         read_lock_bh(&addrconf_hash_lock);
2970         return if6_get_idx(seq, *pos);
2971 }
2972
2973 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2974 {
2975         struct inet6_ifaddr *ifa;
2976
2977         ifa = if6_get_next(seq, v);
2978         ++*pos;
2979         return ifa;
2980 }
2981
2982 static void if6_seq_stop(struct seq_file *seq, void *v)
2983         __releases(addrconf_hash_lock)
2984 {
2985         read_unlock_bh(&addrconf_hash_lock);
2986 }
2987
2988 static int if6_seq_show(struct seq_file *seq, void *v)
2989 {
2990         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2991         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2992                    &ifp->addr,
2993                    ifp->idev->dev->ifindex,
2994                    ifp->prefix_len,
2995                    ifp->scope,
2996                    ifp->flags,
2997                    ifp->idev->dev->name);
2998         return 0;
2999 }
3000
3001 static const struct seq_operations if6_seq_ops = {
3002         .start  = if6_seq_start,
3003         .next   = if6_seq_next,
3004         .show   = if6_seq_show,
3005         .stop   = if6_seq_stop,
3006 };
3007
3008 static int if6_seq_open(struct inode *inode, struct file *file)
3009 {
3010         return seq_open_net(inode, file, &if6_seq_ops,
3011                             sizeof(struct if6_iter_state));
3012 }
3013
3014 static const struct file_operations if6_fops = {
3015         .owner          = THIS_MODULE,
3016         .open           = if6_seq_open,
3017         .read           = seq_read,
3018         .llseek         = seq_lseek,
3019         .release        = seq_release_net,
3020 };
3021
3022 static int if6_proc_net_init(struct net *net)
3023 {
3024         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3025                 return -ENOMEM;
3026         return 0;
3027 }
3028
3029 static void if6_proc_net_exit(struct net *net)
3030 {
3031        proc_net_remove(net, "if_inet6");
3032 }
3033
3034 static struct pernet_operations if6_proc_net_ops = {
3035        .init = if6_proc_net_init,
3036        .exit = if6_proc_net_exit,
3037 };
3038
3039 int __init if6_proc_init(void)
3040 {
3041         return register_pernet_subsys(&if6_proc_net_ops);
3042 }
3043
3044 void if6_proc_exit(void)
3045 {
3046         unregister_pernet_subsys(&if6_proc_net_ops);
3047 }
3048 #endif  /* CONFIG_PROC_FS */
3049
3050 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3051 /* Check if address is a home address configured on any interface. */
3052 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3053 {
3054         int ret = 0;
3055         struct inet6_ifaddr * ifp;
3056         u8 hash = ipv6_addr_hash(addr);
3057         read_lock_bh(&addrconf_hash_lock);
3058         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3059                 if (!net_eq(dev_net(ifp->idev->dev), net))
3060                         continue;
3061                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3062                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3063                         ret = 1;
3064                         break;
3065                 }
3066         }
3067         read_unlock_bh(&addrconf_hash_lock);
3068         return ret;
3069 }
3070 #endif
3071
3072 /*
3073  *      Periodic address status verification
3074  */
3075
3076 static void addrconf_verify(unsigned long foo)
3077 {
3078         struct inet6_ifaddr *ifp;
3079         unsigned long now, next;
3080         int i;
3081
3082         spin_lock_bh(&addrconf_verify_lock);
3083         now = jiffies;
3084         next = now + ADDR_CHECK_FREQUENCY;
3085
3086         del_timer(&addr_chk_timer);
3087
3088         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3089
3090 restart:
3091                 read_lock(&addrconf_hash_lock);
3092                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3093                         unsigned long age;
3094 #ifdef CONFIG_IPV6_PRIVACY
3095                         unsigned long regen_advance;
3096 #endif
3097
3098                         if (ifp->flags & IFA_F_PERMANENT)
3099                                 continue;
3100
3101                         spin_lock(&ifp->lock);
3102                         age = (now - ifp->tstamp) / HZ;
3103
3104 #ifdef CONFIG_IPV6_PRIVACY
3105                         regen_advance = ifp->idev->cnf.regen_max_retry *
3106                                         ifp->idev->cnf.dad_transmits *
3107                                         ifp->idev->nd_parms->retrans_time / HZ;
3108 #endif
3109
3110                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3111                             age >= ifp->valid_lft) {
3112                                 spin_unlock(&ifp->lock);
3113                                 in6_ifa_hold(ifp);
3114                                 read_unlock(&addrconf_hash_lock);
3115                                 ipv6_del_addr(ifp);
3116                                 goto restart;
3117                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3118                                 spin_unlock(&ifp->lock);
3119                                 continue;
3120                         } else if (age >= ifp->prefered_lft) {
3121                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3122                                 int deprecate = 0;
3123
3124                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3125                                         deprecate = 1;
3126                                         ifp->flags |= IFA_F_DEPRECATED;
3127                                 }
3128
3129                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3130                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3131
3132                                 spin_unlock(&ifp->lock);
3133
3134                                 if (deprecate) {
3135                                         in6_ifa_hold(ifp);
3136                                         read_unlock(&addrconf_hash_lock);
3137
3138                                         ipv6_ifa_notify(0, ifp);
3139                                         in6_ifa_put(ifp);
3140                                         goto restart;
3141                                 }
3142 #ifdef CONFIG_IPV6_PRIVACY
3143                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3144                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3145                                 if (age >= ifp->prefered_lft - regen_advance) {
