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/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/in6.h>
  47 #include <linux/netdevice.h>
  48 #include <linux/if_addr.h>
  49 #include <linux/if_arp.h>
  50 #include <linux/if_arcnet.h>
  51 #include <linux/if_infiniband.h>
  52 #include <linux/route.h>
  53 #include <linux/inetdevice.h>
  54 #include <linux/init.h>
  55 #ifdef CONFIG_SYSCTL
  56 #include <linux/sysctl.h>
  57 #endif
  58 #include <linux/capability.h>
  59 #include <linux/delay.h>
  60 #include <linux/notifier.h>
  61 #include <linux/string.h>
  62
  63 #include <net/net_namespace.h>
  64 #include <net/sock.h>
  65 #include <net/snmp.h>
  66
  67 #include <net/ipv6.h>
  68 #include <net/protocol.h>
  69 #include <net/ndisc.h>
  70 #include <net/ip6_route.h>
  71 #include <net/addrconf.h>
  72 #include <net/tcp.h>
  73 #include <net/ip.h>
  74 #include <net/netlink.h>
  75 #include <net/pkt_sched.h>
  76 #include <linux/if_tunnel.h>
  77 #include <linux/rtnetlink.h>
  78
  79 #ifdef CONFIG_IPV6_PRIVACY
  80 #include <linux/random.h>
  81 #endif
  82
  83 #include <asm/uaccess.h>
  84 #include <asm/unaligned.h>
  85
  86 #include <linux/proc_fs.h>
  87 #include <linux/seq_file.h>
  88
  89 /* Set to 3 to get tracing... */
  90 #define ACONF_DEBUG 2
  91
  92 #if ACONF_DEBUG >= 3
  93 #define ADBG(x) printk x
  94 #else
  95 #define ADBG(x)
  96 #endif
  97
  98 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 100
 101 #ifdef CONFIG_SYSCTL
 102 static void addrconf_sysctl_register(struct inet6_dev *idev);
 103 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 104 #else
 105 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 106 {
 107 }
 108
 109 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 110 {
 111 }
 112 #endif
 113
 114 #ifdef CONFIG_IPV6_PRIVACY
 115 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 116 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 117 static void ipv6_regen_rndid(unsigned long data);
 118
 119 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 120 #endif
 121
 122 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 123 static int ipv6_count_addresses(struct inet6_dev *idev);
 124
 125 /*
 126  *      Configured unicast address hash table
 127  */
 128 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 129 static DEFINE_RWLOCK(addrconf_hash_lock);
 130
 131 static void addrconf_verify(unsigned long);
 132
 133 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 134 static DEFINE_SPINLOCK(addrconf_verify_lock);
 135
 136 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 137 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 138
 139 static int addrconf_ifdown(struct net_device *dev, int how);
 140
 141 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 142 static void addrconf_dad_timer(unsigned long data);
 143 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 144 static void addrconf_dad_run(struct inet6_dev *idev);
 145 static void addrconf_rs_timer(unsigned long data);
 146 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 147 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148
 149 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 150                                 struct prefix_info *pinfo);
 151 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 152                               struct net_device *dev);
 153
 154 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 155
 156 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 157         .forwarding             = 0,
 158         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 159         .mtu6                   = IPV6_MIN_MTU,
 160         .accept_ra              = 1,
 161         .accept_redirects       = 1,
 162         .autoconf               = 1,
 163         .force_mld_version      = 0,
 164         .dad_transmits          = 1,
 165         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 166         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 167         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 168 #ifdef CONFIG_IPV6_PRIVACY
 169         .use_tempaddr           = 0,
 170         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 171         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 172         .regen_max_retry        = REGEN_MAX_RETRY,
 173         .max_desync_factor      = MAX_DESYNC_FACTOR,
 174 #endif
 175         .max_addresses          = IPV6_MAX_ADDRESSES,
 176         .accept_ra_defrtr       = 1,
 177         .accept_ra_pinfo        = 1,
 178 #ifdef CONFIG_IPV6_ROUTER_PREF
 179         .accept_ra_rtr_pref     = 1,
 180         .rtr_probe_interval     = 60 * HZ,
 181 #ifdef CONFIG_IPV6_ROUTE_INFO
 182         .accept_ra_rt_info_max_plen = 0,
 183 #endif
 184 #endif
 185         .proxy_ndp              = 0,
 186         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 187         .disable_ipv6           = 0,
 188         .accept_dad             = 1,
 189 };
 190
 191 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 192         .forwarding             = 0,
 193         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 194         .mtu6                   = IPV6_MIN_MTU,
 195         .accept_ra              = 1,
 196         .accept_redirects       = 1,
 197         .autoconf               = 1,
 198         .dad_transmits          = 1,
 199         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 200         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202 #ifdef CONFIG_IPV6_PRIVACY
 203         .use_tempaddr           = 0,
 204         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 205         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 206         .regen_max_retry        = REGEN_MAX_RETRY,
 207         .max_desync_factor      = MAX_DESYNC_FACTOR,
 208 #endif
 209         .max_addresses          = IPV6_MAX_ADDRESSES,
 210         .accept_ra_defrtr       = 1,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_max_plen = 0,
 217 #endif
 218 #endif
 219         .proxy_ndp              = 0,
 220         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 221         .disable_ipv6           = 0,
 222         .accept_dad             = 1,
 223 };
 224
 225 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 226 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 227 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 228 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 229 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 230
 231 /* Check if a valid qdisc is available */
 232 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 233 {
 234         return !qdisc_tx_is_noop(dev);
 235 }
 236
 237 /* Check if a route is valid prefix route */
 238 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 239 {
 240         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 241 }
 242
 243 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 244 {
 245         if (del_timer(&ifp->timer))
 246                 __in6_ifa_put(ifp);
 247 }
 248
 249 enum addrconf_timer_t
 250 {
 251         AC_NONE,
 252         AC_DAD,
 253         AC_RS,
 254 };
 255
 256 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 257                                enum addrconf_timer_t what,
 258                                unsigned long when)
 259 {
 260         if (!del_timer(&ifp->timer))
 261                 in6_ifa_hold(ifp);
 262
 263         switch (what) {
 264         case AC_DAD:
 265                 ifp->timer.function = addrconf_dad_timer;
 266                 break;
 267         case AC_RS:
 268                 ifp->timer.function = addrconf_rs_timer;
 269                 break;
 270         default:;
 271         }
 272         ifp->timer.expires = jiffies + when;
 273         add_timer(&ifp->timer);
 274 }
 275
 276 static int snmp6_alloc_dev(struct inet6_dev *idev)
 277 {
 278         if (snmp_mib_init((void **)idev->stats.ipv6,
 279                           sizeof(struct ipstats_mib)) < 0)
 280                 goto err_ip;
 281         if (snmp_mib_init((void **)idev->stats.icmpv6,
 282                           sizeof(struct icmpv6_mib)) < 0)
 283                 goto err_icmp;
 284         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 285                           sizeof(struct icmpv6msg_mib)) < 0)
 286                 goto err_icmpmsg;
 287
 288         return 0;
 289
 290 err_icmpmsg:
 291         snmp_mib_free((void **)idev->stats.icmpv6);
 292 err_icmp:
 293         snmp_mib_free((void **)idev->stats.ipv6);
 294 err_ip:
 295         return -ENOMEM;
 296 }
 297
 298 static void snmp6_free_dev(struct inet6_dev *idev)
 299 {
 300         snmp_mib_free((void **)idev->stats.icmpv6msg);
 301         snmp_mib_free((void **)idev->stats.icmpv6);
 302         snmp_mib_free((void **)idev->stats.ipv6);
 303 }
 304
 305 /* Nobody refers to this device, we may destroy it. */
 306
 307 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 308 {
 309         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 310         kfree(idev);
 311 }
 312
 313 void in6_dev_finish_destroy(struct inet6_dev *idev)
 314 {
 315         struct net_device *dev = idev->dev;
 316
 317         WARN_ON(idev->addr_list != NULL);
 318         WARN_ON(idev->mc_list != NULL);
 319
 320 #ifdef NET_REFCNT_DEBUG
 321         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 322 #endif
 323         dev_put(dev);
 324         if (!idev->dead) {
 325                 printk("Freeing alive inet6 device %p\n", idev);
 326                 return;
 327         }
 328         snmp6_free_dev(idev);
 329         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 330 }
 331
 332 EXPORT_SYMBOL(in6_dev_finish_destroy);
 333
 334 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 335 {
 336         struct inet6_dev *ndev;
 337
 338         ASSERT_RTNL();
 339
 340         if (dev->mtu < IPV6_MIN_MTU)
 341                 return NULL;
 342
 343         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 344
 345         if (ndev == NULL)
 346                 return NULL;
 347
 348         rwlock_init(&ndev->lock);
 349         ndev->dev = dev;
 350         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 351         ndev->cnf.mtu6 = dev->mtu;
 352         ndev->cnf.sysctl = NULL;
 353         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 354         if (ndev->nd_parms == NULL) {
 355                 kfree(ndev);
 356                 return NULL;
 357         }
 358         if (ndev->cnf.forwarding)
 359                 dev_disable_lro(dev);
 360         /* We refer to the device */
 361         dev_hold(dev);
 362
 363         if (snmp6_alloc_dev(ndev) < 0) {
 364                 ADBG((KERN_WARNING
 365                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 366                         __func__, dev->name));
 367                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 368                 ndev->dead = 1;
 369                 in6_dev_finish_destroy(ndev);
 370                 return NULL;
 371         }
 372
 373         if (snmp6_register_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 ndev->dead = 1;
 379                 in6_dev_finish_destroy(ndev);
 380                 return NULL;
 381         }
 382
 383         /* One reference from device.  We must do this before
 384          * we invoke __ipv6_regen_rndid().
 385          */
 386         in6_dev_hold(ndev);
 387
 388         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 389                 ndev->cnf.accept_dad = -1;
 390
 391 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 392         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 393                 printk(KERN_INFO
 394                        "%s: Disabled Multicast RS\n",
 395                        dev->name);
 396                 ndev->cnf.rtr_solicits = 0;
 397         }
 398 #endif
 399
 400 #ifdef CONFIG_IPV6_PRIVACY
 401         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 402         if ((dev->flags&IFF_LOOPBACK) ||
 403             dev->type == ARPHRD_TUNNEL ||
 404             dev->type == ARPHRD_TUNNEL6 ||
 405             dev->type == ARPHRD_SIT ||
 406             dev->type == ARPHRD_NONE) {
 407                 printk(KERN_INFO
 408                        "%s: Disabled Privacy Extensions\n",
 409                        dev->name);
 410                 ndev->cnf.use_tempaddr = -1;
 411         } else {
 412                 in6_dev_hold(ndev);
 413                 ipv6_regen_rndid((unsigned long) ndev);
 414         }
 415 #endif
 416
 417         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 418                 ndev->if_flags |= IF_READY;
 419
 420         ipv6_mc_init_dev(ndev);
 421         ndev->tstamp = jiffies;
 422         addrconf_sysctl_register(ndev);
 423         /* protected by rtnl_lock */
 424         rcu_assign_pointer(dev->ip6_ptr, ndev);
 425
 426         /* Join all-node multicast group */
 427         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 428
 429         return ndev;
 430 }
 431
 432 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 433 {
 434         struct inet6_dev *idev;
 435
 436         ASSERT_RTNL();
 437
 438         if ((idev = __in6_dev_get(dev)) == NULL) {
 439                 if ((idev = ipv6_add_dev(dev)) == NULL)
 440                         return NULL;
 441         }
 442
 443         if (dev->flags&IFF_UP)
 444                 ipv6_mc_up(idev);
 445         return idev;
 446 }
 447
 448 #ifdef CONFIG_SYSCTL
 449 static void dev_forward_change(struct inet6_dev *idev)
 450 {
 451         struct net_device *dev;
 452         struct inet6_ifaddr *ifa;
 453
 454         if (!idev)
 455                 return;
 456         dev = idev->dev;
 457         if (idev->cnf.forwarding)
 458                 dev_disable_lro(dev);
 459         if (dev && (dev->flags & IFF_MULTICAST)) {
 460                 if (idev->cnf.forwarding)
 461                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 462                 else
 463                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 464         }
 465         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 466                 if (ifa->flags&IFA_F_TENTATIVE)
 467                         continue;
 468                 if (idev->cnf.forwarding)
 469                         addrconf_join_anycast(ifa);
 470                 else
 471                         addrconf_leave_anycast(ifa);
 472         }
 473 }
 474
 475
 476 static void addrconf_forward_change(struct net *net, __s32 newf)
 477 {
 478         struct net_device *dev;
 479         struct inet6_dev *idev;
 480
 481         read_lock(&dev_base_lock);
 482         for_each_netdev(net, dev) {
 483                 rcu_read_lock();
 484                 idev = __in6_dev_get(dev);
 485                 if (idev) {
 486                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 487                         idev->cnf.forwarding = newf;
 488                         if (changed)
 489                                 dev_forward_change(idev);
 490                 }
 491                 rcu_read_unlock();
 492         }
 493         read_unlock(&dev_base_lock);
 494 }
 495
 496 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 497 {
 498         struct net *net;
 499
 500         net = (struct net *)table->extra2;
 501         if (p == &net->ipv6.devconf_dflt->forwarding)
 502                 return 0;
 503
 504         if (!rtnl_trylock())
 505                 return -ERESTARTSYS;
 506
 507         if (p == &net->ipv6.devconf_all->forwarding) {
 508                 __s32 newf = net->ipv6.devconf_all->forwarding;
 509                 net->ipv6.devconf_dflt->forwarding = newf;
 510                 addrconf_forward_change(net, newf);
 511         } else if ((!*p) ^ (!old))
 512                 dev_forward_change((struct inet6_dev *)table->extra1);
 513         rtnl_unlock();
 514
 515         if (*p)
 516                 rt6_purge_dflt_routers(net);
 517         return 1;
 518 }
 519 #endif
 520
 521 /* Nobody refers to this ifaddr, destroy it */
 522
 523 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 524 {
 525         WARN_ON(ifp->if_next != NULL);
 526         WARN_ON(ifp->lst_next != NULL);
 527
 528 #ifdef NET_REFCNT_DEBUG
 529         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 530 #endif
 531
 532         in6_dev_put(ifp->idev);
 533
 534         if (del_timer(&ifp->timer))
 535                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 536
 537         if (!ifp->dead) {
 538                 printk("Freeing alive inet6 address %p\n", ifp);
 539                 return;
 540         }
 541         dst_release(&ifp->rt->u.dst);
 542
 543         kfree(ifp);
 544 }
 545
 546 static void
 547 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 548 {
 549         struct inet6_ifaddr *ifa, **ifap;
 550         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 551
 552         /*
 553          * Each device address list is sorted in order of scope -
 554          * global before linklocal.
 555          */
 556         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 557              ifap = &ifa->if_next) {
 558                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 559                         break;
 560         }
 561
 562         ifp->if_next = *ifap;
 563         *ifap = ifp;
 564 }
 565
 566 /*
 567  *      Hash function taken from net_alias.c
 568  */
 569 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 570 {
 571         __u32 word;
 572
 573         /*
 574          * We perform the hash function over the last 64 bits of the address
 575          * This will include the IEEE address token on links that support it.
