net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90 #include <linux/export.h>
  91
  92 /* Set to 3 to get tracing... */
  93 #define ACONF_DEBUG 2
  94
  95 #if ACONF_DEBUG >= 3
  96 #define ADBG(x) printk x
  97 #else
  98 #define ADBG(x)
  99 #endif
 100
 101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 102
 103 static inline u32 cstamp_delta(unsigned long cstamp)
 104 {
 105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 106 }
 107
 108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 111
 112 #ifdef CONFIG_SYSCTL
 113 static void addrconf_sysctl_register(struct inet6_dev *idev);
 114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 115 #else
 116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 117 {
 118 }
 119
 120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 121 {
 122 }
 123 #endif
 124
 125 #ifdef CONFIG_IPV6_PRIVACY
 126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 128 static void ipv6_regen_rndid(unsigned long data);
 129 #endif
 130
 131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 132 static int ipv6_count_addresses(struct inet6_dev *idev);
 133
 134 /*
 135  *      Configured unicast address hash table
 136  */
 137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 138 static DEFINE_SPINLOCK(addrconf_hash_lock);
 139
 140 static void addrconf_verify(unsigned long);
 141
 142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 143 static DEFINE_SPINLOCK(addrconf_verify_lock);
 144
 145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 147
 148 static void addrconf_type_change(struct net_device *dev,
 149                                  unsigned long event);
 150 static int addrconf_ifdown(struct net_device *dev, int how);
 151
 152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 153 static void addrconf_dad_timer(unsigned long data);
 154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 155 static void addrconf_dad_run(struct inet6_dev *idev);
 156 static void addrconf_rs_timer(unsigned long data);
 157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159
 160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 161                                 struct prefix_info *pinfo);
 162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 163                                struct net_device *dev);
 164
 165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 166
 167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 168         .forwarding             = 0,
 169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 170         .mtu6                   = IPV6_MIN_MTU,
 171         .accept_ra              = 1,
 172         .accept_redirects       = 1,
 173         .autoconf               = 1,
 174         .force_mld_version      = 0,
 175         .dad_transmits          = 1,
 176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 179 #ifdef CONFIG_IPV6_PRIVACY
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185 #endif
 186         .max_addresses          = IPV6_MAX_ADDRESSES,
 187         .accept_ra_defrtr       = 1,
 188         .accept_ra_min_hop_limit= 1,
 189         .accept_ra_pinfo        = 1,
 190 #ifdef CONFIG_IPV6_ROUTER_PREF
 191         .accept_ra_rtr_pref     = 1,
 192         .rtr_probe_interval     = 60 * HZ,
 193 #ifdef CONFIG_IPV6_ROUTE_INFO
 194         .accept_ra_rt_info_max_plen = 0,
 195 #endif
 196 #endif
 197         .proxy_ndp              = 0,
 198         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 199         .disable_ipv6           = 0,
 200         .accept_dad             = 1,
 201 };
 202
 203 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 204         .forwarding             = 0,
 205         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 206         .mtu6                   = IPV6_MIN_MTU,
 207         .accept_ra              = 1,
 208         .accept_redirects       = 1,
 209         .autoconf               = 1,
 210         .dad_transmits          = 1,
 211         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 212         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 213         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 214 #ifdef CONFIG_IPV6_PRIVACY
 215         .use_tempaddr           = 0,
 216         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 217         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 218         .regen_max_retry        = REGEN_MAX_RETRY,
 219         .max_desync_factor      = MAX_DESYNC_FACTOR,
 220 #endif
 221         .max_addresses          = IPV6_MAX_ADDRESSES,
 222         .accept_ra_defrtr       = 1,
 223         .accept_ra_min_hop_limit= 1,
 224         .accept_ra_pinfo        = 1,
 225 #ifdef CONFIG_IPV6_ROUTER_PREF
 226         .accept_ra_rtr_pref     = 1,
 227         .rtr_probe_interval     = 60 * HZ,
 228 #ifdef CONFIG_IPV6_ROUTE_INFO
 229         .accept_ra_rt_info_max_plen = 0,
 230 #endif
 231 #endif
 232         .proxy_ndp              = 0,
 233         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 234         .disable_ipv6           = 0,
 235         .accept_dad             = 1,
 236 };
 237
 238 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 239 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 240 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 241 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 242 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 243
 244 /* Check if a valid qdisc is available */
 245 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 246 {
 247         return !qdisc_tx_is_noop(dev);
 248 }
 249
 250 /* Check if a route is valid prefix route */
 251 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 252 {
 253         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 254 }
 255
 256 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 257 {
 258         if (del_timer(&ifp->timer))
 259                 __in6_ifa_put(ifp);
 260 }
 261
 262 enum addrconf_timer_t {
 263         AC_NONE,
 264         AC_DAD,
 265         AC_RS,
 266 };
 267
 268 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 269                                enum addrconf_timer_t what,
 270                                unsigned long when)
 271 {
 272         if (!del_timer(&ifp->timer))
 273                 in6_ifa_hold(ifp);
 274
 275         switch (what) {
 276         case AC_DAD:
 277                 ifp->timer.function = addrconf_dad_timer;
 278                 break;
 279         case AC_RS:
 280                 ifp->timer.function = addrconf_rs_timer;
 281                 break;
 282         default:
 283                 break;
 284         }
 285         ifp->timer.expires = jiffies + when;
 286         add_timer(&ifp->timer);
 287 }
 288
 289 static int snmp6_alloc_dev(struct inet6_dev *idev)
 290 {
 291         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 292                           sizeof(struct ipstats_mib),
 293                           __alignof__(struct ipstats_mib)) < 0)
 294                 goto err_ip;
 295         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 296                                         GFP_KERNEL);
 297         if (!idev->stats.icmpv6dev)
 298                 goto err_icmp;
 299         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 300                                            GFP_KERNEL);
 301         if (!idev->stats.icmpv6msgdev)
 302                 goto err_icmpmsg;
 303
 304         return 0;
 305
 306 err_icmpmsg:
 307         kfree(idev->stats.icmpv6dev);
 308 err_icmp:
 309         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 310 err_ip:
 311         return -ENOMEM;
 312 }
 313
 314 static void snmp6_free_dev(struct inet6_dev *idev)
 315 {
 316         kfree(idev->stats.icmpv6msgdev);
 317         kfree(idev->stats.icmpv6dev);
 318         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 319 }
 320
 321 /* Nobody refers to this device, we may destroy it. */
 322
 323 void in6_dev_finish_destroy(struct inet6_dev *idev)
 324 {
 325         struct net_device *dev = idev->dev;
 326
 327         WARN_ON(!list_empty(&idev->addr_list));
 328         WARN_ON(idev->mc_list != NULL);
 329
 330 #ifdef NET_REFCNT_DEBUG
 331         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 332 #endif
 333         dev_put(dev);
 334         if (!idev->dead) {
 335                 pr_warning("Freeing alive inet6 device %p\n", idev);
 336                 return;
 337         }
 338         snmp6_free_dev(idev);
 339         kfree_rcu(idev, rcu);
 340 }
 341
 342 EXPORT_SYMBOL(in6_dev_finish_destroy);
 343
 344 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 345 {
 346         struct inet6_dev *ndev;
 347
 348         ASSERT_RTNL();
 349
 350         if (dev->mtu < IPV6_MIN_MTU)
 351                 return NULL;
 352
 353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 354
 355         if (ndev == NULL)
 356                 return NULL;
 357
 358         rwlock_init(&ndev->lock);
 359         ndev->dev = dev;
 360         INIT_LIST_HEAD(&ndev->addr_list);
 361
 362         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 363         ndev->cnf.mtu6 = dev->mtu;
 364         ndev->cnf.sysctl = NULL;
 365         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 366         if (ndev->nd_parms == NULL) {
 367                 kfree(ndev);
 368                 return NULL;
 369         }
 370         if (ndev->cnf.forwarding)
 371                 dev_disable_lro(dev);
 372         /* We refer to the device */
 373         dev_hold(dev);
 374
 375         if (snmp6_alloc_dev(ndev) < 0) {
 376                 ADBG((KERN_WARNING
 377                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 378                         __func__, dev->name));
 379                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 380                 dev_put(dev);
 381                 kfree(ndev);
 382                 return NULL;
 383         }
 384
 385         if (snmp6_register_dev(ndev) < 0) {
 386                 ADBG((KERN_WARNING
 387                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 388                         __func__, dev->name));
 389                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 390                 ndev->dead = 1;
 391                 in6_dev_finish_destroy(ndev);
 392                 return NULL;
 393         }
 394
 395         /* One reference from device.  We must do this before
 396          * we invoke __ipv6_regen_rndid().
 397          */
 398         in6_dev_hold(ndev);
 399
 400         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 401                 ndev->cnf.accept_dad = -1;
 402
 403 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 404         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 405                 printk(KERN_INFO
 406                        "%s: Disabled Multicast RS\n",
 407                        dev->name);
 408                 ndev->cnf.rtr_solicits = 0;
 409         }
 410 #endif
 411
 412 #ifdef CONFIG_IPV6_PRIVACY
 413         INIT_LIST_HEAD(&ndev->tempaddr_list);
 414         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 415         if ((dev->flags&IFF_LOOPBACK) ||
 416             dev->type == ARPHRD_TUNNEL ||
 417             dev->type == ARPHRD_TUNNEL6 ||
 418             dev->type == ARPHRD_SIT ||
 419             dev->type == ARPHRD_NONE) {
 420                 ndev->cnf.use_tempaddr = -1;
 421         } else {
 422                 in6_dev_hold(ndev);
 423                 ipv6_regen_rndid((unsigned long) ndev);
 424         }
 425 #endif
 426
 427         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 428                 ndev->if_flags |= IF_READY;
 429
 430         ipv6_mc_init_dev(ndev);
 431         ndev->tstamp = jiffies;
 432         addrconf_sysctl_register(ndev);
 433         /* protected by rtnl_lock */
 434         rcu_assign_pointer(dev->ip6_ptr, ndev);
 435
 436         /* Join all-node multicast group */
 437         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 438
 439         /* Join all-router multicast group if forwarding is set */
 440         if (ndev->cnf.forwarding && dev && (dev->flags & IFF_MULTICAST))
 441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 442
 443         return ndev;
 444 }
 445
 446 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 447 {
 448         struct inet6_dev *idev;
 449
 450         ASSERT_RTNL();
 451
 452         idev = __in6_dev_get(dev);
 453         if (!idev) {
 454                 idev = ipv6_add_dev(dev);
 455                 if (!idev)
 456                         return NULL;
 457         }
 458
 459         if (dev->flags&IFF_UP)
 460                 ipv6_mc_up(idev);
 461         return idev;
 462 }
 463
 464 #ifdef CONFIG_SYSCTL
 465 static void dev_forward_change(struct inet6_dev *idev)
 466 {
 467         struct net_device *dev;
 468         struct inet6_ifaddr *ifa;
 469
 470         if (!idev)
 471                 return;
 472         dev = idev->dev;
 473         if (idev->cnf.forwarding)
 474                 dev_disable_lro(dev);
 475         if (dev && (dev->flags & IFF_MULTICAST)) {
 476                 if (idev->cnf.forwarding)
 477                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 478                 else
 479                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 480         }
 481
 482         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 483                 if (ifa->flags&IFA_F_TENTATIVE)
 484                         continue;
 485                 if (idev->cnf.forwarding)
 486                         addrconf_join_anycast(ifa);
 487                 else
 488                         addrconf_leave_anycast(ifa);
 489         }
 490 }
 491
 492
 493 static void addrconf_forward_change(struct net *net, __s32 newf)
 494 {
 495         struct net_device *dev;
 496         struct inet6_dev *idev;
 497
 498         for_each_netdev(net, dev) {
 499                 idev = __in6_dev_get(dev);
 500                 if (idev) {
 501                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 502                         idev->cnf.forwarding = newf;
 503                         if (changed)
 504                                 dev_forward_change(idev);
 505                 }
 506         }
 507 }
 508
 509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 510 {
 511         struct net *net;
 512
 513         net = (struct net *)table->extra2;
 514         if (p == &net->ipv6.devconf_dflt->forwarding)
 515                 return 0;
 516
 517         if (!rtnl_trylock()) {
 518                 /* Restore the original values before restarting */
 519                 *p = old;
 520                 return restart_syscall();
 521         }
 522
 523         if (p == &net->ipv6.devconf_all->forwarding) {
 524                 __s32 newf = net->ipv6.devconf_all->forwarding;
 525                 net->ipv6.devconf_dflt->forwarding = newf;
 526                 addrconf_forward_change(net, newf);
 527         } else if ((!*p) ^ (!old))
 528                 dev_forward_change((struct inet6_dev *)table->extra1);
 529         rtnl_unlock();
 530
 531         if (*p)
 532                 rt6_purge_dflt_routers(net);
 533         return 1;
 534 }
 535 #endif
 536
 537 /* Nobody refers to this ifaddr, destroy it */
 538 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 539 {
 540         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 541
 542 #ifdef NET_REFCNT_DEBUG
 543         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 544 #endif
 545
 546         in6_dev_put(ifp->idev);
 547
 548         if (del_timer(&ifp->timer))
 549                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 550
 551         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 552                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 553                 return;
 554         }
 555         dst_release(&ifp->rt->dst);
 556
 557         kfree_rcu(ifp, rcu);
 558 }
 559
 560 static void
 561 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 562 {
 563         struct list_head *p;
 564         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 565
 566         /*
 567          * Each device address list is sorted in order of scope -
 568          * global before linklocal.
 569          */
 570         list_for_each(p, &idev->addr_list) {
 571                 struct inet6_ifaddr *ifa
 572                         = list_entry(p, struct inet6_ifaddr, if_list);
 573                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 574                         break;
 575         }
 576
 577         list_add_tail(&ifp->if_list, p);
 578 }
 579
 580 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 581 {
 582         /*
 583          * We perform the hash function over the last 64 bits of the address
 584          * This will include the IEEE address token on links that support it.
 585          */
 586         return jhash_2words((__force u32)addr->s6_addr32[2],
 587                             (__force u32)addr->s6_addr32[3], 0)
 588                 & (IN6_ADDR_HSIZE - 1);
 589 }
 590
 591 /* On success it returns ifp with increased reference count */
 592
 593 static struct inet6_ifaddr *
 594 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 595               int scope, u32 flags)
 596 {
 597         struct inet6_ifaddr *ifa = NULL;
 598         struct rt6_info *rt;
 599         unsigned int hash;
 600         int err = 0;
 601         int addr_type = ipv6_addr_type(addr);
 602
 603         if (addr_type == IPV6_ADDR_ANY ||
 604             addr_type & IPV6_ADDR_MULTICAST ||
 605             (!(idev->dev->flags & IFF_LOOPBACK) &&
 606              addr_type & IPV6_ADDR_LOOPBACK))
 607                 return ERR_PTR(-EADDRNOTAVAIL);
 608
 609         rcu_read_lock_bh();
 610         if (idev->dead) {
 611                 err = -ENODEV;                  /*XXX*/
 612                 goto out2;
 613         }
 614
 615         if (idev->cnf.disable_ipv6) {
 616                 err = -EACCES;
 617                 goto out2;
 618         }
 619
 620         spin_lock(&addrconf_hash_lock);
 621
 622         /* Ignore adding duplicate addresses on an interface */
 623         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 624                 ADBG(("ipv6_add_addr: already assigned\n"));
 625                 err = -EEXIST;
 626                 goto out;
 627         }
 628
 629         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 630
 631         if (ifa == NULL) {
 632                 ADBG(("ipv6_add_addr: malloc failed\n"));
 633                 err = -ENOBUFS;
 634                 goto out;
 635         }
 636
 637         rt = addrconf_dst_alloc(idev, addr, 0);
 638         if (IS_ERR(rt)) {
 639                 err = PTR_ERR(rt);
 640                 goto out;
 641         }
 642
 643         ipv6_addr_copy(&ifa->addr, addr);
 644
 645         spin_lock_init(&ifa->lock);
 646         spin_lock_init(&ifa->state_lock);
 647         init_timer(&ifa->timer);
 648         INIT_HLIST_NODE(&ifa->addr_lst);
 649         ifa->timer.data = (unsigned long) ifa;
 650         ifa->scope = scope;
 651         ifa->prefix_len = pfxlen;
 652         ifa->flags = flags | IFA_F_TENTATIVE;
 653         ifa->cstamp = ifa->tstamp = jiffies;
 654
 655         ifa->rt = rt;
 656
 657         /*
 658          * part one of RFC 4429, section 3.3
 659          * We should not configure an address as
 660          * optimistic if we do not yet know the link
 661          * layer address of our nexhop router
 662          */
 663
 664         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 665                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 666
 667         ifa->idev = idev;
 668         in6_dev_hold(idev);
 669         /* For caller */
 670         in6_ifa_hold(ifa);
 671
 672         /* Add to big hash table */
 673         hash = ipv6_addr_hash(addr);
 674
 675         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 676         spin_unlock(&addrconf_hash_lock);
 677
 678         write_lock(&idev->lock);
 679         /* Add to inet6_dev unicast addr list. */
 680         ipv6_link_dev_addr(idev, ifa);
 681
 682 #ifdef CONFIG_IPV6_PRIVACY
 683         if (ifa->flags&IFA_F_TEMPORARY) {
 684                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 685                 in6_ifa_hold(ifa);
 686         }
 687 #endif
 688
 689         in6_ifa_hold(ifa);
 690         write_unlock(&idev->lock);
 691 out2:
 692         rcu_read_unlock_bh();
 693
 694         if (likely(err == 0))
 695                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 696         else {
 697                 kfree(ifa);
 698                 ifa = ERR_PTR(err);
 699         }
 700
 701         return ifa;
 702 out:
 703         spin_unlock(&addrconf_hash_lock);
 704         goto out2;
 705 }
 706
 707 /* This function wants to get referenced ifp and releases it before return */
 708
 709 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 710 {
 711         struct inet6_ifaddr *ifa, *ifn;
 712         struct inet6_dev *idev = ifp->idev;
 713         int state;
 714         int deleted = 0, onlink = 0;
 715         unsigned long expires = jiffies;
 716
 717         spin_lock_bh(&ifp->state_lock);
 718         state = ifp->state;
 719         ifp->state = INET6_IFADDR_STATE_DEAD;
 720         spin_unlock_bh(&ifp->state_lock);
 721
 722         if (state == INET6_IFADDR_STATE_DEAD)
 723                 goto out;
 724
 725         spin_lock_bh(&addrconf_hash_lock);
 726         hlist_del_init_rcu(&ifp->addr_lst);
 727         spin_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                 list_del(&ifp->tmp_list);
 733                 if (ifp->ifpub) {
 734                         in6_ifa_put(ifp->ifpub);
 735                         ifp->ifpub = NULL;
 736                 }
 737                 __in6_ifa_put(ifp);
 738         }
 739 #endif
 740
 741         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 742                 if (ifa == ifp) {
 743                         list_del_init(&ifp->if_list);
 744                         __in6_ifa_put(ifp);
 745
 746                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 747                                 break;
 748                         deleted = 1;
 749                         continue;
 750                 } else if (ifp->flags & IFA_F_PERMANENT) {
 751                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 752                                               ifp->prefix_len)) {
 753                                 if (ifa->flags & IFA_F_PERMANENT) {
 754                                         onlink = 1;
 755                                         if (deleted)
 756                                                 break;
 757                                 } else {
 758                                         unsigned long lifetime;
 759
 760                                         if (!onlink)
 761                                                 onlink = -1;
 762
 763                                         spin_lock(&ifa->lock);
 764
 765                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 766                                         /*
 767                                          * Note: Because this address is
 768                                          * not permanent, lifetime <
 769                                          * LONG_MAX / HZ here.
