net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/ipv6.h>
  74 #include <net/protocol.h>
  75 #include <net/ndisc.h>
  76 #include <net/ip6_route.h>
  77 #include <net/addrconf.h>
  78 #include <net/tcp.h>
  79 #include <net/ip.h>
  80 #include <net/netlink.h>
  81 #include <net/pkt_sched.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84
  85 #ifdef CONFIG_IPV6_PRIVACY
  86 #include <linux/random.h>
  87 #endif
  88
  89 #include <linux/uaccess.h>
  90 #include <asm/unaligned.h>
  91
  92 #include <linux/proc_fs.h>
  93 #include <linux/seq_file.h>
  94 #include <linux/export.h>
  95
  96 /* Set to 3 to get tracing... */
  97 #define ACONF_DEBUG 2
  98
  99 #if ACONF_DEBUG >= 3
 100 #define ADBG(x) printk x
 101 #else
 102 #define ADBG(x)
 103 #endif
 104
 105 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 106
 107 static inline u32 cstamp_delta(unsigned long cstamp)
 108 {
 109         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 110 }
 111
 112 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 113 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 114 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 115
 116 #ifdef CONFIG_SYSCTL
 117 static void addrconf_sysctl_register(struct inet6_dev *idev);
 118 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 119 #else
 120 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 121 {
 122 }
 123
 124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 125 {
 126 }
 127 #endif
 128
 129 #ifdef CONFIG_IPV6_PRIVACY
 130 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 131 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 132 static void ipv6_regen_rndid(unsigned long data);
 133 #endif
 134
 135 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 136 static int ipv6_count_addresses(struct inet6_dev *idev);
 137
 138 /*
 139  *      Configured unicast address hash table
 140  */
 141 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 142 static DEFINE_SPINLOCK(addrconf_hash_lock);
 143
 144 static void addrconf_verify(unsigned long);
 145
 146 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 147 static DEFINE_SPINLOCK(addrconf_verify_lock);
 148
 149 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 150 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 151
 152 static void addrconf_type_change(struct net_device *dev,
 153                                  unsigned long event);
 154 static int addrconf_ifdown(struct net_device *dev, int how);
 155
 156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 157 static void addrconf_dad_timer(unsigned long data);
 158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 159 static void addrconf_dad_run(struct inet6_dev *idev);
 160 static void addrconf_rs_timer(unsigned long data);
 161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 163
 164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 165                                 struct prefix_info *pinfo);
 166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 167                                struct net_device *dev);
 168
 169 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 170
 171 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 172         .forwarding             = 0,
 173         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 174         .mtu6                   = IPV6_MIN_MTU,
 175         .accept_ra              = 1,
 176         .accept_redirects       = 1,
 177         .autoconf               = 1,
 178         .force_mld_version      = 0,
 179         .dad_transmits          = 1,
 180         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 181         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 182         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 183 #ifdef CONFIG_IPV6_PRIVACY
 184         .use_tempaddr           = 0,
 185         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 186         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 187         .regen_max_retry        = REGEN_MAX_RETRY,
 188         .max_desync_factor      = MAX_DESYNC_FACTOR,
 189 #endif
 190         .max_addresses          = IPV6_MAX_ADDRESSES,
 191         .accept_ra_defrtr       = 1,
 192         .accept_ra_pinfo        = 1,
 193 #ifdef CONFIG_IPV6_ROUTER_PREF
 194         .accept_ra_rtr_pref     = 1,
 195         .rtr_probe_interval     = 60 * HZ,
 196 #ifdef CONFIG_IPV6_ROUTE_INFO
 197         .accept_ra_rt_info_max_plen = 0,
 198 #endif
 199 #endif
 200         .proxy_ndp              = 0,
 201         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 202         .disable_ipv6           = 0,
 203         .accept_dad             = 1,
 204 };
 205
 206 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 207         .forwarding             = 0,
 208         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 209         .mtu6                   = IPV6_MIN_MTU,
 210         .accept_ra              = 1,
 211         .accept_redirects       = 1,
 212         .autoconf               = 1,
 213         .dad_transmits          = 1,
 214         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 215         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 216         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 217 #ifdef CONFIG_IPV6_PRIVACY
 218         .use_tempaddr           = 0,
 219         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 220         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 221         .regen_max_retry        = REGEN_MAX_RETRY,
 222         .max_desync_factor      = MAX_DESYNC_FACTOR,
 223 #endif
 224         .max_addresses          = IPV6_MAX_ADDRESSES,
 225         .accept_ra_defrtr       = 1,
 226         .accept_ra_pinfo        = 1,
 227 #ifdef CONFIG_IPV6_ROUTER_PREF
 228         .accept_ra_rtr_pref     = 1,
 229         .rtr_probe_interval     = 60 * HZ,
 230 #ifdef CONFIG_IPV6_ROUTE_INFO
 231         .accept_ra_rt_info_max_plen = 0,
 232 #endif
 233 #endif
 234         .proxy_ndp              = 0,
 235         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 236         .disable_ipv6           = 0,
 237         .accept_dad             = 1,
 238 };
 239
 240 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 241 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 242 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 243 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 244 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 245
 246 /* Check if a valid qdisc is available */
 247 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 248 {
 249         return !qdisc_tx_is_noop(dev);
 250 }
 251
 252 /* Check if a route is valid prefix route */
 253 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 254 {
 255         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 256 }
 257
 258 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 259 {
 260         if (del_timer(&ifp->timer))
 261                 __in6_ifa_put(ifp);
 262 }
 263
 264 enum addrconf_timer_t {
 265         AC_NONE,
 266         AC_DAD,
 267         AC_RS,
 268 };
 269
 270 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 271                                enum addrconf_timer_t what,
 272                                unsigned long when)
 273 {
 274         if (!del_timer(&ifp->timer))
 275                 in6_ifa_hold(ifp);
 276
 277         switch (what) {
 278         case AC_DAD:
 279                 ifp->timer.function = addrconf_dad_timer;
 280                 break;
 281         case AC_RS:
 282                 ifp->timer.function = addrconf_rs_timer;
 283                 break;
 284         default:
 285                 break;
 286         }
 287         ifp->timer.expires = jiffies + when;
 288         add_timer(&ifp->timer);
 289 }
 290
 291 static int snmp6_alloc_dev(struct inet6_dev *idev)
 292 {
 293         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 294                           sizeof(struct ipstats_mib),
 295                           __alignof__(struct ipstats_mib)) < 0)
 296                 goto err_ip;
 297         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 298                                         GFP_KERNEL);
 299         if (!idev->stats.icmpv6dev)
 300                 goto err_icmp;
 301         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 302                                            GFP_KERNEL);
 303         if (!idev->stats.icmpv6msgdev)
 304                 goto err_icmpmsg;
 305
 306         return 0;
 307