 576          */
 577
 578         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 579         word ^= (word >> 16);
 580         word ^= (word >> 8);
 581
 582         return ((word ^ (word >> 4)) & 0x0f);
 583 }
 584
 585 /* On success it returns ifp with increased reference count */
 586
 587 static struct inet6_ifaddr *
 588 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 589               int scope, u32 flags)
 590 {
 591         struct inet6_ifaddr *ifa = NULL;
 592         struct rt6_info *rt;
 593         struct net *net = dev_net(idev->dev);
 594         int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6 || net->ipv6.devconf_all->disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         write_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, 0);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ipv6_addr_copy(&ifa->addr, addr);
 639
 640         spin_lock_init(&ifa->lock);
 641         init_timer(&ifa->timer);
 642         ifa->timer.data = (unsigned long) ifa;
 643         ifa->scope = scope;
 644         ifa->prefix_len = pfxlen;
 645         ifa->flags = flags | IFA_F_TENTATIVE;
 646         ifa->cstamp = ifa->tstamp = jiffies;
 647
 648         ifa->rt = rt;
 649
 650         /*
 651          * part one of RFC 4429, section 3.3
 652          * We should not configure an address as
 653          * optimistic if we do not yet know the link
 654          * layer address of our nexhop router
 655          */
 656
 657         if (rt->rt6i_nexthop == NULL)
 658                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 659
 660         ifa->idev = idev;
 661         in6_dev_hold(idev);
 662         /* For caller */
 663         in6_ifa_hold(ifa);
 664
 665         /* Add to big hash table */
 666         hash = ipv6_addr_hash(addr);
 667
 668         ifa->lst_next = inet6_addr_lst[hash];
 669         inet6_addr_lst[hash] = ifa;
 670         in6_ifa_hold(ifa);
 671         write_unlock(&addrconf_hash_lock);
 672
 673         write_lock(&idev->lock);
 674         /* Add to inet6_dev unicast addr list. */
 675         ipv6_link_dev_addr(idev, ifa);
 676
 677 #ifdef CONFIG_IPV6_PRIVACY
 678         if (ifa->flags&IFA_F_TEMPORARY) {
 679                 ifa->tmp_next = idev->tempaddr_list;
 680                 idev->tempaddr_list = ifa;
 681                 in6_ifa_hold(ifa);
 682         }
 683 #endif
 684
 685         in6_ifa_hold(ifa);
 686         write_unlock(&idev->lock);
 687 out2:
 688         rcu_read_unlock_bh();
 689
 690         if (likely(err == 0))
 691                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 692         else {
 693                 kfree(ifa);
 694                 ifa = ERR_PTR(err);
 695         }
 696
 697         return ifa;
 698 out:
 699         write_unlock(&addrconf_hash_lock);
 700         goto out2;
 701 }
 702
 703 /* This function wants to get referenced ifp and releases it before return */
 704
 705 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 706 {
 707         struct inet6_ifaddr *ifa, **ifap;
 708         struct inet6_dev *idev = ifp->idev;
 709         int hash;
 710         int deleted = 0, onlink = 0;
 711         unsigned long expires = jiffies;
 712
 713         hash = ipv6_addr_hash(&ifp->addr);
 714
 715         ifp->dead = 1;
 716
 717         write_lock_bh(&addrconf_hash_lock);
 718         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 719              ifap = &ifa->lst_next) {
 720                 if (ifa == ifp) {
 721                         *ifap = ifa->lst_next;
 722                         __in6_ifa_put(ifp);
 723                         ifa->lst_next = NULL;
 724                         break;
 725                 }
 726         }
 727         write_unlock_bh(&addrconf_hash_lock);
 728
 729         write_lock_bh(&idev->lock);
 730 #ifdef CONFIG_IPV6_PRIVACY
 731         if (ifp->flags&IFA_F_TEMPORARY) {
 732                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 733                      ifap = &ifa->tmp_next) {
 734                         if (ifa == ifp) {
 735                                 *ifap = ifa->tmp_next;
 736                                 if (ifp->ifpub) {
 737                                         in6_ifa_put(ifp->ifpub);
 738                                         ifp->ifpub = NULL;
 739                                 }
 740                                 __in6_ifa_put(ifp);
 741                                 ifa->tmp_next = NULL;
 742                                 break;
 743                         }
 744                 }
 745         }
 746 #endif
 747
 748         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 749                 if (ifa == ifp) {
 750                         *ifap = ifa->if_next;
 751                         __in6_ifa_put(ifp);
 752                         ifa->if_next = NULL;
 753                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 754                                 break;
 755                         deleted = 1;
 756                         continue;
 757                 } else if (ifp->flags & IFA_F_PERMANENT) {
 758                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 759                                               ifp->prefix_len)) {
 760                                 if (ifa->flags & IFA_F_PERMANENT) {
 761                                         onlink = 1;
 762                                         if (deleted)
 763                                                 break;
 764                                 } else {
 765                                         unsigned long lifetime;
 766
 767                                         if (!onlink)
 768                                                 onlink = -1;
 769
 770                                         spin_lock(&ifa->lock);
 771
 772                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 773                                         /*
 774                                          * Note: Because this address is
 775                                          * not permanent, lifetime <
 776                                          * LONG_MAX / HZ here.
 777                                          */
 778                                         if (time_before(expires,
 779                                                         ifa->tstamp + lifetime * HZ))
 780                                                 expires = ifa->tstamp + lifetime * HZ;
 781                                         spin_unlock(&ifa->lock);
 782                                 }
 783                         }
 784                 }
 785                 ifap = &ifa->if_next;
 786         }
 787         write_unlock_bh(&idev->lock);
 788
 789         addrconf_del_timer(ifp);
 790
 791         ipv6_ifa_notify(RTM_DELADDR, ifp);
 792
 793         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 794
 795         /*
 796          * Purge or update corresponding prefix
 797          *
 798          * 1) we don't purge prefix here if address was not permanent.
 799          *    prefix is managed by its own lifetime.
 800          * 2) if there're no addresses, delete prefix.
 801          * 3) if there're still other permanent address(es),
 802          *    corresponding prefix is still permanent.
 803          * 4) otherwise, update prefix lifetime to the
 804          *    longest valid lifetime among the corresponding
 805          *    addresses on the device.
 806          *    Note: subsequent RA will update lifetime.
 807          *
 808          * --yoshfuji
 809          */
 810         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 811                 struct in6_addr prefix;
 812                 struct rt6_info *rt;
 813                 struct net *net = dev_net(ifp->idev->dev);
 814                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 815                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 816
 817                 if (rt && addrconf_is_prefix_route(rt)) {
 818                         if (onlink == 0) {
 819                                 ip6_del_rt(rt);
 820                                 rt = NULL;
 821                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 822                                 rt->rt6i_expires = expires;
 823                                 rt->rt6i_flags |= RTF_EXPIRES;
 824                         }
 825                 }
 826                 dst_release(&rt->u.dst);
 827         }
 828
 829         in6_ifa_put(ifp);
 830 }
 831
 832 #ifdef CONFIG_IPV6_PRIVACY
 833 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 834 {
 835         struct inet6_dev *idev = ifp->idev;
 836         struct in6_addr addr, *tmpaddr;
 837         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 838         unsigned long regen_advance;
 839         int tmp_plen;
 840         int ret = 0;
 841         int max_addresses;
 842         u32 addr_flags;
 843
 844         write_lock(&idev->lock);
 845         if (ift) {
 846                 spin_lock_bh(&ift->lock);
 847                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 848                 spin_unlock_bh(&ift->lock);
 849                 tmpaddr = &addr;
 850         } else {
 851                 tmpaddr = NULL;
 852         }
 853 retry:
 854         in6_dev_hold(idev);
 855         if (idev->cnf.use_tempaddr <= 0) {
 856                 write_unlock(&idev->lock);
 857                 printk(KERN_INFO
 858                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 859                 in6_dev_put(idev);
 860                 ret = -1;
 861                 goto out;
 862         }
 863         spin_lock_bh(&ifp->lock);
 864         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 865                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 866                 spin_unlock_bh(&ifp->lock);
 867                 write_unlock(&idev->lock);
 868                 printk(KERN_WARNING
 869                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 870                 in6_dev_put(idev);
 871                 ret = -1;
 872                 goto out;
 873         }
 874         in6_ifa_hold(ifp);
 875         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 876         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 877                 spin_unlock_bh(&ifp->lock);
 878                 write_unlock(&idev->lock);
 879                 printk(KERN_WARNING
 880                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 881                 in6_ifa_put(ifp);
 882                 in6_dev_put(idev);
 883                 ret = -1;
 884                 goto out;
 885         }
 886         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 887         tmp_valid_lft = min_t(__u32,
 888                               ifp->valid_lft,
 889                               idev->cnf.temp_valid_lft);
 890         tmp_prefered_lft = min_t(__u32,
 891                                  ifp->prefered_lft,
 892                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 893         tmp_plen = ifp->prefix_len;
 894         max_addresses = idev->cnf.max_addresses;
 895         tmp_cstamp = ifp->cstamp;
 896         tmp_tstamp = ifp->tstamp;
 897         spin_unlock_bh(&ifp->lock);
 898
 899         regen_advance = idev->cnf.regen_max_retry *
 900                         idev->cnf.dad_transmits *
 901                         idev->nd_parms->retrans_time / HZ;
 902         write_unlock(&idev->lock);
 903
 904         /* A temporary address is created only if this calculated Preferred
 905          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 906          * an implementation must not create a temporary address with a zero
 907          * Preferred Lifetime.
 908          */
 909         if (tmp_prefered_lft <= regen_advance) {
 910                 in6_ifa_put(ifp);
 911                 in6_dev_put(idev);
 912                 ret = -1;
 913                 goto out;
 914         }
 915
 916         addr_flags = IFA_F_TEMPORARY;
 917         /* set in addrconf_prefix_rcv() */
 918         if (ifp->flags & IFA_F_OPTIMISTIC)
 919                 addr_flags |= IFA_F_OPTIMISTIC;
 920
 921         ift = !max_addresses ||
 922               ipv6_count_addresses(idev) < max_addresses ?
 923                 ipv6_add_addr(idev, &addr, tmp_plen,
 924                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 925                               addr_flags) : NULL;
 926         if (!ift || IS_ERR(ift)) {
 927                 in6_ifa_put(ifp);
 928                 in6_dev_put(idev);
 929                 printk(KERN_INFO
 930                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 931                 tmpaddr = &addr;
 932                 write_lock(&idev->lock);
 933                 goto retry;
 934         }
 935
 936         spin_lock_bh(&ift->lock);
 937         ift->ifpub = ifp;
 938         ift->valid_lft = tmp_valid_lft;
 939         ift->prefered_lft = tmp_prefered_lft;
 940         ift->cstamp = tmp_cstamp;
 941         ift->tstamp = tmp_tstamp;
 942         spin_unlock_bh(&ift->lock);
 943
 944         addrconf_dad_start(ift, 0);
 945         in6_ifa_put(ift);
 946         in6_dev_put(idev);
 947 out:
 948         return ret;
 949 }
 950 #endif
 951
 952 /*
 953  *      Choose an appropriate source address (RFC3484)
 954  */
 955 enum {
 956         IPV6_SADDR_RULE_INIT = 0,
 957         IPV6_SADDR_RULE_LOCAL,
 958         IPV6_SADDR_RULE_SCOPE,
 959         IPV6_SADDR_RULE_PREFERRED,
 960 #ifdef CONFIG_IPV6_MIP6
 961         IPV6_SADDR_RULE_HOA,
 962 #endif
 963         IPV6_SADDR_RULE_OIF,
 964         IPV6_SADDR_RULE_LABEL,
 965 #ifdef CONFIG_IPV6_PRIVACY
 966         IPV6_SADDR_RULE_PRIVACY,
 967 #endif
 968         IPV6_SADDR_RULE_ORCHID,
 969         IPV6_SADDR_RULE_PREFIX,
 970         IPV6_SADDR_RULE_MAX
 971 };
 972
 973 struct ipv6_saddr_score {
 974         int                     rule;
 975         int                     addr_type;
 976         struct inet6_ifaddr     *ifa;
 977         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 978         int                     scopedist;
 979         int                     matchlen;
 980 };
 981
 982 struct ipv6_saddr_dst {
 983         const struct in6_addr *addr;
 984         int ifindex;
 985         int scope;
 986         int label;
 987         unsigned int prefs;
 988 };
 989
 990 static inline int ipv6_saddr_preferred(int type)
 991 {
 992         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 993                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 994                 return 1;
 995         return 0;
 996 }
 997
 998 static int ipv6_get_saddr_eval(struct net *net,
 999                                struct ipv6_saddr_score *score,
1000                                struct ipv6_saddr_dst *dst,
1001                                int i)
1002 {
1003         int ret;
1004
1005         if (i <= score->rule) {
1006                 switch (i) {
1007                 case IPV6_SADDR_RULE_SCOPE:
1008                         ret = score->scopedist;
1009                         break;
1010                 case IPV6_SADDR_RULE_PREFIX:
1011                         ret = score->matchlen;
1012                         break;
1013                 default:
1014                         ret = !!test_bit(i, score->scorebits);
1015                 }
1016                 goto out;
1017         }
1018
1019         switch (i) {
1020         case IPV6_SADDR_RULE_INIT:
1021                 /* Rule 0: remember if hiscore is not ready yet */
1022                 ret = !!score->ifa;
1023                 break;
1024         case IPV6_SADDR_RULE_LOCAL:
1025                 /* Rule 1: Prefer same address */
1026                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1027                 break;
1028         case IPV6_SADDR_RULE_SCOPE:
1029                 /* Rule 2: Prefer appropriate scope
1030                  *
1031                  *      ret
1032                  *       ^
1033                  *    -1 |  d 15
1034                  *    ---+--+-+---> scope
1035                  *       |
1036                  *       |             d is scope of the destination.
1037                  *  B-d  |  \
1038                  *       |   \      <- smaller scope is better if
1039                  *  B-15 |    \        if scope is enough for destinaion.
1040                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1041                  * d-C-1 | /
1042                  *       |/         <- greater is better
1043                  *   -C  /             if scope is not enough for destination.
1044                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1045                  *
1046                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1047                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1048                  * Assume B = 0 and we get C > 29.
1049                  */
1050                 ret = __ipv6_addr_src_scope(score->addr_type);
1051                 if (ret >= dst->scope)
1052                         ret = -ret;
1053                 else
1054                         ret -= 128;     /* 30 is enough */
1055                 score->scopedist = ret;
1056                 break;
1057         case IPV6_SADDR_RULE_PREFERRED:
1058                 /* Rule 3: Avoid deprecated and optimistic addresses */
1059                 ret = ipv6_saddr_preferred(score->addr_type) ||
1060                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1061                 break;
1062 #ifdef CONFIG_IPV6_MIP6
1063         case IPV6_SADDR_RULE_HOA:
1064             {
1065                 /* Rule 4: Prefer home address */
1066                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1067                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1068                 break;
1069             }
1070 #endif
1071         case IPV6_SADDR_RULE_OIF:
1072                 /* Rule 5: Prefer outgoing interface */
1073                 ret = (!dst->ifindex ||
1074                        dst->ifindex == score->ifa->idev->dev->ifindex);
1075                 break;
1076         case IPV6_SADDR_RULE_LABEL:
1077                 /* Rule 6: Prefer matching label */
1078                 ret = ipv6_addr_label(net,
1079                                       &score->ifa->addr, score->addr_type,
1080                                       score->ifa->idev->dev->ifindex) == dst->label;
1081                 break;
1082 #ifdef CONFIG_IPV6_PRIVACY
1083         case IPV6_SADDR_RULE_PRIVACY:
1084             {
1085                 /* Rule 7: Prefer public address
1086                  * Note: prefer temprary address if use_tempaddr >= 2
1087                  */
1088                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1089                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1090                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1091                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1092                 break;
1093             }
1094 #endif
1095         case IPV6_SADDR_RULE_ORCHID:
1096                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1097                  *          non-ORCHID vs non-ORCHID
1098                  */
1099                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1100                         ipv6_addr_orchid(dst->addr));
1101                 break;
1102         case IPV6_SADDR_RULE_PREFIX:
1103                 /* Rule 8: Use longest matching prefix */
1104                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1105                                                        dst->addr);
1106                 break;
1107         default:
1108                 ret = 0;
1109         }
1110
1111         if (ret)
1112                 __set_bit(i, score->scorebits);
1113         score->rule = i;
1114 out:
1115         return ret;
1116 }
1117
1118 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1119                        const struct in6_addr *daddr, unsigned int prefs,
1120                        struct in6_addr *saddr)
1121 {
1122         struct ipv6_saddr_score scores[2],
1123                                 *score = &scores[0], *hiscore = &scores[1];
1124         struct ipv6_saddr_dst dst;
1125         struct net_device *dev;
1126         int dst_type;
1127
1128         dst_type = __ipv6_addr_type(daddr);
1129         dst.addr = daddr;
1130         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1131         dst.scope = __ipv6_addr_src_scope(dst_type);
1132         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1133         dst.prefs = prefs;
1134
1135         hiscore->rule = -1;
1136         hiscore->ifa = NULL;
1137
1138         read_lock(&dev_base_lock);
1139         rcu_read_lock();
1140
1141         for_each_netdev(net, dev) {
1142                 struct inet6_dev *idev;
1143
1144                 /* Candidate Source Address (section 4)
1145                  *  - multicast and link-local destination address,
1146                  *    the set of candidate source address MUST only
1147                  *    include addresses assigned to interfaces
1148                  *    belonging to the same link as the outgoing
1149                  *    interface.
1150                  * (- For site-local destination addresses, the
1151                  *    set of candidate source addresses MUST only
1152                  *    include addresses assigned to interfaces
1153                  *    belonging to the same site as the outgoing
1154                  *    interface.)