3146                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3147                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3148                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3149                                         if (!ifp->regen_count && ifpub) {
3150                                                 ifp->regen_count++;
3151                                                 in6_ifa_hold(ifp);
3152                                                 in6_ifa_hold(ifpub);
3153                                                 spin_unlock(&ifp->lock);
3154                                                 read_unlock(&addrconf_hash_lock);
3155                                                 spin_lock(&ifpub->lock);
3156                                                 ifpub->regen_count = 0;
3157                                                 spin_unlock(&ifpub->lock);
3158                                                 ipv6_create_tempaddr(ifpub, ifp);
3159                                                 in6_ifa_put(ifpub);
3160                                                 in6_ifa_put(ifp);
3161                                                 goto restart;
3162                                         }
3163                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3164                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3165                                 spin_unlock(&ifp->lock);
3166 #endif
3167                         } else {
3168                                 /* ifp->prefered_lft <= ifp->valid_lft */
3169                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3170                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3171                                 spin_unlock(&ifp->lock);
3172                         }
3173                 }
3174                 read_unlock(&addrconf_hash_lock);
3175         }
3176
3177         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3178         add_timer(&addr_chk_timer);
3179         spin_unlock_bh(&addrconf_verify_lock);
3180 }
3181
3182 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3183 {
3184         struct in6_addr *pfx = NULL;
3185
3186         if (addr)
3187                 pfx = nla_data(addr);
3188
3189         if (local) {
3190                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3191                         pfx = NULL;
3192                 else
3193                         pfx = nla_data(local);
3194         }
3195
3196         return pfx;
3197 }
3198
3199 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3200         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3201         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3202         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3203 };
3204
3205 static int
3206 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3207 {
3208         struct net *net = sock_net(skb->sk);
3209         struct ifaddrmsg *ifm;
3210         struct nlattr *tb[IFA_MAX+1];
3211         struct in6_addr *pfx;
3212         int err;
3213
3214         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3215         if (err < 0)
3216                 return err;
3217
3218         ifm = nlmsg_data(nlh);
3219         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3220         if (pfx == NULL)
3221                 return -EINVAL;
3222
3223         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3224 }
3225
3226 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3227                              u32 prefered_lft, u32 valid_lft)
3228 {
3229         u32 flags;
3230         clock_t expires;
3231         unsigned long timeout;
3232
3233         if (!valid_lft || (prefered_lft > valid_lft))
3234                 return -EINVAL;
3235
3236         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3237         if (addrconf_finite_timeout(timeout)) {
3238                 expires = jiffies_to_clock_t(timeout * HZ);
3239                 valid_lft = timeout;
3240                 flags = RTF_EXPIRES;
3241         } else {
3242                 expires = 0;
3243                 flags = 0;
3244                 ifa_flags |= IFA_F_PERMANENT;
3245         }
3246
3247         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3248         if (addrconf_finite_timeout(timeout)) {
3249                 if (timeout == 0)
3250                         ifa_flags |= IFA_F_DEPRECATED;
3251                 prefered_lft = timeout;
3252         }
3253
3254         spin_lock_bh(&ifp->lock);
3255         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3256         ifp->tstamp = jiffies;
3257         ifp->valid_lft = valid_lft;
3258         ifp->prefered_lft = prefered_lft;
3259
3260         spin_unlock_bh(&ifp->lock);
3261         if (!(ifp->flags&IFA_F_TENTATIVE))
3262                 ipv6_ifa_notify(0, ifp);
3263
3264         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3265                               expires, flags);
3266         addrconf_verify(0);
3267
3268         return 0;
3269 }
3270
3271 static int
3272 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3273 {
3274         struct net *net = sock_net(skb->sk);
3275         struct ifaddrmsg *ifm;
3276         struct nlattr *tb[IFA_MAX+1];
3277         struct in6_addr *pfx;
3278         struct inet6_ifaddr *ifa;
3279         struct net_device *dev;
3280         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3281         u8 ifa_flags;
3282         int err;
3283
3284         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3285         if (err < 0)
3286                 return err;
3287
3288         ifm = nlmsg_data(nlh);
3289         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3290         if (pfx == NULL)
3291                 return -EINVAL;
3292
3293         if (tb[IFA_CACHEINFO]) {
3294                 struct ifa_cacheinfo *ci;
3295
3296                 ci = nla_data(tb[IFA_CACHEINFO]);
3297                 valid_lft = ci->ifa_valid;
3298                 preferred_lft = ci->ifa_prefered;
3299         } else {
3300                 preferred_lft = INFINITY_LIFE_TIME;
3301                 valid_lft = INFINITY_LIFE_TIME;
3302         }
3303
3304         dev =  __dev_get_by_index(net, ifm->ifa_index);
3305         if (dev == NULL)
3306                 return -ENODEV;
3307
3308         /* We ignore other flags so far. */
3309         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3310
3311         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3312         if (ifa == NULL) {
3313                 /*
3314                  * It would be best to check for !NLM_F_CREATE here but
3315                  * userspace alreay relies on not having to provide this.