 770                                          */
 771                                         if (time_before(expires,
 772                                                         ifa->tstamp + lifetime * HZ))
 773                                                 expires = ifa->tstamp + lifetime * HZ;
 774                                         spin_unlock(&ifa->lock);
 775                                 }
 776                         }
 777                 }
 778         }
 779         write_unlock_bh(&idev->lock);
 780
 781         addrconf_del_timer(ifp);
 782
 783         ipv6_ifa_notify(RTM_DELADDR, ifp);
 784
 785         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 786
 787         /*
 788          * Purge or update corresponding prefix
 789          *
 790          * 1) we don't purge prefix here if address was not permanent.
 791          *    prefix is managed by its own lifetime.
 792          * 2) if there're no addresses, delete prefix.
 793          * 3) if there're still other permanent address(es),
 794          *    corresponding prefix is still permanent.
 795          * 4) otherwise, update prefix lifetime to the
 796          *    longest valid lifetime among the corresponding
 797          *    addresses on the device.
 798          *    Note: subsequent RA will update lifetime.
 799          *
 800          * --yoshfuji
 801          */
 802         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 803                 struct in6_addr prefix;
 804                 struct rt6_info *rt;
 805                 struct net *net = dev_net(ifp->idev->dev);
 806                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 807                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 808
 809                 if (rt && addrconf_is_prefix_route(rt)) {
 810                         if (onlink == 0) {
 811                                 ip6_del_rt(rt);
 812                                 rt = NULL;
 813                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 814                                 rt->rt6i_expires = expires;
 815                                 rt->rt6i_flags |= RTF_EXPIRES;
 816                         }
 817                 }
 818                 dst_release(&rt->dst);
 819         }
 820
 821         /* clean up prefsrc entries */
 822         rt6_remove_prefsrc(ifp);
 823 out:
 824         in6_ifa_put(ifp);
 825 }
 826
 827 #ifdef CONFIG_IPV6_PRIVACY
 828 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 829 {
 830         struct inet6_dev *idev = ifp->idev;
 831         struct in6_addr addr, *tmpaddr;
 832         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 833         unsigned long regen_advance;
 834         int tmp_plen;
 835         int ret = 0;
 836         int max_addresses;
 837         u32 addr_flags;
 838         unsigned long now = jiffies;
 839
 840         write_lock(&idev->lock);
 841         if (ift) {
 842                 spin_lock_bh(&ift->lock);
 843                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 844                 spin_unlock_bh(&ift->lock);
 845                 tmpaddr = &addr;
 846         } else {
 847                 tmpaddr = NULL;
 848         }
 849 retry:
 850         in6_dev_hold(idev);
 851         if (idev->cnf.use_tempaddr <= 0) {
 852                 write_unlock(&idev->lock);
 853                 printk(KERN_INFO
 854                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 855                 in6_dev_put(idev);
 856                 ret = -1;
 857                 goto out;
 858         }
 859         spin_lock_bh(&ifp->lock);
 860         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 861                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 862                 spin_unlock_bh(&ifp->lock);
 863                 write_unlock(&idev->lock);
 864                 printk(KERN_WARNING
 865                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 866                 in6_dev_put(idev);
 867                 ret = -1;
 868                 goto out;
 869         }
 870         in6_ifa_hold(ifp);
 871         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 872         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 873                 spin_unlock_bh(&ifp->lock);
 874                 write_unlock(&idev->lock);
 875                 printk(KERN_WARNING
 876                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 877                 in6_ifa_put(ifp);
 878                 in6_dev_put(idev);
 879                 ret = -1;
 880                 goto out;
 881         }
 882         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 883         age = (now - ifp->tstamp) / HZ;
 884         tmp_valid_lft = min_t(__u32,
 885                               ifp->valid_lft,
 886                               idev->cnf.temp_valid_lft + age);
 887         tmp_prefered_lft = min_t(__u32,
 888                                  ifp->prefered_lft,
 889                                  idev->cnf.temp_prefered_lft + age -
 890                                  idev->cnf.max_desync_factor);
 891         tmp_plen = ifp->prefix_len;
 892         max_addresses = idev->cnf.max_addresses;
 893         tmp_tstamp = ifp->tstamp;
 894         spin_unlock_bh(&ifp->lock);
 895
 896         regen_advance = idev->cnf.regen_max_retry *
 897                         idev->cnf.dad_transmits *
 898                         idev->nd_parms->retrans_time / HZ;
 899         write_unlock(&idev->lock);
 900
 901         /* A temporary address is created only if this calculated Preferred
 902          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 903          * an implementation must not create a temporary address with a zero
 904          * Preferred Lifetime.
 905          * Use age calculation as in addrconf_verify to avoid unnecessary
 906          * temporary addresses being generated.
 907          */
 908         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 909         if (tmp_prefered_lft <= regen_advance + age) {
 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 = ipv6_add_addr(idev, &addr, tmp_plen,
 922                             ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 923                             addr_flags);
 924         if (IS_ERR(ift)) {
 925                 in6_ifa_put(ifp);
 926                 in6_dev_put(idev);
 927                 printk(KERN_INFO
 928                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 929                 tmpaddr = &addr;
 930                 write_lock(&idev->lock);
 931                 goto retry;
 932         }
 933
 934         spin_lock_bh(&ift->lock);
 935         ift->ifpub = ifp;
 936         ift->valid_lft = tmp_valid_lft;
 937         ift->prefered_lft = tmp_prefered_lft;
 938         ift->cstamp = now;
 939         ift->tstamp = tmp_tstamp;
 940         spin_unlock_bh(&ift->lock);
 941
 942         addrconf_dad_start(ift, 0);
 943         in6_ifa_put(ift);
 944         in6_dev_put(idev);
 945 out:
 946         return ret;
 947 }
 948 #endif
 949
 950 /*
 951  *      Choose an appropriate source address (RFC3484)
 952  */
 953 enum {
 954         IPV6_SADDR_RULE_INIT = 0,
 955         IPV6_SADDR_RULE_LOCAL,
 956         IPV6_SADDR_RULE_SCOPE,
 957         IPV6_SADDR_RULE_PREFERRED,
 958 #ifdef CONFIG_IPV6_MIP6
 959         IPV6_SADDR_RULE_HOA,
 960 #endif
 961         IPV6_SADDR_RULE_OIF,
 962         IPV6_SADDR_RULE_LABEL,
 963 #ifdef CONFIG_IPV6_PRIVACY
 964         IPV6_SADDR_RULE_PRIVACY,
 965 #endif
 966         IPV6_SADDR_RULE_ORCHID,
 967         IPV6_SADDR_RULE_PREFIX,
 968         IPV6_SADDR_RULE_MAX
 969 };
 970
 971 struct ipv6_saddr_score {
 972         int                     rule;
 973         int                     addr_type;
 974         struct inet6_ifaddr     *ifa;
 975         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 976         int                     scopedist;
 977         int                     matchlen;
 978 };
 979
 980 struct ipv6_saddr_dst {
 981         const struct in6_addr *addr;
 982         int ifindex;
 983         int scope;
 984         int label;
 985         unsigned int prefs;
 986 };
 987
 988 static inline int ipv6_saddr_preferred(int type)
 989 {
 990         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 991                 return 1;
 992         return 0;
 993 }
 994
 995 static int ipv6_get_saddr_eval(struct net *net,
 996                                struct ipv6_saddr_score *score,
 997                                struct ipv6_saddr_dst *dst,
 998                                int i)
 999 {
1000         int ret;
1001
1002         if (i <= score->rule) {
1003                 switch (i) {
1004                 case IPV6_SADDR_RULE_SCOPE:
1005                         ret = score->scopedist;
1006                         break;
1007                 case IPV6_SADDR_RULE_PREFIX:
1008                         ret = score->matchlen;
1009                         break;
1010                 default:
1011                         ret = !!test_bit(i, score->scorebits);
1012                 }
1013                 goto out;
1014         }
1015
1016         switch (i) {
1017         case IPV6_SADDR_RULE_INIT:
1018                 /* Rule 0: remember if hiscore is not ready yet */
1019                 ret = !!score->ifa;
1020                 break;
1021         case IPV6_SADDR_RULE_LOCAL:
1022                 /* Rule 1: Prefer same address */
1023                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1024                 break;
1025         case IPV6_SADDR_RULE_SCOPE:
1026                 /* Rule 2: Prefer appropriate scope
1027                  *
1028                  *      ret
1029                  *       ^
1030                  *    -1 |  d 15
1031                  *    ---+--+-+---> scope
1032                  *       |
1033                  *       |             d is scope of the destination.
1034                  *  B-d  |  \
1035                  *       |   \      <- smaller scope is better if
1036                  *  B-15 |    \        if scope is enough for destinaion.
1037                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1038                  * d-C-1 | /
1039                  *       |/         <- greater is better
1040                  *   -C  /             if scope is not enough for destination.
1041                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1042                  *
1043                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1044                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1045                  * Assume B = 0 and we get C > 29.
1046                  */
1047                 ret = __ipv6_addr_src_scope(score->addr_type);
1048                 if (ret >= dst->scope)
1049                         ret = -ret;
1050                 else
1051                         ret -= 128;     /* 30 is enough */
1052                 score->scopedist = ret;
1053                 break;
1054         case IPV6_SADDR_RULE_PREFERRED:
1055                 /* Rule 3: Avoid deprecated and optimistic addresses */
1056                 ret = ipv6_saddr_preferred(score->addr_type) ||
1057                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1058                 break;
1059 #ifdef CONFIG_IPV6_MIP6
1060         case IPV6_SADDR_RULE_HOA:
1061             {
1062                 /* Rule 4: Prefer home address */
1063                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1064                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1065                 break;
1066             }
1067 #endif
1068         case IPV6_SADDR_RULE_OIF:
1069                 /* Rule 5: Prefer outgoing interface */
1070                 ret = (!dst->ifindex ||
1071                        dst->ifindex == score->ifa->idev->dev->ifindex);
1072                 break;
1073         case IPV6_SADDR_RULE_LABEL:
1074                 /* Rule 6: Prefer matching label */
1075                 ret = ipv6_addr_label(net,
1076                                       &score->ifa->addr, score->addr_type,
1077                                       score->ifa->idev->dev->ifindex) == dst->label;
1078                 break;
1079 #ifdef CONFIG_IPV6_PRIVACY
1080         case IPV6_SADDR_RULE_PRIVACY:
1081             {
1082                 /* Rule 7: Prefer public address
1083                  * Note: prefer temporary address if use_tempaddr >= 2
1084                  */
1085                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1086                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1087                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1088                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1089                 break;
1090             }
1091 #endif
1092         case IPV6_SADDR_RULE_ORCHID:
1093                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1094                  *          non-ORCHID vs non-ORCHID
1095                  */
1096                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1097                         ipv6_addr_orchid(dst->addr));
1098                 break;
1099         case IPV6_SADDR_RULE_PREFIX:
1100                 /* Rule 8: Use longest matching prefix */
1101                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1102                                                        dst->addr);
1103                 break;
1104         default:
1105                 ret = 0;
1106         }
1107
1108         if (ret)
1109                 __set_bit(i, score->scorebits);
1110         score->rule = i;
1111 out:
1112         return ret;
1113 }
1114
1115 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1116                        const struct in6_addr *daddr, unsigned int prefs,
1117                        struct in6_addr *saddr)
1118 {
1119         struct ipv6_saddr_score scores[2],
1120                                 *score = &scores[0], *hiscore = &scores[1];
1121         struct ipv6_saddr_dst dst;
1122         struct net_device *dev;
1123         int dst_type;
1124
1125         dst_type = __ipv6_addr_type(daddr);
1126         dst.addr = daddr;
1127         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1128         dst.scope = __ipv6_addr_src_scope(dst_type);
1129         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1130         dst.prefs = prefs;
1131
1132         hiscore->rule = -1;
1133         hiscore->ifa = NULL;
1134
1135         rcu_read_lock();
1136
1137         for_each_netdev_rcu(net, dev) {
1138                 struct inet6_dev *idev;
1139
1140                 /* Candidate Source Address (section 4)
1141                  *  - multicast and link-local destination address,
1142                  *    the set of candidate source address MUST only
1143                  *    include addresses assigned to interfaces
1144                  *    belonging to the same link as the outgoing
1145                  *    interface.
1146                  * (- For site-local destination addresses, the
1147                  *    set of candidate source addresses MUST only
1148                  *    include addresses assigned to interfaces
1149                  *    belonging to the same site as the outgoing
1150                  *    interface.)
1151                  */
1152                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1153                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1154                     dst.ifindex && dev->ifindex != dst.ifindex)
1155                         continue;
1156
1157                 idev = __in6_dev_get(dev);
1158                 if (!idev)
1159                         continue;
1160
1161                 read_lock_bh(&idev->lock);
1162                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1163                         int i;
1164
1165                         /*
1166                          * - Tentative Address (RFC2462 section 5.4)
1167                          *  - A tentative address is not considered
1168                          *    "assigned to an interface" in the traditional
1169                          *    sense, unless it is also flagged as optimistic.
1170                          * - Candidate Source Address (section 4)
1171                          *  - In any case, anycast addresses, multicast
1172                          *    addresses, and the unspecified address MUST
1173                          *    NOT be included in a candidate set.
1174                          */
1175                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1176                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1177                                 continue;
1178
1179                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1180
1181                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1182                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1183                                 LIMIT_NETDEBUG(KERN_DEBUG
1184                                                "ADDRCONF: unspecified / multicast address "
1185                                                "assigned as unicast address on %s",
1186                                                dev->name);
1187                                 continue;
1188                         }
1189
1190                         score->rule = -1;
1191                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1192
1193                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1194                                 int minihiscore, miniscore;
1195
1196                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1197                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1198
1199                                 if (minihiscore > miniscore) {
1200                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1201                                             score->scopedist > 0) {
1202                                                 /*
1203                                                  * special case:
1204                                                  * each remaining entry
1205                                                  * has too small (not enough)
1206                                                  * scope, because ifa entries
1207                                                  * are sorted by their scope
1208                                                  * values.