1155                  */
1156                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1157                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1158                     dst.ifindex && dev->ifindex != dst.ifindex)
1159                         continue;
1160
1161                 idev = __in6_dev_get(dev);
1162                 if (!idev)
1163                         continue;
1164
1165                 read_lock_bh(&idev->lock);
1166                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1167                         int i;
1168
1169                         /*
1170                          * - Tentative Address (RFC2462 section 5.4)
1171                          *  - A tentative address is not considered
1172                          *    "assigned to an interface" in the traditional
1173                          *    sense, unless it is also flagged as optimistic.
1174                          * - Candidate Source Address (section 4)
1175                          *  - In any case, anycast addresses, multicast
1176                          *    addresses, and the unspecified address MUST
1177                          *    NOT be included in a candidate set.
1178                          */
1179                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1180                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1181                                 continue;
1182
1183                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1184
1185                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1186                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1187                                 LIMIT_NETDEBUG(KERN_DEBUG
1188                                                "ADDRCONF: unspecified / multicast address "
1189                                                "assigned as unicast address on %s",
1190                                                dev->name);
1191                                 continue;
1192                         }
1193
1194                         score->rule = -1;
1195                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1196
1197                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1198                                 int minihiscore, miniscore;
1199
1200                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1201                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1202
1203                                 if (minihiscore > miniscore) {
1204                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1205                                             score->scopedist > 0) {
1206                                                 /*
1207                                                  * special case:
1208                                                  * each remaining entry
1209                                                  * has too small (not enough)
1210                                                  * scope, because ifa entries
1211                                                  * are sorted by their scope
1212                                                  * values.
1213                                                  */
1214                                                 goto try_nextdev;
1215                                         }
1216                                         break;
1217                                 } else if (minihiscore < miniscore) {
1218                                         struct ipv6_saddr_score *tmp;
1219
1220                                         if (hiscore->ifa)
1221                                                 in6_ifa_put(hiscore->ifa);
1222
1223                                         in6_ifa_hold(score->ifa);
1224
1225                                         tmp = hiscore;
1226                                         hiscore = score;
1227                                         score = tmp;
1228
1229                                         /* restore our iterator */
1230                                         score->ifa = hiscore->ifa;
1231
1232                                         break;
1233                                 }
1234                         }
1235                 }
1236 try_nextdev:
1237                 read_unlock_bh(&idev->lock);
1238         }
1239         rcu_read_unlock();
1240         read_unlock(&dev_base_lock);
1241
1242         if (!hiscore->ifa)
1243                 return -EADDRNOTAVAIL;
1244
1245         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1246         in6_ifa_put(hiscore->ifa);
1247         return 0;
1248 }
1249
1250 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1251
1252 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1253                     unsigned char banned_flags)
1254 {
1255         struct inet6_dev *idev;
1256         int err = -EADDRNOTAVAIL;
1257
1258         rcu_read_lock();
1259         if ((idev = __in6_dev_get(dev)) != NULL) {
1260                 struct inet6_ifaddr *ifp;
1261
1262                 read_lock_bh(&idev->lock);
1263                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1264                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1265                                 ipv6_addr_copy(addr, &ifp->addr);
1266                                 err = 0;
1267                                 break;
1268                         }
1269                 }
1270                 read_unlock_bh(&idev->lock);
1271         }
1272         rcu_read_unlock();
1273         return err;
1274 }
1275
1276 static int ipv6_count_addresses(struct inet6_dev *idev)
1277 {
1278         int cnt = 0;
1279         struct inet6_ifaddr *ifp;
1280
1281         read_lock_bh(&idev->lock);
1282         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1283                 cnt++;
1284         read_unlock_bh(&idev->lock);
1285         return cnt;
1286 }
1287
1288 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1289                   struct net_device *dev, int strict)
1290 {
1291         struct inet6_ifaddr * ifp;
1292         u8 hash = ipv6_addr_hash(addr);
1293
1294         read_lock_bh(&addrconf_hash_lock);
1295         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1296                 if (!net_eq(dev_net(ifp->idev->dev), net))
1297                         continue;
1298                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1299                     !(ifp->flags&IFA_F_TENTATIVE)) {
1300                         if (dev == NULL || ifp->idev->dev == dev ||
1301                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1302                                 break;
1303                 }
1304         }
1305         read_unlock_bh(&addrconf_hash_lock);
1306         return ifp != NULL;
1307 }
1308 EXPORT_SYMBOL(ipv6_chk_addr);
1309
1310 static
1311 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1312                        struct net_device *dev)
1313 {
1314         struct inet6_ifaddr * ifp;
1315         u8 hash = ipv6_addr_hash(addr);
1316
1317         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1318                 if (!net_eq(dev_net(ifp->idev->dev), net))
1319                         continue;
1320                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1321                         if (dev == NULL || ifp->idev->dev == dev)
1322                                 break;
1323                 }
1324         }
1325         return ifp != NULL;
1326 }
1327
1328 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1329 {
1330         struct inet6_dev *idev;
1331         struct inet6_ifaddr *ifa;
1332         int     onlink;
1333
1334         onlink = 0;
1335         rcu_read_lock();
1336         idev = __in6_dev_get(dev);
1337         if (idev) {
1338                 read_lock_bh(&idev->lock);
1339                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1340                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1341                                                    ifa->prefix_len);
1342                         if (onlink)
1343                                 break;
1344                 }
1345                 read_unlock_bh(&idev->lock);
1346         }
1347         rcu_read_unlock();
1348         return onlink;
1349 }
1350
1351 EXPORT_SYMBOL(ipv6_chk_prefix);
1352
1353 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1354                                      struct net_device *dev, int strict)
1355 {
1356         struct inet6_ifaddr * ifp;
1357         u8 hash = ipv6_addr_hash(addr);
1358
1359         read_lock_bh(&addrconf_hash_lock);
1360         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1361                 if (!net_eq(dev_net(ifp->idev->dev), net))
1362                         continue;
1363                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1364                         if (dev == NULL || ifp->idev->dev == dev ||
1365                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1366                                 in6_ifa_hold(ifp);
1367                                 break;
1368                         }
1369                 }
1370         }
1371         read_unlock_bh(&addrconf_hash_lock);
1372
1373         return ifp;
1374 }
1375
1376 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1377 {
1378         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1379         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1380         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1381         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1382         int sk_ipv6only = ipv6_only_sock(sk);
1383         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1384         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1385         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1386
1387         if (!sk2_rcv_saddr && !sk_ipv6only)
1388                 return 1;
1389
1390         if (addr_type2 == IPV6_ADDR_ANY &&
1391             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1392                 return 1;
1393
1394         if (addr_type == IPV6_ADDR_ANY &&
1395             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1396                 return 1;
1397
1398         if (sk2_rcv_saddr6 &&
1399             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1400                 return 1;
1401
1402         if (addr_type == IPV6_ADDR_MAPPED &&
1403             !sk2_ipv6only &&
1404             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1405                 return 1;
1406
1407         return 0;
1408 }
1409
1410 /* Gets referenced address, destroys ifaddr */
1411
1412 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1413 {
1414         if (ifp->flags&IFA_F_PERMANENT) {
1415                 spin_lock_bh(&ifp->lock);
1416                 addrconf_del_timer(ifp);
1417                 ifp->flags |= IFA_F_TENTATIVE;
1418                 spin_unlock_bh(&ifp->lock);
1419                 in6_ifa_put(ifp);
1420 #ifdef CONFIG_IPV6_PRIVACY
1421         } else if (ifp->flags&IFA_F_TEMPORARY) {
1422                 struct inet6_ifaddr *ifpub;
1423                 spin_lock_bh(&ifp->lock);
1424                 ifpub = ifp->ifpub;
1425                 if (ifpub) {
1426                         in6_ifa_hold(ifpub);
1427                         spin_unlock_bh(&ifp->lock);
1428                         ipv6_create_tempaddr(ifpub, ifp);
1429                         in6_ifa_put(ifpub);
1430                 } else {
1431                         spin_unlock_bh(&ifp->lock);
1432                 }
1433                 ipv6_del_addr(ifp);
1434 #endif
1435         } else
1436                 ipv6_del_addr(ifp);
1437 }
1438
1439 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1440 {
1441         struct inet6_dev *idev = ifp->idev;
1442
1443         if (net_ratelimit())
1444                 printk(KERN_INFO "%s: IPv6 duplicate address detected!\n",
1445                         ifp->idev->dev->name);
1446
1447         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1448                 struct in6_addr addr;
1449
1450                 addr.s6_addr32[0] = htonl(0xfe800000);
1451                 addr.s6_addr32[1] = 0;
1452
1453                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1454                     ipv6_addr_equal(&ifp->addr, &addr)) {
1455                         /* DAD failed for link-local based on MAC address */
1456                         idev->cnf.disable_ipv6 = 1;
1457
1458                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1459                                 ifp->idev->dev->name);
1460                 }
1461         }
1462
1463         addrconf_dad_stop(ifp);
1464 }
1465
1466 /* Join to solicited addr multicast group. */
1467
1468 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1469 {
1470         struct in6_addr maddr;
1471
1472         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1473                 return;
1474
1475         addrconf_addr_solict_mult(addr, &maddr);
1476         ipv6_dev_mc_inc(dev, &maddr);
1477 }
1478
1479 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1480 {
1481         struct in6_addr maddr;
1482
1483         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1484                 return;
1485
1486         addrconf_addr_solict_mult(addr, &maddr);
1487         __ipv6_dev_mc_dec(idev, &maddr);
1488 }
1489
1490 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1491 {
1492         struct in6_addr addr;
1493         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1494         if (ipv6_addr_any(&addr))
1495                 return;
1496         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1497 }
1498
1499 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1500 {
1501         struct in6_addr addr;
1502         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1503         if (ipv6_addr_any(&addr))
1504                 return;
1505         __ipv6_dev_ac_dec(ifp->idev, &addr);
1506 }
1507
1508 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1509 {
1510         if (dev->addr_len != ETH_ALEN)
1511                 return -1;
1512         memcpy(eui, dev->dev_addr, 3);
1513         memcpy(eui + 5, dev->dev_addr + 3, 3);
1514
1515         /*
1516          * The zSeries OSA network cards can be shared among various
1517          * OS instances, but the OSA cards have only one MAC address.
1518          * This leads to duplicate address conflicts in conjunction
1519          * with IPv6 if more than one instance uses the same card.
1520          *
1521          * The driver for these cards can deliver a unique 16-bit
1522          * identifier for each instance sharing the same card.  It is
1523          * placed instead of 0xFFFE in the interface identifier.  The
1524          * "u" bit of the interface identifier is not inverted in this
1525          * case.  Hence the resulting interface identifier has local
1526          * scope according to RFC2373.
1527          */
1528         if (dev->dev_id) {
1529                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1530                 eui[4] = dev->dev_id & 0xFF;
1531         } else {
1532                 eui[3] = 0xFF;
1533                 eui[4] = 0xFE;
1534                 eui[0] ^= 2;
1535         }
1536         return 0;
1537 }
1538
1539 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1540 {
1541         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1542         if (dev->addr_len != ARCNET_ALEN)
1543                 return -1;
1544         memset(eui, 0, 7);
1545         eui[7] = *(u8*)dev->dev_addr;
1546         return 0;
1547 }
1548
1549 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1550 {
1551         if (dev->addr_len != INFINIBAND_ALEN)
1552                 return -1;
1553         memcpy(eui, dev->dev_addr + 12, 8);
1554         eui[0] |= 2;
1555         return 0;
1556 }
1557
1558 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1559 {
1560         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1561                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1562                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1563                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1564                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1565                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1566         eui[1] = 0;
1567         eui[2] = 0x5E;
1568         eui[3] = 0xFE;
1569         memcpy(eui + 4, &addr, 4);
1570         return 0;
1571 }
1572 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1573
1574 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1575 {
1576         if (dev->priv_flags & IFF_ISATAP)
1577                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1578         return -1;
1579 }
1580
1581 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1582 {
1583         switch (dev->type) {
1584         case ARPHRD_ETHER:
1585         case ARPHRD_FDDI:
1586         case ARPHRD_IEEE802_TR:
1587                 return addrconf_ifid_eui48(eui, dev);
1588         case ARPHRD_ARCNET:
1589                 return addrconf_ifid_arcnet(eui, dev);
1590         case ARPHRD_INFINIBAND:
1591                 return addrconf_ifid_infiniband(eui, dev);
1592         case ARPHRD_SIT:
1593                 return addrconf_ifid_sit(eui, dev);
1594         }
1595         return -1;
1596 }
1597
1598 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1599 {
1600         int err = -1;
1601         struct inet6_ifaddr *ifp;
1602
1603         read_lock_bh(&idev->lock);
1604         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1605                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1606                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1607                         err = 0;
1608                         break;
1609                 }
1610         }
1611         read_unlock_bh(&idev->lock);
1612         return err;
1613 }
1614
1615 #ifdef CONFIG_IPV6_PRIVACY
1616 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1617 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1618 {
1619 regen:
1620         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1621         idev->rndid[0] &= ~0x02;
1622
1623         /*
1624          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1625          * check if generated address is not inappropriate
1626          *
1627          *  - Reserved subnet anycast (RFC 2526)
1628          *      11111101 11....11 1xxxxxxx
1629          *  - ISATAP (RFC4214) 6.1
1630          *      00-00-5E-FE-xx-xx-xx-xx
1631          *  - value 0
1632          *  - XXX: already assigned to an address on the device
1633          */
1634         if (idev->rndid[0] == 0xfd &&
1635             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1636             (idev->rndid[7]&0x80))
1637                 goto regen;
1638         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1639                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1640                         goto regen;
1641                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1642                         goto regen;
1643         }
1644
1645         return 0;
1646 }
1647
1648 static void ipv6_regen_rndid(unsigned long data)
1649 {
1650         struct inet6_dev *idev = (struct inet6_dev *) data;
1651         unsigned long expires;
1652
1653         rcu_read_lock_bh();
1654         write_lock_bh(&idev->lock);
1655
1656         if (idev->dead)
1657                 goto out;
1658
1659         if (__ipv6_regen_rndid(idev) < 0)
1660                 goto out;
1661
1662         expires = jiffies +
1663                 idev->cnf.temp_prefered_lft * HZ -
1664                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1665         if (time_before(expires, jiffies)) {
1666                 printk(KERN_WARNING
1667                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1668                         idev->dev->name);
1669                 goto out;
1670         }
1671
1672         if (!mod_timer(&idev->regen_timer, expires))
1673                 in6_dev_hold(idev);
1674
1675 out:
1676         write_unlock_bh(&idev->lock);
1677         rcu_read_unlock_bh();
1678         in6_dev_put(idev);
1679 }
1680
1681 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1682         int ret = 0;
1683
1684         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1685                 ret = __ipv6_regen_rndid(idev);
1686         return ret;
1687 }
1688 #endif
1689
1690 /*
1691  *      Add prefix route.
1692  */
1693
1694 static void
1695 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1696                       unsigned long expires, u32 flags)
1697 {
1698         struct fib6_config cfg = {
1699                 .fc_table = RT6_TABLE_PREFIX,
1700                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1701                 .fc_ifindex = dev->ifindex,
1702                 .fc_expires = expires,
1703                 .fc_dst_len = plen,
1704                 .fc_flags = RTF_UP | flags,
1705                 .fc_nlinfo.nl_net = dev_net(dev),
1706                 .fc_protocol = RTPROT_KERNEL,
1707         };
1708
1709         ipv6_addr_copy(&cfg.fc_dst, pfx);
1710
1711         /* Prevent useless cloning on PtP SIT.
1712            This thing is done here expecting that the whole
1713            class of non-broadcast devices need not cloning.