3316                  */
3317                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3318                                       ifm->ifa_prefixlen, ifa_flags,
3319                                       preferred_lft, valid_lft);
3320         }
3321
3322         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3323             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3324                 err = -EEXIST;
3325         else
3326                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3327
3328         in6_ifa_put(ifa);
3329
3330         return err;
3331 }
3332
3333 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3334                           u8 scope, int ifindex)
3335 {
3336         struct ifaddrmsg *ifm;
3337
3338         ifm = nlmsg_data(nlh);
3339         ifm->ifa_family = AF_INET6;
3340         ifm->ifa_prefixlen = prefixlen;
3341         ifm->ifa_flags = flags;
3342         ifm->ifa_scope = scope;
3343         ifm->ifa_index = ifindex;
3344 }
3345
3346 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3347                          unsigned long tstamp, u32 preferred, u32 valid)
3348 {
3349         struct ifa_cacheinfo ci;
3350
3351         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3352                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3353         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3354                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3355         ci.ifa_prefered = preferred;
3356         ci.ifa_valid = valid;
3357
3358         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3359 }
3360
3361 static inline int rt_scope(int ifa_scope)
3362 {
3363         if (ifa_scope & IFA_HOST)
3364                 return RT_SCOPE_HOST;
3365         else if (ifa_scope & IFA_LINK)
3366                 return RT_SCOPE_LINK;
3367         else if (ifa_scope & IFA_SITE)
3368                 return RT_SCOPE_SITE;
3369         else
3370                 return RT_SCOPE_UNIVERSE;
3371 }
3372
3373 static inline int inet6_ifaddr_msgsize(void)
3374 {
3375         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3376                + nla_total_size(16) /* IFA_ADDRESS */
3377                + nla_total_size(sizeof(struct ifa_cacheinfo));
3378 }
3379
3380 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3381                              u32 pid, u32 seq, int event, unsigned int flags)
3382 {
3383         struct nlmsghdr  *nlh;
3384         u32 preferred, valid;
3385
3386         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3387         if (nlh == NULL)
3388                 return -EMSGSIZE;
3389
3390         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3391                       ifa->idev->dev->ifindex);
3392
3393         if (!(ifa->flags&IFA_F_PERMANENT)) {
3394                 preferred = ifa->prefered_lft;
3395                 valid = ifa->valid_lft;
3396                 if (preferred != INFINITY_LIFE_TIME) {
3397                         long tval = (jiffies - ifa->tstamp)/HZ;
3398                         preferred -= tval;
3399                         if (valid != INFINITY_LIFE_TIME)
3400                                 valid -= tval;
3401                 }
3402         } else {
3403                 preferred = INFINITY_LIFE_TIME;
3404                 valid = INFINITY_LIFE_TIME;
3405         }
3406
3407         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3408             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3409                 nlmsg_cancel(skb, nlh);
3410                 return -EMSGSIZE;
3411         }
3412
3413         return nlmsg_end(skb, nlh);
3414 }
3415
3416 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3417                                 u32 pid, u32 seq, int event, u16 flags)
3418 {
3419         struct nlmsghdr  *nlh;
3420         u8 scope = RT_SCOPE_UNIVERSE;
3421         int ifindex = ifmca->idev->dev->ifindex;
3422
3423         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3424                 scope = RT_SCOPE_SITE;
3425
3426         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3427         if (nlh == NULL)
3428                 return -EMSGSIZE;
3429
3430         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3431         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3432             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3433                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3434                 nlmsg_cancel(skb, nlh);
3435                 return -EMSGSIZE;
3436         }
3437
3438         return nlmsg_end(skb, nlh);
3439 }
3440
3441 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3442                                 u32 pid, u32 seq, int event, unsigned int flags)
3443 {
3444         struct nlmsghdr  *nlh;
3445         u8 scope = RT_SCOPE_UNIVERSE;
3446         int ifindex = ifaca->aca_idev->dev->ifindex;
3447
3448         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3449                 scope = RT_SCOPE_SITE;
3450
3451         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3452         if (nlh == NULL)
3453                 return -EMSGSIZE;
3454
3455         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3456         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3457             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3458                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3459                 nlmsg_cancel(skb, nlh);
3460                 return -EMSGSIZE;
3461         }
3462
3463         return nlmsg_end(skb, nlh);
3464 }
3465
3466 enum addr_type_t
3467 {
3468         UNICAST_ADDR,
3469         MULTICAST_ADDR,
3470         ANYCAST_ADDR,
3471 };
3472
3473 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3474                            enum addr_type_t type)
3475 {
3476         int idx, ip_idx;
3477         int s_idx, s_ip_idx;
3478         int err = 1;
3479         struct net_device *dev;
3480         struct inet6_dev *idev = NULL;
3481         struct inet6_ifaddr *ifa;
3482         struct ifmcaddr6 *ifmca;
3483         struct ifacaddr6 *ifaca;
3484         struct net *net = sock_net(skb->sk);
3485
3486         s_idx = cb->args[0];
3487         s_ip_idx = ip_idx = cb->args[1];
3488
3489         idx = 0;
3490         for_each_netdev(net, dev) {
3491                 if (idx < s_idx)
3492                         goto cont;
3493                 if (idx > s_idx)
3494                         s_ip_idx = 0;
3495                 ip_idx = 0;
3496                 if ((idev = in6_dev_get(dev)) == NULL)
3497                         goto cont;
3498                 read_lock_bh(&idev->lock);
3499                 switch (type) {
3500                 case UNICAST_ADDR:
3501                         /* unicast address incl. temp addr */
3502                         for (ifa = idev->addr_list; ifa;
3503                              ifa = ifa->if_next, ip_idx++) {
3504                                 if (ip_idx < s_ip_idx)
3505                                         continue;
3506                                 err = inet6_fill_ifaddr(skb, ifa,
3507                                                         NETLINK_CB(cb->skb).pid,
3508                                                         cb->nlh->nlmsg_seq,
3509                                                         RTM_NEWADDR,
3510                                                         NLM_F_MULTI);
3511                         }
3512                         break;
3513                 case MULTICAST_ADDR:
3514                         /* multicast address */
3515                         for (ifmca = idev->mc_list; ifmca;
3516                              ifmca = ifmca->next, ip_idx++) {
3517                                 if (ip_idx < s_ip_idx)
3518                                         continue;
3519                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3520                                                           NETLINK_CB(cb->skb).pid,
3521                                                           cb->nlh->nlmsg_seq,
3522                                                           RTM_GETMULTICAST,
3523                                                           NLM_F_MULTI);
3524                         }
3525                         break;
3526                 case ANYCAST_ADDR:
3527                         /* anycast address */
3528                         for (ifaca = idev->ac_list; ifaca;