1209                                                  */
1210                                                 goto try_nextdev;
1211                                         }
1212                                         break;
1213                                 } else if (minihiscore < miniscore) {
1214                                         if (hiscore->ifa)
1215                                                 in6_ifa_put(hiscore->ifa);
1216
1217                                         in6_ifa_hold(score->ifa);
1218
1219                                         swap(hiscore, score);
1220
1221                                         /* restore our iterator */
1222                                         score->ifa = hiscore->ifa;
1223
1224                                         break;
1225                                 }
1226                         }
1227                 }
1228 try_nextdev:
1229                 read_unlock_bh(&idev->lock);
1230         }
1231         rcu_read_unlock();
1232
1233         if (!hiscore->ifa)
1234                 return -EADDRNOTAVAIL;
1235
1236         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1237         in6_ifa_put(hiscore->ifa);
1238         return 0;
1239 }
1240 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1241
1242 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1243                       unsigned char banned_flags)
1244 {
1245         struct inet6_ifaddr *ifp;
1246         int err = -EADDRNOTAVAIL;
1247
1248         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1249                 if (ifp->scope == IFA_LINK &&
1250                     !(ifp->flags & banned_flags)) {
1251                         ipv6_addr_copy(addr, &ifp->addr);
1252                         err = 0;
1253                         break;
1254                 }
1255         }
1256         return err;
1257 }
1258
1259 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1260                     unsigned char banned_flags)
1261 {
1262         struct inet6_dev *idev;
1263         int err = -EADDRNOTAVAIL;
1264
1265         rcu_read_lock();
1266         idev = __in6_dev_get(dev);
1267         if (idev) {
1268                 read_lock_bh(&idev->lock);
1269                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
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         list_for_each_entry(ifp, &idev->addr_list, if_list)
1283                 cnt++;
1284         read_unlock_bh(&idev->lock);
1285         return cnt;
1286 }
1287
1288 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1289                   struct net_device *dev, int strict)
1290 {
1291         struct inet6_ifaddr *ifp;
1292         struct hlist_node *node;
1293         unsigned int hash = ipv6_addr_hash(addr);
1294
1295         rcu_read_lock_bh();
1296         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1297                 if (!net_eq(dev_net(ifp->idev->dev), net))
1298                         continue;
1299                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1300                     !(ifp->flags&IFA_F_TENTATIVE) &&
1301                     (dev == NULL || ifp->idev->dev == dev ||
1302                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1303                         rcu_read_unlock_bh();
1304                         return 1;
1305                 }
1306         }
1307
1308         rcu_read_unlock_bh();
1309         return 0;
1310 }
1311 EXPORT_SYMBOL(ipv6_chk_addr);
1312
1313 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1314                                struct net_device *dev)
1315 {
1316         unsigned int hash = ipv6_addr_hash(addr);
1317         struct inet6_ifaddr *ifp;
1318         struct hlist_node *node;
1319
1320         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1321                 if (!net_eq(dev_net(ifp->idev->dev), net))
1322                         continue;
1323                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1324                         if (dev == NULL || ifp->idev->dev == dev)
1325                                 return true;
1326                 }
1327         }
1328         return false;
1329 }
1330
1331 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1332 {
1333         struct inet6_dev *idev;
1334         struct inet6_ifaddr *ifa;
1335         int     onlink;
1336
1337         onlink = 0;
1338         rcu_read_lock();
1339         idev = __in6_dev_get(dev);
1340         if (idev) {
1341                 read_lock_bh(&idev->lock);
1342                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1343                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1344                                                    ifa->prefix_len);
1345                         if (onlink)
1346                                 break;
1347                 }
1348                 read_unlock_bh(&idev->lock);
1349         }
1350         rcu_read_unlock();
1351         return onlink;
1352 }
1353
1354 EXPORT_SYMBOL(ipv6_chk_prefix);
1355
1356 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1357                                      struct net_device *dev, int strict)
1358 {
1359         struct inet6_ifaddr *ifp, *result = NULL;
1360         unsigned int hash = ipv6_addr_hash(addr);
1361         struct hlist_node *node;
1362
1363         rcu_read_lock_bh();
1364         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1365                 if (!net_eq(dev_net(ifp->idev->dev), net))
1366                         continue;
1367                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1368                         if (dev == NULL || ifp->idev->dev == dev ||
1369                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1370                                 result = ifp;
1371                                 in6_ifa_hold(ifp);
1372                                 break;
1373                         }
1374                 }
1375         }
1376         rcu_read_unlock_bh();
1377
1378         return result;
1379 }
1380
1381 /* Gets referenced address, destroys ifaddr */
1382
1383 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1384 {
1385         if (dad_failed)
1386                 ifp->flags |= IFA_F_DADFAILED;
1387
1388 #ifdef CONFIG_IPV6_PRIVACY
1389         if (ifp->flags&IFA_F_TEMPORARY) {
1390                 struct inet6_ifaddr *ifpub;
1391                 spin_lock_bh(&ifp->lock);
1392                 ifpub = ifp->ifpub;
1393                 if (ifpub) {
1394                         in6_ifa_hold(ifpub);
1395                         spin_unlock_bh(&ifp->lock);
1396                         ipv6_create_tempaddr(ifpub, ifp);
1397                         in6_ifa_put(ifpub);
1398                 } else {
1399                         spin_unlock_bh(&ifp->lock);
1400                 }
1401                 ipv6_del_addr(ifp);
1402         } else
1403 #endif
1404         if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1405                 spin_lock_bh(&ifp->lock);
1406                 addrconf_del_timer(ifp);
1407                 ifp->flags |= IFA_F_TENTATIVE;
1408                 spin_unlock_bh(&ifp->lock);
1409                 if (dad_failed)
1410                         ipv6_ifa_notify(0, ifp);
1411                 in6_ifa_put(ifp);
1412         } else
1413                 ipv6_del_addr(ifp);
1414 }
1415
1416 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1417 {
1418         int err = -ENOENT;
1419
1420         spin_lock(&ifp->state_lock);
1421         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1422                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1423                 err = 0;
1424         }
1425         spin_unlock(&ifp->state_lock);
1426
1427         return err;
1428 }
1429
1430 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1431 {
1432         struct inet6_dev *idev = ifp->idev;
1433
1434         if (addrconf_dad_end(ifp)) {
1435                 in6_ifa_put(ifp);
1436                 return;
1437         }
1438
1439         if (net_ratelimit())
1440                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1441                         ifp->idev->dev->name, &ifp->addr);
1442
1443         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1444                 struct in6_addr addr;
1445
1446                 addr.s6_addr32[0] = htonl(0xfe800000);
1447                 addr.s6_addr32[1] = 0;
1448
1449                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1450                     ipv6_addr_equal(&ifp->addr, &addr)) {
1451                         /* DAD failed for link-local based on MAC address */
1452                         idev->cnf.disable_ipv6 = 1;
1453
1454                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1455                                 ifp->idev->dev->name);
1456                 }
1457         }
1458
1459         addrconf_dad_stop(ifp, 1);
1460 }
1461
1462 /* Join to solicited addr multicast group. */
1463
1464 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1465 {
1466         struct in6_addr maddr;
1467
1468         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1469                 return;
1470
1471         addrconf_addr_solict_mult(addr, &maddr);
1472         ipv6_dev_mc_inc(dev, &maddr);
1473 }
1474
1475 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1476 {
1477         struct in6_addr maddr;
1478
1479         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1480                 return;
1481
1482         addrconf_addr_solict_mult(addr, &maddr);
1483         __ipv6_dev_mc_dec(idev, &maddr);
1484 }
1485
1486 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1487 {
1488         struct in6_addr addr;
1489         if (ifp->prefix_len == 127) /* RFC 6164 */
1490                 return;
1491         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1492         if (ipv6_addr_any(&addr))
1493                 return;
1494         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1495 }
1496
1497 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1498 {
1499         struct in6_addr addr;
1500         if (ifp->prefix_len == 127) /* RFC 6164 */
1501                 return;
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 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1559 {
1560         if (addr == 0)
1561                 return -1;
1562         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1563                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1564                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1565                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1566                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1567                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1568         eui[1] = 0;
1569         eui[2] = 0x5E;
1570         eui[3] = 0xFE;
1571         memcpy(eui + 4, &addr, 4);
1572         return 0;
1573 }
1574
1575 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1576 {
1577         if (dev->priv_flags & IFF_ISATAP)
1578                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1579         return -1;
1580 }
1581
1582 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1583 {
1584         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1585 }
1586
1587 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1588 {
1589         switch (dev->type) {
1590         case ARPHRD_ETHER:
1591         case ARPHRD_FDDI:
1592         case ARPHRD_IEEE802_TR:
1593                 return addrconf_ifid_eui48(eui, dev);
1594         case ARPHRD_ARCNET:
1595                 return addrconf_ifid_arcnet(eui, dev);
1596         case ARPHRD_INFINIBAND:
1597                 return addrconf_ifid_infiniband(eui, dev);
1598         case ARPHRD_SIT:
1599                 return addrconf_ifid_sit(eui, dev);
1600         case ARPHRD_IPGRE:
1601                 return addrconf_ifid_gre(eui, dev);
1602         }
1603         return -1;
1604 }
1605
1606 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1607 {
1608         int err = -1;
1609         struct inet6_ifaddr *ifp;
1610
1611         read_lock_bh(&idev->lock);
1612         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1613                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1614                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1615                         err = 0;
1616                         break;
1617                 }
1618         }
1619         read_unlock_bh(&idev->lock);
1620         return err;
1621 }
1622
1623 #ifdef CONFIG_IPV6_PRIVACY
1624 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1625 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1626 {
1627 regen:
1628         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1629         idev->rndid[0] &= ~0x02;
1630
1631         /*
1632          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1633          * check if generated address is not inappropriate
1634          *
1635          *  - Reserved subnet anycast (RFC 2526)
1636          *      11111101 11....11 1xxxxxxx
1637          *  - ISATAP (RFC4214) 6.1
1638          *      00-00-5E-FE-xx-xx-xx-xx
1639          *  - value 0
1640          *  - XXX: already assigned to an address on the device
1641          */
1642         if (idev->rndid[0] == 0xfd &&
1643             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1644             (idev->rndid[7]&0x80))
1645                 goto regen;
1646         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1647                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1648                         goto regen;
1649                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1650                         goto regen;
1651         }
1652
1653         return 0;
1654 }
1655
1656 static void ipv6_regen_rndid(unsigned long data)
1657 {
1658         struct inet6_dev *idev = (struct inet6_dev *) data;
1659         unsigned long expires;
1660
1661         rcu_read_lock_bh();
1662         write_lock_bh(&idev->lock);
1663
1664         if (idev->dead)
1665                 goto out;
1666
1667         if (__ipv6_regen_rndid(idev) < 0)
1668                 goto out;
1669
1670         expires = jiffies +
1671                 idev->cnf.temp_prefered_lft * HZ -
1672                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1673                 idev->cnf.max_desync_factor * HZ;
1674         if (time_before(expires, jiffies)) {
1675                 printk(KERN_WARNING
1676                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1677                         idev->dev->name);
1678                 goto out;
1679         }
1680
1681         if (!mod_timer(&idev->regen_timer, expires))
1682                 in6_dev_hold(idev);
1683
1684 out:
1685         write_unlock_bh(&idev->lock);
1686         rcu_read_unlock_bh();
1687         in6_dev_put(idev);
1688 }
1689
1690 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1691         int ret = 0;
1692
1693         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1694                 ret = __ipv6_regen_rndid(idev);
1695         return ret;
1696 }
1697 #endif
1698
1699 /*
1700  *      Add prefix route.
1701  */
1702
1703 static void
1704 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1705                       unsigned long expires, u32 flags)
1706 {
1707         struct fib6_config cfg = {
1708                 .fc_table = RT6_TABLE_PREFIX,
1709                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1710                 .fc_ifindex = dev->ifindex,
1711                 .fc_expires = expires,
1712                 .fc_dst_len = plen,
1713                 .fc_flags = RTF_UP | flags,
1714                 .fc_nlinfo.nl_net = dev_net(dev),
1715                 .fc_protocol = RTPROT_KERNEL,
1716         };
1717
1718         ipv6_addr_copy(&cfg.fc_dst, pfx);
1719
1720         /* Prevent useless cloning on PtP SIT.
1721            This thing is done here expecting that the whole
1722            class of non-broadcast devices need not cloning.
1723          */
1724 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1725         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1726                 cfg.fc_flags |= RTF_NONEXTHOP;
1727 #endif
1728
1729         ip6_route_add(&cfg);
1730 }
1731
1732
1733 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1734                                                   int plen,
1735                                                   const struct net_device *dev,
1736                                                   u32 flags, u32 noflags)
1737 {
1738         struct fib6_node *fn;
1739         struct rt6_info *rt = NULL;
1740         struct fib6_table *table;
1741
1742         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1743         if (table == NULL)
1744                 return NULL;
1745
1746         write_lock_bh(&table->tb6_lock);
1747         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1748         if (!fn)
1749                 goto out;
1750         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1751                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1752                         continue;
1753                 if ((rt->rt6i_flags & flags) != flags)
1754                         continue;
1755                 if ((rt->rt6i_flags & noflags) != 0)
1756                         continue;
1757                 dst_hold(&rt->dst);
1758                 break;
1759         }
1760 out:
1761         write_unlock_bh(&table->tb6_lock);
1762         return rt;
1763 }
1764
1765
1766 /* Create "default" multicast route to the interface */
1767
1768 static void addrconf_add_mroute(struct net_device *dev)
1769 {
1770         struct fib6_config cfg = {
1771                 .fc_table = RT6_TABLE_LOCAL,
1772                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1773                 .fc_ifindex = dev->ifindex,
1774                 .fc_dst_len = 8,
1775                 .fc_flags = RTF_UP,
1776                 .fc_nlinfo.nl_net = dev_net(dev),
1777         };
1778
1779         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1780
1781         ip6_route_add(&cfg);
1782 }
1783
1784 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1785 static void sit_route_add(struct net_device *dev)
1786 {
1787         struct fib6_config cfg = {
1788                 .fc_table = RT6_TABLE_MAIN,
1789                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1790                 .fc_ifindex = dev->ifindex,
1791                 .fc_dst_len = 96,
1792                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1793                 .fc_nlinfo.nl_net = dev_net(dev),
1794         };
1795
1796         /* prefix length - 96 bits "::d.d.d.d" */
1797         ip6_route_add(&cfg);
1798 }
1799 #endif
1800
1801 static void addrconf_add_lroute(struct net_device *dev)
1802 {
1803         struct in6_addr addr;
1804
1805         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1806         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1807 }
1808
1809 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1810 {
1811         struct inet6_dev *idev;
1812
1813         ASSERT_RTNL();
1814
1815         idev = ipv6_find_idev(dev);
1816         if (!idev)
1817                 return ERR_PTR(-ENOBUFS);
1818
1819         if (idev->cnf.disable_ipv6)
1820                 return ERR_PTR(-EACCES);
1821
1822         /* Add default multicast route */
1823         if (!(dev->flags & IFF_LOOPBACK))
1824                 addrconf_add_mroute(dev);
1825
1826         /* Add link local route */
1827         addrconf_add_lroute(dev);
1828         return idev;
1829 }
1830
1831 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1832 {
1833         struct prefix_info *pinfo;
1834         __u32 valid_lft;
1835         __u32 prefered_lft;
1836         int addr_type;
1837         struct inet6_dev *in6_dev;
1838         struct net *net = dev_net(dev);
1839
1840         pinfo = (struct prefix_info *) opt;
1841
1842         if (len < sizeof(struct prefix_info)) {
1843                 ADBG(("addrconf: prefix option too short\n"));
1844                 return;
1845         }
1846
1847         /*
1848          *      Validation checks ([ADDRCONF], page 19)
1849          */
1850
1851         addr_type = ipv6_addr_type(&pinfo->prefix);
1852
1853         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1854                 return;
1855
1856         valid_lft = ntohl(pinfo->valid);
1857         prefered_lft = ntohl(pinfo->prefered);
1858
1859         if (prefered_lft > valid_lft) {
1860                 if (net_ratelimit())
1861                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1862                 return;
1863         }
1864
1865         in6_dev = in6_dev_get(dev);
1866
1867         if (in6_dev == NULL) {
1868                 if (net_ratelimit())
1869                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1870                 return;
1871         }
1872
1873         /*
1874          *      Two things going on here:
1875          *      1) Add routes for on-link prefixes
1876          *      2) Configure prefixes with the auto flag set
1877          */
1878
1879         if (pinfo->onlink) {
1880                 struct rt6_info *rt;
1881                 unsigned long rt_expires;
1882
1883                 /* Avoid arithmetic overflow. Really, we could
1884                  * save rt_expires in seconds, likely valid_lft,
1885                  * but it would require division in fib gc, that it
1886                  * not good.
1887                  */
1888                 if (HZ > USER_HZ)
1889                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1890                 else
1891                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1892
1893                 if (addrconf_finite_timeout(rt_expires))
1894                         rt_expires *= HZ;
1895
1896                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1897                                                pinfo->prefix_len,
1898                                                dev,
1899                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1900                                                RTF_GATEWAY | RTF_DEFAULT);
1901
1902                 if (rt) {
1903                         /* Autoconf prefix route */
1904                         if (valid_lft == 0) {
1905                                 ip6_del_rt(rt);
1906                                 rt = NULL;
1907                         } else if (addrconf_finite_timeout(rt_expires)) {
1908                                 /* not infinity */
1909                                 rt->rt6i_expires = jiffies + rt_expires;
1910                                 rt->rt6i_flags |= RTF_EXPIRES;
1911                         } else {
1912                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1913                                 rt->rt6i_expires = 0;
1914                         }
1915                 } else if (valid_lft) {
1916                         clock_t expires = 0;
1917                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1918                         if (addrconf_finite_timeout(rt_expires)) {
1919                                 /* not infinity */
1920                                 flags |= RTF_EXPIRES;
1921                                 expires = jiffies_to_clock_t(rt_expires);
1922                         }
1923                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1924                                               dev, expires, flags);
1925                 }
1926                 if (rt)
1927                         dst_release(&rt->dst);
1928         }
1929
1930         /* Try to figure out our local address for this prefix */
1931
1932         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1933                 struct inet6_ifaddr * ifp;
1934                 struct in6_addr addr;
1935                 int create = 0, update_lft = 0;
1936
1937                 if (pinfo->prefix_len == 64) {
1938                         memcpy(&addr, &pinfo->prefix, 8);
1939                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1940                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1941                                 in6_dev_put(in6_dev);
1942                                 return;
1943                         }
1944                         goto ok;
1945                 }
1946                 if (net_ratelimit())
1947                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1948                                pinfo->prefix_len);
1949                 in6_dev_put(in6_dev);
1950                 return;
1951
1952 ok:
1953
1954                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1955
1956                 if (ifp == NULL && valid_lft) {
1957                         int max_addresses = in6_dev->cnf.max_addresses;
1958                         u32 addr_flags = 0;
1959
1960 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1961                         if (in6_dev->cnf.optimistic_dad &&
1962                             !net->ipv6.devconf_all->forwarding)
1963                                 addr_flags = IFA_F_OPTIMISTIC;
1964 #endif
1965
1966                         /* Do not allow to create too much of autoconfigured
1967                          * addresses; this would be too easy way to crash kernel.