 308 err_icmpmsg:
 309         kfree(idev->stats.icmpv6dev);
 310 err_icmp:
 311         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 312 err_ip:
 313         return -ENOMEM;
 314 }
 315
 316 static void snmp6_free_dev(struct inet6_dev *idev)
 317 {
 318         kfree(idev->stats.icmpv6msgdev);
 319         kfree(idev->stats.icmpv6dev);
 320         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 321 }
 322
 323 /* Nobody refers to this device, we may destroy it. */
 324
 325 void in6_dev_finish_destroy(struct inet6_dev *idev)
 326 {
 327         struct net_device *dev = idev->dev;
 328
 329         WARN_ON(!list_empty(&idev->addr_list));
 330         WARN_ON(idev->mc_list != NULL);
 331
 332 #ifdef NET_REFCNT_DEBUG
 333         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 334 #endif
 335         dev_put(dev);
 336         if (!idev->dead) {
 337                 pr_warn("Freeing alive inet6 device %p\n", idev);
 338                 return;
 339         }
 340         snmp6_free_dev(idev);
 341         kfree_rcu(idev, rcu);
 342 }
 343 EXPORT_SYMBOL(in6_dev_finish_destroy);
 344
 345 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 346 {
 347         struct inet6_dev *ndev;
 348
 349         ASSERT_RTNL();
 350
 351         if (dev->mtu < IPV6_MIN_MTU)
 352                 return NULL;
 353
 354         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 355
 356         if (ndev == NULL)
 357                 return NULL;
 358
 359         rwlock_init(&ndev->lock);
 360         ndev->dev = dev;
 361         INIT_LIST_HEAD(&ndev->addr_list);
 362
 363         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 364         ndev->cnf.mtu6 = dev->mtu;
 365         ndev->cnf.sysctl = NULL;
 366         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 367         if (ndev->nd_parms == NULL) {
 368                 kfree(ndev);
 369                 return NULL;
 370         }
 371         if (ndev->cnf.forwarding)
 372                 dev_disable_lro(dev);
 373         /* We refer to the device */
 374         dev_hold(dev);
 375
 376         if (snmp6_alloc_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 dev_put(dev);
 382                 kfree(ndev);
 383                 return NULL;
 384         }
 385
 386         if (snmp6_register_dev(ndev) < 0) {
 387                 ADBG((KERN_WARNING
 388                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 389                         __func__, dev->name));
 390                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 391                 ndev->dead = 1;
 392                 in6_dev_finish_destroy(ndev);
 393                 return NULL;
 394         }
 395
 396         /* One reference from device.  We must do this before
 397          * we invoke __ipv6_regen_rndid().
 398          */
 399         in6_dev_hold(ndev);
 400
 401         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 402                 ndev->cnf.accept_dad = -1;
 403
 404 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 405         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 406                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 407                 ndev->cnf.rtr_solicits = 0;
 408         }
 409 #endif
 410
 411 #ifdef CONFIG_IPV6_PRIVACY
 412         INIT_LIST_HEAD(&ndev->tempaddr_list);
 413         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 414         if ((dev->flags&IFF_LOOPBACK) ||
 415             dev->type == ARPHRD_TUNNEL ||
 416             dev->type == ARPHRD_TUNNEL6 ||
 417             dev->type == ARPHRD_SIT ||
 418             dev->type == ARPHRD_NONE) {
 419                 ndev->cnf.use_tempaddr = -1;
 420         } else {
 421                 in6_dev_hold(ndev);
 422                 ipv6_regen_rndid((unsigned long) ndev);
 423         }
 424 #endif
 425
 426         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 427                 ndev->if_flags |= IF_READY;
 428
 429         ipv6_mc_init_dev(ndev);
 430         ndev->tstamp = jiffies;
 431         addrconf_sysctl_register(ndev);
 432         /* protected by rtnl_lock */
 433         rcu_assign_pointer(dev->ip6_ptr, ndev);
 434
 435         /* Join all-node multicast group */
 436         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 437
 438         /* Join all-router multicast group if forwarding is set */
 439         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 440                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 441
 442         return ndev;
 443 }
 444
 445 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 446 {
 447         struct inet6_dev *idev;
 448
 449         ASSERT_RTNL();
 450
 451         idev = __in6_dev_get(dev);
 452         if (!idev) {
 453                 idev = ipv6_add_dev(dev);
 454                 if (!idev)
 455                         return NULL;
 456         }
 457
 458         if (dev->flags&IFF_UP)
 459                 ipv6_mc_up(idev);
 460         return idev;
 461 }
 462
 463 #ifdef CONFIG_SYSCTL
 464 static void dev_forward_change(struct inet6_dev *idev)
 465 {
 466         struct net_device *dev;
 467         struct inet6_ifaddr *ifa;
 468
 469         if (!idev)
 470                 return;
 471         dev = idev->dev;
 472         if (idev->cnf.forwarding)
 473                 dev_disable_lro(dev);
 474         if (dev && (dev->flags & IFF_MULTICAST)) {
 475                 if (idev->cnf.forwarding)
 476                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 477                 else
 478                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 479         }
 480
 481         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 482                 if (ifa->flags&IFA_F_TENTATIVE)
 483                         continue;
 484                 if (idev->cnf.forwarding)
 485                         addrconf_join_anycast(ifa);
 486                 else
 487                         addrconf_leave_anycast(ifa);
 488         }
 489 }
 490
 491
 492 static void addrconf_forward_change(struct net *net, __s32 newf)
 493 {
 494         struct net_device *dev;
 495         struct inet6_dev *idev;
 496
 497         for_each_netdev(net, dev) {
 498                 idev = __in6_dev_get(dev);
 499                 if (idev) {
 500                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 501                         idev->cnf.forwarding = newf;
 502                         if (changed)
 503                                 dev_forward_change(idev);
 504                 }
 505         }
 506 }
 507
 508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 509 {
 510         struct net *net;
 511         int old;
 512
 513         if (!rtnl_trylock())
 514                 return restart_syscall();
 515
 516         net = (struct net *)table->extra2;
 517         old = *p;
 518         *p = newf;
 519
 520         if (p == &net->ipv6.devconf_dflt->forwarding) {
 521                 rtnl_unlock();
 522                 return 0;
 523         }
 524
 525         if (p == &net->ipv6.devconf_all->forwarding) {
 526                 net->ipv6.devconf_dflt->forwarding = newf;
 527                 addrconf_forward_change(net, newf);
 528         } else if ((!newf) ^ (!old))
 529                 dev_forward_change((struct inet6_dev *)table->extra1);
 530         rtnl_unlock();
 531
 532         if (newf)
 533                 rt6_purge_dflt_routers(net);
 534         return 1;
 535 }
 536 #endif
 537
 538 /* Nobody refers to this ifaddr, destroy it */
 539 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 540 {
 541         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 542
 543 #ifdef NET_REFCNT_DEBUG
 544         pr_debug("%s\n", __func__);
 545 #endif
 546
 547         in6_dev_put(ifp->idev);
 548
 549         if (del_timer(&ifp->timer))
 550                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 551
 552         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 553                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 554                 return;
 555         }
 556         dst_release(&ifp->rt->dst);
 557
 558         kfree_rcu(ifp, rcu);
 559 }
 560
 561 static void
 562 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 563 {
 564         struct list_head *p;
 565         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 566
 567         /*
 568          * Each device address list is sorted in order of scope -
 569          * global before linklocal.