1714          */
1715 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1716         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1717                 cfg.fc_flags |= RTF_NONEXTHOP;
1718 #endif
1719
1720         ip6_route_add(&cfg);
1721 }
1722
1723 /* Create "default" multicast route to the interface */
1724
1725 static void addrconf_add_mroute(struct net_device *dev)
1726 {
1727         struct fib6_config cfg = {
1728                 .fc_table = RT6_TABLE_LOCAL,
1729                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1730                 .fc_ifindex = dev->ifindex,
1731                 .fc_dst_len = 8,
1732                 .fc_flags = RTF_UP,
1733                 .fc_nlinfo.nl_net = dev_net(dev),
1734         };
1735
1736         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1737
1738         ip6_route_add(&cfg);
1739 }
1740
1741 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1742 static void sit_route_add(struct net_device *dev)
1743 {
1744         struct fib6_config cfg = {
1745                 .fc_table = RT6_TABLE_MAIN,
1746                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1747                 .fc_ifindex = dev->ifindex,
1748                 .fc_dst_len = 96,
1749                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1750                 .fc_nlinfo.nl_net = dev_net(dev),
1751         };
1752
1753         /* prefix length - 96 bits "::d.d.d.d" */
1754         ip6_route_add(&cfg);
1755 }
1756 #endif
1757
1758 static void addrconf_add_lroute(struct net_device *dev)
1759 {
1760         struct in6_addr addr;
1761
1762         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1763         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1764 }
1765
1766 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1767 {
1768         struct inet6_dev *idev;
1769
1770         ASSERT_RTNL();
1771
1772         if ((idev = ipv6_find_idev(dev)) == NULL)
1773                 return NULL;
1774
1775         /* Add default multicast route */
1776         addrconf_add_mroute(dev);
1777
1778         /* Add link local route */
1779         addrconf_add_lroute(dev);
1780         return idev;
1781 }
1782
1783 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1784 {
1785         struct prefix_info *pinfo;
1786         __u32 valid_lft;
1787         __u32 prefered_lft;
1788         int addr_type;
1789         struct inet6_dev *in6_dev;
1790
1791         pinfo = (struct prefix_info *) opt;
1792
1793         if (len < sizeof(struct prefix_info)) {
1794                 ADBG(("addrconf: prefix option too short\n"));
1795                 return;
1796         }
1797
1798         /*
1799          *      Validation checks ([ADDRCONF], page 19)
1800          */
1801
1802         addr_type = ipv6_addr_type(&pinfo->prefix);
1803
1804         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1805                 return;
1806
1807         valid_lft = ntohl(pinfo->valid);
1808         prefered_lft = ntohl(pinfo->prefered);
1809
1810         if (prefered_lft > valid_lft) {
1811                 if (net_ratelimit())
1812                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1813                 return;
1814         }
1815
1816         in6_dev = in6_dev_get(dev);
1817
1818         if (in6_dev == NULL) {
1819                 if (net_ratelimit())
1820                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1821                 return;
1822         }
1823
1824         /*
1825          *      Two things going on here:
1826          *      1) Add routes for on-link prefixes
1827          *      2) Configure prefixes with the auto flag set
1828          */
1829
1830         if (pinfo->onlink) {
1831                 struct rt6_info *rt;
1832                 unsigned long rt_expires;
1833
1834                 /* Avoid arithmetic overflow. Really, we could
1835                  * save rt_expires in seconds, likely valid_lft,
1836                  * but it would require division in fib gc, that it
1837                  * not good.
1838                  */
1839                 if (HZ > USER_HZ)
1840                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1841                 else
1842                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1843
1844                 if (addrconf_finite_timeout(rt_expires))
1845                         rt_expires *= HZ;
1846
1847                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1848                                 dev->ifindex, 1);
1849
1850                 if (rt && addrconf_is_prefix_route(rt)) {
1851                         /* Autoconf prefix route */
1852                         if (valid_lft == 0) {
1853                                 ip6_del_rt(rt);
1854                                 rt = NULL;
1855                         } else if (addrconf_finite_timeout(rt_expires)) {
1856                                 /* not infinity */
1857                                 rt->rt6i_expires = jiffies + rt_expires;
1858                                 rt->rt6i_flags |= RTF_EXPIRES;
1859                         } else {
1860                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1861                                 rt->rt6i_expires = 0;
1862                         }
1863                 } else if (valid_lft) {
1864                         clock_t expires = 0;
1865                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1866                         if (addrconf_finite_timeout(rt_expires)) {
1867                                 /* not infinity */
1868                                 flags |= RTF_EXPIRES;
1869                                 expires = jiffies_to_clock_t(rt_expires);
1870                         }
1871                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1872                                               dev, expires, flags);
1873                 }
1874                 if (rt)
1875                         dst_release(&rt->u.dst);
1876         }
1877
1878         /* Try to figure out our local address for this prefix */
1879
1880         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1881                 struct inet6_ifaddr * ifp;
1882                 struct in6_addr addr;
1883                 int create = 0, update_lft = 0;
1884                 struct net *net = dev_net(dev);
1885
1886                 if (pinfo->prefix_len == 64) {
1887                         memcpy(&addr, &pinfo->prefix, 8);
1888                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1889                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1890                                 in6_dev_put(in6_dev);
1891                                 return;
1892                         }
1893                         goto ok;
1894                 }
1895                 if (net_ratelimit())
1896                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1897                                pinfo->prefix_len);
1898                 in6_dev_put(in6_dev);
1899                 return;
1900
1901 ok:
1902
1903                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1904
1905                 if (ifp == NULL && valid_lft) {
1906                         int max_addresses = in6_dev->cnf.max_addresses;
1907                         u32 addr_flags = 0;
1908
1909 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1910                         if (in6_dev->cnf.optimistic_dad &&
1911                             !net->ipv6.devconf_all->forwarding)
1912                                 addr_flags = IFA_F_OPTIMISTIC;
1913 #endif
1914
1915                         /* Do not allow to create too much of autoconfigured
1916                          * addresses; this would be too easy way to crash kernel.
1917                          */
1918                         if (!max_addresses ||
1919                             ipv6_count_addresses(in6_dev) < max_addresses)
1920                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1921                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1922                                                     addr_flags);
1923
1924                         if (!ifp || IS_ERR(ifp)) {
1925                                 in6_dev_put(in6_dev);
1926                                 return;
1927                         }
1928
1929                         update_lft = create = 1;
1930                         ifp->cstamp = jiffies;
1931                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1932                 }
1933
1934                 if (ifp) {
1935                         int flags;
1936                         unsigned long now;
1937 #ifdef CONFIG_IPV6_PRIVACY
1938                         struct inet6_ifaddr *ift;
1939 #endif
1940                         u32 stored_lft;
1941
1942                         /* update lifetime (RFC2462 5.5.3 e) */
1943                         spin_lock(&ifp->lock);
1944                         now = jiffies;
1945                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1946                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1947                         else
1948                                 stored_lft = 0;
1949                         if (!update_lft && stored_lft) {
1950                                 if (valid_lft > MIN_VALID_LIFETIME ||
1951                                     valid_lft > stored_lft)
1952                                         update_lft = 1;
1953                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1954                                         /* valid_lft <= stored_lft is always true */
1955                                         /* XXX: IPsec */
1956                                         update_lft = 0;
1957                                 } else {
1958                                         valid_lft = MIN_VALID_LIFETIME;
1959                                         if (valid_lft < prefered_lft)
1960                                                 prefered_lft = valid_lft;
1961                                         update_lft = 1;
1962                                 }
1963                         }
1964
1965                         if (update_lft) {
1966                                 ifp->valid_lft = valid_lft;
1967                                 ifp->prefered_lft = prefered_lft;
1968                                 ifp->tstamp = now;
1969                                 flags = ifp->flags;
1970                                 ifp->flags &= ~IFA_F_DEPRECATED;
1971                                 spin_unlock(&ifp->lock);
1972
1973                                 if (!(flags&IFA_F_TENTATIVE))
1974                                         ipv6_ifa_notify(0, ifp);
1975                         } else
1976                                 spin_unlock(&ifp->lock);
1977
1978 #ifdef CONFIG_IPV6_PRIVACY
1979                         read_lock_bh(&in6_dev->lock);
1980                         /* update all temporary addresses in the list */
1981                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1982                                 /*
1983                                  * When adjusting the lifetimes of an existing
1984                                  * temporary address, only lower the lifetimes.
1985                                  * Implementations must not increase the
1986                                  * lifetimes of an existing temporary address
1987                                  * when processing a Prefix Information Option.
1988                                  */
1989                                 if (ifp != ift->ifpub)
1990                                         continue;
1991
1992                                 spin_lock(&ift->lock);
1993                                 flags = ift->flags;
1994                                 if (ift->valid_lft > valid_lft &&
1995                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1996                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1997                                 if (ift->prefered_lft > prefered_lft &&
1998                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1999                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2000                                 spin_unlock(&ift->lock);
2001                                 if (!(flags&IFA_F_TENTATIVE))
2002                                         ipv6_ifa_notify(0, ift);
2003                         }
2004
2005                         if (create && in6_dev->cnf.use_tempaddr > 0) {
2006                                 /*
2007                                  * When a new public address is created as described in [ADDRCONF],
2008                                  * also create a new temporary address.
2009                                  */
2010                                 read_unlock_bh(&in6_dev->lock);
2011                                 ipv6_create_tempaddr(ifp, NULL);
2012                         } else {
2013                                 read_unlock_bh(&in6_dev->lock);
2014                         }
2015 #endif
2016                         in6_ifa_put(ifp);
2017                         addrconf_verify(0);
2018                 }
2019         }
2020         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2021         in6_dev_put(in6_dev);
2022 }
2023
2024 /*
2025  *      Set destination address.
2026  *      Special case for SIT interfaces where we create a new "virtual"
2027  *      device.
2028  */
2029 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2030 {
2031         struct in6_ifreq ireq;
2032         struct net_device *dev;
2033         int err = -EINVAL;
2034
2035         rtnl_lock();
2036
2037         err = -EFAULT;
2038         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2039                 goto err_exit;
2040
2041         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2042
2043         err = -ENODEV;
2044         if (dev == NULL)
2045                 goto err_exit;
2046
2047 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2048         if (dev->type == ARPHRD_SIT) {
2049                 const struct net_device_ops *ops = dev->netdev_ops;
2050                 struct ifreq ifr;
2051                 struct ip_tunnel_parm p;
2052
2053                 err = -EADDRNOTAVAIL;
2054                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2055                         goto err_exit;
2056
2057                 memset(&p, 0, sizeof(p));
2058                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2059                 p.iph.saddr = 0;
2060                 p.iph.version = 4;
2061                 p.iph.ihl = 5;
2062                 p.iph.protocol = IPPROTO_IPV6;
2063                 p.iph.ttl = 64;
2064                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2065
2066                 if (ops->ndo_do_ioctl) {
2067                         mm_segment_t oldfs = get_fs();
2068
2069                         set_fs(KERNEL_DS);
2070                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2071                         set_fs(oldfs);
2072                 } else
2073                         err = -EOPNOTSUPP;
2074
2075                 if (err == 0) {
2076                         err = -ENOBUFS;
2077                         dev = __dev_get_by_name(net, p.name);
2078                         if (!dev)
2079                                 goto err_exit;
2080                         err = dev_open(dev);
2081                 }
2082         }
2083 #endif
2084
2085 err_exit:
2086         rtnl_unlock();
2087         return err;
2088 }
2089
2090 /*
2091  *      Manual configuration of address on an interface
2092  */
2093 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2094                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2095                           __u32 valid_lft)
2096 {
2097         struct inet6_ifaddr *ifp;
2098         struct inet6_dev *idev;
2099         struct net_device *dev;
2100         int scope;
2101         u32 flags;
2102         clock_t expires;
2103         unsigned long timeout;
2104
2105         ASSERT_RTNL();
2106
2107         if (plen > 128)
2108                 return -EINVAL;
2109
2110         /* check the lifetime */
2111         if (!valid_lft || prefered_lft > valid_lft)
2112                 return -EINVAL;
2113
2114         dev = __dev_get_by_index(net, ifindex);
2115         if (!dev)
2116                 return -ENODEV;
2117
2118         if ((idev = addrconf_add_dev(dev)) == NULL)
2119                 return -ENOBUFS;
2120
2121         scope = ipv6_addr_scope(pfx);
2122
2123         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2124         if (addrconf_finite_timeout(timeout)) {
2125                 expires = jiffies_to_clock_t(timeout * HZ);
2126                 valid_lft = timeout;
2127                 flags = RTF_EXPIRES;
2128         } else {
2129                 expires = 0;
2130                 flags = 0;
2131                 ifa_flags |= IFA_F_PERMANENT;
2132         }
2133
2134         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2135         if (addrconf_finite_timeout(timeout)) {
2136                 if (timeout == 0)
2137                         ifa_flags |= IFA_F_DEPRECATED;
2138                 prefered_lft = timeout;
2139         }
2140
2141         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2142
2143         if (!IS_ERR(ifp)) {
2144                 spin_lock_bh(&ifp->lock);
2145                 ifp->valid_lft = valid_lft;
2146                 ifp->prefered_lft = prefered_lft;
2147                 ifp->tstamp = jiffies;
2148                 spin_unlock_bh(&ifp->lock);
2149
2150                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2151                                       expires, flags);
2152                 /*
2153                  * Note that section 3.1 of RFC 4429 indicates
2154                  * that the Optimistic flag should not be set for
2155                  * manually configured addresses
2156                  */
2157                 addrconf_dad_start(ifp, 0);
2158                 in6_ifa_put(ifp);
2159                 addrconf_verify(0);
2160                 return 0;
2161         }
2162
2163         return PTR_ERR(ifp);
2164 }
2165
2166 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2167                           unsigned int plen)
2168 {
2169         struct inet6_ifaddr *ifp;
2170         struct inet6_dev *idev;
2171         struct net_device *dev;
2172
2173         if (plen > 128)
2174                 return -EINVAL;
2175
2176         dev = __dev_get_by_index(net, ifindex);
2177         if (!dev)
2178                 return -ENODEV;
2179
2180         if ((idev = __in6_dev_get(dev)) == NULL)
2181                 return -ENXIO;
2182
2183         read_lock_bh(&idev->lock);
2184         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2185                 if (ifp->prefix_len == plen &&
2186                     ipv6_addr_equal(pfx, &ifp->addr)) {
2187                         in6_ifa_hold(ifp);
2188                         read_unlock_bh(&idev->lock);
2189
2190                         ipv6_del_addr(ifp);
2191
2192                         /* If the last address is deleted administratively,
2193                            disable IPv6 on this interface.