1968                          */
1969                         if (!max_addresses ||
1970                             ipv6_count_addresses(in6_dev) < max_addresses)
1971                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1972                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1973                                                     addr_flags);
1974
1975                         if (!ifp || IS_ERR(ifp)) {
1976                                 in6_dev_put(in6_dev);
1977                                 return;
1978                         }
1979
1980                         update_lft = create = 1;
1981                         ifp->cstamp = jiffies;
1982                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1983                 }
1984
1985                 if (ifp) {
1986                         int flags;
1987                         unsigned long now;
1988 #ifdef CONFIG_IPV6_PRIVACY
1989                         struct inet6_ifaddr *ift;
1990 #endif
1991                         u32 stored_lft;
1992
1993                         /* update lifetime (RFC2462 5.5.3 e) */
1994                         spin_lock(&ifp->lock);
1995                         now = jiffies;
1996                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1997                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1998                         else
1999                                 stored_lft = 0;
2000                         if (!update_lft && stored_lft) {
2001                                 if (valid_lft > MIN_VALID_LIFETIME ||
2002                                     valid_lft > stored_lft)
2003                                         update_lft = 1;
2004                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2005                                         /* valid_lft <= stored_lft is always true */
2006                                         /*
2007                                          * RFC 4862 Section 5.5.3e:
2008                                          * "Note that the preferred lifetime of
2009                                          *  the corresponding address is always
2010                                          *  reset to the Preferred Lifetime in
2011                                          *  the received Prefix Information
2012                                          *  option, regardless of whether the
2013                                          *  valid lifetime is also reset or
2014                                          *  ignored."
2015                                          *
2016                                          *  So if the preferred lifetime in
2017                                          *  this advertisement is different
2018                                          *  than what we have stored, but the
2019                                          *  valid lifetime is invalid, just
2020                                          *  reset prefered_lft.
2021                                          *
2022                                          *  We must set the valid lifetime
2023                                          *  to the stored lifetime since we'll
2024                                          *  be updating the timestamp below,
2025                                          *  else we'll set it back to the
2026                                          *  minimum.
2027                                          */
2028                                         if (prefered_lft != ifp->prefered_lft) {
2029                                                 valid_lft = stored_lft;
2030                                                 update_lft = 1;
2031                                         }
2032                                 } else {
2033                                         valid_lft = MIN_VALID_LIFETIME;
2034                                         if (valid_lft < prefered_lft)
2035                                                 prefered_lft = valid_lft;
2036                                         update_lft = 1;
2037                                 }
2038                         }
2039
2040                         if (update_lft) {
2041                                 ifp->valid_lft = valid_lft;
2042                                 ifp->prefered_lft = prefered_lft;
2043                                 ifp->tstamp = now;
2044                                 flags = ifp->flags;
2045                                 ifp->flags &= ~IFA_F_DEPRECATED;
2046                                 spin_unlock(&ifp->lock);
2047
2048                                 if (!(flags&IFA_F_TENTATIVE))
2049                                         ipv6_ifa_notify(0, ifp);
2050                         } else
2051                                 spin_unlock(&ifp->lock);
2052
2053 #ifdef CONFIG_IPV6_PRIVACY
2054                         read_lock_bh(&in6_dev->lock);
2055                         /* update all temporary addresses in the list */
2056                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2057                                             tmp_list) {
2058                                 int age, max_valid, max_prefered;
2059
2060                                 if (ifp != ift->ifpub)
2061                                         continue;
2062
2063                                 /*
2064                                  * RFC 4941 section 3.3:
2065                                  * If a received option will extend the lifetime
2066                                  * of a public address, the lifetimes of
2067                                  * temporary addresses should be extended,
2068                                  * subject to the overall constraint that no
2069                                  * temporary addresses should ever remain
2070                                  * "valid" or "preferred" for a time longer than
2071                                  * (TEMP_VALID_LIFETIME) or
2072                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2073                                  * respectively.
2074                                  */
2075                                 age = (now - ift->cstamp) / HZ;
2076                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2077                                 if (max_valid < 0)
2078                                         max_valid = 0;
2079
2080                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2081                                                in6_dev->cnf.max_desync_factor -
2082                                                age;
2083                                 if (max_prefered < 0)
2084                                         max_prefered = 0;
2085
2086                                 if (valid_lft > max_valid)
2087                                         valid_lft = max_valid;
2088
2089                                 if (prefered_lft > max_prefered)
2090                                         prefered_lft = max_prefered;
2091
2092                                 spin_lock(&ift->lock);
2093                                 flags = ift->flags;
2094                                 ift->valid_lft = valid_lft;
2095                                 ift->prefered_lft = prefered_lft;
2096                                 ift->tstamp = now;
2097                                 if (prefered_lft > 0)
2098                                         ift->flags &= ~IFA_F_DEPRECATED;
2099
2100                                 spin_unlock(&ift->lock);
2101                                 if (!(flags&IFA_F_TENTATIVE))
2102                                         ipv6_ifa_notify(0, ift);
2103                         }
2104
2105                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2106                                 /*
2107                                  * When a new public address is created as
2108                                  * described in [ADDRCONF], also create a new
2109                                  * temporary address. Also create a temporary
2110                                  * address if it's enabled but no temporary
2111                                  * address currently exists.
2112                                  */
2113                                 read_unlock_bh(&in6_dev->lock);
2114                                 ipv6_create_tempaddr(ifp, NULL);
2115                         } else {
2116                                 read_unlock_bh(&in6_dev->lock);
2117                         }
2118 #endif
2119                         in6_ifa_put(ifp);
2120                         addrconf_verify(0);
2121                 }
2122         }
2123         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2124         in6_dev_put(in6_dev);
2125 }
2126
2127 /*
2128  *      Set destination address.
2129  *      Special case for SIT interfaces where we create a new "virtual"
2130  *      device.
2131  */
2132 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2133 {
2134         struct in6_ifreq ireq;
2135         struct net_device *dev;
2136         int err = -EINVAL;
2137
2138         rtnl_lock();
2139
2140         err = -EFAULT;
2141         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2142                 goto err_exit;
2143
2144         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2145
2146         err = -ENODEV;
2147         if (dev == NULL)
2148                 goto err_exit;
2149
2150 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2151         if (dev->type == ARPHRD_SIT) {
2152                 const struct net_device_ops *ops = dev->netdev_ops;
2153                 struct ifreq ifr;
2154                 struct ip_tunnel_parm p;
2155
2156                 err = -EADDRNOTAVAIL;
2157                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2158                         goto err_exit;
2159
2160                 memset(&p, 0, sizeof(p));
2161                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2162                 p.iph.saddr = 0;
2163                 p.iph.version = 4;
2164                 p.iph.ihl = 5;
2165                 p.iph.protocol = IPPROTO_IPV6;
2166                 p.iph.ttl = 64;
2167                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2168
2169                 if (ops->ndo_do_ioctl) {
2170                         mm_segment_t oldfs = get_fs();
2171
2172                         set_fs(KERNEL_DS);
2173                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2174                         set_fs(oldfs);
2175                 } else
2176                         err = -EOPNOTSUPP;
2177
2178                 if (err == 0) {
2179                         err = -ENOBUFS;
2180                         dev = __dev_get_by_name(net, p.name);
2181                         if (!dev)
2182                                 goto err_exit;
2183                         err = dev_open(dev);
2184                 }
2185         }
2186 #endif
2187
2188 err_exit:
2189         rtnl_unlock();
2190         return err;
2191 }
2192
2193 /*
2194  *      Manual configuration of address on an interface
2195  */
2196 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2197                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2198                           __u32 valid_lft)
2199 {
2200         struct inet6_ifaddr *ifp;
2201         struct inet6_dev *idev;
2202         struct net_device *dev;
2203         int scope;
2204         u32 flags;
2205         clock_t expires;
2206         unsigned long timeout;
2207
2208         ASSERT_RTNL();
2209
2210         if (plen > 128)
2211                 return -EINVAL;
2212
2213         /* check the lifetime */
2214         if (!valid_lft || prefered_lft > valid_lft)
2215                 return -EINVAL;
2216
2217         dev = __dev_get_by_index(net, ifindex);
2218         if (!dev)
2219                 return -ENODEV;
2220
2221         idev = addrconf_add_dev(dev);
2222         if (IS_ERR(idev))
2223                 return PTR_ERR(idev);
2224
2225         scope = ipv6_addr_scope(pfx);
2226
2227         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2228         if (addrconf_finite_timeout(timeout)) {
2229                 expires = jiffies_to_clock_t(timeout * HZ);
2230                 valid_lft = timeout;
2231                 flags = RTF_EXPIRES;
2232         } else {
2233                 expires = 0;
2234                 flags = 0;
2235                 ifa_flags |= IFA_F_PERMANENT;
2236         }
2237
2238         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2239         if (addrconf_finite_timeout(timeout)) {
2240                 if (timeout == 0)
2241                         ifa_flags |= IFA_F_DEPRECATED;
2242                 prefered_lft = timeout;
2243         }
2244
2245         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2246
2247         if (!IS_ERR(ifp)) {
2248                 spin_lock_bh(&ifp->lock);
2249                 ifp->valid_lft = valid_lft;
2250                 ifp->prefered_lft = prefered_lft;
2251                 ifp->tstamp = jiffies;
2252                 spin_unlock_bh(&ifp->lock);
2253
2254                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2255                                       expires, flags);
2256                 /*
2257                  * Note that section 3.1 of RFC 4429 indicates
2258                  * that the Optimistic flag should not be set for
2259                  * manually configured addresses
2260                  */
2261                 addrconf_dad_start(ifp, 0);
2262                 in6_ifa_put(ifp);
2263                 addrconf_verify(0);
2264                 return 0;
2265         }
2266
2267         return PTR_ERR(ifp);
2268 }
2269
2270 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2271                           unsigned int plen)
2272 {
2273         struct inet6_ifaddr *ifp;
2274         struct inet6_dev *idev;
2275         struct net_device *dev;
2276
2277         if (plen > 128)
2278                 return -EINVAL;
2279
2280         dev = __dev_get_by_index(net, ifindex);
2281         if (!dev)
2282                 return -ENODEV;
2283
2284         if ((idev = __in6_dev_get(dev)) == NULL)
2285                 return -ENXIO;
2286
2287         read_lock_bh(&idev->lock);
2288         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2289                 if (ifp->prefix_len == plen &&
2290                     ipv6_addr_equal(pfx, &ifp->addr)) {
2291                         in6_ifa_hold(ifp);
2292                         read_unlock_bh(&idev->lock);
2293
2294                         ipv6_del_addr(ifp);
2295
2296                         /* If the last address is deleted administratively,
2297                            disable IPv6 on this interface.
2298                          */
2299                         if (list_empty(&idev->addr_list))
2300                                 addrconf_ifdown(idev->dev, 1);
2301                         return 0;
2302                 }
2303         }
2304         read_unlock_bh(&idev->lock);
2305         return -EADDRNOTAVAIL;
2306 }
2307
2308
2309 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2310 {
2311         struct in6_ifreq ireq;
2312         int err;
2313
2314         if (!capable(CAP_NET_ADMIN))
2315                 return -EPERM;
2316
2317         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2318                 return -EFAULT;
2319
2320         rtnl_lock();
2321         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2322                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2323                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2324         rtnl_unlock();
2325         return err;
2326 }
2327
2328 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2329 {
2330         struct in6_ifreq ireq;
2331         int err;
2332
2333         if (!capable(CAP_NET_ADMIN))
2334                 return -EPERM;
2335
2336         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2337                 return -EFAULT;
2338
2339         rtnl_lock();
2340         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2341                              ireq.ifr6_prefixlen);
2342         rtnl_unlock();
2343         return err;
2344 }
2345
2346 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2347                      int plen, int scope)
2348 {
2349         struct inet6_ifaddr *ifp;
2350
2351         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2352         if (!IS_ERR(ifp)) {
2353                 spin_lock_bh(&ifp->lock);
2354                 ifp->flags &= ~IFA_F_TENTATIVE;
2355                 spin_unlock_bh(&ifp->lock);
2356                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2357                 in6_ifa_put(ifp);
2358         }
2359 }
2360
2361 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2362 static void sit_add_v4_addrs(struct inet6_dev *idev)
2363 {
2364         struct in6_addr addr;
2365         struct net_device *dev;
2366         struct net *net = dev_net(idev->dev);
2367         int scope;
2368
2369         ASSERT_RTNL();
2370
2371         memset(&addr, 0, sizeof(struct in6_addr));
2372         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2373
2374         if (idev->dev->flags&IFF_POINTOPOINT) {
2375                 addr.s6_addr32[0] = htonl(0xfe800000);
2376                 scope = IFA_LINK;
2377         } else {
2378                 scope = IPV6_ADDR_COMPATv4;
2379         }
2380
2381         if (addr.s6_addr32[3]) {
2382                 add_addr(idev, &addr, 128, scope);
2383                 return;
2384         }
2385
2386         for_each_netdev(net, dev) {
2387                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2388                 if (in_dev && (dev->flags & IFF_UP)) {
2389                         struct in_ifaddr * ifa;
2390
2391                         int flag = scope;
2392
2393                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2394                                 int plen;
2395
2396                                 addr.s6_addr32[3] = ifa->ifa_local;
2397
2398                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2399                                         continue;
2400                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2401                                         if (idev->dev->flags&IFF_POINTOPOINT)
2402                                                 continue;
2403                                         flag |= IFA_HOST;
2404                                 }
2405                                 if (idev->dev->flags&IFF_POINTOPOINT)
2406                                         plen = 64;
2407                                 else
2408                                         plen = 96;
2409
2410                                 add_addr(idev, &addr, plen, flag);
2411                         }
2412                 }
2413         }
2414 }
2415 #endif
2416
2417 static void init_loopback(struct net_device *dev)
2418 {
2419         struct inet6_dev  *idev;
2420         struct net_device *sp_dev;
2421         struct inet6_ifaddr *sp_ifa;
2422         struct rt6_info *sp_rt;
2423
2424         /* ::1 */
2425
2426         ASSERT_RTNL();
2427
2428         if ((idev = ipv6_find_idev(dev)) == NULL) {
2429                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2430                 return;
2431         }
2432
2433         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2434
2435         /* Add routes to other interface's IPv6 addresses */
2436         for_each_netdev(dev_net(dev), sp_dev) {
2437                 if (!strcmp(sp_dev->name, dev->name))
2438                         continue;
2439
2440                 idev = __in6_dev_get(sp_dev);
2441                 if (!idev)
2442                         continue;
2443
2444                 read_lock_bh(&idev->lock);
2445                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2446
2447                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2448                                 continue;
2449
2450                         if (sp_ifa->rt) {
2451                                 /* This dst has been added to garbage list when
2452                                  * lo device down, release this obsolete dst and
2453                                  * reallocate a new router for ifa.