 570          */
 571         list_for_each(p, &idev->addr_list) {
 572                 struct inet6_ifaddr *ifa
 573                         = list_entry(p, struct inet6_ifaddr, if_list);
 574                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 575                         break;
 576         }
 577
 578         list_add_tail(&ifp->if_list, p);
 579 }
 580
 581 static u32 inet6_addr_hash(const struct in6_addr *addr)
 582 {
 583         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, false);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ifa->addr = *addr;
 639
 640         spin_lock_init(&ifa->lock);
 641         spin_lock_init(&ifa->state_lock);
 642         init_timer(&ifa->timer);
 643         INIT_HLIST_NODE(&ifa->addr_lst);
 644         ifa->timer.data = (unsigned long) ifa;
 645         ifa->scope = scope;
 646         ifa->prefix_len = pfxlen;
 647         ifa->flags = flags | IFA_F_TENTATIVE;
 648         ifa->cstamp = ifa->tstamp = jiffies;
 649
 650         ifa->rt = rt;
 651
 652         ifa->idev = idev;
 653         in6_dev_hold(idev);
 654         /* For caller */
 655         in6_ifa_hold(ifa);
 656
 657         /* Add to big hash table */
 658         hash = inet6_addr_hash(addr);
 659
 660         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 661         spin_unlock(&addrconf_hash_lock);
 662
 663         write_lock(&idev->lock);
 664         /* Add to inet6_dev unicast addr list. */
 665         ipv6_link_dev_addr(idev, ifa);
 666
 667 #ifdef CONFIG_IPV6_PRIVACY
 668         if (ifa->flags&IFA_F_TEMPORARY) {
 669                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 670                 in6_ifa_hold(ifa);
 671         }
 672 #endif
 673
 674         in6_ifa_hold(ifa);
 675         write_unlock(&idev->lock);
 676 out2:
 677         rcu_read_unlock_bh();
 678
 679         if (likely(err == 0))
 680                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 681         else {
 682                 kfree(ifa);
 683                 ifa = ERR_PTR(err);
 684         }
 685
 686         return ifa;
 687 out:
 688         spin_unlock(&addrconf_hash_lock);
 689         goto out2;
 690 }
 691
 692 /* This function wants to get referenced ifp and releases it before return */
 693
 694 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 695 {
 696         struct inet6_ifaddr *ifa, *ifn;
 697         struct inet6_dev *idev = ifp->idev;
 698         int state;
 699         int deleted = 0, onlink = 0;
 700         unsigned long expires = jiffies;
 701
 702         spin_lock_bh(&ifp->state_lock);
 703         state = ifp->state;
 704         ifp->state = INET6_IFADDR_STATE_DEAD;
 705         spin_unlock_bh(&ifp->state_lock);
 706
 707         if (state == INET6_IFADDR_STATE_DEAD)
 708                 goto out;
 709
 710         spin_lock_bh(&addrconf_hash_lock);
 711         hlist_del_init_rcu(&ifp->addr_lst);
 712         spin_unlock_bh(&addrconf_hash_lock);
 713
 714         write_lock_bh(&idev->lock);
 715 #ifdef CONFIG_IPV6_PRIVACY
 716         if (ifp->flags&IFA_F_TEMPORARY) {
 717                 list_del(&ifp->tmp_list);
 718                 if (ifp->ifpub) {
 719                         in6_ifa_put(ifp->ifpub);
 720                         ifp->ifpub = NULL;
 721                 }
 722                 __in6_ifa_put(ifp);
 723         }
 724 #endif
 725
 726         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 727                 if (ifa == ifp) {
 728                         list_del_init(&ifp->if_list);
 729                         __in6_ifa_put(ifp);
 730
 731                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 732                                 break;
 733                         deleted = 1;
 734                         continue;
 735                 } else if (ifp->flags & IFA_F_PERMANENT) {
 736                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 737                                               ifp->prefix_len)) {
 738                                 if (ifa->flags & IFA_F_PERMANENT) {
 739                                         onlink = 1;
 740                                         if (deleted)
 741                                                 break;
 742                                 } else {
 743                                         unsigned long lifetime;
 744
 745                                         if (!onlink)
 746                                                 onlink = -1;
 747
 748                                         spin_lock(&ifa->lock);
 749
 750                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 751                                         /*
 752                                          * Note: Because this address is
 753                                          * not permanent, lifetime <
 754                                          * LONG_MAX / HZ here.
 755                                          */
 756                                         if (time_before(expires,
 757                                                         ifa->tstamp + lifetime * HZ))
 758                                                 expires = ifa->tstamp + lifetime * HZ;
 759                                         spin_unlock(&ifa->lock);
 760                                 }
 761                         }
 762                 }
 763         }
 764         write_unlock_bh(&idev->lock);
 765
 766         addrconf_del_timer(ifp);
 767
 768         ipv6_ifa_notify(RTM_DELADDR, ifp);
 769
 770         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 771
 772         /*
 773          * Purge or update corresponding prefix
 774          *
 775          * 1) we don't purge prefix here if address was not permanent.
 776          *    prefix is managed by its own lifetime.
 777          * 2) if there're no addresses, delete prefix.
 778          * 3) if there're still other permanent address(es),
 779          *    corresponding prefix is still permanent.
 780          * 4) otherwise, update prefix lifetime to the
 781          *    longest valid lifetime among the corresponding
 782          *    addresses on the device.
 783          *    Note: subsequent RA will update lifetime.