2194                          */
2195                         if (idev->addr_list == NULL)
2196                                 addrconf_ifdown(idev->dev, 1);
2197                         return 0;
2198                 }
2199         }
2200         read_unlock_bh(&idev->lock);
2201         return -EADDRNOTAVAIL;
2202 }
2203
2204
2205 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2206 {
2207         struct in6_ifreq ireq;
2208         int err;
2209
2210         if (!capable(CAP_NET_ADMIN))
2211                 return -EPERM;
2212
2213         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2214                 return -EFAULT;
2215
2216         rtnl_lock();
2217         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2218                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2219                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2220         rtnl_unlock();
2221         return err;
2222 }
2223
2224 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2225 {
2226         struct in6_ifreq ireq;
2227         int err;
2228
2229         if (!capable(CAP_NET_ADMIN))
2230                 return -EPERM;
2231
2232         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2233                 return -EFAULT;
2234
2235         rtnl_lock();
2236         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2237                              ireq.ifr6_prefixlen);
2238         rtnl_unlock();
2239         return err;
2240 }
2241
2242 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2243                      int plen, int scope)
2244 {
2245         struct inet6_ifaddr *ifp;
2246
2247         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2248         if (!IS_ERR(ifp)) {
2249                 spin_lock_bh(&ifp->lock);
2250                 ifp->flags &= ~IFA_F_TENTATIVE;
2251                 spin_unlock_bh(&ifp->lock);
2252                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2253                 in6_ifa_put(ifp);
2254         }
2255 }
2256
2257 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2258 static void sit_add_v4_addrs(struct inet6_dev *idev)
2259 {
2260         struct in6_addr addr;
2261         struct net_device *dev;
2262         struct net *net = dev_net(idev->dev);
2263         int scope;
2264
2265         ASSERT_RTNL();
2266
2267         memset(&addr, 0, sizeof(struct in6_addr));
2268         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2269
2270         if (idev->dev->flags&IFF_POINTOPOINT) {
2271                 addr.s6_addr32[0] = htonl(0xfe800000);
2272                 scope = IFA_LINK;
2273         } else {
2274                 scope = IPV6_ADDR_COMPATv4;
2275         }
2276
2277         if (addr.s6_addr32[3]) {
2278                 add_addr(idev, &addr, 128, scope);
2279                 return;
2280         }
2281
2282         for_each_netdev(net, dev) {
2283                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2284                 if (in_dev && (dev->flags & IFF_UP)) {
2285                         struct in_ifaddr * ifa;
2286
2287                         int flag = scope;
2288
2289                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2290                                 int plen;
2291
2292                                 addr.s6_addr32[3] = ifa->ifa_local;
2293
2294                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2295                                         continue;
2296                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2297                                         if (idev->dev->flags&IFF_POINTOPOINT)
2298                                                 continue;
2299                                         flag |= IFA_HOST;
2300                                 }
2301                                 if (idev->dev->flags&IFF_POINTOPOINT)
2302                                         plen = 64;
2303                                 else
2304                                         plen = 96;
2305
2306                                 add_addr(idev, &addr, plen, flag);
2307                         }
2308                 }
2309         }
2310 }
2311 #endif
2312
2313 static void init_loopback(struct net_device *dev)
2314 {
2315         struct inet6_dev  *idev;
2316
2317         /* ::1 */
2318
2319         ASSERT_RTNL();
2320
2321         if ((idev = ipv6_find_idev(dev)) == NULL) {
2322                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2323                 return;
2324         }
2325
2326         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2327 }
2328
2329 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2330 {
2331         struct inet6_ifaddr * ifp;
2332         u32 addr_flags = IFA_F_PERMANENT;
2333
2334 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2335         if (idev->cnf.optimistic_dad &&
2336             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2337                 addr_flags |= IFA_F_OPTIMISTIC;
2338 #endif
2339
2340
2341         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2342         if (!IS_ERR(ifp)) {
2343                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2344                 addrconf_dad_start(ifp, 0);
2345                 in6_ifa_put(ifp);
2346         }
2347 }
2348
2349 static void addrconf_dev_config(struct net_device *dev)
2350 {
2351         struct in6_addr addr;
2352         struct inet6_dev    * idev;
2353
2354         ASSERT_RTNL();
2355
2356         if ((dev->type != ARPHRD_ETHER) &&
2357             (dev->type != ARPHRD_FDDI) &&
2358             (dev->type != ARPHRD_IEEE802_TR) &&
2359             (dev->type != ARPHRD_ARCNET) &&
2360             (dev->type != ARPHRD_INFINIBAND)) {
2361                 /* Alas, we support only Ethernet autoconfiguration. */
2362                 return;
2363         }
2364
2365         idev = addrconf_add_dev(dev);
2366         if (idev == NULL)
2367                 return;
2368
2369         memset(&addr, 0, sizeof(struct in6_addr));
2370         addr.s6_addr32[0] = htonl(0xFE800000);
2371
2372         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2373                 addrconf_add_linklocal(idev, &addr);
2374 }
2375
2376 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2377 static void addrconf_sit_config(struct net_device *dev)
2378 {
2379         struct inet6_dev *idev;
2380
2381         ASSERT_RTNL();
2382
2383         /*
2384          * Configure the tunnel with one of our IPv4
2385          * addresses... we should configure all of
2386          * our v4 addrs in the tunnel
2387          */
2388
2389         if ((idev = ipv6_find_idev(dev)) == NULL) {
2390                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2391                 return;
2392         }
2393
2394         if (dev->priv_flags & IFF_ISATAP) {
2395                 struct in6_addr addr;
2396
2397                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2398                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2399                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2400                         addrconf_add_linklocal(idev, &addr);
2401                 return;
2402         }
2403
2404         sit_add_v4_addrs(idev);
2405
2406         if (dev->flags&IFF_POINTOPOINT) {
2407                 addrconf_add_mroute(dev);
2408                 addrconf_add_lroute(dev);
2409         } else
2410                 sit_route_add(dev);
2411 }
2412 #endif
2413
2414 static inline int
2415 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2416 {
2417         struct in6_addr lladdr;
2418
2419         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2420                 addrconf_add_linklocal(idev, &lladdr);
2421                 return 0;
2422         }
2423         return -1;
2424 }
2425
2426 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2427 {
2428         struct net_device *link_dev;
2429         struct net *net = dev_net(idev->dev);
2430
2431         /* first try to inherit the link-local address from the link device */
2432         if (idev->dev->iflink &&
2433             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2434                 if (!ipv6_inherit_linklocal(idev, link_dev))
2435                         return;
2436         }
2437         /* then try to inherit it from any device */
2438         for_each_netdev(net, link_dev) {
2439                 if (!ipv6_inherit_linklocal(idev, link_dev))
2440                         return;
2441         }
2442         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2443 }
2444
2445 /*
2446  * Autoconfigure tunnel with a link-local address so routing protocols,
2447  * DHCPv6, MLD etc. can be run over the virtual link
2448  */
2449
2450 static void addrconf_ip6_tnl_config(struct net_device *dev)
2451 {
2452         struct inet6_dev *idev;
2453
2454         ASSERT_RTNL();
2455
2456         if ((idev = addrconf_add_dev(dev)) == NULL) {
2457                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2458                 return;
2459         }
2460         ip6_tnl_add_linklocal(idev);
2461 }
2462
2463 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2464                            void * data)
2465 {
2466         struct net_device *dev = (struct net_device *) data;
2467         struct inet6_dev *idev = __in6_dev_get(dev);
2468         int run_pending = 0;
2469         int err;
2470
2471         switch(event) {
2472         case NETDEV_REGISTER:
2473                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2474                         idev = ipv6_add_dev(dev);
2475                         if (!idev)
2476                                 return notifier_from_errno(-ENOMEM);
2477                 }
2478                 break;
2479         case NETDEV_UP:
2480         case NETDEV_CHANGE:
2481                 if (dev->flags & IFF_SLAVE)
2482                         break;
2483
2484                 if (event == NETDEV_UP) {
2485                         if (!addrconf_qdisc_ok(dev)) {
2486                                 /* device is not ready yet. */
2487                                 printk(KERN_INFO
2488                                         "ADDRCONF(NETDEV_UP): %s: "
2489                                         "link is not ready\n",
2490                                         dev->name);
2491                                 break;
2492                         }
2493
2494                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2495                                 idev = ipv6_add_dev(dev);
2496
2497                         if (idev) {
2498                                 idev->if_flags |= IF_READY;
2499                                 run_pending = 1;
2500                         }
2501                 } else {
2502                         if (!addrconf_qdisc_ok(dev)) {
2503                                 /* device is still not ready. */
2504                                 break;
2505                         }
2506
2507                         if (idev) {
2508                                 if (idev->if_flags & IF_READY) {
2509                                         /* device is already configured. */
2510                                         break;
2511                                 }
2512                                 idev->if_flags |= IF_READY;
2513                         }
2514
2515                         printk(KERN_INFO
2516                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2517                                         "link becomes ready\n",
2518                                         dev->name);
2519
2520                         run_pending = 1;
2521                 }
2522
2523                 switch(dev->type) {
2524 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2525                 case ARPHRD_SIT:
2526                         addrconf_sit_config(dev);
2527                         break;
2528 #endif
2529                 case ARPHRD_TUNNEL6:
2530                         addrconf_ip6_tnl_config(dev);
2531                         break;
2532                 case ARPHRD_LOOPBACK:
2533                         init_loopback(dev);
2534                         break;
2535
2536                 default:
2537                         addrconf_dev_config(dev);
2538                         break;
2539                 }
2540                 if (idev) {
2541                         if (run_pending)
2542                                 addrconf_dad_run(idev);
2543
2544                         /* If the MTU changed during the interface down, when the
2545                            interface up, the changed MTU must be reflected in the
2546                            idev as well as routers.
2547                          */
2548                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2549                                 rt6_mtu_change(dev, dev->mtu);
2550                                 idev->cnf.mtu6 = dev->mtu;
2551                         }
2552                         idev->tstamp = jiffies;
2553                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2554                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2555                            stop IPv6 on this interface.
2556                          */
2557                         if (dev->mtu < IPV6_MIN_MTU)
2558                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2559                 }
2560                 break;
2561
2562         case NETDEV_CHANGEMTU:
2563                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2564                         rt6_mtu_change(dev, dev->mtu);
2565                         idev->cnf.mtu6 = dev->mtu;
2566                         break;
2567                 }
2568
2569                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2570                         idev = ipv6_add_dev(dev);
2571                         if (idev)
2572                                 break;
2573                 }
2574
2575                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2576
2577         case NETDEV_DOWN:
2578         case NETDEV_UNREGISTER:
2579                 /*
2580                  *      Remove all addresses from this interface.
2581                  */
2582                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2583                 break;
2584
2585         case NETDEV_CHANGENAME:
2586                 if (idev) {
2587                         snmp6_unregister_dev(idev);
2588                         addrconf_sysctl_unregister(idev);
2589                         addrconf_sysctl_register(idev);
2590                         err = snmp6_register_dev(idev);
2591                         if (err)
2592                                 return notifier_from_errno(err);
2593                 }
2594                 break;
2595         }
2596
2597         return NOTIFY_OK;
2598 }
2599
2600 /*
2601  *      addrconf module should be notified of a device going up
2602  */
2603 static struct notifier_block ipv6_dev_notf = {
2604         .notifier_call = addrconf_notify,
2605         .priority = 0
2606 };
2607
2608 static int addrconf_ifdown(struct net_device *dev, int how)
2609 {
2610         struct inet6_dev *idev;
2611         struct inet6_ifaddr *ifa, **bifa;
2612         struct net *net = dev_net(dev);
2613         int i;
2614
2615         ASSERT_RTNL();
2616
2617         rt6_ifdown(net, dev);
2618         neigh_ifdown(&nd_tbl, dev);
2619
2620         idev = __in6_dev_get(dev);
2621         if (idev == NULL)
2622                 return -ENODEV;
2623
2624         /* Step 1: remove reference to ipv6 device from parent device.
2625                    Do not dev_put!
2626          */
2627         if (how) {
2628                 idev->dead = 1;
2629
2630                 /* protected by rtnl_lock */
2631                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2632
2633                 /* Step 1.5: remove snmp6 entry */
2634                 snmp6_unregister_dev(idev);
2635
2636         }
2637
2638         /* Step 2: clear hash table */
2639         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2640                 bifa = &inet6_addr_lst[i];
2641
2642                 write_lock_bh(&addrconf_hash_lock);
2643                 while ((ifa = *bifa) != NULL) {
2644                         if (ifa->idev == idev) {
2645                                 *bifa = ifa->lst_next;
2646                                 ifa->lst_next = NULL;
2647                                 addrconf_del_timer(ifa);
2648                                 in6_ifa_put(ifa);
2649                                 continue;
2650                         }
2651                         bifa = &ifa->lst_next;
2652                 }
2653                 write_unlock_bh(&addrconf_hash_lock);
2654         }
2655
2656         write_lock_bh(&idev->lock);
2657
2658         /* Step 3: clear flags for stateless addrconf */
2659         if (!how)
2660                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2661
2662         /* Step 4: clear address list */
2663 #ifdef CONFIG_IPV6_PRIVACY
2664         if (how && del_timer(&idev->regen_timer))
2665                 in6_dev_put(idev);
2666
2667         /* clear tempaddr list */
2668         while ((ifa = idev->tempaddr_list) != NULL) {
2669                 idev->tempaddr_list = ifa->tmp_next;
2670                 ifa->tmp_next = NULL;
2671                 ifa->dead = 1;
2672                 write_unlock_bh(&idev->lock);
2673                 spin_lock_bh(&ifa->lock);
2674
2675                 if (ifa->ifpub) {
2676                         in6_ifa_put(ifa->ifpub);
2677                         ifa->ifpub = NULL;
2678                 }
2679                 spin_unlock_bh(&ifa->lock);
2680                 in6_ifa_put(ifa);
2681                 write_lock_bh(&idev->lock);
2682         }
2683 #endif
2684         while ((ifa = idev->addr_list) != NULL) {
2685                 idev->addr_list = ifa->if_next;
2686                 ifa->if_next = NULL;
2687                 ifa->dead = 1;
2688                 addrconf_del_timer(ifa);
2689                 write_unlock_bh(&idev->lock);
2690
2691                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2692                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2693                 in6_ifa_put(ifa);
2694
2695                 write_lock_bh(&idev->lock);
2696         }
2697         write_unlock_bh(&idev->lock);
2698
2699         /* Step 5: Discard multicast list */
2700
2701         if (how)
2702                 ipv6_mc_destroy_dev(idev);
2703         else
2704                 ipv6_mc_down(idev);
2705
2706         idev->tstamp = jiffies;
2707
2708         /* Shot the device (if unregistered) */
2709
2710         if (how) {
2711                 addrconf_sysctl_unregister(idev);
2712                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2713                 neigh_ifdown(&nd_tbl, dev);
2714                 in6_dev_put(idev);
2715         }
2716         return 0;
2717 }
2718
2719 static void addrconf_rs_timer(unsigned long data)
2720 {
2721         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2722
2723         if (ifp->idev->cnf.forwarding)
2724                 goto out;
2725
2726         if (ifp->idev->if_flags & IF_RA_RCVD) {
2727                 /*
2728                  *      Announcement received after solicitation
2729                  *      was sent
2730                  */
2731                 goto out;
2732         }
2733
2734         spin_lock(&ifp->lock);
2735         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2736                 /* The wait after the last probe can be shorter */
2737                 addrconf_mod_timer(ifp, AC_RS,
2738                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2739                                    ifp->idev->cnf.rtr_solicit_delay :
2740                                    ifp->idev->cnf.rtr_solicit_interval);
2741                 spin_unlock(&ifp->lock);
2742
2743                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2744         } else {
2745                 spin_unlock(&ifp->lock);
2746                 /*
2747                  * Note: we do not support deprecated "all on-link"
2748                  * assumption any longer.
2749                  */
2750                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2751                        ifp->idev->dev->name);
2752         }
2753
2754 out:
2755         in6_ifa_put(ifp);
2756 }
2757
2758 /*
2759  *      Duplicate Address Detection
2760  */
2761 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2762 {
2763         unsigned long rand_num;
2764         struct inet6_dev *idev = ifp->idev;
2765
2766         if (ifp->flags & IFA_F_OPTIMISTIC)
2767                 rand_num = 0;
2768         else
2769                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2770
2771         ifp->probes = idev->cnf.dad_transmits;
2772         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2773 }
2774
2775 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2776 {
2777         struct inet6_dev *idev = ifp->idev;
2778         struct net_device *dev = idev->dev;
2779
2780         addrconf_join_solict(dev, &ifp->addr);
2781
2782         net_srandom(ifp->addr.s6_addr32[3]);
2783
2784         read_lock_bh(&idev->lock);
2785         if (ifp->dead)
2786                 goto out;
2787         spin_lock_bh(&ifp->lock);
2788
2789         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2790             idev->cnf.accept_dad < 1 ||
2791             !(ifp->flags&IFA_F_TENTATIVE) ||
2792             ifp->flags & IFA_F_NODAD) {
2793                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2794                 spin_unlock_bh(&ifp->lock);
2795                 read_unlock_bh(&idev->lock);
2796
2797                 addrconf_dad_completed(ifp);
2798                 return;
2799         }
2800
2801         if (!(idev->if_flags & IF_READY)) {
2802                 spin_unlock_bh(&ifp->lock);
2803                 read_unlock_bh(&idev->lock);
2804                 /*
2805                  * If the defice is not ready:
2806                  * - keep it tentative if it is a permanent address.
2807                  * - otherwise, kill it.
2808                  */
2809                 in6_ifa_hold(ifp);
2810                 addrconf_dad_stop(ifp);
2811                 return;
2812         }
2813
2814         /*
2815          * Optimistic nodes can start receiving
2816          * Frames right away
2817          */
2818         if(ifp->flags & IFA_F_OPTIMISTIC)
2819                 ip6_ins_rt(ifp->rt);
2820
2821         addrconf_dad_kick(ifp);
2822         spin_unlock_bh(&ifp->lock);
2823 out:
2824         read_unlock_bh(&idev->lock);
2825 }
2826
2827 static void addrconf_dad_timer(unsigned long data)
2828 {
2829         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2830         struct inet6_dev *idev = ifp->idev;
2831         struct in6_addr mcaddr;
2832
2833         read_lock_bh(&idev->lock);
2834         if (idev->dead) {
2835                 read_unlock_bh(&idev->lock);
2836                 goto out;
2837         }
2838         spin_lock_bh(&ifp->lock);
2839         if (ifp->probes == 0) {
2840                 /*
2841                  * DAD was successful
2842                  */
2843
2844                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2845                 spin_unlock_bh(&ifp->lock);
2846                 read_unlock_bh(&idev->lock);
2847
2848                 addrconf_dad_completed(ifp);
2849
2850                 goto out;
2851         }
2852
2853         ifp->probes--;
2854         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2855         spin_unlock_bh(&ifp->lock);
2856         read_unlock_bh(&idev->lock);
2857
2858         /* send a neighbour solicitation for our addr */
2859         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2860         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2861 out:
2862         in6_ifa_put(ifp);
2863 }
2864
2865 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2866 {
2867         struct net_device *     dev = ifp->idev->dev;
2868
2869         /*
2870          *      Configure the address for reception. Now it is valid.