2454                                  */
2455                                 if (sp_ifa->rt->dst.obsolete > 0) {
2456                                         dst_release(&sp_ifa->rt->dst);
2457                                         sp_ifa->rt = NULL;
2458                                 } else {
2459                                         continue;
2460                                 }
2461                         }
2462
2463                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2464
2465                         /* Failure cases are ignored */
2466                         if (!IS_ERR(sp_rt)) {
2467                                 sp_ifa->rt = sp_rt;
2468                                 ip6_ins_rt(sp_rt);
2469                         }
2470                 }
2471                 read_unlock_bh(&idev->lock);
2472         }
2473 }
2474
2475 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2476 {
2477         struct inet6_ifaddr * ifp;
2478         u32 addr_flags = IFA_F_PERMANENT;
2479
2480 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2481         if (idev->cnf.optimistic_dad &&
2482             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2483                 addr_flags |= IFA_F_OPTIMISTIC;
2484 #endif
2485
2486
2487         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2488         if (!IS_ERR(ifp)) {
2489                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2490                 addrconf_dad_start(ifp, 0);
2491                 in6_ifa_put(ifp);
2492         }
2493 }
2494
2495 static void addrconf_dev_config(struct net_device *dev)
2496 {
2497         struct in6_addr addr;
2498         struct inet6_dev    * idev;
2499
2500         ASSERT_RTNL();
2501
2502         if ((dev->type != ARPHRD_ETHER) &&
2503             (dev->type != ARPHRD_FDDI) &&
2504             (dev->type != ARPHRD_IEEE802_TR) &&
2505             (dev->type != ARPHRD_ARCNET) &&
2506             (dev->type != ARPHRD_INFINIBAND)) {
2507                 /* Alas, we support only Ethernet autoconfiguration. */
2508                 return;
2509         }
2510
2511         idev = addrconf_add_dev(dev);
2512         if (IS_ERR(idev))
2513                 return;
2514
2515         memset(&addr, 0, sizeof(struct in6_addr));
2516         addr.s6_addr32[0] = htonl(0xFE800000);
2517
2518         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2519                 addrconf_add_linklocal(idev, &addr);
2520 }
2521
2522 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2523 static void addrconf_sit_config(struct net_device *dev)
2524 {
2525         struct inet6_dev *idev;
2526
2527         ASSERT_RTNL();
2528
2529         /*
2530          * Configure the tunnel with one of our IPv4
2531          * addresses... we should configure all of
2532          * our v4 addrs in the tunnel
2533          */
2534
2535         if ((idev = ipv6_find_idev(dev)) == NULL) {
2536                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2537                 return;
2538         }
2539
2540         if (dev->priv_flags & IFF_ISATAP) {
2541                 struct in6_addr addr;
2542
2543                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2544                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2545                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2546                         addrconf_add_linklocal(idev, &addr);
2547                 return;
2548         }
2549
2550         sit_add_v4_addrs(idev);
2551
2552         if (dev->flags&IFF_POINTOPOINT) {
2553                 addrconf_add_mroute(dev);
2554                 addrconf_add_lroute(dev);
2555         } else
2556                 sit_route_add(dev);
2557 }
2558 #endif
2559
2560 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2561 static void addrconf_gre_config(struct net_device *dev)
2562 {
2563         struct inet6_dev *idev;
2564         struct in6_addr addr;
2565
2566         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2567
2568         ASSERT_RTNL();
2569
2570         if ((idev = ipv6_find_idev(dev)) == NULL) {
2571                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2572                 return;
2573         }
2574
2575         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2576         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2577
2578         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2579                 addrconf_add_linklocal(idev, &addr);
2580 }
2581 #endif
2582
2583 static inline int
2584 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2585 {
2586         struct in6_addr lladdr;
2587
2588         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2589                 addrconf_add_linklocal(idev, &lladdr);
2590                 return 0;
2591         }
2592         return -1;
2593 }
2594
2595 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2596 {
2597         struct net_device *link_dev;
2598         struct net *net = dev_net(idev->dev);
2599
2600         /* first try to inherit the link-local address from the link device */
2601         if (idev->dev->iflink &&
2602             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2603                 if (!ipv6_inherit_linklocal(idev, link_dev))
2604                         return;
2605         }
2606         /* then try to inherit it from any device */
2607         for_each_netdev(net, link_dev) {
2608                 if (!ipv6_inherit_linklocal(idev, link_dev))
2609                         return;
2610         }
2611         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2612 }
2613
2614 /*
2615  * Autoconfigure tunnel with a link-local address so routing protocols,
2616  * DHCPv6, MLD etc. can be run over the virtual link
2617  */
2618
2619 static void addrconf_ip6_tnl_config(struct net_device *dev)
2620 {
2621         struct inet6_dev *idev;
2622
2623         ASSERT_RTNL();
2624
2625         idev = addrconf_add_dev(dev);
2626         if (IS_ERR(idev)) {
2627                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2628                 return;
2629         }
2630         ip6_tnl_add_linklocal(idev);
2631 }
2632
2633 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2634                            void * data)
2635 {
2636         struct net_device *dev = (struct net_device *) data;
2637         struct inet6_dev *idev = __in6_dev_get(dev);
2638         struct net *net = dev_net(dev);
2639         int run_pending = 0;
2640         int err;
2641
2642         switch (event) {
2643         case NETDEV_REGISTER:
2644                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2645                         idev = ipv6_add_dev(dev);
2646                         if (!idev)
2647                                 return notifier_from_errno(-ENOMEM);
2648                 }
2649                 break;
2650
2651         case NETDEV_UP:
2652         case NETDEV_CHANGE:
2653                 if (dev->flags & IFF_SLAVE)
2654                         break;
2655
2656                 if (event == NETDEV_UP) {
2657                         if (!addrconf_qdisc_ok(dev)) {
2658                                 /* device is not ready yet. */
2659                                 printk(KERN_INFO
2660                                         "ADDRCONF(NETDEV_UP): %s: "
2661                                         "link is not ready\n",
2662                                         dev->name);
2663                                 break;
2664                         }
2665
2666                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2667                                 idev = ipv6_add_dev(dev);
2668
2669                         if (idev) {
2670                                 idev->if_flags |= IF_READY;
2671                                 run_pending = 1;
2672                         }
2673                 } else {
2674                         if (!addrconf_qdisc_ok(dev)) {
2675                                 /* device is still not ready. */
2676                                 break;
2677                         }
2678
2679                         if (idev) {
2680                                 if (idev->if_flags & IF_READY)
2681                                         /* device is already configured. */
2682                                         break;
2683                                 idev->if_flags |= IF_READY;
2684                         }
2685
2686                         printk(KERN_INFO
2687                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2688                                         "link becomes ready\n",
2689                                         dev->name);
2690
2691                         run_pending = 1;
2692                 }
2693
2694                 switch (dev->type) {
2695 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2696                 case ARPHRD_SIT:
2697                         addrconf_sit_config(dev);
2698                         break;
2699 #endif
2700 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2701                 case ARPHRD_IPGRE:
2702                         addrconf_gre_config(dev);
2703                         break;
2704 #endif
2705                 case ARPHRD_TUNNEL6:
2706                         addrconf_ip6_tnl_config(dev);
2707                         break;
2708                 case ARPHRD_LOOPBACK:
2709                         init_loopback(dev);
2710                         break;
2711
2712                 default:
2713                         addrconf_dev_config(dev);
2714                         break;
2715                 }
2716
2717                 if (idev) {
2718                         if (run_pending)
2719                                 addrconf_dad_run(idev);
2720
2721                         /*
2722                          * If the MTU changed during the interface down,
2723                          * when the interface up, the changed MTU must be
2724                          * reflected in the idev as well as routers.
2725                          */
2726                         if (idev->cnf.mtu6 != dev->mtu &&
2727                             dev->mtu >= IPV6_MIN_MTU) {
2728                                 rt6_mtu_change(dev, dev->mtu);
2729                                 idev->cnf.mtu6 = dev->mtu;
2730                         }
2731                         idev->tstamp = jiffies;
2732                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2733
2734                         /*
2735                          * If the changed mtu during down is lower than
2736                          * IPV6_MIN_MTU stop IPv6 on this interface.
2737                          */
2738                         if (dev->mtu < IPV6_MIN_MTU)
2739                                 addrconf_ifdown(dev, dev != net->loopback_dev);
2740                 }
2741                 break;
2742
2743         case NETDEV_CHANGEMTU:
2744                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2745                         rt6_mtu_change(dev, dev->mtu);
2746                         idev->cnf.mtu6 = dev->mtu;
2747                         break;
2748                 }
2749
2750                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2751                         idev = ipv6_add_dev(dev);
2752                         if (idev)
2753                                 break;
2754                 }
2755
2756                 /*
2757                  * MTU falled under IPV6_MIN_MTU.
2758                  * Stop IPv6 on this interface.
2759                  */
2760                 addrconf_ifdown(dev, dev != net->loopback_dev);
2761                 break;
2762
2763         case NETDEV_DOWN:
2764         case NETDEV_UNREGISTER:
2765                 /*
2766                  *      Remove all addresses from this interface.
2767                  */
2768                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2769                 break;
2770
2771         case NETDEV_CHANGENAME:
2772                 if (idev) {
2773                         snmp6_unregister_dev(idev);
2774                         addrconf_sysctl_unregister(idev);
2775                         addrconf_sysctl_register(idev);
2776                         err = snmp6_register_dev(idev);
2777                         if (err)
2778                                 return notifier_from_errno(err);
2779                 }
2780                 break;
2781
2782         case NETDEV_PRE_TYPE_CHANGE:
2783         case NETDEV_POST_TYPE_CHANGE:
2784                 addrconf_type_change(dev, event);
2785                 break;
2786         }
2787
2788         return NOTIFY_OK;
2789 }
2790
2791 /*
2792  *      addrconf module should be notified of a device going up
2793  */
2794 static struct notifier_block ipv6_dev_notf = {
2795         .notifier_call = addrconf_notify,
2796 };
2797
2798 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2799 {
2800         struct inet6_dev *idev;
2801         ASSERT_RTNL();
2802
2803         idev = __in6_dev_get(dev);
2804
2805         if (event == NETDEV_POST_TYPE_CHANGE)
2806                 ipv6_mc_remap(idev);
2807         else if (event == NETDEV_PRE_TYPE_CHANGE)
2808                 ipv6_mc_unmap(idev);
2809 }
2810
2811 static int addrconf_ifdown(struct net_device *dev, int how)
2812 {
2813         struct net *net = dev_net(dev);
2814         struct inet6_dev *idev;
2815         struct inet6_ifaddr *ifa;
2816         int state, i;
2817
2818         ASSERT_RTNL();
2819
2820         rt6_ifdown(net, dev);
2821         neigh_ifdown(&nd_tbl, dev);
2822
2823         idev = __in6_dev_get(dev);
2824         if (idev == NULL)
2825                 return -ENODEV;
2826
2827         /*
2828          * Step 1: remove reference to ipv6 device from parent device.
2829          *         Do not dev_put!
2830          */
2831         if (how) {
2832                 idev->dead = 1;
2833
2834                 /* protected by rtnl_lock */
2835                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2836
2837                 /* Step 1.5: remove snmp6 entry */
2838                 snmp6_unregister_dev(idev);
2839
2840         }
2841
2842         /* Step 2: clear hash table */
2843         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2844                 struct hlist_head *h = &inet6_addr_lst[i];
2845                 struct hlist_node *n;
2846
2847                 spin_lock_bh(&addrconf_hash_lock);
2848         restart:
2849                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2850                         if (ifa->idev == idev) {
2851                                 hlist_del_init_rcu(&ifa->addr_lst);
2852                                 addrconf_del_timer(ifa);
2853                                 goto restart;
2854                         }
2855                 }
2856                 spin_unlock_bh(&addrconf_hash_lock);
2857         }
2858
2859         write_lock_bh(&idev->lock);
2860
2861         /* Step 2: clear flags for stateless addrconf */
2862         if (!how)
2863                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2864
2865 #ifdef CONFIG_IPV6_PRIVACY
2866         if (how && del_timer(&idev->regen_timer))
2867                 in6_dev_put(idev);
2868
2869         /* Step 3: clear tempaddr list */
2870         while (!list_empty(&idev->tempaddr_list)) {
2871                 ifa = list_first_entry(&idev->tempaddr_list,
2872                                        struct inet6_ifaddr, tmp_list);
2873                 list_del(&ifa->tmp_list);
2874                 write_unlock_bh(&idev->lock);
2875                 spin_lock_bh(&ifa->lock);
2876
2877                 if (ifa->ifpub) {
2878                         in6_ifa_put(ifa->ifpub);
2879                         ifa->ifpub = NULL;
2880                 }
2881                 spin_unlock_bh(&ifa->lock);
2882                 in6_ifa_put(ifa);
2883                 write_lock_bh(&idev->lock);
2884         }
2885 #endif
2886
2887         while (!list_empty(&idev->addr_list)) {
2888                 ifa = list_first_entry(&idev->addr_list,
2889                                        struct inet6_ifaddr, if_list);
2890                 addrconf_del_timer(ifa);
2891
2892                 list_del(&ifa->if_list);
2893
2894                 write_unlock_bh(&idev->lock);
2895
2896                 spin_lock_bh(&ifa->state_lock);
2897                 state = ifa->state;
2898                 ifa->state = INET6_IFADDR_STATE_DEAD;
2899                 spin_unlock_bh(&ifa->state_lock);
2900
2901                 if (state != INET6_IFADDR_STATE_DEAD) {
2902                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2903                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2904                 }
2905                 in6_ifa_put(ifa);
2906
2907                 write_lock_bh(&idev->lock);
2908         }
2909
2910         write_unlock_bh(&idev->lock);
2911
2912         /* Step 5: Discard multicast list */
2913         if (how)
2914                 ipv6_mc_destroy_dev(idev);
2915         else
2916                 ipv6_mc_down(idev);
2917
2918         idev->tstamp = jiffies;
2919
2920         /* Last: Shot the device (if unregistered) */
2921         if (how) {
2922                 addrconf_sysctl_unregister(idev);
2923                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2924                 neigh_ifdown(&nd_tbl, dev);
2925                 in6_dev_put(idev);
2926         }
2927         return 0;
2928 }
2929
2930 static void addrconf_rs_timer(unsigned long data)
2931 {
2932         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2933         struct inet6_dev *idev = ifp->idev;
2934
2935         read_lock(&idev->lock);
2936         if (idev->dead || !(idev->if_flags & IF_READY))
2937                 goto out;
2938
2939         if (idev->cnf.forwarding)
2940                 goto out;
2941
2942         /* Announcement received after solicitation was sent */
2943         if (idev->if_flags & IF_RA_RCVD)
2944                 goto out;
2945
2946         spin_lock(&ifp->lock);
2947         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2948                 /* The wait after the last probe can be shorter */
2949                 addrconf_mod_timer(ifp, AC_RS,
2950                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2951                                    idev->cnf.rtr_solicit_delay :
2952                                    idev->cnf.rtr_solicit_interval);
2953                 spin_unlock(&ifp->lock);
2954
2955                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2956         } else {
2957                 spin_unlock(&ifp->lock);
2958                 /*
2959                  * Note: we do not support deprecated "all on-link"
2960                  * assumption any longer.
2961                  */
2962                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2963                        idev->dev->name);
2964         }
2965
2966 out:
2967         read_unlock(&idev->lock);
2968         in6_ifa_put(ifp);
2969 }
2970
2971 /*
2972  *      Duplicate Address Detection
2973  */
2974 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2975 {
2976         unsigned long rand_num;
2977         struct inet6_dev *idev = ifp->idev;
2978
2979         if (ifp->flags & IFA_F_OPTIMISTIC)
2980                 rand_num = 0;
2981         else
2982                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2983
2984         ifp->probes = idev->cnf.dad_transmits;
2985         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2986 }
2987
2988 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2989 {
2990         struct inet6_dev *idev = ifp->idev;
2991         struct net_device *dev = idev->dev;
2992
2993         addrconf_join_solict(dev, &ifp->addr);
2994
2995         net_srandom(ifp->addr.s6_addr32[3]);
2996
2997         read_lock_bh(&idev->lock);
2998         spin_lock(&ifp->lock);
2999         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3000                 goto out;
3001
3002         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3003             idev->cnf.accept_dad < 1 ||
3004             !(ifp->flags&IFA_F_TENTATIVE) ||
3005             ifp->flags & IFA_F_NODAD) {
3006                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3007                 spin_unlock(&ifp->lock);
3008                 read_unlock_bh(&idev->lock);
3009
3010                 addrconf_dad_completed(ifp);
3011                 return;
3012         }
3013
3014         if (!(idev->if_flags & IF_READY)) {
3015                 spin_unlock(&ifp->lock);
3016                 read_unlock_bh(&idev->lock);
3017                 /*
3018                  * If the device is not ready:
3019                  * - keep it tentative if it is a permanent address.
3020                  * - otherwise, kill it.
3021                  */
3022                 in6_ifa_hold(ifp);
3023                 addrconf_dad_stop(ifp, 0);
3024                 return;
3025         }
3026
3027         /*
3028          * Optimistic nodes can start receiving
3029          * Frames right away
3030          */
3031         if (ifp->flags & IFA_F_OPTIMISTIC)
3032                 ip6_ins_rt(ifp->rt);
3033
3034         addrconf_dad_kick(ifp);
3035 out:
3036         spin_unlock(&ifp->lock);
3037         read_unlock_bh(&idev->lock);
3038 }
3039
3040 static void addrconf_dad_timer(unsigned long data)
3041 {
3042         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3043         struct inet6_dev *idev = ifp->idev;
3044         struct in6_addr mcaddr;
3045
3046         if (!ifp->probes && addrconf_dad_end(ifp))
3047                 goto out;
3048
3049         read_lock(&idev->lock);
3050         if (idev->dead || !(idev->if_flags & IF_READY)) {
3051                 read_unlock(&idev->lock);
3052                 goto out;
3053         }
3054
3055         spin_lock(&ifp->lock);
3056         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3057                 spin_unlock(&ifp->lock);
3058                 read_unlock(&idev->lock);
3059                 goto out;
3060         }
3061
3062         if (ifp->probes == 0) {
3063                 /*
3064                  * DAD was successful
3065                  */
3066
3067                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3068                 spin_unlock(&ifp->lock);
3069                 read_unlock(&idev->lock);
3070
3071                 addrconf_dad_completed(ifp);
3072
3073                 goto out;
3074         }
3075
3076         ifp->probes--;
3077         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3078         spin_unlock(&ifp->lock);
3079         read_unlock(&idev->lock);
3080
3081         /* send a neighbour solicitation for our addr */
3082         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3083         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3084 out:
3085         in6_ifa_put(ifp);
3086 }
3087
3088 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3089 {
3090         struct net_device *dev = ifp->idev->dev;
3091
3092         /*
3093          *      Configure the address for reception. Now it is valid.
3094          */
3095
3096         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3097
3098         /* If added prefix is link local and we are prepared to process
3099            router advertisements, start sending router solicitations.