3529                              ifaca = ifaca->aca_next, ip_idx++) {
3530                                 if (ip_idx < s_ip_idx)
3531                                         continue;
3532                                 err = inet6_fill_ifacaddr(skb, ifaca,
3533                                                           NETLINK_CB(cb->skb).pid,
3534                                                           cb->nlh->nlmsg_seq,
3535                                                           RTM_GETANYCAST,
3536                                                           NLM_F_MULTI);
3537                         }
3538                         break;
3539                 default:
3540                         break;
3541                 }
3542                 read_unlock_bh(&idev->lock);
3543                 in6_dev_put(idev);
3544
3545                 if (err <= 0)
3546                         break;
3547 cont:
3548                 idx++;
3549         }
3550         cb->args[0] = idx;
3551         cb->args[1] = ip_idx;
3552         return skb->len;
3553 }
3554
3555 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3556 {
3557         enum addr_type_t type = UNICAST_ADDR;
3558
3559         return inet6_dump_addr(skb, cb, type);
3560 }
3561
3562 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3563 {
3564         enum addr_type_t type = MULTICAST_ADDR;
3565
3566         return inet6_dump_addr(skb, cb, type);
3567 }
3568
3569
3570 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3571 {
3572         enum addr_type_t type = ANYCAST_ADDR;
3573
3574         return inet6_dump_addr(skb, cb, type);
3575 }
3576
3577 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3578                              void *arg)
3579 {
3580         struct net *net = sock_net(in_skb->sk);
3581         struct ifaddrmsg *ifm;
3582         struct nlattr *tb[IFA_MAX+1];
3583         struct in6_addr *addr = NULL;
3584         struct net_device *dev = NULL;
3585         struct inet6_ifaddr *ifa;
3586         struct sk_buff *skb;
3587         int err;
3588
3589         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3590         if (err < 0)
3591                 goto errout;
3592
3593         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3594         if (addr == NULL) {
3595                 err = -EINVAL;
3596                 goto errout;
3597         }
3598
3599         ifm = nlmsg_data(nlh);
3600         if (ifm->ifa_index)
3601                 dev = __dev_get_by_index(net, ifm->ifa_index);
3602
3603         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3604                 err = -EADDRNOTAVAIL;
3605                 goto errout;
3606         }
3607
3608         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3609                 err = -ENOBUFS;
3610                 goto errout_ifa;
3611         }
3612
3613         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3614                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3615         if (err < 0) {
3616                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3617                 WARN_ON(err == -EMSGSIZE);
3618                 kfree_skb(skb);
3619                 goto errout_ifa;
3620         }
3621         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3622 errout_ifa:
3623         in6_ifa_put(ifa);
3624 errout:
3625         return err;
3626 }
3627
3628 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3629 {
3630         struct sk_buff *skb;
3631         struct net *net = dev_net(ifa->idev->dev);
3632         int err = -ENOBUFS;
3633
3634         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3635         if (skb == NULL)
3636                 goto errout;
3637
3638         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3639         if (err < 0) {
3640                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3641                 WARN_ON(err == -EMSGSIZE);
3642                 kfree_skb(skb);
3643                 goto errout;
3644         }
3645         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3646         return;
3647 errout:
3648         if (err < 0)
3649                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3650 }
3651
3652 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3653                                 __s32 *array, int bytes)
3654 {
3655         BUG_ON(bytes < (DEVCONF_MAX * 4));
3656
3657         memset(array, 0, bytes);
3658         array[DEVCONF_FORWARDING] = cnf->forwarding;
3659         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3660         array[DEVCONF_MTU6] = cnf->mtu6;
3661         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3662         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3663         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3664         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3665         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3666         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3667         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3668         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3669 #ifdef CONFIG_IPV6_PRIVACY
3670         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3671         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3672         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3673         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3674         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3675 #endif
3676         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3677         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3678         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3679 #ifdef CONFIG_IPV6_ROUTER_PREF
3680         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3681         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3682 #ifdef CONFIG_IPV6_ROUTE_INFO
3683         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3684 #endif
3685 #endif
3686         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3687         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3688 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3689         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3690 #endif
3691 #ifdef CONFIG_IPV6_MROUTE
3692         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3693 #endif
3694         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3695         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3696 }
3697
3698 static inline size_t inet6_if_nlmsg_size(void)
3699 {
3700         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3701                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3702                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3703                + nla_total_size(4) /* IFLA_MTU */
3704                + nla_total_size(4) /* IFLA_LINK */
3705                + nla_total_size( /* IFLA_PROTINFO */
3706                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3707                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3708                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3709                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3710                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3711                  );
3712 }
3713
3714 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3715                                       int bytes)
3716 {
3717         int i;
3718         int pad = bytes - sizeof(u64) * items;
3719         BUG_ON(pad < 0);
3720
3721         /* Use put_unaligned() because stats may not be aligned for u64. */
3722         put_unaligned(items, &stats[0]);
3723         for (i = 1; i < items; i++)
3724                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3725
3726         memset(&stats[items], 0, pad);
3727 }
3728
3729 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3730                              int bytes)
3731 {
3732         switch(attrtype) {
3733         case IFLA_INET6_STATS:
3734                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3735                 break;
3736         case IFLA_INET6_ICMP6STATS:
3737                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3738                 break;
3739         }
3740 }
3741
3742 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3743                              u32 pid, u32 seq, int event, unsigned int flags)