2871          */
2872
2873         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2874
2875         /* If added prefix is link local and forwarding is off,
2876            start sending router solicitations.
2877          */
2878
2879         if (ifp->idev->cnf.forwarding == 0 &&
2880             ifp->idev->cnf.rtr_solicits > 0 &&
2881             (dev->flags&IFF_LOOPBACK) == 0 &&
2882             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2883                 /*
2884                  *      If a host as already performed a random delay
2885                  *      [...] as part of DAD [...] there is no need
2886                  *      to delay again before sending the first RS
2887                  */
2888                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2889
2890                 spin_lock_bh(&ifp->lock);
2891                 ifp->probes = 1;
2892                 ifp->idev->if_flags |= IF_RS_SENT;
2893                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2894                 spin_unlock_bh(&ifp->lock);
2895         }
2896 }
2897
2898 static void addrconf_dad_run(struct inet6_dev *idev) {
2899         struct inet6_ifaddr *ifp;
2900
2901         read_lock_bh(&idev->lock);
2902         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2903                 spin_lock_bh(&ifp->lock);
2904                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2905                         spin_unlock_bh(&ifp->lock);
2906                         continue;
2907                 }
2908                 spin_unlock_bh(&ifp->lock);
2909                 addrconf_dad_kick(ifp);
2910         }
2911         read_unlock_bh(&idev->lock);
2912 }
2913
2914 #ifdef CONFIG_PROC_FS
2915 struct if6_iter_state {
2916         struct seq_net_private p;
2917         int bucket;
2918 };
2919
2920 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2921 {
2922         struct inet6_ifaddr *ifa = NULL;
2923         struct if6_iter_state *state = seq->private;
2924         struct net *net = seq_file_net(seq);
2925
2926         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2927                 ifa = inet6_addr_lst[state->bucket];
2928
2929                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2930                         ifa = ifa->lst_next;
2931                 if (ifa)
2932                         break;
2933         }
2934         return ifa;
2935 }
2936
2937 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2938 {
2939         struct if6_iter_state *state = seq->private;
2940         struct net *net = seq_file_net(seq);
2941
2942         ifa = ifa->lst_next;
2943 try_again:
2944         if (ifa) {
2945                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2946                         ifa = ifa->lst_next;
2947                         goto try_again;
2948                 }
2949         }
2950
2951         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2952                 ifa = inet6_addr_lst[state->bucket];
2953                 goto try_again;
2954         }
2955
2956         return ifa;
2957 }
2958
2959 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2960 {
2961         struct inet6_ifaddr *ifa = if6_get_first(seq);
2962
2963         if (ifa)
2964                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2965                         --pos;
2966         return pos ? NULL : ifa;
2967 }
2968
2969 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2970         __acquires(addrconf_hash_lock)
2971 {
2972         read_lock_bh(&addrconf_hash_lock);
2973         return if6_get_idx(seq, *pos);
2974 }
2975
2976 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2977 {
2978         struct inet6_ifaddr *ifa;
2979
2980         ifa = if6_get_next(seq, v);
2981         ++*pos;
2982         return ifa;
2983 }
2984
2985 static void if6_seq_stop(struct seq_file *seq, void *v)
2986         __releases(addrconf_hash_lock)
2987 {
2988         read_unlock_bh(&addrconf_hash_lock);
2989 }
2990
2991 static int if6_seq_show(struct seq_file *seq, void *v)
2992 {
2993         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2994         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2995                    &ifp->addr,
2996                    ifp->idev->dev->ifindex,
2997                    ifp->prefix_len,
2998                    ifp->scope,
2999                    ifp->flags,
3000                    ifp->idev->dev->name);
3001         return 0;
3002 }
3003
3004 static const struct seq_operations if6_seq_ops = {
3005         .start  = if6_seq_start,
3006         .next   = if6_seq_next,
3007         .show   = if6_seq_show,
3008         .stop   = if6_seq_stop,
3009 };
3010
3011 static int if6_seq_open(struct inode *inode, struct file *file)
3012 {
3013         return seq_open_net(inode, file, &if6_seq_ops,
3014                             sizeof(struct if6_iter_state));
3015 }
3016
3017 static const struct file_operations if6_fops = {
3018         .owner          = THIS_MODULE,
3019         .open           = if6_seq_open,
3020         .read           = seq_read,
3021         .llseek         = seq_lseek,
3022         .release        = seq_release_net,
3023 };
3024
3025 static int if6_proc_net_init(struct net *net)
3026 {
3027         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3028                 return -ENOMEM;
3029         return 0;
3030 }
3031
3032 static void if6_proc_net_exit(struct net *net)
3033 {
3034        proc_net_remove(net, "if_inet6");
3035 }
3036
3037 static struct pernet_operations if6_proc_net_ops = {
3038        .init = if6_proc_net_init,
3039        .exit = if6_proc_net_exit,
3040 };
3041
3042 int __init if6_proc_init(void)
3043 {
3044         return register_pernet_subsys(&if6_proc_net_ops);
3045 }
3046
3047 void if6_proc_exit(void)
3048 {
3049         unregister_pernet_subsys(&if6_proc_net_ops);
3050 }
3051 #endif  /* CONFIG_PROC_FS */
3052
3053 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3054 /* Check if address is a home address configured on any interface. */
3055 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3056 {
3057         int ret = 0;
3058         struct inet6_ifaddr * ifp;
3059         u8 hash = ipv6_addr_hash(addr);
3060         read_lock_bh(&addrconf_hash_lock);
3061         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3062                 if (!net_eq(dev_net(ifp->idev->dev), net))
3063                         continue;
3064                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3065                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3066                         ret = 1;
3067                         break;
3068                 }
3069         }
3070         read_unlock_bh(&addrconf_hash_lock);
3071         return ret;
3072 }
3073 #endif
3074
3075 /*
3076  *      Periodic address status verification
3077  */
3078
3079 static void addrconf_verify(unsigned long foo)
3080 {
3081         struct inet6_ifaddr *ifp;
3082         unsigned long now, next;
3083         int i;
3084
3085         spin_lock_bh(&addrconf_verify_lock);
3086         now = jiffies;
3087         next = now + ADDR_CHECK_FREQUENCY;
3088
3089         del_timer(&addr_chk_timer);
3090
3091         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3092
3093 restart:
3094                 read_lock(&addrconf_hash_lock);
3095                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3096                         unsigned long age;
3097 #ifdef CONFIG_IPV6_PRIVACY
3098                         unsigned long regen_advance;
3099 #endif
3100
3101                         if (ifp->flags & IFA_F_PERMANENT)
3102                                 continue;
3103
3104                         spin_lock(&ifp->lock);
3105                         age = (now - ifp->tstamp) / HZ;
3106
3107 #ifdef CONFIG_IPV6_PRIVACY
3108                         regen_advance = ifp->idev->cnf.regen_max_retry *
3109                                         ifp->idev->cnf.dad_transmits *
3110                                         ifp->idev->nd_parms->retrans_time / HZ;
3111 #endif
3112
3113                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3114                             age >= ifp->valid_lft) {
3115                                 spin_unlock(&ifp->lock);
3116                                 in6_ifa_hold(ifp);
3117                                 read_unlock(&addrconf_hash_lock);
3118                                 ipv6_del_addr(ifp);
3119                                 goto restart;
3120                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3121                                 spin_unlock(&ifp->lock);
3122                                 continue;
3123                         } else if (age >= ifp->prefered_lft) {
3124                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3125                                 int deprecate = 0;
3126
3127                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3128                                         deprecate = 1;
3129                                         ifp->flags |= IFA_F_DEPRECATED;
3130                                 }
3131
3132                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3133                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3134
3135                                 spin_unlock(&ifp->lock);
3136
3137                                 if (deprecate) {
3138                                         in6_ifa_hold(ifp);
3139                                         read_unlock(&addrconf_hash_lock);
3140
3141                                         ipv6_ifa_notify(0, ifp);
3142                                         in6_ifa_put(ifp);
3143                                         goto restart;
3144                                 }
3145 #ifdef CONFIG_IPV6_PRIVACY
3146                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3147                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3148                                 if (age >= ifp->prefered_lft - regen_advance) {
3149                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3150                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3151                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3152                                         if (!ifp->regen_count && ifpub) {
3153                                                 ifp->regen_count++;
3154                                                 in6_ifa_hold(ifp);
3155                                                 in6_ifa_hold(ifpub);
3156                                                 spin_unlock(&ifp->lock);
3157                                                 read_unlock(&addrconf_hash_lock);
3158                                                 spin_lock(&ifpub->lock);
3159                                                 ifpub->regen_count = 0;
3160                                                 spin_unlock(&ifpub->lock);
3161                                                 ipv6_create_tempaddr(ifpub, ifp);
3162                                                 in6_ifa_put(ifpub);
3163                                                 in6_ifa_put(ifp);
3164                                                 goto restart;
3165                                         }
3166                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3167                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3168                                 spin_unlock(&ifp->lock);
3169 #endif
3170                         } else {
3171                                 /* ifp->prefered_lft <= ifp->valid_lft */
3172                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3173                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3174                                 spin_unlock(&ifp->lock);
3175                         }
3176                 }
3177                 read_unlock(&addrconf_hash_lock);
3178         }
3179
3180         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3181         add_timer(&addr_chk_timer);
3182         spin_unlock_bh(&addrconf_verify_lock);
3183 }
3184
3185 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3186 {
3187         struct in6_addr *pfx = NULL;
3188
3189         if (addr)
3190                 pfx = nla_data(addr);
3191
3192         if (local) {
3193                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3194                         pfx = NULL;
3195                 else
3196                         pfx = nla_data(local);
3197         }
3198
3199         return pfx;
3200 }
3201
3202 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3203         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3204         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3205         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3206 };
3207
3208 static int
3209 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3210 {
3211         struct net *net = sock_net(skb->sk);
3212         struct ifaddrmsg *ifm;
3213         struct nlattr *tb[IFA_MAX+1];
3214         struct in6_addr *pfx;
3215         int err;
3216
3217         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3218         if (err < 0)
3219                 return err;
3220
3221         ifm = nlmsg_data(nlh);
3222         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3223         if (pfx == NULL)
3224                 return -EINVAL;
3225
3226         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3227 }
3228
3229 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3230                              u32 prefered_lft, u32 valid_lft)
3231 {
3232         u32 flags;
3233         clock_t expires;
3234         unsigned long timeout;
3235
3236         if (!valid_lft || (prefered_lft > valid_lft))
3237                 return -EINVAL;
3238
3239         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3240         if (addrconf_finite_timeout(timeout)) {
3241                 expires = jiffies_to_clock_t(timeout * HZ);
3242                 valid_lft = timeout;
3243                 flags = RTF_EXPIRES;
3244         } else {
3245                 expires = 0;
3246                 flags = 0;
3247                 ifa_flags |= IFA_F_PERMANENT;
3248         }
3249
3250         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3251         if (addrconf_finite_timeout(timeout)) {
3252                 if (timeout == 0)
3253                         ifa_flags |= IFA_F_DEPRECATED;
3254                 prefered_lft = timeout;
3255         }
3256
3257         spin_lock_bh(&ifp->lock);
3258         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3259         ifp->tstamp = jiffies;
3260         ifp->valid_lft = valid_lft;
3261         ifp->prefered_lft = prefered_lft;
3262
3263         spin_unlock_bh(&ifp->lock);
3264         if (!(ifp->flags&IFA_F_TENTATIVE))
3265                 ipv6_ifa_notify(0, ifp);
3266
3267         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3268                               expires, flags);
3269         addrconf_verify(0);
3270
3271         return 0;
3272 }
3273
3274 static int
3275 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3276 {
3277         struct net *net = sock_net(skb->sk);
3278         struct ifaddrmsg *ifm;
3279         struct nlattr *tb[IFA_MAX+1];
3280         struct in6_addr *pfx;
3281         struct inet6_ifaddr *ifa;
3282         struct net_device *dev;
3283         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3284         u8 ifa_flags;
3285         int err;
3286
3287         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3288         if (err < 0)
3289                 return err;
3290
3291         ifm = nlmsg_data(nlh);
3292         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3293         if (pfx == NULL)
3294                 return -EINVAL;
3295
3296         if (tb[IFA_CACHEINFO]) {
3297                 struct ifa_cacheinfo *ci;
3298
3299                 ci = nla_data(tb[IFA_CACHEINFO]);
3300                 valid_lft = ci->ifa_valid;
3301                 preferred_lft = ci->ifa_prefered;
3302         } else {
3303                 preferred_lft = INFINITY_LIFE_TIME;
3304                 valid_lft = INFINITY_LIFE_TIME;
3305         }
3306
3307         dev =  __dev_get_by_index(net, ifm->ifa_index);
3308         if (dev == NULL)
3309                 return -ENODEV;
3310
3311         /* We ignore other flags so far. */
3312         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3313
3314         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3315         if (ifa == NULL) {
3316                 /*
3317                  * It would be best to check for !NLM_F_CREATE here but
3318                  * userspace alreay relies on not having to provide this.