3100          */
3101
3102         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3103              ifp->idev->cnf.accept_ra == 2) &&
3104             ifp->idev->cnf.rtr_solicits > 0 &&
3105             (dev->flags&IFF_LOOPBACK) == 0 &&
3106             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3107                 /*
3108                  *      If a host as already performed a random delay
3109                  *      [...] as part of DAD [...] there is no need
3110                  *      to delay again before sending the first RS
3111                  */
3112                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3113
3114                 spin_lock_bh(&ifp->lock);
3115                 ifp->probes = 1;
3116                 ifp->idev->if_flags |= IF_RS_SENT;
3117                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3118                 spin_unlock_bh(&ifp->lock);
3119         }
3120 }
3121
3122 static void addrconf_dad_run(struct inet6_dev *idev)
3123 {
3124         struct inet6_ifaddr *ifp;
3125
3126         read_lock_bh(&idev->lock);
3127         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3128                 spin_lock(&ifp->lock);
3129                 if (ifp->flags & IFA_F_TENTATIVE &&
3130                     ifp->state == INET6_IFADDR_STATE_DAD)
3131                         addrconf_dad_kick(ifp);
3132                 spin_unlock(&ifp->lock);
3133         }
3134         read_unlock_bh(&idev->lock);
3135 }
3136
3137 #ifdef CONFIG_PROC_FS
3138 struct if6_iter_state {
3139         struct seq_net_private p;
3140         int bucket;
3141 };
3142
3143 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3144 {
3145         struct inet6_ifaddr *ifa = NULL;
3146         struct if6_iter_state *state = seq->private;
3147         struct net *net = seq_file_net(seq);
3148
3149         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3150                 struct hlist_node *n;
3151                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3152                                          addr_lst)
3153                         if (net_eq(dev_net(ifa->idev->dev), net))
3154                                 return ifa;
3155         }
3156         return NULL;
3157 }
3158
3159 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3160                                          struct inet6_ifaddr *ifa)
3161 {
3162         struct if6_iter_state *state = seq->private;
3163         struct net *net = seq_file_net(seq);
3164         struct hlist_node *n = &ifa->addr_lst;
3165
3166         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3167                 if (net_eq(dev_net(ifa->idev->dev), net))
3168                         return ifa;
3169
3170         while (++state->bucket < IN6_ADDR_HSIZE) {
3171                 hlist_for_each_entry_rcu_bh(ifa, n,
3172                                      &inet6_addr_lst[state->bucket], addr_lst) {
3173                         if (net_eq(dev_net(ifa->idev->dev), net))
3174                                 return ifa;
3175                 }
3176         }
3177
3178         return NULL;
3179 }
3180
3181 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3182 {
3183         struct inet6_ifaddr *ifa = if6_get_first(seq);
3184
3185         if (ifa)
3186                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3187                         --pos;
3188         return pos ? NULL : ifa;
3189 }
3190
3191 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3192         __acquires(rcu_bh)
3193 {
3194         rcu_read_lock_bh();
3195         return if6_get_idx(seq, *pos);
3196 }
3197
3198 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3199 {
3200         struct inet6_ifaddr *ifa;
3201
3202         ifa = if6_get_next(seq, v);
3203         ++*pos;
3204         return ifa;
3205 }
3206
3207 static void if6_seq_stop(struct seq_file *seq, void *v)
3208         __releases(rcu_bh)
3209 {
3210         rcu_read_unlock_bh();
3211 }
3212
3213 static int if6_seq_show(struct seq_file *seq, void *v)
3214 {
3215         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3216         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3217                    &ifp->addr,
3218                    ifp->idev->dev->ifindex,
3219                    ifp->prefix_len,
3220                    ifp->scope,
3221                    ifp->flags,
3222                    ifp->idev->dev->name);
3223         return 0;
3224 }
3225
3226 static const struct seq_operations if6_seq_ops = {
3227         .start  = if6_seq_start,
3228         .next   = if6_seq_next,
3229         .show   = if6_seq_show,
3230         .stop   = if6_seq_stop,
3231 };
3232
3233 static int if6_seq_open(struct inode *inode, struct file *file)
3234 {
3235         return seq_open_net(inode, file, &if6_seq_ops,
3236                             sizeof(struct if6_iter_state));
3237 }
3238
3239 static const struct file_operations if6_fops = {
3240         .owner          = THIS_MODULE,
3241         .open           = if6_seq_open,
3242         .read           = seq_read,
3243         .llseek         = seq_lseek,
3244         .release        = seq_release_net,
3245 };
3246
3247 static int __net_init if6_proc_net_init(struct net *net)
3248 {
3249         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3250                 return -ENOMEM;
3251         return 0;
3252 }
3253
3254 static void __net_exit if6_proc_net_exit(struct net *net)
3255 {
3256        proc_net_remove(net, "if_inet6");
3257 }
3258
3259 static struct pernet_operations if6_proc_net_ops = {
3260        .init = if6_proc_net_init,
3261        .exit = if6_proc_net_exit,
3262 };
3263
3264 int __init if6_proc_init(void)
3265 {
3266         return register_pernet_subsys(&if6_proc_net_ops);
3267 }
3268
3269 void if6_proc_exit(void)
3270 {
3271         unregister_pernet_subsys(&if6_proc_net_ops);
3272 }
3273 #endif  /* CONFIG_PROC_FS */
3274
3275 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3276 /* Check if address is a home address configured on any interface. */
3277 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3278 {
3279         int ret = 0;
3280         struct inet6_ifaddr *ifp = NULL;
3281         struct hlist_node *n;
3282         unsigned int hash = ipv6_addr_hash(addr);
3283
3284         rcu_read_lock_bh();
3285         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3286                 if (!net_eq(dev_net(ifp->idev->dev), net))
3287                         continue;
3288                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3289                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3290                         ret = 1;
3291                         break;
3292                 }
3293         }
3294         rcu_read_unlock_bh();
3295         return ret;
3296 }
3297 #endif
3298
3299 /*
3300  *      Periodic address status verification
3301  */
3302
3303 static void addrconf_verify(unsigned long foo)
3304 {
3305         unsigned long now, next, next_sec, next_sched;
3306         struct inet6_ifaddr *ifp;
3307         struct hlist_node *node;
3308         int i;
3309
3310         rcu_read_lock_bh();
3311         spin_lock(&addrconf_verify_lock);
3312         now = jiffies;
3313         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3314
3315         del_timer(&addr_chk_timer);
3316
3317         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3318 restart:
3319                 hlist_for_each_entry_rcu_bh(ifp, node,
3320                                          &inet6_addr_lst[i], addr_lst) {
3321                         unsigned long age;
3322
3323                         if (ifp->flags & IFA_F_PERMANENT)
3324                                 continue;
3325
3326                         spin_lock(&ifp->lock);
3327                         /* We try to batch several events at once. */
3328                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3329
3330                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3331                             age >= ifp->valid_lft) {
3332                                 spin_unlock(&ifp->lock);
3333                                 in6_ifa_hold(ifp);
3334                                 ipv6_del_addr(ifp);
3335                                 goto restart;
3336                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3337                                 spin_unlock(&ifp->lock);
3338                                 continue;
3339                         } else if (age >= ifp->prefered_lft) {
3340                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3341                                 int deprecate = 0;
3342
3343                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3344                                         deprecate = 1;
3345                                         ifp->flags |= IFA_F_DEPRECATED;
3346                                 }
3347
3348                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3349                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3350
3351                                 spin_unlock(&ifp->lock);
3352
3353                                 if (deprecate) {
3354                                         in6_ifa_hold(ifp);
3355
3356                                         ipv6_ifa_notify(0, ifp);
3357                                         in6_ifa_put(ifp);
3358                                         goto restart;
3359                                 }
3360 #ifdef CONFIG_IPV6_PRIVACY
3361                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3362                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3363                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3364                                         ifp->idev->cnf.dad_transmits *
3365                                         ifp->idev->nd_parms->retrans_time / HZ;
3366
3367                                 if (age >= ifp->prefered_lft - regen_advance) {
3368                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3369                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3370                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3371                                         if (!ifp->regen_count && ifpub) {
3372                                                 ifp->regen_count++;
3373                                                 in6_ifa_hold(ifp);
3374                                                 in6_ifa_hold(ifpub);
3375                                                 spin_unlock(&ifp->lock);
3376
3377                                                 spin_lock(&ifpub->lock);
3378                                                 ifpub->regen_count = 0;
3379                                                 spin_unlock(&ifpub->lock);
3380                                                 ipv6_create_tempaddr(ifpub, ifp);
3381                                                 in6_ifa_put(ifpub);
3382                                                 in6_ifa_put(ifp);
3383                                                 goto restart;
3384                                         }
3385                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3386                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3387                                 spin_unlock(&ifp->lock);
3388 #endif
3389                         } else {
3390                                 /* ifp->prefered_lft <= ifp->valid_lft */
3391                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3392                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3393                                 spin_unlock(&ifp->lock);
3394                         }
3395                 }
3396         }
3397
3398         next_sec = round_jiffies_up(next);
3399         next_sched = next;
3400
3401         /* If rounded timeout is accurate enough, accept it. */
3402         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3403                 next_sched = next_sec;
3404
3405         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3406         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3407                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3408
3409         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3410               now, next, next_sec, next_sched));
3411
3412         addr_chk_timer.expires = next_sched;
3413         add_timer(&addr_chk_timer);
3414         spin_unlock(&addrconf_verify_lock);
3415         rcu_read_unlock_bh();
3416 }
3417
3418 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3419 {
3420         struct in6_addr *pfx = NULL;
3421
3422         if (addr)
3423                 pfx = nla_data(addr);
3424
3425         if (local) {
3426                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3427                         pfx = NULL;
3428                 else
3429                         pfx = nla_data(local);
3430         }
3431
3432         return pfx;
3433 }
3434
3435 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3436         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3437         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3438         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3439 };
3440
3441 static int
3442 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3443 {
3444         struct net *net = sock_net(skb->sk);
3445         struct ifaddrmsg *ifm;
3446         struct nlattr *tb[IFA_MAX+1];
3447         struct in6_addr *pfx;
3448         int err;
3449
3450         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3451         if (err < 0)
3452                 return err;
3453
3454         ifm = nlmsg_data(nlh);
3455         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3456         if (pfx == NULL)
3457                 return -EINVAL;
3458
3459         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3460 }
3461
3462 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3463                              u32 prefered_lft, u32 valid_lft)
3464 {
3465         u32 flags;
3466         clock_t expires;
3467         unsigned long timeout;
3468
3469         if (!valid_lft || (prefered_lft > valid_lft))
3470                 return -EINVAL;
3471
3472         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3473         if (addrconf_finite_timeout(timeout)) {
3474                 expires = jiffies_to_clock_t(timeout * HZ);
3475                 valid_lft = timeout;
3476                 flags = RTF_EXPIRES;
3477         } else {
3478                 expires = 0;
3479                 flags = 0;
3480                 ifa_flags |= IFA_F_PERMANENT;
3481         }
3482
3483         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3484         if (addrconf_finite_timeout(timeout)) {
3485                 if (timeout == 0)
3486                         ifa_flags |= IFA_F_DEPRECATED;
3487                 prefered_lft = timeout;
3488         }
3489
3490         spin_lock_bh(&ifp->lock);
3491         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3492         ifp->tstamp = jiffies;
3493         ifp->valid_lft = valid_lft;
3494         ifp->prefered_lft = prefered_lft;
3495
3496         spin_unlock_bh(&ifp->lock);
3497         if (!(ifp->flags&IFA_F_TENTATIVE))
3498                 ipv6_ifa_notify(0, ifp);
3499
3500         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3501                               expires, flags);
3502         addrconf_verify(0);
3503
3504         return 0;
3505 }
3506
3507 static int
3508 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3509 {
3510         struct net *net = sock_net(skb->sk);
3511         struct ifaddrmsg *ifm;
3512         struct nlattr *tb[IFA_MAX+1];
3513         struct in6_addr *pfx;
3514         struct inet6_ifaddr *ifa;
3515         struct net_device *dev;
3516         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3517         u8 ifa_flags;
3518         int err;
3519
3520         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3521         if (err < 0)
3522                 return err;
3523
3524         ifm = nlmsg_data(nlh);
3525         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3526         if (pfx == NULL)
3527                 return -EINVAL;
3528
3529         if (tb[IFA_CACHEINFO]) {
3530                 struct ifa_cacheinfo *ci;
3531
3532                 ci = nla_data(tb[IFA_CACHEINFO]);
3533                 valid_lft = ci->ifa_valid;
3534                 preferred_lft = ci->ifa_prefered;
3535         } else {
3536                 preferred_lft = INFINITY_LIFE_TIME;
3537                 valid_lft = INFINITY_LIFE_TIME;
3538         }
3539
3540         dev =  __dev_get_by_index(net, ifm->ifa_index);
3541         if (dev == NULL)
3542                 return -ENODEV;
3543
3544         /* We ignore other flags so far. */
3545         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3546
3547         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3548         if (ifa == NULL) {
3549                 /*
3550                  * It would be best to check for !NLM_F_CREATE here but
3551                  * userspace alreay relies on not having to provide this.
3552                  */
3553                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3554                                       ifm->ifa_prefixlen, ifa_flags,
3555                                       preferred_lft, valid_lft);
3556         }
3557
3558         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3559             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3560                 err = -EEXIST;
3561         else
3562                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3563
3564         in6_ifa_put(ifa);
3565
3566         return err;
3567 }
3568
3569 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3570                           u8 scope, int ifindex)
3571 {
3572         struct ifaddrmsg *ifm;
3573
3574         ifm = nlmsg_data(nlh);
3575         ifm->ifa_family = AF_INET6;
3576         ifm->ifa_prefixlen = prefixlen;
3577         ifm->ifa_flags = flags;
3578         ifm->ifa_scope = scope;
3579         ifm->ifa_index = ifindex;
3580 }
3581
3582 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3583                          unsigned long tstamp, u32 preferred, u32 valid)
3584 {
3585         struct ifa_cacheinfo ci;
3586
3587         ci.cstamp = cstamp_delta(cstamp);
3588         ci.tstamp = cstamp_delta(tstamp);
3589         ci.ifa_prefered = preferred;
3590         ci.ifa_valid = valid;
3591
3592         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3593 }
3594
3595 static inline int rt_scope(int ifa_scope)
3596 {
3597         if (ifa_scope & IFA_HOST)
3598                 return RT_SCOPE_HOST;
3599         else if (ifa_scope & IFA_LINK)
3600                 return RT_SCOPE_LINK;
3601         else if (ifa_scope & IFA_SITE)
3602                 return RT_SCOPE_SITE;
3603         else
3604                 return RT_SCOPE_UNIVERSE;
3605 }
3606
3607 static inline int inet6_ifaddr_msgsize(void)
3608 {
3609         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3610                + nla_total_size(16) /* IFA_ADDRESS */
3611                + nla_total_size(sizeof(struct ifa_cacheinfo));
3612 }
3613
3614 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3615                              u32 pid, u32 seq, int event, unsigned int flags)
3616 {
3617         struct nlmsghdr  *nlh;
3618         u32 preferred, valid;
3619
3620         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3621         if (nlh == NULL)
3622                 return -EMSGSIZE;
3623
3624         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3625                       ifa->idev->dev->ifindex);
3626
3627         if (!(ifa->flags&IFA_F_PERMANENT)) {
3628                 preferred = ifa->prefered_lft;
3629                 valid = ifa->valid_lft;
3630                 if (preferred != INFINITY_LIFE_TIME) {
3631                         long tval = (jiffies - ifa->tstamp)/HZ;
3632                         if (preferred > tval)
3633                                 preferred -= tval;
3634                         else
3635                                 preferred = 0;
3636                         if (valid != INFINITY_LIFE_TIME) {
3637                                 if (valid > tval)
3638                                         valid -= tval;
3639                                 else
3640                                         valid = 0;
3641                         }
3642                 }
3643         } else {
3644                 preferred = INFINITY_LIFE_TIME;
3645                 valid = INFINITY_LIFE_TIME;
3646         }
3647
3648         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3649             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3650                 nlmsg_cancel(skb, nlh);
3651                 return -EMSGSIZE;
3652         }
3653
3654         return nlmsg_end(skb, nlh);
3655 }
3656
3657 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3658                                 u32 pid, u32 seq, int event, u16 flags)
3659 {
3660         struct nlmsghdr  *nlh;
3661         u8 scope = RT_SCOPE_UNIVERSE;
3662         int ifindex = ifmca->idev->dev->ifindex;
3663
3664         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3665                 scope = RT_SCOPE_SITE;
3666
3667         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3668         if (nlh == NULL)
3669                 return -EMSGSIZE;
3670
3671         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3672         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3673             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3674                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3675                 nlmsg_cancel(skb, nlh);
3676                 return -EMSGSIZE;
3677         }
3678
3679         return nlmsg_end(skb, nlh);
3680 }
3681
3682 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3683                                 u32 pid, u32 seq, int event, unsigned int flags)
3684 {
3685         struct nlmsghdr  *nlh;
3686         u8 scope = RT_SCOPE_UNIVERSE;
3687         int ifindex = ifaca->aca_idev->dev->ifindex;
3688
3689         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3690                 scope = RT_SCOPE_SITE;
3691
3692         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3693         if (nlh == NULL)
3694                 return -EMSGSIZE;
3695
3696         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3697         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3698             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3699                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3700                 nlmsg_cancel(skb, nlh);
3701                 return -EMSGSIZE;
3702         }
3703
3704         return nlmsg_end(skb, nlh);
3705 }
3706
3707 enum addr_type_t {
3708         UNICAST_ADDR,
3709         MULTICAST_ADDR,
3710         ANYCAST_ADDR,
3711 };
3712
3713 /* called with rcu_read_lock() */
3714 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3715                           struct netlink_callback *cb, enum addr_type_t type,
3716                           int s_ip_idx, int *p_ip_idx)
3717 {
3718         struct ifmcaddr6 *ifmca;
3719         struct ifacaddr6 *ifaca;