3744 {
3745         struct net_device *dev = idev->dev;
3746         struct nlattr *nla;
3747         struct ifinfomsg *hdr;
3748         struct nlmsghdr *nlh;
3749         void *protoinfo;
3750         struct ifla_cacheinfo ci;
3751
3752         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3753         if (nlh == NULL)
3754                 return -EMSGSIZE;
3755
3756         hdr = nlmsg_data(nlh);
3757         hdr->ifi_family = AF_INET6;
3758         hdr->__ifi_pad = 0;
3759         hdr->ifi_type = dev->type;
3760         hdr->ifi_index = dev->ifindex;
3761         hdr->ifi_flags = dev_get_flags(dev);
3762         hdr->ifi_change = 0;
3763
3764         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3765
3766         if (dev->addr_len)
3767                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3768
3769         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3770         if (dev->ifindex != dev->iflink)
3771                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3772
3773         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3774         if (protoinfo == NULL)
3775                 goto nla_put_failure;
3776
3777         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3778
3779         ci.max_reasm_len = IPV6_MAXPLEN;
3780         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3781                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3782         ci.reachable_time = idev->nd_parms->reachable_time;
3783         ci.retrans_time = idev->nd_parms->retrans_time;
3784         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3785
3786         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3787         if (nla == NULL)
3788                 goto nla_put_failure;
3789         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3790
3791         /* XXX - MC not implemented */
3792
3793         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3794         if (nla == NULL)
3795                 goto nla_put_failure;
3796         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3797
3798         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3799         if (nla == NULL)
3800                 goto nla_put_failure;
3801         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3802
3803         nla_nest_end(skb, protoinfo);
3804         return nlmsg_end(skb, nlh);
3805
3806 nla_put_failure:
3807         nlmsg_cancel(skb, nlh);
3808         return -EMSGSIZE;
3809 }
3810
3811 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3812 {
3813         struct net *net = sock_net(skb->sk);
3814         int idx, err;
3815         int s_idx = cb->args[0];
3816         struct net_device *dev;
3817         struct inet6_dev *idev;
3818
3819         read_lock(&dev_base_lock);
3820         idx = 0;
3821         for_each_netdev(net, dev) {
3822                 if (idx < s_idx)
3823                         goto cont;
3824                 if ((idev = in6_dev_get(dev)) == NULL)
3825                         goto cont;
3826                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3827                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3828                 in6_dev_put(idev);
3829                 if (err <= 0)
3830                         break;
3831 cont:
3832                 idx++;
3833         }
3834         read_unlock(&dev_base_lock);
3835         cb->args[0] = idx;
3836
3837         return skb->len;
3838 }
3839
3840 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3841 {
3842         struct sk_buff *skb;
3843         struct net *net = dev_net(idev->dev);
3844         int err = -ENOBUFS;
3845
3846         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3847         if (skb == NULL)
3848                 goto errout;
3849
3850         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3851         if (err < 0) {
3852                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3853                 WARN_ON(err == -EMSGSIZE);
3854                 kfree_skb(skb);
3855                 goto errout;
3856         }
3857         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3858         return;
3859 errout:
3860         if (err < 0)
3861                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3862 }
3863
3864 static inline size_t inet6_prefix_nlmsg_size(void)
3865 {
3866         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3867                + nla_total_size(sizeof(struct in6_addr))
3868                + nla_total_size(sizeof(struct prefix_cacheinfo));
3869 }
3870
3871 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3872                              struct prefix_info *pinfo, u32 pid, u32 seq,
3873                              int event, unsigned int flags)
3874 {
3875         struct prefixmsg *pmsg;
3876         struct nlmsghdr *nlh;
3877         struct prefix_cacheinfo ci;
3878
3879         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3880         if (nlh == NULL)
3881                 return -EMSGSIZE;
3882
3883         pmsg = nlmsg_data(nlh);
3884         pmsg->prefix_family = AF_INET6;
3885         pmsg->prefix_pad1 = 0;
3886         pmsg->prefix_pad2 = 0;
3887         pmsg->prefix_ifindex = idev->dev->ifindex;
3888         pmsg->prefix_len = pinfo->prefix_len;
3889         pmsg->prefix_type = pinfo->type;
3890         pmsg->prefix_pad3 = 0;
3891         pmsg->prefix_flags = 0;
3892         if (pinfo->onlink)
3893                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3894         if (pinfo->autoconf)
3895                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3896
3897         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3898
3899         ci.preferred_time = ntohl(pinfo->prefered);
3900         ci.valid_time = ntohl(pinfo->valid);
3901         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3902
3903         return nlmsg_end(skb, nlh);
3904
3905 nla_put_failure:
3906         nlmsg_cancel(skb, nlh);
3907         return -EMSGSIZE;
3908 }
3909
3910 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3911                          struct prefix_info *pinfo)
3912 {
3913         struct sk_buff *skb;
3914         struct net *net = dev_net(idev->dev);
3915         int err = -ENOBUFS;
3916
3917         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3918         if (skb == NULL)
3919                 goto errout;
3920
3921         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3922         if (err < 0) {
3923                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3924                 WARN_ON(err == -EMSGSIZE);
3925                 kfree_skb(skb);
3926                 goto errout;
3927         }
3928         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3929         return;
3930 errout:
3931         if (err < 0)
3932                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3933 }
3934
3935 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3936 {
3937         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3938
3939         switch (event) {
3940         case RTM_NEWADDR:
3941                 /*
3942                  * If the address was optimistic
3943                  * we inserted the route at the start of
3944                  * our DAD process, so we don't need
3945                  * to do it again
3946                  */
3947                 if (!(ifp->rt->rt6i_node))
3948                         ip6_ins_rt(ifp->rt);
3949                 if (ifp->idev->cnf.forwarding)
3950                         addrconf_join_anycast(ifp);
3951                 break;
3952         case RTM_DELADDR:
3953                 if (ifp->idev->cnf.forwarding)
3954                         addrconf_leave_anycast(ifp);
3955                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3956                 dst_hold(&ifp->rt->u.dst);
3957                 if (ip6_del_rt(ifp->rt))
3958                         dst_free(&ifp->rt->u.dst);
3959                 break;
3960         }
3961 }
3962
3963 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3964 {
3965         rcu_read_lock_bh();
3966         if (likely(ifp->idev->dead == 0))
3967                 __ipv6_ifa_notify(event, ifp);
3968         rcu_read_unlock_bh();
3969 }
3970
3971 #ifdef CONFIG_SYSCTL
3972
3973 static
3974 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3975                            void __user *buffer, size_t *lenp, loff_t *ppos)