3319                  */
3320                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3321                                       ifm->ifa_prefixlen, ifa_flags,
3322                                       preferred_lft, valid_lft);
3323         }
3324
3325         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3326             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3327                 err = -EEXIST;
3328         else
3329                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3330
3331         in6_ifa_put(ifa);
3332
3333         return err;
3334 }
3335
3336 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3337                           u8 scope, int ifindex)
3338 {
3339         struct ifaddrmsg *ifm;
3340
3341         ifm = nlmsg_data(nlh);
3342         ifm->ifa_family = AF_INET6;
3343         ifm->ifa_prefixlen = prefixlen;
3344         ifm->ifa_flags = flags;
3345         ifm->ifa_scope = scope;
3346         ifm->ifa_index = ifindex;
3347 }
3348
3349 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3350                          unsigned long tstamp, u32 preferred, u32 valid)
3351 {
3352         struct ifa_cacheinfo ci;
3353
3354         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3355                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3356         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3357                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3358         ci.ifa_prefered = preferred;
3359         ci.ifa_valid = valid;
3360
3361         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3362 }
3363
3364 static inline int rt_scope(int ifa_scope)
3365 {
3366         if (ifa_scope & IFA_HOST)
3367                 return RT_SCOPE_HOST;
3368         else if (ifa_scope & IFA_LINK)
3369                 return RT_SCOPE_LINK;
3370         else if (ifa_scope & IFA_SITE)
3371                 return RT_SCOPE_SITE;
3372         else
3373                 return RT_SCOPE_UNIVERSE;
3374 }
3375
3376 static inline int inet6_ifaddr_msgsize(void)
3377 {
3378         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3379                + nla_total_size(16) /* IFA_ADDRESS */
3380                + nla_total_size(sizeof(struct ifa_cacheinfo));
3381 }
3382
3383 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3384                              u32 pid, u32 seq, int event, unsigned int flags)
3385 {
3386         struct nlmsghdr  *nlh;
3387         u32 preferred, valid;
3388
3389         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3390         if (nlh == NULL)
3391                 return -EMSGSIZE;
3392
3393         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3394                       ifa->idev->dev->ifindex);
3395
3396         if (!(ifa->flags&IFA_F_PERMANENT)) {
3397                 preferred = ifa->prefered_lft;
3398                 valid = ifa->valid_lft;
3399                 if (preferred != INFINITY_LIFE_TIME) {
3400                         long tval = (jiffies - ifa->tstamp)/HZ;
3401                         preferred -= tval;
3402                         if (valid != INFINITY_LIFE_TIME)
3403                                 valid -= tval;
3404                 }
3405         } else {
3406                 preferred = INFINITY_LIFE_TIME;
3407                 valid = INFINITY_LIFE_TIME;
3408         }
3409
3410         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3411             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3412                 nlmsg_cancel(skb, nlh);
3413                 return -EMSGSIZE;
3414         }
3415
3416         return nlmsg_end(skb, nlh);
3417 }
3418
3419 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3420                                 u32 pid, u32 seq, int event, u16 flags)
3421 {
3422         struct nlmsghdr  *nlh;
3423         u8 scope = RT_SCOPE_UNIVERSE;
3424         int ifindex = ifmca->idev->dev->ifindex;
3425
3426         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3427                 scope = RT_SCOPE_SITE;
3428
3429         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3430         if (nlh == NULL)
3431                 return -EMSGSIZE;
3432
3433         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3434         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3435             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3436                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3437                 nlmsg_cancel(skb, nlh);
3438                 return -EMSGSIZE;
3439         }
3440
3441         return nlmsg_end(skb, nlh);
3442 }
3443
3444 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3445                                 u32 pid, u32 seq, int event, unsigned int flags)
3446 {
3447         struct nlmsghdr  *nlh;
3448         u8 scope = RT_SCOPE_UNIVERSE;
3449         int ifindex = ifaca->aca_idev->dev->ifindex;
3450
3451         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3452                 scope = RT_SCOPE_SITE;
3453
3454         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3455         if (nlh == NULL)
3456                 return -EMSGSIZE;
3457
3458         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3459         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3460             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3461                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3462                 nlmsg_cancel(skb, nlh);
3463                 return -EMSGSIZE;
3464         }
3465
3466         return nlmsg_end(skb, nlh);
3467 }
3468
3469 enum addr_type_t
3470 {
3471         UNICAST_ADDR,
3472         MULTICAST_ADDR,
3473         ANYCAST_ADDR,
3474 };
3475
3476 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3477                            enum addr_type_t type)
3478 {
3479         int idx, ip_idx;
3480         int s_idx, s_ip_idx;
3481         int err = 1;
3482         struct net_device *dev;
3483         struct inet6_dev *idev = NULL;
3484         struct inet6_ifaddr *ifa;
3485         struct ifmcaddr6 *ifmca;
3486         struct ifacaddr6 *ifaca;
3487         struct net *net = sock_net(skb->sk);
3488
3489         s_idx = cb->args[0];
3490         s_ip_idx = ip_idx = cb->args[1];
3491
3492         idx = 0;
3493         for_each_netdev(net, dev) {
3494                 if (idx < s_idx)
3495                         goto cont;
3496                 if (idx > s_idx)
3497                         s_ip_idx = 0;
3498                 ip_idx = 0;
3499                 if ((idev = in6_dev_get(dev)) == NULL)
3500                         goto cont;
3501                 read_lock_bh(&idev->lock);
3502                 switch (type) {
3503                 case UNICAST_ADDR:
3504                         /* unicast address incl. temp addr */
3505                         for (ifa = idev->addr_list; ifa;
3506                              ifa = ifa->if_next, ip_idx++) {
3507                                 if (ip_idx < s_ip_idx)
3508                                         continue;
3509                                 err = inet6_fill_ifaddr(skb, ifa,
3510                                                         NETLINK_CB(cb->skb).pid,
3511                                                         cb->nlh->nlmsg_seq,
3512                                                         RTM_NEWADDR,
3513                                                         NLM_F_MULTI);
3514                         }
3515                         break;
3516                 case MULTICAST_ADDR:
3517                         /* multicast address */
3518                         for (ifmca = idev->mc_list; ifmca;
3519                              ifmca = ifmca->next, ip_idx++) {
3520                                 if (ip_idx < s_ip_idx)
3521                                         continue;
3522                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3523                                                           NETLINK_CB(cb->skb).pid,
3524                                                           cb->nlh->nlmsg_seq,
3525                                                           RTM_GETMULTICAST,
3526                                                           NLM_F_MULTI);
3527                         }
3528                         break;
3529                 case ANYCAST_ADDR:
3530                         /* anycast address */
3531                         for (ifaca = idev->ac_list; ifaca;
3532                              ifaca = ifaca->aca_next, ip_idx++) {
3533                                 if (ip_idx < s_ip_idx)
3534                                         continue;
3535                                 err = inet6_fill_ifacaddr(skb, ifaca,
3536                                                           NETLINK_CB(cb->skb).pid,
3537                                                           cb->nlh->nlmsg_seq,
3538                                                           RTM_GETANYCAST,
3539                                                           NLM_F_MULTI);
3540                         }
3541                         break;
3542                 default:
3543                         break;
3544                 }
3545                 read_unlock_bh(&idev->lock);
3546                 in6_dev_put(idev);
3547
3548                 if (err <= 0)
3549                         break;
3550 cont:
3551                 idx++;
3552         }
3553         cb->args[0] = idx;
3554         cb->args[1] = ip_idx;
3555         return skb->len;
3556 }
3557
3558 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3559 {
3560         enum addr_type_t type = UNICAST_ADDR;
3561
3562         return inet6_dump_addr(skb, cb, type);
3563 }
3564
3565 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3566 {
3567         enum addr_type_t type = MULTICAST_ADDR;
3568
3569         return inet6_dump_addr(skb, cb, type);
3570 }
3571
3572
3573 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3574 {
3575         enum addr_type_t type = ANYCAST_ADDR;
3576
3577         return inet6_dump_addr(skb, cb, type);
3578 }
3579
3580 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3581                              void *arg)
3582 {
3583         struct net *net = sock_net(in_skb->sk);
3584         struct ifaddrmsg *ifm;
3585         struct nlattr *tb[IFA_MAX+1];
3586         struct in6_addr *addr = NULL;
3587         struct net_device *dev = NULL;
3588         struct inet6_ifaddr *ifa;
3589         struct sk_buff *skb;
3590         int err;
3591
3592         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3593         if (err < 0)
3594                 goto errout;
3595
3596         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3597         if (addr == NULL) {
3598                 err = -EINVAL;
3599                 goto errout;
3600         }
3601
3602         ifm = nlmsg_data(nlh);
3603         if (ifm->ifa_index)
3604                 dev = __dev_get_by_index(net, ifm->ifa_index);
3605
3606         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3607                 err = -EADDRNOTAVAIL;
3608                 goto errout;
3609         }
3610
3611         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3612                 err = -ENOBUFS;
3613                 goto errout_ifa;
3614         }
3615
3616         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3617                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3618         if (err < 0) {
3619                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3620                 WARN_ON(err == -EMSGSIZE);
3621                 kfree_skb(skb);
3622                 goto errout_ifa;
3623         }
3624         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3625 errout_ifa:
3626         in6_ifa_put(ifa);
3627 errout:
3628         return err;
3629 }
3630
3631 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3632 {
3633         struct sk_buff *skb;
3634         struct net *net = dev_net(ifa->idev->dev);
3635         int err = -ENOBUFS;
3636
3637         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3638         if (skb == NULL)
3639                 goto errout;
3640
3641         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3642         if (err < 0) {
3643                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3644                 WARN_ON(err == -EMSGSIZE);
3645                 kfree_skb(skb);
3646                 goto errout;
3647         }
3648         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3649         return;
3650 errout:
3651         if (err < 0)
3652                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3653 }
3654
3655 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3656                                 __s32 *array, int bytes)
3657 {
3658         BUG_ON(bytes < (DEVCONF_MAX * 4));
3659
3660         memset(array, 0, bytes);
3661         array[DEVCONF_FORWARDING] = cnf->forwarding;
3662         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3663         array[DEVCONF_MTU6] = cnf->mtu6;
3664         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3665         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3666         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3667         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3668         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3669         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3670         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3671         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3672 #ifdef CONFIG_IPV6_PRIVACY
3673         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3674         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3675         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3676         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3677         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3678 #endif
3679         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3680         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3681         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3682 #ifdef CONFIG_IPV6_ROUTER_PREF
3683         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3684         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3685 #ifdef CONFIG_IPV6_ROUTE_INFO
3686         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3687 #endif
3688 #endif
3689         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3690         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3691 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3692         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3693 #endif
3694 #ifdef CONFIG_IPV6_MROUTE
3695         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3696 #endif
3697         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3698         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3699 }
3700
3701 static inline size_t inet6_if_nlmsg_size(void)
3702 {
3703         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3704                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3705                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3706                + nla_total_size(4) /* IFLA_MTU */
3707                + nla_total_size(4) /* IFLA_LINK */
3708                + nla_total_size( /* IFLA_PROTINFO */
3709                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3710                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3711                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3712                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3713                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3714                  );
3715 }
3716
3717 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3718                                       int bytes)
3719 {
3720         int i;
3721         int pad = bytes - sizeof(u64) * items;
3722         BUG_ON(pad < 0);
3723
3724         /* Use put_unaligned() because stats may not be aligned for u64. */
3725         put_unaligned(items, &stats[0]);
3726         for (i = 1; i < items; i++)
3727                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3728
3729         memset(&stats[items], 0, pad);
3730 }
3731
3732 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3733                              int bytes)
3734 {
3735         switch(attrtype) {
3736         case IFLA_INET6_STATS:
3737                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3738                 break;
3739         case IFLA_INET6_ICMP6STATS:
3740                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3741                 break;
3742         }
3743 }
3744
3745 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3746                              u32 pid, u32 seq, int event, unsigned int flags)
3747 {
3748         struct net_device *dev = idev->dev;
3749         struct nlattr *nla;
3750         struct ifinfomsg *hdr;
3751         struct nlmsghdr *nlh;
3752         void *protoinfo;
3753         struct ifla_cacheinfo ci;
3754
3755         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3756         if (nlh == NULL)
3757                 return -EMSGSIZE;
3758
3759         hdr = nlmsg_data(nlh);
3760         hdr->ifi_family = AF_INET6;
3761         hdr->__ifi_pad = 0;
3762         hdr->ifi_type = dev->type;
3763         hdr->ifi_index = dev->ifindex;
3764         hdr->ifi_flags = dev_get_flags(dev);
3765         hdr->ifi_change = 0;
3766
3767         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3768
3769         if (dev->addr_len)
3770                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3771
3772         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3773         if (dev->ifindex != dev->iflink)
3774                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3775
3776         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3777         if (protoinfo == NULL)
3778                 goto nla_put_failure;
3779
3780         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3781
3782         ci.max_reasm_len = IPV6_MAXPLEN;
3783         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3784                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3785         ci.reachable_time = idev->nd_parms->reachable_time;
3786         ci.retrans_time = idev->nd_parms->retrans_time;
3787         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3788
3789         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3790         if (nla == NULL)
3791                 goto nla_put_failure;
3792         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3793
3794         /* XXX - MC not implemented */
3795
3796         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3797         if (nla == NULL)
3798                 goto nla_put_failure;
3799         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3800
3801         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3802         if (nla == NULL)
3803                 goto nla_put_failure;
3804         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3805
3806         nla_nest_end(skb, protoinfo);
3807         return nlmsg_end(skb, nlh);
3808
3809 nla_put_failure:
3810         nlmsg_cancel(skb, nlh);
3811         return -EMSGSIZE;
3812 }
3813
3814 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3815 {
3816         struct net *net = sock_net(skb->sk);
3817         int idx, err;
3818         int s_idx = cb->args[0];
3819         struct net_device *dev;
3820         struct inet6_dev *idev;
3821
3822         read_lock(&dev_base_lock);
3823         idx = 0;
3824         for_each_netdev(net, dev) {
3825                 if (idx < s_idx)
3826                         goto cont;
3827                 if ((idev = in6_dev_get(dev)) == NULL)
3828                         goto cont;
3829                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3830                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3831                 in6_dev_put(idev);
3832                 if (err <= 0)
3833                         break;
3834 cont:
3835                 idx++;
3836         }
3837         read_unlock(&dev_base_lock);
3838         cb->args[0] = idx;
3839
3840         return skb->len;
3841 }
3842
3843 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3844 {
3845         struct sk_buff *skb;
3846         struct net *net = dev_net(idev->dev);
3847         int err = -ENOBUFS;
3848
3849         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3850         if (skb == NULL)
3851                 goto errout;
3852
3853         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3854         if (err < 0) {
3855                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3856                 WARN_ON(err == -EMSGSIZE);
3857                 kfree_skb(skb);
3858                 goto errout;
3859         }
3860         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3861         return;
3862 errout:
3863         if (err < 0)
3864                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3865 }
3866
3867 static inline size_t inet6_prefix_nlmsg_size(void)
3868 {
3869         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3870                + nla_total_size(sizeof(struct in6_addr))
3871                + nla_total_size(sizeof(struct prefix_cacheinfo));
3872 }
3873
3874 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3875                              struct prefix_info *pinfo, u32 pid, u32 seq,
3876                              int event, unsigned int flags)
3877 {
3878         struct prefixmsg *pmsg;
3879         struct nlmsghdr *nlh;
3880         struct prefix_cacheinfo ci;
3881
3882         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3883         if (nlh == NULL)
3884                 return -EMSGSIZE;
3885
3886         pmsg = nlmsg_data(nlh);
3887         pmsg->prefix_family = AF_INET6;
3888         pmsg->prefix_pad1 = 0;
3889         pmsg->prefix_pad2 = 0;
3890         pmsg->prefix_ifindex = idev->dev->ifindex;
3891         pmsg->prefix_len = pinfo->prefix_len;
3892         pmsg->prefix_type = pinfo->type;
3893         pmsg->prefix_pad3 = 0;
3894         pmsg->prefix_flags = 0;
3895         if (pinfo->onlink)
3896                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3897         if (pinfo->autoconf)
3898                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3899
3900         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3901
3902         ci.preferred_time = ntohl(pinfo->prefered);
3903         ci.valid_time = ntohl(pinfo->valid);
3904         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3905
3906         return nlmsg_end(skb, nlh);
3907
3908 nla_put_failure:
3909         nlmsg_cancel(skb, nlh);
3910         return -EMSGSIZE;
3911 }
3912
3913 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3914                          struct prefix_info *pinfo)
3915 {
3916         struct sk_buff *skb;
3917         struct net *net = dev_net(idev->dev);
3918         int err = -ENOBUFS;
3919
3920         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3921         if (skb == NULL)
3922                 goto errout;
3923
3924         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3925         if (err < 0) {
3926                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3927                 WARN_ON(err == -EMSGSIZE);
3928                 kfree_skb(skb);
3929                 goto errout;
3930         }
3931         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3932         return;
3933 errout:
3934         if (err < 0)
3935                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3936 }
3937
3938 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3939 {
3940         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3941
3942         switch (event) {
3943         case RTM_NEWADDR:
3944                 /*
3945                  * If the address was optimistic
3946                  * we inserted the route at the start of
3947                  * our DAD process, so we don't need
3948                  * to do it again
3949                  */
3950                 if (!(ifp->rt->rt6i_node))
3951                         ip6_ins_rt(ifp->rt);
3952                 if (ifp->idev->cnf.forwarding)
3953                         addrconf_join_anycast(ifp);
3954                 break;
3955         case RTM_DELADDR:
3956                 if (ifp->idev->cnf.forwarding)
3957                         addrconf_leave_anycast(ifp);
3958                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3959                 dst_hold(&ifp->rt->u.dst);
3960                 if (ip6_del_rt(ifp->rt))
3961                         dst_free(&ifp->rt->u.dst);
3962                 break;
3963         }
3964 }
3965
3966 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3967 {
3968         rcu_read_lock_bh();
3969         if (likely(ifp->idev->dead == 0))
3970                 __ipv6_ifa_notify(event, ifp);