3720         int err = 1;
3721         int ip_idx = *p_ip_idx;
3722
3723         read_lock_bh(&idev->lock);
3724         switch (type) {
3725         case UNICAST_ADDR: {
3726                 struct inet6_ifaddr *ifa;
3727
3728                 /* unicast address incl. temp addr */
3729                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3730                         if (++ip_idx < s_ip_idx)
3731                                 continue;
3732                         err = inet6_fill_ifaddr(skb, ifa,
3733                                                 NETLINK_CB(cb->skb).pid,
3734                                                 cb->nlh->nlmsg_seq,
3735                                                 RTM_NEWADDR,
3736                                                 NLM_F_MULTI);
3737                         if (err <= 0)
3738                                 break;
3739                 }
3740                 break;
3741         }
3742         case MULTICAST_ADDR:
3743                 /* multicast address */
3744                 for (ifmca = idev->mc_list; ifmca;
3745                      ifmca = ifmca->next, ip_idx++) {
3746                         if (ip_idx < s_ip_idx)
3747                                 continue;
3748                         err = inet6_fill_ifmcaddr(skb, ifmca,
3749                                                   NETLINK_CB(cb->skb).pid,
3750                                                   cb->nlh->nlmsg_seq,
3751                                                   RTM_GETMULTICAST,
3752                                                   NLM_F_MULTI);
3753                         if (err <= 0)
3754                                 break;
3755                 }
3756                 break;
3757         case ANYCAST_ADDR:
3758                 /* anycast address */
3759                 for (ifaca = idev->ac_list; ifaca;
3760                      ifaca = ifaca->aca_next, ip_idx++) {
3761                         if (ip_idx < s_ip_idx)
3762                                 continue;
3763                         err = inet6_fill_ifacaddr(skb, ifaca,
3764                                                   NETLINK_CB(cb->skb).pid,
3765                                                   cb->nlh->nlmsg_seq,
3766                                                   RTM_GETANYCAST,
3767                                                   NLM_F_MULTI);
3768                         if (err <= 0)
3769                                 break;
3770                 }
3771                 break;
3772         default:
3773                 break;
3774         }
3775         read_unlock_bh(&idev->lock);
3776         *p_ip_idx = ip_idx;
3777         return err;
3778 }
3779
3780 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3781                            enum addr_type_t type)
3782 {
3783         struct net *net = sock_net(skb->sk);
3784         int h, s_h;
3785         int idx, ip_idx;
3786         int s_idx, s_ip_idx;
3787         struct net_device *dev;
3788         struct inet6_dev *idev;
3789         struct hlist_head *head;
3790         struct hlist_node *node;
3791
3792         s_h = cb->args[0];
3793         s_idx = idx = cb->args[1];
3794         s_ip_idx = ip_idx = cb->args[2];
3795
3796         rcu_read_lock();
3797         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3798                 idx = 0;
3799                 head = &net->dev_index_head[h];
3800                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3801                         if (idx < s_idx)
3802                                 goto cont;
3803                         if (h > s_h || idx > s_idx)
3804                                 s_ip_idx = 0;
3805                         ip_idx = 0;
3806                         idev = __in6_dev_get(dev);
3807                         if (!idev)
3808                                 goto cont;
3809
3810                         if (in6_dump_addrs(idev, skb, cb, type,
3811                                            s_ip_idx, &ip_idx) <= 0)
3812                                 goto done;
3813 cont:
3814                         idx++;
3815                 }
3816         }
3817 done:
3818         rcu_read_unlock();
3819         cb->args[0] = h;
3820         cb->args[1] = idx;
3821         cb->args[2] = ip_idx;
3822
3823         return skb->len;
3824 }
3825
3826 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3827 {
3828         enum addr_type_t type = UNICAST_ADDR;
3829
3830         return inet6_dump_addr(skb, cb, type);
3831 }
3832
3833 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3834 {
3835         enum addr_type_t type = MULTICAST_ADDR;
3836
3837         return inet6_dump_addr(skb, cb, type);
3838 }
3839
3840
3841 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3842 {
3843         enum addr_type_t type = ANYCAST_ADDR;
3844
3845         return inet6_dump_addr(skb, cb, type);
3846 }
3847
3848 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3849                              void *arg)
3850 {
3851         struct net *net = sock_net(in_skb->sk);
3852         struct ifaddrmsg *ifm;
3853         struct nlattr *tb[IFA_MAX+1];
3854         struct in6_addr *addr = NULL;
3855         struct net_device *dev = NULL;
3856         struct inet6_ifaddr *ifa;
3857         struct sk_buff *skb;
3858         int err;
3859
3860         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3861         if (err < 0)
3862                 goto errout;
3863
3864         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3865         if (addr == NULL) {
3866                 err = -EINVAL;
3867                 goto errout;
3868         }
3869
3870         ifm = nlmsg_data(nlh);
3871         if (ifm->ifa_index)
3872                 dev = __dev_get_by_index(net, ifm->ifa_index);
3873
3874         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3875         if (!ifa) {
3876                 err = -EADDRNOTAVAIL;
3877                 goto errout;
3878         }
3879
3880         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3881         if (!skb) {
3882                 err = -ENOBUFS;
3883                 goto errout_ifa;
3884         }
3885
3886         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3887                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3888         if (err < 0) {
3889                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3890                 WARN_ON(err == -EMSGSIZE);
3891                 kfree_skb(skb);
3892                 goto errout_ifa;
3893         }
3894         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3895 errout_ifa:
3896         in6_ifa_put(ifa);
3897 errout:
3898         return err;
3899 }
3900
3901 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3902 {
3903         struct sk_buff *skb;
3904         struct net *net = dev_net(ifa->idev->dev);
3905         int err = -ENOBUFS;
3906
3907         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3908         if (skb == NULL)
3909                 goto errout;
3910
3911         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3912         if (err < 0) {
3913                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3914                 WARN_ON(err == -EMSGSIZE);
3915                 kfree_skb(skb);
3916                 goto errout;
3917         }
3918         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3919         return;
3920 errout:
3921         if (err < 0)
3922                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3923 }
3924
3925 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3926                                 __s32 *array, int bytes)
3927 {
3928         BUG_ON(bytes < (DEVCONF_MAX * 4));
3929
3930         memset(array, 0, bytes);
3931         array[DEVCONF_FORWARDING] = cnf->forwarding;
3932         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3933         array[DEVCONF_MTU6] = cnf->mtu6;
3934         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3935         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3936         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3937         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3938         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3939         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3940                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3941         array[DEVCONF_RTR_SOLICIT_DELAY] =
3942                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3943         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3944 #ifdef CONFIG_IPV6_PRIVACY
3945         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3946         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3947         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3948         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3949         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3950 #endif
3951         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3952         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3953         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
3954         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3955 #ifdef CONFIG_IPV6_ROUTER_PREF
3956         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3957         array[DEVCONF_RTR_PROBE_INTERVAL] =
3958                 jiffies_to_msecs(cnf->rtr_probe_interval);
3959 #ifdef CONFIG_IPV6_ROUTE_INFO
3960         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3961 #endif
3962 #endif
3963         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3964         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3965 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3966         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3967 #endif
3968 #ifdef CONFIG_IPV6_MROUTE
3969         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3970 #endif
3971         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3972         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3973         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3974 }
3975
3976 static inline size_t inet6_ifla6_size(void)
3977 {
3978         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3979              + nla_total_size(sizeof(struct ifla_cacheinfo))
3980              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3981              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3982              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3983 }
3984
3985 static inline size_t inet6_if_nlmsg_size(void)
3986 {
3987         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3988                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3989                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3990                + nla_total_size(4) /* IFLA_MTU */
3991                + nla_total_size(4) /* IFLA_LINK */
3992                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3993 }
3994
3995 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3996                                       int items, int bytes)
3997 {
3998         int i;
3999         int pad = bytes - sizeof(u64) * items;
4000         BUG_ON(pad < 0);
4001
4002         /* Use put_unaligned() because stats may not be aligned for u64. */
4003         put_unaligned(items, &stats[0]);
4004         for (i = 1; i < items; i++)
4005                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4006
4007         memset(&stats[items], 0, pad);
4008 }
4009
4010 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4011                                       int items, int bytes, size_t syncpoff)
4012 {
4013         int i;
4014         int pad = bytes - sizeof(u64) * items;
4015         BUG_ON(pad < 0);
4016
4017         /* Use put_unaligned() because stats may not be aligned for u64. */
4018         put_unaligned(items, &stats[0]);
4019         for (i = 1; i < items; i++)
4020                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4021
4022         memset(&stats[items], 0, pad);
4023 }
4024
4025 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4026                              int bytes)
4027 {
4028         switch (attrtype) {
4029         case IFLA_INET6_STATS:
4030                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4031                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4032                 break;
4033         case IFLA_INET6_ICMP6STATS:
4034                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4035                 break;
4036         }
4037 }
4038
4039 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4040 {
4041         struct nlattr *nla;
4042         struct ifla_cacheinfo ci;
4043
4044         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
4045
4046         ci.max_reasm_len = IPV6_MAXPLEN;
4047         ci.tstamp = cstamp_delta(idev->tstamp);
4048         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4049         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4050         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
4051
4052         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4053         if (nla == NULL)
4054                 goto nla_put_failure;
4055         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4056
4057         /* XXX - MC not implemented */
4058
4059         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4060         if (nla == NULL)
4061                 goto nla_put_failure;
4062         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4063
4064         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4065         if (nla == NULL)
4066                 goto nla_put_failure;
4067         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4068
4069         return 0;
4070
4071 nla_put_failure:
4072         return -EMSGSIZE;
4073 }
4074
4075 static size_t inet6_get_link_af_size(const struct net_device *dev)
4076 {
4077         if (!__in6_dev_get(dev))
4078                 return 0;
4079
4080         return inet6_ifla6_size();
4081 }
4082
4083 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4084 {
4085         struct inet6_dev *idev = __in6_dev_get(dev);
4086
4087         if (!idev)
4088                 return -ENODATA;
4089
4090         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4091                 return -EMSGSIZE;
4092
4093         return 0;
4094 }
4095
4096 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4097                              u32 pid, u32 seq, int event, unsigned int flags)
4098 {
4099         struct net_device *dev = idev->dev;
4100         struct ifinfomsg *hdr;
4101         struct nlmsghdr *nlh;
4102         void *protoinfo;
4103
4104         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4105         if (nlh == NULL)
4106                 return -EMSGSIZE;
4107
4108         hdr = nlmsg_data(nlh);
4109         hdr->ifi_family = AF_INET6;
4110         hdr->__ifi_pad = 0;
4111         hdr->ifi_type = dev->type;
4112         hdr->ifi_index = dev->ifindex;
4113         hdr->ifi_flags = dev_get_flags(dev);
4114         hdr->ifi_change = 0;
4115
4116         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4117
4118         if (dev->addr_len)
4119                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4120
4121         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4122         if (dev->ifindex != dev->iflink)
4123                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4124
4125         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4126         if (protoinfo == NULL)
4127                 goto nla_put_failure;
4128
4129         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4130                 goto nla_put_failure;
4131
4132         nla_nest_end(skb, protoinfo);
4133         return nlmsg_end(skb, nlh);
4134
4135 nla_put_failure:
4136         nlmsg_cancel(skb, nlh);
4137         return -EMSGSIZE;
4138 }
4139
4140 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4141 {
4142         struct net *net = sock_net(skb->sk);
4143         int h, s_h;
4144         int idx = 0, s_idx;
4145         struct net_device *dev;
4146         struct inet6_dev *idev;
4147         struct hlist_head *head;
4148         struct hlist_node *node;
4149
4150         s_h = cb->args[0];
4151         s_idx = cb->args[1];
4152
4153         rcu_read_lock();
4154         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4155                 idx = 0;
4156                 head = &net->dev_index_head[h];
4157                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4158                         if (idx < s_idx)
4159                                 goto cont;
4160                         idev = __in6_dev_get(dev);
4161                         if (!idev)
4162                                 goto cont;
4163                         if (inet6_fill_ifinfo(skb, idev,
4164                                               NETLINK_CB(cb->skb).pid,
4165                                               cb->nlh->nlmsg_seq,
4166                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4167                                 goto out;
4168 cont:
4169                         idx++;
4170                 }
4171         }
4172 out:
4173         rcu_read_unlock();
4174         cb->args[1] = idx;
4175         cb->args[0] = h;
4176
4177         return skb->len;
4178 }
4179
4180 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4181 {
4182         struct sk_buff *skb;
4183         struct net *net = dev_net(idev->dev);
4184         int err = -ENOBUFS;
4185
4186         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4187         if (skb == NULL)
4188                 goto errout;
4189
4190         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4191         if (err < 0) {
4192                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4193                 WARN_ON(err == -EMSGSIZE);
4194                 kfree_skb(skb);
4195                 goto errout;
4196         }
4197         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4198         return;
4199 errout:
4200         if (err < 0)
4201                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4202 }
4203
4204 static inline size_t inet6_prefix_nlmsg_size(void)
4205 {
4206         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4207                + nla_total_size(sizeof(struct in6_addr))
4208                + nla_total_size(sizeof(struct prefix_cacheinfo));
4209 }
4210
4211 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4212                              struct prefix_info *pinfo, u32 pid, u32 seq,
4213                              int event, unsigned int flags)
4214 {
4215         struct prefixmsg *pmsg;
4216         struct nlmsghdr *nlh;
4217         struct prefix_cacheinfo ci;
4218
4219         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4220         if (nlh == NULL)
4221                 return -EMSGSIZE;
4222
4223         pmsg = nlmsg_data(nlh);
4224         pmsg->prefix_family = AF_INET6;
4225         pmsg->prefix_pad1 = 0;
4226         pmsg->prefix_pad2 = 0;
4227         pmsg->prefix_ifindex = idev->dev->ifindex;
4228         pmsg->prefix_len = pinfo->prefix_len;
4229         pmsg->prefix_type = pinfo->type;
4230         pmsg->prefix_pad3 = 0;
4231         pmsg->prefix_flags = 0;
4232         if (pinfo->onlink)
4233                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4234         if (pinfo->autoconf)
4235                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4236
4237         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4238
4239         ci.preferred_time = ntohl(pinfo->prefered);
4240         ci.valid_time = ntohl(pinfo->valid);
4241         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4242
4243         return nlmsg_end(skb, nlh);
4244
4245 nla_put_failure:
4246         nlmsg_cancel(skb, nlh);
4247         return -EMSGSIZE;
4248 }
4249
4250 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4251                          struct prefix_info *pinfo)
4252 {
4253         struct sk_buff *skb;
4254         struct net *net = dev_net(idev->dev);
4255         int err = -ENOBUFS;
4256
4257         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4258         if (skb == NULL)
4259                 goto errout;
4260
4261         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4262         if (err < 0) {
4263                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4264                 WARN_ON(err == -EMSGSIZE);
4265                 kfree_skb(skb);
4266                 goto errout;
4267         }
4268         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4269         return;
4270 errout:
4271         if (err < 0)
4272                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4273 }
4274
4275 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4276 {
4277         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4278
4279         switch (event) {
4280         case RTM_NEWADDR:
4281                 /*
4282                  * If the address was optimistic
4283                  * we inserted the route at the start of
4284                  * our DAD process, so we don't need
4285                  * to do it again
4286                  */
4287                 if (!rcu_access_pointer(ifp->rt->rt6i_node))
4288                         ip6_ins_rt(ifp->rt);
4289                 if (ifp->idev->cnf.forwarding)
4290                         addrconf_join_anycast(ifp);
4291                 break;
4292         case RTM_DELADDR:
4293                 if (ifp->idev->cnf.forwarding)
4294                         addrconf_leave_anycast(ifp);
4295                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4296                 dst_hold(&ifp->rt->dst);
4297
4298                 if (ip6_del_rt(ifp->rt))
4299                         dst_free(&ifp->rt->dst);
4300                 break;
4301         }
4302 }
4303
4304 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4305 {
4306         rcu_read_lock_bh();
4307         if (likely(ifp->idev->dead == 0))
4308                 __ipv6_ifa_notify(event, ifp);
4309         rcu_read_unlock_bh();
4310 }
4311
4312 #ifdef CONFIG_SYSCTL
4313
4314 static
4315 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4316                            void __user *buffer, size_t *lenp, loff_t *ppos)
4317 {
4318         int *valp = ctl->data;
4319         int val = *valp;
4320         loff_t pos = *ppos;
4321         int ret;
4322
4323         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4324
4325         if (write)
4326                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4327         if (ret)
4328                 *ppos = pos;
4329         return ret;
4330 }
4331
4332 static
4333 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
4334                         void __user *buffer, size_t *lenp, loff_t *ppos)
4335 {
4336         struct inet6_dev *idev = ctl->extra1;
4337         int min_mtu = IPV6_MIN_MTU;
4338         struct ctl_table lctl;
4339
4340         lctl = *ctl;
4341         lctl.extra1 = &min_mtu;
4342         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
4343
4344         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
4345 }
4346
4347 static void dev_disable_change(struct inet6_dev *idev)
4348 {