3976 {
3977         int *valp = ctl->data;
3978         int val = *valp;
3979         int ret;
3980
3981         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3982
3983         if (write)
3984                 ret = addrconf_fixup_forwarding(ctl, valp, val);
3985         return ret;
3986 }
3987
3988 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3989                                             void __user *oldval,
3990                                             size_t __user *oldlenp,
3991                                             void __user *newval, size_t newlen)
3992 {
3993         int *valp = table->data;
3994         int val = *valp;
3995         int new;
3996
3997         if (!newval || !newlen)
3998                 return 0;
3999         if (newlen != sizeof(int))
4000                 return -EINVAL;
4001         if (get_user(new, (int __user *)newval))
4002                 return -EFAULT;
4003         if (new == *valp)
4004                 return 0;
4005         if (oldval && oldlenp) {
4006                 size_t len;
4007                 if (get_user(len, oldlenp))
4008                         return -EFAULT;
4009                 if (len) {
4010                         if (len > table->maxlen)
4011                                 len = table->maxlen;
4012                         if (copy_to_user(oldval, valp, len))
4013                                 return -EFAULT;
4014                         if (put_user(len, oldlenp))
4015                                 return -EFAULT;
4016                 }
4017         }
4018
4019         *valp = new;
4020         return addrconf_fixup_forwarding(table, valp, val);
4021 }
4022
4023 static struct addrconf_sysctl_table
4024 {
4025         struct ctl_table_header *sysctl_header;
4026         ctl_table addrconf_vars[DEVCONF_MAX+1];
4027         char *dev_name;
4028 } addrconf_sysctl __read_mostly = {
4029         .sysctl_header = NULL,
4030         .addrconf_vars = {
4031                 {
4032                         .ctl_name       =       NET_IPV6_FORWARDING,
4033                         .procname       =       "forwarding",
4034                         .data           =       &ipv6_devconf.forwarding,
4035                         .maxlen         =       sizeof(int),
4036                         .mode           =       0644,
4037                         .proc_handler   =       addrconf_sysctl_forward,
4038                         .strategy       =       addrconf_sysctl_forward_strategy,
4039                 },
4040                 {
4041                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4042                         .procname       =       "hop_limit",
4043                         .data           =       &ipv6_devconf.hop_limit,
4044                         .maxlen         =       sizeof(int),
4045                         .mode           =       0644,
4046                         .proc_handler   =       proc_dointvec,
4047                 },
4048                 {
4049                         .ctl_name       =       NET_IPV6_MTU,
4050                         .procname       =       "mtu",
4051                         .data           =       &ipv6_devconf.mtu6,
4052                         .maxlen         =       sizeof(int),
4053                         .mode           =       0644,
4054                         .proc_handler   =       proc_dointvec,
4055                 },
4056                 {
4057                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4058                         .procname       =       "accept_ra",
4059                         .data           =       &ipv6_devconf.accept_ra,
4060                         .maxlen         =       sizeof(int),
4061                         .mode           =       0644,
4062                         .proc_handler   =       proc_dointvec,
4063                 },
4064                 {
4065                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4066                         .procname       =       "accept_redirects",
4067                         .data           =       &ipv6_devconf.accept_redirects,
4068                         .maxlen         =       sizeof(int),
4069                         .mode           =       0644,
4070                         .proc_handler   =       proc_dointvec,
4071                 },
4072                 {
4073                         .ctl_name       =       NET_IPV6_AUTOCONF,
4074                         .procname       =       "autoconf",
4075                         .data           =       &ipv6_devconf.autoconf,
4076                         .maxlen         =       sizeof(int),
4077                         .mode           =       0644,
4078                         .proc_handler   =       proc_dointvec,
4079                 },
4080                 {
4081                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4082                         .procname       =       "dad_transmits",
4083                         .data           =       &ipv6_devconf.dad_transmits,
4084                         .maxlen         =       sizeof(int),
4085                         .mode           =       0644,
4086                         .proc_handler   =       proc_dointvec,
4087                 },
4088                 {
4089                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4090                         .procname       =       "router_solicitations",
4091                         .data           =       &ipv6_devconf.rtr_solicits,
4092                         .maxlen         =       sizeof(int),
4093                         .mode           =       0644,
4094                         .proc_handler   =       proc_dointvec,
4095                 },
4096                 {
4097                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4098                         .procname       =       "router_solicitation_interval",
4099                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4100                         .maxlen         =       sizeof(int),
4101                         .mode           =       0644,
4102                         .proc_handler   =       proc_dointvec_jiffies,
4103                         .strategy       =       sysctl_jiffies,
4104                 },
4105                 {
4106                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4107                         .procname       =       "router_solicitation_delay",
4108                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4109                         .maxlen         =       sizeof(int),
4110                         .mode           =       0644,
4111                         .proc_handler   =       proc_dointvec_jiffies,
4112                         .strategy       =       sysctl_jiffies,
4113                 },
4114                 {
4115                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4116                         .procname       =       "force_mld_version",
4117                         .data           =       &ipv6_devconf.force_mld_version,
4118                         .maxlen         =       sizeof(int),
4119                         .mode           =       0644,
4120                         .proc_handler   =       proc_dointvec,
4121                 },
4122 #ifdef CONFIG_IPV6_PRIVACY
4123                 {
4124                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4125                         .procname       =       "use_tempaddr",
4126                         .data           =       &ipv6_devconf.use_tempaddr,
4127                         .maxlen         =       sizeof(int),
4128                         .mode           =       0644,
4129                         .proc_handler   =       proc_dointvec,
4130                 },
4131                 {
4132                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4133                         .procname       =       "temp_valid_lft",
4134                         .data           =       &ipv6_devconf.temp_valid_lft,
4135                         .maxlen         =       sizeof(int),
4136                         .mode           =       0644,
4137                         .proc_handler   =       proc_dointvec,
4138                 },
4139                 {
4140                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4141                         .procname       =       "temp_prefered_lft",