 784          *
 785          * --yoshfuji
 786          */
 787         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 788                 struct in6_addr prefix;
 789                 struct rt6_info *rt;
 790                 struct net *net = dev_net(ifp->idev->dev);
 791                 struct flowi6 fl6 = {};
 792
 793                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 794                 fl6.flowi6_oif = ifp->idev->dev->ifindex;
 795                 fl6.daddr = prefix;
 796                 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
 797                                                          RT6_LOOKUP_F_IFACE);
 798
 799                 if (rt != net->ipv6.ip6_null_entry &&
 800                     addrconf_is_prefix_route(rt)) {
 801                         if (onlink == 0) {
 802                                 ip6_del_rt(rt);
 803                                 rt = NULL;
 804                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 805                                 rt6_set_expires(rt, expires);
 806                         }
 807                 }
 808                 dst_release(&rt->dst);
 809         }
 810
 811         /* clean up prefsrc entries */
 812         rt6_remove_prefsrc(ifp);
 813 out:
 814         in6_ifa_put(ifp);
 815 }
 816
 817 #ifdef CONFIG_IPV6_PRIVACY
 818 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 819 {
 820         struct inet6_dev *idev = ifp->idev;
 821         struct in6_addr addr, *tmpaddr;
 822         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 823         unsigned long regen_advance;
 824         int tmp_plen;
 825         int ret = 0;
 826         int max_addresses;
 827         u32 addr_flags;
 828         unsigned long now = jiffies;
 829
 830         write_lock(&idev->lock);
 831         if (ift) {
 832                 spin_lock_bh(&ift->lock);
 833                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 834                 spin_unlock_bh(&ift->lock);
 835                 tmpaddr = &addr;
 836         } else {
 837                 tmpaddr = NULL;
 838         }
 839 retry:
 840         in6_dev_hold(idev);
 841         if (idev->cnf.use_tempaddr <= 0) {
 842                 write_unlock(&idev->lock);
 843                 pr_info("%s: use_tempaddr is disabled\n", __func__);
 844                 in6_dev_put(idev);
 845                 ret = -1;
 846                 goto out;
 847         }
 848         spin_lock_bh(&ifp->lock);
 849         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 850                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 851                 spin_unlock_bh(&ifp->lock);
 852                 write_unlock(&idev->lock);
 853                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
 854                         __func__);
 855                 in6_dev_put(idev);
 856                 ret = -1;
 857                 goto out;
 858         }
 859         in6_ifa_hold(ifp);
 860         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 861         __ipv6_try_regen_rndid(idev, tmpaddr);
 862         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 863         age = (now - ifp->tstamp) / HZ;
 864         tmp_valid_lft = min_t(__u32,
 865                               ifp->valid_lft,
 866                               idev->cnf.temp_valid_lft + age);
 867         tmp_prefered_lft = min_t(__u32,
 868                                  ifp->prefered_lft,
 869                                  idev->cnf.temp_prefered_lft + age -
 870                                  idev->cnf.max_desync_factor);
 871         tmp_plen = ifp->prefix_len;
 872         max_addresses = idev->cnf.max_addresses;
 873         tmp_tstamp = ifp->tstamp;
 874         spin_unlock_bh(&ifp->lock);
 875
 876         regen_advance = idev->cnf.regen_max_retry *
 877                         idev->cnf.dad_transmits *
 878                         idev->nd_parms->retrans_time / HZ;
 879         write_unlock(&idev->lock);
 880
 881         /* A temporary address is created only if this calculated Preferred
 882          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 883          * an implementation must not create a temporary address with a zero
 884          * Preferred Lifetime.
 885          */
 886         if (tmp_prefered_lft <= regen_advance) {
 887                 in6_ifa_put(ifp);
 888                 in6_dev_put(idev);
 889                 ret = -1;
 890                 goto out;
 891         }
 892
 893         addr_flags = IFA_F_TEMPORARY;
 894         /* set in addrconf_prefix_rcv() */
 895         if (ifp->flags & IFA_F_OPTIMISTIC)
 896                 addr_flags |= IFA_F_OPTIMISTIC;
 897
 898         ift = !max_addresses ||
 899               ipv6_count_addresses(idev) < max_addresses ?
 900                 ipv6_add_addr(idev, &addr, tmp_plen,
 901                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 902                               addr_flags) : NULL;
 903         if (!ift || IS_ERR(ift)) {
 904                 in6_ifa_put(ifp);
 905                 in6_dev_put(idev);
 906                 pr_info("%s: retry temporary address regeneration\n", __func__);
 907                 tmpaddr = &addr;
 908                 write_lock(&idev->lock);
 909                 goto retry;
 910         }
 911
 912         spin_lock_bh(&ift->lock);
 913         ift->ifpub = ifp;
 914         ift->valid_lft = tmp_valid_lft;
 915         ift->prefered_lft = tmp_prefered_lft;
 916         ift->cstamp = now;
 917         ift->tstamp = tmp_tstamp;
 918         spin_unlock_bh(&ift->lock);
 919
 920         addrconf_dad_start(ift);
 921         in6_ifa_put(ift);
 922         in6_dev_put(idev);
 923 out:
 924         return ret;
 925 }
 926 #endif
 927
 928 /*
 929  *      Choose an appropriate source address (RFC3484)
 930  */
 931 enum {
 932         IPV6_SADDR_RULE_INIT = 0,
 933         IPV6_SADDR_RULE_LOCAL,
 934         IPV6_SADDR_RULE_SCOPE,
 935         IPV6_SADDR_RULE_PREFERRED,
 936 #ifdef CONFIG_IPV6_MIP6
 937         IPV6_SADDR_RULE_HOA,
 938 #endif
 939         IPV6_SADDR_RULE_OIF,
 940         IPV6_SADDR_RULE_LABEL,
 941 #ifdef CONFIG_IPV6_PRIVACY
 942         IPV6_SADDR_RULE_PRIVACY,
 943 #endif
 944         IPV6_SADDR_RULE_ORCHID,
 945         IPV6_SADDR_RULE_PREFIX,
 946         IPV6_SADDR_RULE_MAX
 947 };
 948
 949 struct ipv6_saddr_score {
 950         int                     rule;
 951         int                     addr_type;
 952         struct inet6_ifaddr     *ifa;
 953         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 954         int                     scopedist;
 955         int                     matchlen;
 956 };
 957
 958 struct ipv6_saddr_dst {
 959         const struct in6_addr *addr;
 960         int ifindex;
 961         int scope;
 962         int label;
 963         unsigned int prefs;
 964 };
 965
 966 static inline int ipv6_saddr_preferred(int type)
 967 {
 968         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 969                 return 1;
 970         return 0;
 971 }
 972
 973 static int ipv6_get_saddr_eval(struct net *net,
 974                                struct ipv6_saddr_score *score,
 975                                struct ipv6_saddr_dst *dst,
 976                                int i)
 977 {
 978         int ret;
 979
 980         if (i <= score->rule) {
 981                 switch (i) {
 982                 case IPV6_SADDR_RULE_SCOPE:
 983                         ret = score->scopedist;
 984                         break;
 985                 case IPV6_SADDR_RULE_PREFIX:
 986                         ret = score->matchlen;
 987                         break;
 988                 default:
 989                         ret = !!test_bit(i, score->scorebits);
 990                 }
 991                 goto out;
 992         }
 993
 994         switch (i) {
 995         case IPV6_SADDR_RULE_INIT:
 996                 /* Rule 0: remember if hiscore is not ready yet */
 997                 ret = !!score->ifa;
 998                 break;
 999         case IPV6_SADDR_RULE_LOCAL:
1000                 /* Rule 1: Prefer same address */
1001                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1002                 break;
1003         case IPV6_SADDR_RULE_SCOPE:
1004                 /* Rule 2: Prefer appropriate scope
1005                  *
1006                  *      ret
1007                  *       ^
1008                  *    -1 |  d 15
1009                  *    ---+--+-+---> scope
1010                  *       |
1011                  *       |             d is scope of the destination.