4349         if (!idev || !idev->dev)
4350                 return;
4351
4352         if (idev->cnf.disable_ipv6)
4353                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4354         else
4355                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4356 }
4357
4358 static void addrconf_disable_change(struct net *net, __s32 newf)
4359 {
4360         struct net_device *dev;
4361         struct inet6_dev *idev;
4362
4363         for_each_netdev(net, dev) {
4364                 idev = __in6_dev_get(dev);
4365                 if (idev) {
4366                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4367                         idev->cnf.disable_ipv6 = newf;
4368                         if (changed)
4369                                 dev_disable_change(idev);
4370                 }
4371         }
4372 }
4373
4374 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4375 {
4376         struct net *net;
4377
4378         net = (struct net *)table->extra2;
4379
4380         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4381                 return 0;
4382
4383         if (!rtnl_trylock()) {
4384                 /* Restore the original values before restarting */
4385                 *p = old;
4386                 return restart_syscall();
4387         }
4388
4389         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4390                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4391                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4392                 addrconf_disable_change(net, newf);
4393         } else if ((!*p) ^ (!old))
4394                 dev_disable_change((struct inet6_dev *)table->extra1);
4395
4396         rtnl_unlock();
4397         return 0;
4398 }
4399
4400 static
4401 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4402                             void __user *buffer, size_t *lenp, loff_t *ppos)
4403 {
4404         int *valp = ctl->data;
4405         int val = *valp;
4406         loff_t pos = *ppos;
4407         int ret;
4408
4409         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4410
4411         if (write)
4412                 ret = addrconf_disable_ipv6(ctl, valp, val);
4413         if (ret)
4414                 *ppos = pos;
4415         return ret;
4416 }
4417
4418 static struct addrconf_sysctl_table
4419 {
4420         struct ctl_table_header *sysctl_header;
4421         ctl_table addrconf_vars[DEVCONF_MAX+1];
4422         char *dev_name;
4423 } addrconf_sysctl __read_mostly = {
4424         .sysctl_header = NULL,
4425         .addrconf_vars = {
4426                 {
4427                         .procname       = "forwarding",
4428                         .data           = &ipv6_devconf.forwarding,
4429                         .maxlen         = sizeof(int),
4430                         .mode           = 0644,
4431                         .proc_handler   = addrconf_sysctl_forward,
4432                 },
4433                 {
4434                         .procname       = "hop_limit",
4435                         .data           = &ipv6_devconf.hop_limit,
4436                         .maxlen         = sizeof(int),
4437                         .mode           = 0644,
4438                         .proc_handler   = proc_dointvec,
4439                 },
4440                 {
4441                         .procname       = "mtu",
4442                         .data           = &ipv6_devconf.mtu6,
4443                         .maxlen         = sizeof(int),
4444                         .mode           = 0644,
4445                         .proc_handler   = addrconf_sysctl_mtu,
4446                 },
4447                 {
4448                         .procname       = "accept_ra",
4449                         .data           = &ipv6_devconf.accept_ra,
4450                         .maxlen         = sizeof(int),
4451                         .mode           = 0644,
4452                         .proc_handler   = proc_dointvec,
4453                 },
4454                 {
4455                         .procname       = "accept_redirects",
4456                         .data           = &ipv6_devconf.accept_redirects,
4457                         .maxlen         = sizeof(int),
4458                         .mode           = 0644,
4459                         .proc_handler   = proc_dointvec,
4460                 },
4461                 {
4462                         .procname       = "autoconf",
4463                         .data           = &ipv6_devconf.autoconf,
4464                         .maxlen         = sizeof(int),
4465                         .mode           = 0644,
4466                         .proc_handler   = proc_dointvec,
4467                 },
4468                 {
4469                         .procname       = "dad_transmits",
4470                         .data           = &ipv6_devconf.dad_transmits,
4471                         .maxlen         = sizeof(int),
4472                         .mode           = 0644,
4473                         .proc_handler   = proc_dointvec,
4474                 },
4475                 {
4476                         .procname       = "router_solicitations",
4477                         .data           = &ipv6_devconf.rtr_solicits,
4478                         .maxlen         = sizeof(int),
4479                         .mode           = 0644,
4480                         .proc_handler   = proc_dointvec,
4481                 },
4482                 {
4483                         .procname       = "router_solicitation_interval",
4484                         .data           = &ipv6_devconf.rtr_solicit_interval,
4485                         .maxlen         = sizeof(int),
4486                         .mode           = 0644,
4487                         .proc_handler   = proc_dointvec_jiffies,
4488                 },
4489                 {
4490                         .procname       = "router_solicitation_delay",
4491                         .data           = &ipv6_devconf.rtr_solicit_delay,
4492                         .maxlen         = sizeof(int),
4493                         .mode           = 0644,
4494                         .proc_handler   = proc_dointvec_jiffies,
4495                 },
4496                 {
4497                         .procname       = "force_mld_version",
4498                         .data           = &ipv6_devconf.force_mld_version,
4499                         .maxlen         = sizeof(int),
4500                         .mode           = 0644,
4501                         .proc_handler   = proc_dointvec,
4502                 },
4503 #ifdef CONFIG_IPV6_PRIVACY
4504                 {
4505                         .procname       = "use_tempaddr",
4506                         .data           = &ipv6_devconf.use_tempaddr,
4507                         .maxlen         = sizeof(int),
4508                         .mode           = 0644,
4509                         .proc_handler   = proc_dointvec,
4510                 },
4511                 {
4512                         .procname       = "temp_valid_lft",
4513                         .data           = &ipv6_devconf.temp_valid_lft,
4514                         .maxlen         = sizeof(int),
4515                         .mode           = 0644,
4516                         .proc_handler   = proc_dointvec,
4517                 },
4518                 {
4519                         .procname       = "temp_prefered_lft",
4520                         .data           = &ipv6_devconf.temp_prefered_lft,
4521                         .maxlen         = sizeof(int),
4522                         .mode           = 0644,
4523                         .proc_handler   = proc_dointvec,
4524                 },
4525                 {
4526                         .procname       = "regen_max_retry",
4527                         .data           = &ipv6_devconf.regen_max_retry,
4528                         .maxlen         = sizeof(int),
4529                         .mode           = 0644,
4530                         .proc_handler   = proc_dointvec,
4531                 },
4532                 {
4533                         .procname       = "max_desync_factor",
4534                         .data           = &ipv6_devconf.max_desync_factor,
4535                         .maxlen         = sizeof(int),
4536                         .mode           = 0644,
4537                         .proc_handler   = proc_dointvec,
4538                 },
4539 #endif
4540                 {
4541                         .procname       = "max_addresses",
4542                         .data           = &ipv6_devconf.max_addresses,
4543                         .maxlen         = sizeof(int),
4544                         .mode           = 0644,
4545                         .proc_handler   = proc_dointvec,
4546                 },
4547                 {
4548                         .procname       = "accept_ra_defrtr",
4549                         .data           = &ipv6_devconf.accept_ra_defrtr,
4550                         .maxlen         = sizeof(int),
4551                         .mode           = 0644,
4552                         .proc_handler   = proc_dointvec,
4553                 },
4554                 {
4555                         .procname       = "accept_ra_min_hop_limit",
4556                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
4557                         .maxlen         = sizeof(int),
4558                         .mode           = 0644,
4559                         .proc_handler   = proc_dointvec,
4560                 },
4561                 {
4562                         .procname       = "accept_ra_pinfo",
4563                         .data           = &ipv6_devconf.accept_ra_pinfo,
4564                         .maxlen         = sizeof(int),
4565                         .mode           = 0644,
4566                         .proc_handler   = proc_dointvec,
4567                 },
4568 #ifdef CONFIG_IPV6_ROUTER_PREF
4569                 {
4570                         .procname       = "accept_ra_rtr_pref",
4571                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4572                         .maxlen         = sizeof(int),
4573                         .mode           = 0644,
4574                         .proc_handler   = proc_dointvec,
4575                 },
4576                 {
4577                         .procname       = "router_probe_interval",
4578                         .data           = &ipv6_devconf.rtr_probe_interval,
4579                         .maxlen         = sizeof(int),
4580                         .mode           = 0644,
4581                         .proc_handler   = proc_dointvec_jiffies,
4582                 },
4583 #ifdef CONFIG_IPV6_ROUTE_INFO
4584                 {
4585                         .procname       = "accept_ra_rt_info_max_plen",
4586                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4587                         .maxlen         = sizeof(int),
4588                         .mode           = 0644,
4589                         .proc_handler   = proc_dointvec,
4590                 },
4591 #endif
4592 #endif
4593                 {
4594                         .procname       = "proxy_ndp",
4595                         .data           = &ipv6_devconf.proxy_ndp,
4596                         .maxlen         = sizeof(int),
4597                         .mode           = 0644,
4598                         .proc_handler   = proc_dointvec,
4599                 },
4600                 {
4601                         .procname       = "accept_source_route",
4602                         .data           = &ipv6_devconf.accept_source_route,
4603                         .maxlen         = sizeof(int),
4604                         .mode           = 0644,
4605                         .proc_handler   = proc_dointvec,
4606                 },
4607 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4608                 {
4609                         .procname       = "optimistic_dad",
4610                         .data           = &ipv6_devconf.optimistic_dad,
4611                         .maxlen         = sizeof(int),
4612                         .mode           = 0644,
4613                         .proc_handler   = proc_dointvec,
4614
4615                 },
4616 #endif
4617 #ifdef CONFIG_IPV6_MROUTE
4618                 {
4619                         .procname       = "mc_forwarding",
4620                         .data           = &ipv6_devconf.mc_forwarding,
4621                         .maxlen         = sizeof(int),
4622                         .mode           = 0444,
4623                         .proc_handler   = proc_dointvec,
4624                 },
4625 #endif
4626                 {
4627                         .procname       = "disable_ipv6",
4628                         .data           = &ipv6_devconf.disable_ipv6,
4629                         .maxlen         = sizeof(int),
4630                         .mode           = 0644,
4631                         .proc_handler   = addrconf_sysctl_disable,
4632                 },
4633                 {
4634                         .procname       = "accept_dad",
4635                         .data           = &ipv6_devconf.accept_dad,
4636                         .maxlen         = sizeof(int),
4637                         .mode           = 0644,
4638                         .proc_handler   = proc_dointvec,
4639                 },
4640                 {
4641                         .procname       = "force_tllao",
4642                         .data           = &ipv6_devconf.force_tllao,
4643                         .maxlen         = sizeof(int),
4644                         .mode           = 0644,
4645                         .proc_handler   = proc_dointvec
4646                 },
4647                 {
4648                         /* sentinel */
4649                 }
4650         },
4651 };
4652
4653 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4654                 struct inet6_dev *idev, struct ipv6_devconf *p)
4655 {
4656         int i;
4657         struct addrconf_sysctl_table *t;
4658
4659 #define ADDRCONF_CTL_PATH_DEV   3
4660
4661         struct ctl_path addrconf_ctl_path[] = {
4662                 { .procname = "net", },
4663                 { .procname = "ipv6", },
4664                 { .procname = "conf", },
4665                 { /* to be set */ },
4666                 { },
4667         };
4668
4669
4670         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4671         if (t == NULL)
4672                 goto out;
4673
4674         for (i = 0; t->addrconf_vars[i].data; i++) {
4675                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4676                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4677                 t->addrconf_vars[i].extra2 = net;
4678         }
4679
4680         /*
4681          * Make a copy of dev_name, because '.procname' is regarded as const
4682          * by sysctl and we wouldn't want anyone to change it under our feet
4683          * (see SIOCSIFNAME).
4684          */
4685         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4686         if (!t->dev_name)
4687                 goto free;
4688
4689         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4690
4691         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4692                         t->addrconf_vars);
4693         if (t->sysctl_header == NULL)
4694                 goto free_procname;
4695
4696         p->sysctl = t;
4697         return 0;
4698
4699 free_procname:
4700         kfree(t->dev_name);
4701 free:
4702         kfree(t);
4703 out:
4704         return -ENOBUFS;
4705 }
4706
4707 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4708 {
4709         struct addrconf_sysctl_table *t;
4710
4711         if (p->sysctl == NULL)
4712                 return;
4713
4714         t = p->sysctl;
4715         p->sysctl = NULL;
4716         unregister_net_sysctl_table(t->sysctl_header);
4717         kfree(t->dev_name);
4718         kfree(t);
4719 }
4720
4721 static void addrconf_sysctl_register(struct inet6_dev *idev)
4722 {
4723         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4724                               &ndisc_ifinfo_sysctl_change);
4725         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4726                                         idev, &idev->cnf);
4727 }
4728
4729 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4730 {
4731         __addrconf_sysctl_unregister(&idev->cnf);
4732         neigh_sysctl_unregister(idev->nd_parms);
4733 }
4734
4735
4736 #endif
4737
4738 static int __net_init addrconf_init_net(struct net *net)
4739 {
4740         int err = -ENOMEM;
4741         struct ipv6_devconf *all, *dflt;
4742
4743         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
4744         if (all == NULL)
4745                 goto err_alloc_all;
4746
4747         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4748         if (dflt == NULL)
4749                 goto err_alloc_dflt;
4750
4751         /* these will be inherited by all namespaces */
4752         dflt->autoconf = ipv6_defaults.autoconf;
4753         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4754
4755         net->ipv6.devconf_all = all;
4756         net->ipv6.devconf_dflt = dflt;
4757
4758 #ifdef CONFIG_SYSCTL
4759         err = __addrconf_sysctl_register(net, "all", NULL, all);
4760         if (err < 0)
4761                 goto err_reg_all;
4762
4763         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4764         if (err < 0)
4765                 goto err_reg_dflt;
4766 #endif
4767         return 0;
4768
4769 #ifdef CONFIG_SYSCTL
4770 err_reg_dflt:
4771         __addrconf_sysctl_unregister(all);
4772 err_reg_all:
4773         kfree(dflt);
4774 #endif
4775 err_alloc_dflt:
4776         kfree(all);
4777 err_alloc_all:
4778         return err;
4779 }
4780
4781 static void __net_exit addrconf_exit_net(struct net *net)
4782 {
4783 #ifdef CONFIG_SYSCTL
4784         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4785         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4786 #endif
4787         if (!net_eq(net, &init_net)) {
4788                 kfree(net->ipv6.devconf_dflt);
4789                 kfree(net->ipv6.devconf_all);
4790         }
4791 }
4792
4793 static struct pernet_operations addrconf_ops = {
4794         .init = addrconf_init_net,
4795         .exit = addrconf_exit_net,
4796 };
4797
4798 /*
4799  *      Device notifier
4800  */
4801
4802 int register_inet6addr_notifier(struct notifier_block *nb)
4803 {
4804         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4805 }
4806 EXPORT_SYMBOL(register_inet6addr_notifier);
4807
4808 int unregister_inet6addr_notifier(struct notifier_block *nb)
4809 {
4810         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4811 }
4812 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4813
4814 static struct rtnl_af_ops inet6_ops = {
4815         .family           = AF_INET6,
4816         .fill_link_af     = inet6_fill_link_af,
4817         .get_link_af_size = inet6_get_link_af_size,
4818 };
4819
4820 /*
4821  *      Init / cleanup code
4822  */
4823
4824 int __init addrconf_init(void)
4825 {
4826         int i, err;
4827
4828         err = ipv6_addr_label_init();
4829         if (err < 0) {
4830                 printk(KERN_CRIT "IPv6 Addrconf:"
4831                        " cannot initialize default policy table: %d.\n", err);
4832                 goto out;
4833         }
4834
4835         err = register_pernet_subsys(&addrconf_ops);
4836         if (err < 0)
4837                 goto out_addrlabel;
4838
4839         /* The addrconf netdev notifier requires that loopback_dev
4840          * has it's ipv6 private information allocated and setup
4841          * before it can bring up and give link-local addresses
4842          * to other devices which are up.
4843          *
4844          * Unfortunately, loopback_dev is not necessarily the first
4845          * entry in the global dev_base list of net devices.  In fact,
4846          * it is likely to be the very last entry on that list.
4847          * So this causes the notifier registry below to try and
4848          * give link-local addresses to all devices besides loopback_dev
4849          * first, then loopback_dev, which cases all the non-loopback_dev
4850          * devices to fail to get a link-local address.
4851          *
4852          * So, as a temporary fix, allocate the ipv6 structure for
4853          * loopback_dev first by hand.
4854          * Longer term, all of the dependencies ipv6 has upon the loopback
4855          * device and it being up should be removed.
4856          */
4857         rtnl_lock();
4858         if (!ipv6_add_dev(init_net.loopback_dev))
4859                 err = -ENOMEM;
4860         rtnl_unlock();
4861         if (err)
4862                 goto errlo;
4863
4864         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4865                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4866
4867         register_netdevice_notifier(&ipv6_dev_notf);
4868
4869         addrconf_verify(0);
4870
4871         err = rtnl_af_register(&inet6_ops);
4872         if (err < 0)
4873                 goto errout_af;
4874
4875         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4876                               NULL);
4877         if (err < 0)
4878                 goto errout;
4879
4880         /* Only the first call to __rtnl_register can fail */
4881         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4882         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4883         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4884                         inet6_dump_ifaddr, NULL);
4885         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4886                         inet6_dump_ifmcaddr, NULL);
4887         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4888                         inet6_dump_ifacaddr, NULL);
4889
4890         ipv6_addr_label_rtnl_register();
4891
4892         return 0;
4893 errout:
4894         rtnl_af_unregister(&inet6_ops);
4895 errout_af:
4896         unregister_netdevice_notifier(&ipv6_dev_notf);
4897 errlo:
4898         unregister_pernet_subsys(&addrconf_ops);
4899 out_addrlabel:
4900         ipv6_addr_label_cleanup();
4901 out:
4902         return err;
4903 }
4904
4905 void addrconf_cleanup(void)
4906 {
4907         struct net_device *dev;
4908         int i;
4909
4910         unregister_netdevice_notifier(&ipv6_dev_notf);
4911         unregister_pernet_subsys(&addrconf_ops);
4912         ipv6_addr_label_cleanup();
4913
4914         rtnl_lock();
4915
4916         __rtnl_af_unregister(&inet6_ops);
4917
4918         /* clean dev list */
4919         for_each_netdev(&init_net, dev) {
4920                 if (__in6_dev_get(dev) == NULL)
4921                         continue;
4922                 addrconf_ifdown(dev, 1);
4923         }
4924         addrconf_ifdown(init_net.loopback_dev, 2);
4925
4926         /*
4927          *      Check hash table.
4928          */
4929         spin_lock_bh(&addrconf_hash_lock);
4930         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4931                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4932         spin_unlock_bh(&addrconf_hash_lock);
4933
4934         del_timer(&addr_chk_timer);
4935         rtnl_unlock();
4936 }