3971         rcu_read_unlock_bh();
3972 }
3973
3974 #ifdef CONFIG_SYSCTL
3975
3976 static
3977 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3978                            void __user *buffer, size_t *lenp, loff_t *ppos)
3979 {
3980         int *valp = ctl->data;
3981         int val = *valp;
3982         int ret;
3983
3984         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3985
3986         if (write)
3987                 ret = addrconf_fixup_forwarding(ctl, valp, val);
3988         return ret;
3989 }
3990
3991 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3992                                             void __user *oldval,
3993                                             size_t __user *oldlenp,
3994                                             void __user *newval, size_t newlen)
3995 {
3996         int *valp = table->data;
3997         int val = *valp;
3998         int new;
3999
4000         if (!newval || !newlen)
4001                 return 0;
4002         if (newlen != sizeof(int))
4003                 return -EINVAL;
4004         if (get_user(new, (int __user *)newval))
4005                 return -EFAULT;
4006         if (new == *valp)
4007                 return 0;
4008         if (oldval && oldlenp) {
4009                 size_t len;
4010                 if (get_user(len, oldlenp))
4011                         return -EFAULT;
4012                 if (len) {
4013                         if (len > table->maxlen)
4014                                 len = table->maxlen;
4015                         if (copy_to_user(oldval, valp, len))
4016                                 return -EFAULT;
4017                         if (put_user(len, oldlenp))
4018                                 return -EFAULT;
4019                 }
4020         }
4021
4022         *valp = new;
4023         return addrconf_fixup_forwarding(table, valp, val);
4024 }
4025
4026 static struct addrconf_sysctl_table
4027 {
4028         struct ctl_table_header *sysctl_header;
4029         ctl_table addrconf_vars[DEVCONF_MAX+1];
4030         char *dev_name;
4031 } addrconf_sysctl __read_mostly = {
4032         .sysctl_header = NULL,
4033         .addrconf_vars = {
4034                 {
4035                         .ctl_name       =       NET_IPV6_FORWARDING,
4036                         .procname       =       "forwarding",
4037                         .data           =       &ipv6_devconf.forwarding,
4038                         .maxlen         =       sizeof(int),
4039                         .mode           =       0644,
4040                         .proc_handler   =       addrconf_sysctl_forward,
4041                         .strategy       =       addrconf_sysctl_forward_strategy,
4042                 },
4043                 {
4044                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4045                         .procname       =       "hop_limit",
4046                         .data           =       &ipv6_devconf.hop_limit,
4047                         .maxlen         =       sizeof(int),
4048                         .mode           =       0644,
4049                         .proc_handler   =       proc_dointvec,
4050                 },
4051                 {
4052                         .ctl_name       =       NET_IPV6_MTU,
4053                         .procname       =       "mtu",
4054                         .data           =       &ipv6_devconf.mtu6,
4055                         .maxlen         =       sizeof(int),
4056                         .mode           =       0644,
4057                         .proc_handler   =       proc_dointvec,
4058                 },
4059                 {
4060                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4061                         .procname       =       "accept_ra",
4062                         .data           =       &ipv6_devconf.accept_ra,
4063                         .maxlen         =       sizeof(int),
4064                         .mode           =       0644,
4065                         .proc_handler   =       proc_dointvec,
4066                 },
4067                 {
4068                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4069                         .procname       =       "accept_redirects",
4070                         .data           =       &ipv6_devconf.accept_redirects,
4071                         .maxlen         =       sizeof(int),
4072                         .mode           =       0644,
4073                         .proc_handler   =       proc_dointvec,
4074                 },
4075                 {
4076                         .ctl_name       =       NET_IPV6_AUTOCONF,
4077                         .procname       =       "autoconf",
4078                         .data           =       &ipv6_devconf.autoconf,
4079                         .maxlen         =       sizeof(int),
4080                         .mode           =       0644,
4081                         .proc_handler   =       proc_dointvec,
4082                 },
4083                 {
4084                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4085                         .procname       =       "dad_transmits",
4086                         .data           =       &ipv6_devconf.dad_transmits,
4087                         .maxlen         =       sizeof(int),
4088                         .mode           =       0644,
4089                         .proc_handler   =       proc_dointvec,
4090                 },
4091                 {
4092                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4093                         .procname       =       "router_solicitations",
4094                         .data           =       &ipv6_devconf.rtr_solicits,
4095                         .maxlen         =       sizeof(int),
4096                         .mode           =       0644,
4097                         .proc_handler   =       proc_dointvec,
4098                 },
4099                 {
4100                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4101                         .procname       =       "router_solicitation_interval",
4102                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4103                         .maxlen         =       sizeof(int),
4104                         .mode           =       0644,
4105                         .proc_handler   =       proc_dointvec_jiffies,
4106                         .strategy       =       sysctl_jiffies,
4107                 },
4108                 {
4109                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4110                         .procname       =       "router_solicitation_delay",
4111                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4112                         .maxlen         =       sizeof(int),
4113                         .mode           =       0644,
4114                         .proc_handler   =       proc_dointvec_jiffies,
4115                         .strategy       =       sysctl_jiffies,
4116                 },
4117                 {
4118                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4119                         .procname       =       "force_mld_version",
4120                         .data           =       &ipv6_devconf.force_mld_version,
4121                         .maxlen         =       sizeof(int),
4122                         .mode           =       0644,
4123                         .proc_handler   =       proc_dointvec,
4124                 },
4125 #ifdef CONFIG_IPV6_PRIVACY
4126                 {
4127                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4128                         .procname       =       "use_tempaddr",
4129                         .data           =       &ipv6_devconf.use_tempaddr,
4130                         .maxlen         =       sizeof(int),
4131                         .mode           =       0644,
4132                         .proc_handler   =       proc_dointvec,
4133                 },
4134                 {
4135                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4136                         .procname       =       "temp_valid_lft",
4137                         .data           =       &ipv6_devconf.temp_valid_lft,
4138                         .maxlen         =       sizeof(int),
4139                         .mode           =       0644,
4140                         .proc_handler   =       proc_dointvec,
4141                 },
4142                 {
4143                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4144                         .procname       =       "temp_prefered_lft",
4145                         .data           =       &ipv6_devconf.temp_prefered_lft,
4146                         .maxlen         =       sizeof(int),
4147                         .mode           =       0644,
4148                         .proc_handler   =       proc_dointvec,
4149                 },
4150                 {
4151                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4152                         .procname       =       "regen_max_retry",
4153                         .data           =       &ipv6_devconf.regen_max_retry,
4154                         .maxlen         =       sizeof(int),
4155                         .mode           =       0644,
4156                         .proc_handler   =       proc_dointvec,
4157                 },
4158                 {
4159                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4160                         .procname       =       "max_desync_factor",
4161                         .data           =       &ipv6_devconf.max_desync_factor,
4162                         .maxlen         =       sizeof(int),
4163                         .mode           =       0644,
4164                         .proc_handler   =       proc_dointvec,
4165                 },
4166 #endif
4167                 {
4168                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4169                         .procname       =       "max_addresses",
4170                         .data           =       &ipv6_devconf.max_addresses,
4171                         .maxlen         =       sizeof(int),
4172                         .mode           =       0644,
4173                         .proc_handler   =       proc_dointvec,
4174                 },
4175                 {
4176                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4177                         .procname       =       "accept_ra_defrtr",
4178                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4179                         .maxlen         =       sizeof(int),
4180                         .mode           =       0644,
4181                         .proc_handler   =       proc_dointvec,
4182                 },
4183                 {
4184                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4185                         .procname       =       "accept_ra_pinfo",
4186                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4187                         .maxlen         =       sizeof(int),
4188                         .mode           =       0644,
4189                         .proc_handler   =       proc_dointvec,
4190                 },
4191 #ifdef CONFIG_IPV6_ROUTER_PREF
4192                 {
4193                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4194                         .procname       =       "accept_ra_rtr_pref",
4195                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4196                         .maxlen         =       sizeof(int),
4197                         .mode           =       0644,
4198                         .proc_handler   =       proc_dointvec,
4199                 },
4200                 {
4201                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4202                         .procname       =       "router_probe_interval",
4203                         .data           =       &ipv6_devconf.rtr_probe_interval,
4204                         .maxlen         =       sizeof(int),
4205                         .mode           =       0644,
4206                         .proc_handler   =       proc_dointvec_jiffies,
4207                         .strategy       =       sysctl_jiffies,
4208                 },
4209 #ifdef CONFIG_IPV6_ROUTE_INFO
4210                 {
4211                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4212                         .procname       =       "accept_ra_rt_info_max_plen",
4213                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4214                         .maxlen         =       sizeof(int),
4215                         .mode           =       0644,
4216                         .proc_handler   =       proc_dointvec,
4217                 },
4218 #endif
4219 #endif
4220                 {
4221                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4222                         .procname       =       "proxy_ndp",
4223                         .data           =       &ipv6_devconf.proxy_ndp,
4224                         .maxlen         =       sizeof(int),
4225                         .mode           =       0644,
4226                         .proc_handler   =       proc_dointvec,
4227                 },
4228                 {
4229                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4230                         .procname       =       "accept_source_route",
4231                         .data           =       &ipv6_devconf.accept_source_route,
4232                         .maxlen         =       sizeof(int),
4233                         .mode           =       0644,
4234                         .proc_handler   =       proc_dointvec,
4235                 },
4236 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4237                 {
4238                         .ctl_name       =       CTL_UNNUMBERED,
4239                         .procname       =       "optimistic_dad",
4240                         .data           =       &ipv6_devconf.optimistic_dad,
4241                         .maxlen         =       sizeof(int),
4242                         .mode           =       0644,
4243                         .proc_handler   =       proc_dointvec,
4244
4245                 },
4246 #endif
4247 #ifdef CONFIG_IPV6_MROUTE
4248                 {
4249                         .ctl_name       =       CTL_UNNUMBERED,
4250                         .procname       =       "mc_forwarding",
4251                         .data           =       &ipv6_devconf.mc_forwarding,
4252                         .maxlen         =       sizeof(int),
4253                         .mode           =       0444,
4254                         .proc_handler   =       proc_dointvec,
4255                 },
4256 #endif
4257                 {
4258                         .ctl_name       =       CTL_UNNUMBERED,
4259                         .procname       =       "disable_ipv6",
4260                         .data           =       &ipv6_devconf.disable_ipv6,
4261                         .maxlen         =       sizeof(int),
4262                         .mode           =       0644,
4263                         .proc_handler   =       proc_dointvec,
4264                 },
4265                 {
4266                         .ctl_name       =       CTL_UNNUMBERED,
4267                         .procname       =       "accept_dad",
4268                         .data           =       &ipv6_devconf.accept_dad,
4269                         .maxlen         =       sizeof(int),
4270                         .mode           =       0644,
4271                         .proc_handler   =       proc_dointvec,
4272                 },
4273                 {
4274                         .ctl_name       =       0,      /* sentinel */
4275                 }
4276         },
4277 };
4278
4279 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4280                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4281 {
4282         int i;
4283         struct addrconf_sysctl_table *t;
4284
4285 #define ADDRCONF_CTL_PATH_DEV   3
4286
4287         struct ctl_path addrconf_ctl_path[] = {
4288                 { .procname = "net", .ctl_name = CTL_NET, },
4289                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4290                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4291                 { /* to be set */ },
4292                 { },
4293         };
4294
4295
4296         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4297         if (t == NULL)
4298                 goto out;
4299
4300         for (i=0; t->addrconf_vars[i].data; i++) {
4301                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4302                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4303                 t->addrconf_vars[i].extra2 = net;
4304         }
4305
4306         /*
4307          * Make a copy of dev_name, because '.procname' is regarded as const
4308          * by sysctl and we wouldn't want anyone to change it under our feet
4309          * (see SIOCSIFNAME).
4310          */
4311         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4312         if (!t->dev_name)
4313                 goto free;
4314
4315         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4316         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4317
4318         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4319                         t->addrconf_vars);
4320         if (t->sysctl_header == NULL)
4321                 goto free_procname;
4322
4323         p->sysctl = t;
4324         return 0;
4325
4326 free_procname:
4327         kfree(t->dev_name);
4328 free:
4329         kfree(t);
4330 out:
4331         return -ENOBUFS;
4332 }
4333
4334 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4335 {
4336         struct addrconf_sysctl_table *t;
4337
4338         if (p->sysctl == NULL)
4339                 return;
4340
4341         t = p->sysctl;
4342         p->sysctl = NULL;
4343         unregister_sysctl_table(t->sysctl_header);
4344         kfree(t->dev_name);
4345         kfree(t);
4346 }
4347
4348 static void addrconf_sysctl_register(struct inet6_dev *idev)
4349 {
4350         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4351                               NET_IPV6_NEIGH, "ipv6",
4352                               &ndisc_ifinfo_sysctl_change,
4353                               ndisc_ifinfo_sysctl_strategy);
4354         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4355                         idev->dev->ifindex, idev, &idev->cnf);
4356 }
4357
4358 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4359 {
4360         __addrconf_sysctl_unregister(&idev->cnf);
4361         neigh_sysctl_unregister(idev->nd_parms);
4362 }
4363
4364
4365 #endif
4366
4367 static int addrconf_init_net(struct net *net)
4368 {
4369         int err;
4370         struct ipv6_devconf *all, *dflt;
4371
4372         err = -ENOMEM;
4373         all = &ipv6_devconf;
4374         dflt = &ipv6_devconf_dflt;
4375
4376         if (net != &init_net) {
4377                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4378                 if (all == NULL)
4379                         goto err_alloc_all;
4380
4381                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4382                 if (dflt == NULL)
4383                         goto err_alloc_dflt;
4384         }
4385
4386         net->ipv6.devconf_all = all;
4387         net->ipv6.devconf_dflt = dflt;
4388
4389 #ifdef CONFIG_SYSCTL
4390         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4391                         NULL, all);
4392         if (err < 0)
4393                 goto err_reg_all;
4394
4395         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4396                         NULL, dflt);
4397         if (err < 0)
4398                 goto err_reg_dflt;
4399 #endif
4400         return 0;
4401
4402 #ifdef CONFIG_SYSCTL
4403 err_reg_dflt:
4404         __addrconf_sysctl_unregister(all);
4405 err_reg_all:
4406         kfree(dflt);
4407 #endif
4408 err_alloc_dflt:
4409         kfree(all);
4410 err_alloc_all:
4411         return err;
4412 }
4413
4414 static void addrconf_exit_net(struct net *net)
4415 {
4416 #ifdef CONFIG_SYSCTL
4417         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4418         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4419 #endif
4420         if (net != &init_net) {
4421                 kfree(net->ipv6.devconf_dflt);
4422                 kfree(net->ipv6.devconf_all);
4423         }
4424 }
4425
4426 static struct pernet_operations addrconf_ops = {
4427         .init = addrconf_init_net,
4428         .exit = addrconf_exit_net,
4429 };
4430
4431 /*
4432  *      Device notifier
4433  */
4434
4435 int register_inet6addr_notifier(struct notifier_block *nb)
4436 {
4437         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4438 }
4439
4440 EXPORT_SYMBOL(register_inet6addr_notifier);
4441
4442 int unregister_inet6addr_notifier(struct notifier_block *nb)
4443 {
4444         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4445 }
4446
4447 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4448
4449 /*
4450  *      Init / cleanup code
4451  */
4452
4453 int __init addrconf_init(void)
4454 {
4455         int err;
4456
4457         if ((err = ipv6_addr_label_init()) < 0) {
4458                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4459                         err);
4460                 return err;
4461         }
4462
4463         register_pernet_subsys(&addrconf_ops);
4464
4465         /* The addrconf netdev notifier requires that loopback_dev
4466          * has it's ipv6 private information allocated and setup
4467          * before it can bring up and give link-local addresses
4468          * to other devices which are up.
4469          *
4470          * Unfortunately, loopback_dev is not necessarily the first
4471          * entry in the global dev_base list of net devices.  In fact,
4472          * it is likely to be the very last entry on that list.
4473          * So this causes the notifier registry below to try and
4474          * give link-local addresses to all devices besides loopback_dev
4475          * first, then loopback_dev, which cases all the non-loopback_dev
4476          * devices to fail to get a link-local address.
4477          *
4478          * So, as a temporary fix, allocate the ipv6 structure for
4479          * loopback_dev first by hand.
4480          * Longer term, all of the dependencies ipv6 has upon the loopback
4481          * device and it being up should be removed.
4482          */
4483         rtnl_lock();
4484         if (!ipv6_add_dev(init_net.loopback_dev))
4485                 err = -ENOMEM;
4486         rtnl_unlock();
4487         if (err)
4488                 goto errlo;
4489
4490         register_netdevice_notifier(&ipv6_dev_notf);
4491
4492         addrconf_verify(0);
4493
4494         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4495         if (err < 0)
4496                 goto errout;
4497
4498         /* Only the first call to __rtnl_register can fail */
4499         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4500         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4501         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4502         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4503         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4504
4505         ipv6_addr_label_rtnl_register();
4506
4507         return 0;
4508 errout:
4509         unregister_netdevice_notifier(&ipv6_dev_notf);
4510 errlo:
4511         unregister_pernet_subsys(&addrconf_ops);
4512
4513         return err;
4514 }
4515
4516 void addrconf_cleanup(void)
4517 {
4518         struct inet6_ifaddr *ifa;
4519         struct net_device *dev;
4520         int i;
4521
4522         unregister_netdevice_notifier(&ipv6_dev_notf);
4523         unregister_pernet_subsys(&addrconf_ops);
4524
4525         rtnl_lock();
4526
4527         /* clean dev list */
4528         for_each_netdev(&init_net, dev) {
4529                 if (__in6_dev_get(dev) == NULL)
4530                         continue;
4531                 addrconf_ifdown(dev, 1);
4532         }
4533         addrconf_ifdown(init_net.loopback_dev, 2);
4534
4535         /*
4536          *      Check hash table.
4537          */
4538         write_lock_bh(&addrconf_hash_lock);
4539         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4540                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4541                         struct inet6_ifaddr *bifa;
4542
4543                         bifa = ifa;
4544                         ifa = ifa->lst_next;
4545                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4546                         /* Do not free it; something is wrong.
4547                            Now we can investigate it with debugger.
4548                          */
4549                 }
4550         }
4551         write_unlock_bh(&addrconf_hash_lock);
4552
4553         del_timer(&addr_chk_timer);
4554         rtnl_unlock();
4555 }