4142                         .data           =       &ipv6_devconf.temp_prefered_lft,
4143                         .maxlen         =       sizeof(int),
4144                         .mode           =       0644,
4145                         .proc_handler   =       proc_dointvec,
4146                 },
4147                 {
4148                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4149                         .procname       =       "regen_max_retry",
4150                         .data           =       &ipv6_devconf.regen_max_retry,
4151                         .maxlen         =       sizeof(int),
4152                         .mode           =       0644,
4153                         .proc_handler   =       proc_dointvec,
4154                 },
4155                 {
4156                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4157                         .procname       =       "max_desync_factor",
4158                         .data           =       &ipv6_devconf.max_desync_factor,
4159                         .maxlen         =       sizeof(int),
4160                         .mode           =       0644,
4161                         .proc_handler   =       proc_dointvec,
4162                 },
4163 #endif
4164                 {
4165                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4166                         .procname       =       "max_addresses",
4167                         .data           =       &ipv6_devconf.max_addresses,
4168                         .maxlen         =       sizeof(int),
4169                         .mode           =       0644,
4170                         .proc_handler   =       proc_dointvec,
4171                 },
4172                 {
4173                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4174                         .procname       =       "accept_ra_defrtr",
4175                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4176                         .maxlen         =       sizeof(int),
4177                         .mode           =       0644,
4178                         .proc_handler   =       proc_dointvec,
4179                 },
4180                 {
4181                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4182                         .procname       =       "accept_ra_pinfo",
4183                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4184                         .maxlen         =       sizeof(int),
4185                         .mode           =       0644,
4186                         .proc_handler   =       proc_dointvec,
4187                 },
4188 #ifdef CONFIG_IPV6_ROUTER_PREF
4189                 {
4190                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4191                         .procname       =       "accept_ra_rtr_pref",
4192                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4193                         .maxlen         =       sizeof(int),
4194                         .mode           =       0644,
4195                         .proc_handler   =       proc_dointvec,
4196                 },
4197                 {
4198                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4199                         .procname       =       "router_probe_interval",
4200                         .data           =       &ipv6_devconf.rtr_probe_interval,
4201                         .maxlen         =       sizeof(int),
4202                         .mode           =       0644,
4203                         .proc_handler   =       proc_dointvec_jiffies,
4204                         .strategy       =       sysctl_jiffies,
4205                 },
4206 #ifdef CONFIG_IPV6_ROUTE_INFO
4207                 {
4208                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4209                         .procname       =       "accept_ra_rt_info_max_plen",
4210                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4211                         .maxlen         =       sizeof(int),
4212                         .mode           =       0644,
4213                         .proc_handler   =       proc_dointvec,
4214                 },
4215 #endif
4216 #endif
4217                 {
4218                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4219                         .procname       =       "proxy_ndp",
4220                         .data           =       &ipv6_devconf.proxy_ndp,
4221                         .maxlen         =       sizeof(int),
4222                         .mode           =       0644,
4223                         .proc_handler   =       proc_dointvec,
4224                 },
4225                 {
4226                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4227                         .procname       =       "accept_source_route",
4228                         .data           =       &ipv6_devconf.accept_source_route,
4229                         .maxlen         =       sizeof(int),
4230                         .mode           =       0644,
4231                         .proc_handler   =       proc_dointvec,
4232                 },
4233 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4234                 {
4235                         .ctl_name       =       CTL_UNNUMBERED,
4236                         .procname       =       "optimistic_dad",
4237                         .data           =       &ipv6_devconf.optimistic_dad,
4238                         .maxlen         =       sizeof(int),
4239                         .mode           =       0644,
4240                         .proc_handler   =       proc_dointvec,
4241
4242                 },
4243 #endif
4244 #ifdef CONFIG_IPV6_MROUTE
4245                 {
4246                         .ctl_name       =       CTL_UNNUMBERED,
4247                         .procname       =       "mc_forwarding",
4248                         .data           =       &ipv6_devconf.mc_forwarding,
4249                         .maxlen         =       sizeof(int),
4250                         .mode           =       0444,
4251                         .proc_handler   =       proc_dointvec,
4252                 },
4253 #endif
4254                 {
4255                         .ctl_name       =       CTL_UNNUMBERED,
4256                         .procname       =       "disable_ipv6",
4257                         .data           =       &ipv6_devconf.disable_ipv6,
4258                         .maxlen         =       sizeof(int),
4259                         .mode           =       0644,
4260                         .proc_handler   =       proc_dointvec,
4261                 },
4262                 {
4263                         .ctl_name       =       CTL_UNNUMBERED,
4264                         .procname       =       "accept_dad",
4265                         .data           =       &ipv6_devconf.accept_dad,
4266                         .maxlen         =       sizeof(int),
4267                         .mode           =       0644,
4268                         .proc_handler   =       proc_dointvec,
4269                 },
4270                 {
4271                         .ctl_name       =       0,      /* sentinel */
4272                 }
4273         },
4274 };
4275
4276 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4277                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4278 {
4279         int i;
4280         struct addrconf_sysctl_table *t;
4281
4282 #define ADDRCONF_CTL_PATH_DEV   3
4283
4284         struct ctl_path addrconf_ctl_path[] = {
4285                 { .procname = "net", .ctl_name = CTL_NET, },
4286                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4287                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4288                 { /* to be set */ },
4289                 { },
4290         };
4291
4292
4293         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4294         if (t == NULL)
4295                 goto out;
4296
4297         for (i=0; t->addrconf_vars[i].data; i++) {
4298                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4299                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4300                 t->addrconf_vars[i].extra2 = net;
4301         }
4302
4303         /*
4304          * Make a copy of dev_name, because '.procname' is regarded as const
4305          * by sysctl and we wouldn't want anyone to change it under our feet
4306          * (see SIOCSIFNAME).