1012                  *  B-d  |  \
1013                  *       |   \      <- smaller scope is better if
1014                  *  B-15 |    \        if scope is enough for destinaion.
1015                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1016                  * d-C-1 | /
1017                  *       |/         <- greater is better
1018                  *   -C  /             if scope is not enough for destination.
1019                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1020                  *
1021                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1022                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1023                  * Assume B = 0 and we get C > 29.
1024                  */
1025                 ret = __ipv6_addr_src_scope(score->addr_type);
1026                 if (ret >= dst->scope)
1027                         ret = -ret;
1028                 else
1029                         ret -= 128;     /* 30 is enough */
1030                 score->scopedist = ret;
1031                 break;
1032         case IPV6_SADDR_RULE_PREFERRED:
1033                 /* Rule 3: Avoid deprecated and optimistic addresses */
1034                 ret = ipv6_saddr_preferred(score->addr_type) ||
1035                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1036                 break;
1037 #ifdef CONFIG_IPV6_MIP6
1038         case IPV6_SADDR_RULE_HOA:
1039             {
1040                 /* Rule 4: Prefer home address */
1041                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1042                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1043                 break;
1044             }
1045 #endif
1046         case IPV6_SADDR_RULE_OIF:
1047                 /* Rule 5: Prefer outgoing interface */
1048                 ret = (!dst->ifindex ||
1049                        dst->ifindex == score->ifa->idev->dev->ifindex);
1050                 break;
1051         case IPV6_SADDR_RULE_LABEL:
1052                 /* Rule 6: Prefer matching label */
1053                 ret = ipv6_addr_label(net,
1054                                       &score->ifa->addr, score->addr_type,
1055                                       score->ifa->idev->dev->ifindex) == dst->label;
1056                 break;
1057 #ifdef CONFIG_IPV6_PRIVACY
1058         case IPV6_SADDR_RULE_PRIVACY:
1059             {
1060                 /* Rule 7: Prefer public address
1061                  * Note: prefer temporary address if use_tempaddr >= 2
1062                  */
1063                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1064                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1065                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1066                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1067                 break;
1068             }
1069 #endif
1070         case IPV6_SADDR_RULE_ORCHID:
1071                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1072                  *          non-ORCHID vs non-ORCHID
1073                  */
1074                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1075                         ipv6_addr_orchid(dst->addr));
1076                 break;
1077         case IPV6_SADDR_RULE_PREFIX:
1078                 /* Rule 8: Use longest matching prefix */
1079                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1080                 if (ret > score->ifa->prefix_len)
1081                         ret = score->ifa->prefix_len;
1082                 score->matchlen = ret;
1083                 break;
1084         default:
1085                 ret = 0;
1086         }
1087
1088         if (ret)
1089                 __set_bit(i, score->scorebits);
1090         score->rule = i;
1091 out:
1092         return ret;
1093 }
1094
1095 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1096                        const struct in6_addr *daddr, unsigned int prefs,
1097                        struct in6_addr *saddr)
1098 {
1099         struct ipv6_saddr_score scores[2],
1100                                 *score = &scores[0], *hiscore = &scores[1];
1101         struct ipv6_saddr_dst dst;
1102         struct net_device *dev;
1103         int dst_type;
1104
1105         dst_type = __ipv6_addr_type(daddr);
1106         dst.addr = daddr;
1107         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1108         dst.scope = __ipv6_addr_src_scope(dst_type);
1109         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1110         dst.prefs = prefs;
1111
1112         hiscore->rule = -1;
1113         hiscore->ifa = NULL;
1114
1115         rcu_read_lock();
1116
1117         for_each_netdev_rcu(net, dev) {
1118                 struct inet6_dev *idev;
1119
1120                 /* Candidate Source Address (section 4)
1121                  *  - multicast and link-local destination address,
1122                  *    the set of candidate source address MUST only
1123                  *    include addresses assigned to interfaces
1124                  *    belonging to the same link as the outgoing
1125                  *    interface.
1126                  * (- For site-local destination addresses, the
1127                  *    set of candidate source addresses MUST only
1128                  *    include addresses assigned to interfaces
1129                  *    belonging to the same site as the outgoing
1130                  *    interface.)
1131                  */
1132                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1133                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1134                     dst.ifindex && dev->ifindex != dst.ifindex)
1135                         continue;
1136
1137                 idev = __in6_dev_get(dev);
1138                 if (!idev)
1139                         continue;
1140
1141                 read_lock_bh(&idev->lock);
1142                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1143                         int i;
1144
1145                         /*
1146                          * - Tentative Address (RFC2462 section 5.4)
1147                          *  - A tentative address is not considered
1148                          *    "assigned to an interface" in the traditional
1149                          *    sense, unless it is also flagged as optimistic.
1150                          * - Candidate Source Address (section 4)
1151                          *  - In any case, anycast addresses, multicast
1152                          *    addresses, and the unspecified address MUST
1153                          *    NOT be included in a candidate set.