4307          */
4308         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4309         if (!t->dev_name)
4310                 goto free;
4311
4312         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4313         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4314
4315         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4316                         t->addrconf_vars);
4317         if (t->sysctl_header == NULL)
4318                 goto free_procname;
4319
4320         p->sysctl = t;
4321         return 0;
4322
4323 free_procname:
4324         kfree(t->dev_name);
4325 free:
4326         kfree(t);
4327 out:
4328         return -ENOBUFS;
4329 }
4330
4331 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4332 {
4333         struct addrconf_sysctl_table *t;
4334
4335         if (p->sysctl == NULL)
4336                 return;
4337
4338         t = p->sysctl;
4339         p->sysctl = NULL;
4340         unregister_sysctl_table(t->sysctl_header);
4341         kfree(t->dev_name);
4342         kfree(t);
4343 }
4344
4345 static void addrconf_sysctl_register(struct inet6_dev *idev)
4346 {
4347         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4348                               NET_IPV6_NEIGH, "ipv6",
4349                               &ndisc_ifinfo_sysctl_change,
4350                               ndisc_ifinfo_sysctl_strategy);
4351         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4352                         idev->dev->ifindex, idev, &idev->cnf);
4353 }
4354
4355 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4356 {
4357         __addrconf_sysctl_unregister(&idev->cnf);
4358         neigh_sysctl_unregister(idev->nd_parms);
4359 }
4360
4361
4362 #endif
4363
4364 static int addrconf_init_net(struct net *net)
4365 {
4366         int err;
4367         struct ipv6_devconf *all, *dflt;
4368
4369         err = -ENOMEM;
4370         all = &ipv6_devconf;
4371         dflt = &ipv6_devconf_dflt;
4372
4373         if (net != &init_net) {
4374                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4375                 if (all == NULL)
4376                         goto err_alloc_all;
4377
4378                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4379                 if (dflt == NULL)
4380                         goto err_alloc_dflt;
4381         }
4382
4383         net->ipv6.devconf_all = all;
4384         net->ipv6.devconf_dflt = dflt;
4385
4386 #ifdef CONFIG_SYSCTL
4387         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4388                         NULL, all);
4389         if (err < 0)
4390                 goto err_reg_all;
4391
4392         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4393                         NULL, dflt);
4394         if (err < 0)
4395                 goto err_reg_dflt;
4396 #endif
4397         return 0;
4398
4399 #ifdef CONFIG_SYSCTL
4400 err_reg_dflt:
4401         __addrconf_sysctl_unregister(all);
4402 err_reg_all:
4403         kfree(dflt);
4404 #endif
4405 err_alloc_dflt:
4406         kfree(all);
4407 err_alloc_all:
4408         return err;
4409 }
4410
4411 static void addrconf_exit_net(struct net *net)
4412 {
4413 #ifdef CONFIG_SYSCTL
4414         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4415         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4416 #endif
4417         if (net != &init_net) {
4418                 kfree(net->ipv6.devconf_dflt);
4419                 kfree(net->ipv6.devconf_all);
4420         }
4421 }
4422
4423 static struct pernet_operations addrconf_ops = {
4424         .init = addrconf_init_net,
4425         .exit = addrconf_exit_net,
4426 };
4427
4428 /*
4429  *      Device notifier
4430  */
4431
4432 int register_inet6addr_notifier(struct notifier_block *nb)
4433 {
4434         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4435 }
4436
4437 EXPORT_SYMBOL(register_inet6addr_notifier);
4438
4439 int unregister_inet6addr_notifier(struct notifier_block *nb)
4440 {
4441         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4442 }
4443
4444 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4445
4446 /*
4447  *      Init / cleanup code
4448  */
4449
4450 int __init addrconf_init(void)
4451 {
4452         int err;
4453
4454         if ((err = ipv6_addr_label_init()) < 0) {
4455                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4456                         err);
4457                 return err;
4458         }
4459
4460         register_pernet_subsys(&addrconf_ops);
4461
4462         /* The addrconf netdev notifier requires that loopback_dev
4463          * has it's ipv6 private information allocated and setup
4464          * before it can bring up and give link-local addresses
4465          * to other devices which are up.
4466          *
4467          * Unfortunately, loopback_dev is not necessarily the first
4468          * entry in the global dev_base list of net devices.  In fact,
4469          * it is likely to be the very last entry on that list.
4470          * So this causes the notifier registry below to try and
4471          * give link-local addresses to all devices besides loopback_dev
4472          * first, then loopback_dev, which cases all the non-loopback_dev
4473          * devices to fail to get a link-local address.
4474          *
4475          * So, as a temporary fix, allocate the ipv6 structure for
4476          * loopback_dev first by hand.
4477          * Longer term, all of the dependencies ipv6 has upon the loopback
4478          * device and it being up should be removed.
4479          */
4480         rtnl_lock();
4481         if (!ipv6_add_dev(init_net.loopback_dev))
4482                 err = -ENOMEM;
4483         rtnl_unlock();
4484         if (err)
4485                 goto errlo;
4486
4487         register_netdevice_notifier(&ipv6_dev_notf);
4488
4489         addrconf_verify(0);
4490
4491         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4492         if (err < 0)
4493                 goto errout;
4494
4495         /* Only the first call to __rtnl_register can fail */
4496         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4497         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4498         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4499         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4500         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4501
4502         ipv6_addr_label_rtnl_register();
4503
4504         return 0;
4505 errout:
4506         unregister_netdevice_notifier(&ipv6_dev_notf);
4507 errlo:
4508         unregister_pernet_subsys(&addrconf_ops);
4509
4510         return err;
4511 }
4512
4513 void addrconf_cleanup(void)
4514 {
4515         struct inet6_ifaddr *ifa;
4516         struct net_device *dev;
4517         int i;
4518
4519         unregister_netdevice_notifier(&ipv6_dev_notf);
4520         unregister_pernet_subsys(&addrconf_ops);
4521
4522         rtnl_lock();
4523
4524         /* clean dev list */
4525         for_each_netdev(&init_net, dev) {
4526                 if (__in6_dev_get(dev) == NULL)
4527                         continue;
4528                 addrconf_ifdown(dev, 1);
4529         }
4530         addrconf_ifdown(init_net.loopback_dev, 2);
4531
4532         /*
4533          *      Check hash table.
4534          */
4535         write_lock_bh(&addrconf_hash_lock);
4536         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4537                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4538                         struct inet6_ifaddr *bifa;
4539
4540                         bifa = ifa;
4541                         ifa = ifa->lst_next;
4542                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4543                         /* Do not free it; something is wrong.
4544                            Now we can investigate it with debugger.
4545                          */
4546                 }
4547         }
4548         write_unlock_bh(&addrconf_hash_lock);
4549
4550         del_timer(&addr_chk_timer);
4551         rtnl_unlock();
4552 }