1154                          */
1155                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1156                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1157                                 continue;
1158
1159                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1160
1161                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1162                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1163                                 LIMIT_NETDEBUG(KERN_DEBUG
1164                                                "ADDRCONF: unspecified / multicast address "
1165                                                "assigned as unicast address on %s",
1166                                                dev->name);
1167                                 continue;
1168                         }
1169
1170                         score->rule = -1;
1171                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1172
1173                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1174                                 int minihiscore, miniscore;
1175
1176                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1177                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1178
1179                                 if (minihiscore > miniscore) {
1180                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1181                                             score->scopedist > 0) {
1182                                                 /*
1183                                                  * special case:
1184                                                  * each remaining entry
1185                                                  * has too small (not enough)
1186                                                  * scope, because ifa entries
1187                                                  * are sorted by their scope
1188                                                  * values.
1189                                                  */
1190                                                 goto try_nextdev;
1191                                         }
1192                                         break;
1193                                 } else if (minihiscore < miniscore) {
1194                                         if (hiscore->ifa)
1195                                                 in6_ifa_put(hiscore->ifa);
1196
1197                                         in6_ifa_hold(score->ifa);
1198
1199                                         swap(hiscore, score);
1200
1201                                         /* restore our iterator */
1202                                         score->ifa = hiscore->ifa;
1203
1204                                         break;
1205                                 }
1206                         }
1207                 }
1208 try_nextdev:
1209                 read_unlock_bh(&idev->lock);
1210         }
1211         rcu_read_unlock();
1212
1213         if (!hiscore->ifa)
1214                 return -EADDRNOTAVAIL;
1215
1216         *saddr = hiscore->ifa->addr;
1217         in6_ifa_put(hiscore->ifa);
1218         return 0;
1219 }
1220 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1221
1222 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1223                     unsigned char banned_flags)
1224 {
1225         struct inet6_dev *idev;
1226         int err = -EADDRNOTAVAIL;
1227
1228         rcu_read_lock();
1229         idev = __in6_dev_get(dev);
1230         if (idev) {
1231                 struct inet6_ifaddr *ifp;
1232
1233                 read_lock_bh(&idev->lock);
1234                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1235                         if (ifp->scope == IFA_LINK &&
1236                             !(ifp->flags & banned_flags)) {
1237                                 *addr = ifp->addr;
1238                                 err = 0;
1239                                 break;
1240                         }
1241                 }
1242                 read_unlock_bh(&idev->lock);
1243         }
1244         rcu_read_unlock();
1245         return err;
1246 }
1247
1248 static int ipv6_count_addresses(struct inet6_dev *idev)
1249 {
1250         int cnt = 0;
1251         struct inet6_ifaddr *ifp;
1252
1253         read_lock_bh(&idev->lock);
1254         list_for_each_entry(ifp, &idev->addr_list, if_list)
1255                 cnt++;
1256         read_unlock_bh(&idev->lock);
1257         return cnt;
1258 }
1259
1260 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1261                   struct net_device *dev, int strict)
1262 {
1263         struct inet6_ifaddr *ifp;
1264         struct hlist_node *node;
1265         unsigned int hash = inet6_addr_hash(addr);
1266
1267         rcu_read_lock_bh();
1268         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1269                 if (!net_eq(dev_net(ifp->idev->dev), net))
1270                         continue;
1271                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1272                     !(ifp->flags&IFA_F_TENTATIVE) &&
1273                     (dev == NULL || ifp->idev->dev == dev ||
1274                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1275                         rcu_read_unlock_bh();
1276                         return 1;
1277                 }
1278         }
1279
1280         rcu_read_unlock_bh();
1281         return 0;
1282 }
1283 EXPORT_SYMBOL(ipv6_chk_addr);
1284
1285 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1286                                struct net_device *dev)
1287 {
1288         unsigned int hash = inet6_addr_hash(addr);
1289         struct inet6_ifaddr *ifp;
1290         struct hlist_node *node;
1291
1292         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1293                 if (!net_eq(dev_net(ifp->idev->dev), net))
1294                         continue;
1295                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1296                         if (dev == NULL || ifp->idev->dev == dev)
1297                                 return true;
1298                 }
1299         }
1300         return false;
1301 }
1302
1303 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1304 {
1305         struct inet6_dev *idev;
1306         struct inet6_ifaddr *ifa;
1307         int     onlink;
1308
1309         onlink = 0;
1310         rcu_read_lock();
1311         idev = __in6_dev_get(dev);
1312         if (idev) {
1313                 read_lock_bh(&idev->lock);
1314                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1315                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1316                                                    ifa->prefix_len);
1317                         if (onlink)
1318                                 break;
1319                 }
1320                 read_unlock_bh(&idev->lock);
1321         }
1322         rcu_read_unlock();
1323         return onlink;
1324 }
1325 EXPORT_SYMBOL(ipv6_chk_prefix);
1326
1327 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1328                                      struct net_device *dev, int strict)
1329 {
1330         struct inet6_ifaddr *ifp, *result = NULL;
1331         unsigned int hash = inet6_addr_hash(addr);
1332         struct hlist_node *node;
1333
1334         rcu_read_lock_bh();
1335         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1336                 if (!net_eq(dev_net(ifp->idev->dev), net))
1337                         continue;
1338                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1339                         if (dev == NULL || ifp->idev->dev == dev ||
1340                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1341                                 result = ifp;
1342                                 in6_ifa_hold(ifp);
1343                                 break;
1344                         }
1345                 }
1346         }
1347         rcu_read_unlock_bh();
1348
1349         return result;
1350 }
1351
1352 /* Gets referenced address, destroys ifaddr */
1353
1354 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1355 {
1356         if (ifp->flags&IFA_F_PERMANENT) {
1357                 spin_lock_bh(&ifp->lock);
1358                 addrconf_del_timer(ifp);
1359                 ifp->flags |= IFA_F_TENTATIVE;
1360                 if (dad_failed)
1361                         ifp->flags |= IFA_F_DADFAILED;
1362                 spin_unlock_bh(&ifp->lock);
1363                 if (dad_failed)
1364                         ipv6_ifa_notify(0, ifp);
1365                 in6_ifa_put(ifp);
1366 #ifdef CONFIG_IPV6_PRIVACY
1367         } else if (ifp->flags&IFA_F_TEMPORARY) {
1368                 struct inet6_ifaddr *ifpub;
1369                 spin_lock_bh(&ifp->lock);
1370                 ifpub = ifp->ifpub;
1371                 if (ifpub) {
1372                         in6_ifa_hold(ifpub);
1373                         spin_unlock_bh(&ifp->lock);
1374                         ipv6_create_tempaddr(ifpub, ifp);
1375                         in6_ifa_put(ifpub);
1376                 } else {
1377                         spin_unlock_bh(&ifp->lock);
1378                 }
1379                 ipv6_del_addr(ifp);
1380 #endif
1381         } else
1382                 ipv6_del_addr(ifp);
1383 }
1384
1385 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1386 {
1387         int err = -ENOENT;
1388
1389         spin_lock(&ifp->state_lock);
1390         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1391                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1392                 err = 0;
1393         }
1394         spin_unlock(&ifp->state_lock);
1395
1396         return err;
1397 }
1398
1399 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1400 {
1401         struct inet6_dev *idev = ifp->idev;
1402
1403         if (addrconf_dad_end(ifp)) {
1404                 in6_ifa_put(ifp);
1405                 return;
1406         }
1407
1408         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1409                              ifp->idev->dev->name, &ifp->addr);
1410
1411         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1412                 struct in6_addr addr;
1413
1414                 addr.s6_addr32[0] = htonl(0xfe800000);
1415                 addr.s6_addr32[1] = 0;
1416
1417                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1418                     ipv6_addr_equal(&ifp->addr, &addr)) {
1419                         /* DAD failed for link-local based on MAC address */
1420                         idev->cnf.disable_ipv6 = 1;
1421
1422                         pr_info("%s: IPv6 being disabled!\n",
1423                                 ifp->idev->dev->name);
1424                 }
1425         }
1426
1427         addrconf_dad_stop(ifp, 1);
1428 }
1429
1430 /* Join to solicited addr multicast group. */
1431
1432 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1433 {
1434         struct in6_addr maddr;
1435
1436         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1437                 return;
1438
1439         addrconf_addr_solict_mult(addr, &maddr);
1440         ipv6_dev_mc_inc(dev, &maddr);
1441 }
1442
1443 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1444 {
1445         struct in6_addr maddr;
1446
1447         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1448                 return;
1449
1450         addrconf_addr_solict_mult(addr, &maddr);
1451         __ipv6_dev_mc_dec(idev, &maddr);
1452 }
1453
1454 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1455 {
1456         struct in6_addr addr;
1457         if (ifp->prefix_len == 127) /* RFC 6164 */
1458                 return;
1459         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1460         if (ipv6_addr_any(&addr))
1461                 return;
1462         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1463 }
1464
1465 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1466 {
1467         struct in6_addr addr;
1468         if (ifp->prefix_len == 127) /* RFC 6164 */
1469                 return;
1470         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1471         if (ipv6_addr_any(&addr))
1472                 return;
1473         __ipv6_dev_ac_dec(ifp->idev, &addr);
1474 }
1475
1476 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1477 {
1478         if (dev->addr_len != ETH_ALEN)
1479                 return -1;
1480         memcpy(eui, dev->dev_addr, 3);
1481         memcpy(eui + 5, dev->dev_addr + 3, 3);
1482
1483         /*
1484          * The zSeries OSA network cards can be shared among various
1485          * OS instances, but the OSA cards have only one MAC address.
1486          * This leads to duplicate address conflicts in conjunction
1487          * with IPv6 if more than one instance uses the same card.
1488          *
1489          * The driver for these cards can deliver a unique 16-bit
1490          * identifier for each instance sharing the same card.  It is
1491          * placed instead of 0xFFFE in the interface identifier.  The
1492          * "u" bit of the interface identifier is not inverted in this
1493          * case.  Hence the resulting interface identifier has local
1494          * scope according to RFC2373.
1495          */
1496         if (dev->dev_id) {
1497                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1498                 eui[4] = dev->dev_id & 0xFF;
1499         } else {
1500                 eui[3] = 0xFF;
1501                 eui[4] = 0xFE;
1502                 eui[0] ^= 2;
1503         }
1504         return 0;
1505 }
1506
1507 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1508 {
1509         if (dev->addr_len != IEEE802154_ADDR_LEN)
1510                 return -1;
1511         memcpy(eui, dev->dev_addr, 8);
1512         return 0;
1513 }
1514
1515 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1516 {
1517         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1518         if (dev->addr_len != ARCNET_ALEN)
1519                 return -1;
1520         memset(eui, 0, 7);
1521         eui[7] = *(u8 *)dev->dev_addr;
1522         return 0;
1523 }
1524
1525 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1526 {
1527         if (dev->addr_len != INFINIBAND_ALEN)
1528                 return -1;
1529         memcpy(eui, dev->dev_addr + 12, 8);
1530         eui[0] |= 2;
1531         return 0;
1532 }
1533
1534 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1535 {
1536         if (addr == 0)
1537                 return -1;
1538         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1539                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1540                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1541                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1542                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1543                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1544         eui[1] = 0;
1545         eui[2] = 0x5E;
1546         eui[3] = 0xFE;
1547         memcpy(eui + 4, &addr, 4);
1548         return 0;
1549 }
1550
1551 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1552 {
1553         if (dev->priv_flags & IFF_ISATAP)
1554                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1555         return -1;
1556 }
1557
1558 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1559 {
1560         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1561 }
1562
1563 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1564 {
1565         switch (dev->type) {
1566         case ARPHRD_ETHER:
1567         case ARPHRD_FDDI:
1568                 return addrconf_ifid_eui48(eui, dev);
1569         case ARPHRD_ARCNET:
1570                 return addrconf_ifid_arcnet(eui, dev);
1571         case ARPHRD_INFINIBAND:
1572                 return addrconf_ifid_infiniband(eui, dev);
1573         case ARPHRD_SIT:
1574                 return addrconf_ifid_sit(eui, dev);
1575         case ARPHRD_IPGRE:
1576                 return addrconf_ifid_gre(eui, dev);
1577         case ARPHRD_IEEE802154:
1578                 return addrconf_ifid_eui64(eui, dev);
1579         }
1580         return -1;
1581 }
1582
1583 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1584 {
1585         int err = -1;
1586         struct inet6_ifaddr *ifp;
1587
1588         read_lock_bh(&idev->lock);
1589         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1590                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1591                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1592                         err = 0;
1593                         break;
1594                 }
1595         }
1596         read_unlock_bh(&idev->lock);
1597         return err;
1598 }
1599
1600 #ifdef CONFIG_IPV6_PRIVACY
1601 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1602 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1603 {
1604 regen:
1605         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1606         idev->rndid[0] &= ~0x02;
1607
1608         /*
1609          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1610          * check if generated address is not inappropriate
1611          *
1612          *  - Reserved subnet anycast (RFC 2526)
1613          *      11111101 11....11 1xxxxxxx
1614          *  - ISATAP (RFC4214) 6.1
1615          *      00-00-5E-FE-xx-xx-xx-xx
1616          *  - value 0
1617          *  - XXX: already assigned to an address on the device
1618          */
1619         if (idev->rndid[0] == 0xfd &&
1620             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1621             (idev->rndid[7]&0x80))
1622                 goto regen;
1623         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1624                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1625                         goto regen;
1626                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1627                         goto regen;
1628         }
1629 }
1630
1631 static void ipv6_regen_rndid(unsigned long data)
1632 {
1633         struct inet6_dev *idev = (struct inet6_dev *) data;
1634         unsigned long expires;
1635
1636         rcu_read_lock_bh();
1637         write_lock_bh(&idev->lock);
1638
1639         if (idev->dead)
1640                 goto out;
1641
1642         __ipv6_regen_rndid(idev);
1643
1644         expires = jiffies +
1645                 idev->cnf.temp_prefered_lft * HZ -
1646                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1647                 idev->cnf.max_desync_factor * HZ;
1648         if (time_before(expires, jiffies)) {
1649                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1650                         __func__, idev->dev->name);
1651                 goto out;
1652         }
1653
1654         if (!mod_timer(&idev->regen_timer, expires))
1655                 in6_dev_hold(idev);
1656
1657 out:
1658         write_unlock_bh(&idev->lock);
1659         rcu_read_unlock_bh();
1660         in6_dev_put(idev);
1661 }
1662
1663 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1664 {
1665         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1666                 __ipv6_regen_rndid(idev);
1667 }
1668 #endif
1669
1670 /*
1671  *      Add prefix route.
1672  */
1673
1674 static void
1675 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1676                       unsigned long expires, u32 flags)
1677 {
1678         struct fib6_config cfg = {
1679                 .fc_table = RT6_TABLE_PREFIX,
1680                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1681                 .fc_ifindex = dev->ifindex,
1682                 .fc_expires = expires,
1683                 .fc_dst_len = plen,
1684                 .fc_flags = RTF_UP | flags,
1685                 .fc_nlinfo.nl_net = dev_net(dev),
1686                 .fc_protocol = RTPROT_KERNEL,
1687         };
1688
1689         cfg.fc_dst = *pfx;
1690
1691         /* Prevent useless cloning on PtP SIT.
1692            This thing is done here expecting that the whole
1693            class of non-broadcast devices need not cloning.
1694          */
1695 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1696         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1697                 cfg.fc_flags |= RTF_NONEXTHOP;
1698 #endif
1699
1700         ip6_route_add(&cfg);
1701 }
1702
1703
1704 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1705                                                   int plen,
1706                                                   const struct net_device *dev,
1707                                                   u32 flags, u32 noflags)
1708 {
1709         struct fib6_node *fn;
1710         struct rt6_info *rt = NULL;
1711         struct fib6_table *table;
1712
1713         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1714         if (table == NULL)
1715                 return NULL;
1716
1717         read_lock_bh(&table->tb6_lock);
1718         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1719         if (!fn)
1720                 goto out;
1721         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1722                 if (rt->dst.dev->ifindex != dev->ifindex)
1723                         continue;
1724                 if ((rt->rt6i_flags & flags) != flags)
1725                         continue;
1726                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1727                         continue;
1728                 dst_hold(&rt->dst);
1729                 break;
1730         }
1731 out:
1732         read_unlock_bh(&table->tb6_lock);
1733         return rt;
1734 }
1735
1736
1737 /* Create "default" multicast route to the interface */
1738
1739 static void addrconf_add_mroute(struct net_device *dev)
1740 {
1741         struct fib6_config cfg = {
1742                 .fc_table = RT6_TABLE_LOCAL,
1743                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1744                 .fc_ifindex = dev->ifindex,
1745                 .fc_dst_len = 8,
1746                 .fc_flags = RTF_UP,
1747                 .fc_nlinfo.nl_net = dev_net(dev),
1748         };
1749
1750         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1751
1752         ip6_route_add(&cfg);
1753 }
1754
1755 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1756 static void sit_route_add(struct net_device *dev)
1757 {
1758         struct fib6_config cfg = {
1759                 .fc_table = RT6_TABLE_MAIN,
1760                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1761                 .fc_ifindex = dev->ifindex,
1762                 .fc_dst_len = 96,
1763                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1764                 .fc_nlinfo.nl_net = dev_net(dev),
1765         };
1766
1767         /* prefix length - 96 bits "::d.d.d.d" */
1768         ip6_route_add(&cfg);
1769 }
1770 #endif
1771
1772 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1773 {
1774         struct inet6_dev *idev;
1775
1776         ASSERT_RTNL();
1777
1778         idev = ipv6_find_idev(dev);
1779         if (!idev)
1780                 return ERR_PTR(-ENOBUFS);
1781
1782         if (idev->cnf.disable_ipv6)
1783                 return ERR_PTR(-EACCES);
1784
1785         /* Add default multicast route */
1786         if (!(dev->flags & IFF_LOOPBACK))
1787                 addrconf_add_mroute(dev);
1788
1789         return idev;
1790 }
1791
1792 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1793 {
1794         struct prefix_info *pinfo;
1795         __u32 valid_lft;
1796         __u32 prefered_lft;
1797         int addr_type;
1798         struct inet6_dev *in6_dev;
1799         struct net *net = dev_net(dev);
1800
1801         pinfo = (struct prefix_info *) opt;
1802
1803         if (len < sizeof(struct prefix_info)) {
1804                 ADBG(("addrconf: prefix option too short\n"));
1805                 return;
1806         }
1807
1808         /*
1809          *      Validation checks ([ADDRCONF], page 19)
1810          */
1811
1812         addr_type = ipv6_addr_type(&pinfo->prefix);
1813
1814         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1815                 return;
1816
1817         valid_lft = ntohl(pinfo->valid);
1818         prefered_lft = ntohl(pinfo->prefered);
1819
1820         if (prefered_lft > valid_lft) {
1821                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1822                 return;
1823         }
1824
1825         in6_dev = in6_dev_get(dev);
1826
1827         if (in6_dev == NULL) {
1828                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1829                                     dev->name);
1830                 return;
1831         }
1832
1833         /*
1834          *      Two things going on here:
1835          *      1) Add routes for on-link prefixes
1836          *      2) Configure prefixes with the auto flag set
1837          */
1838
1839         if (pinfo->onlink) {
1840                 struct rt6_info *rt;
1841                 unsigned long rt_expires;
1842
1843                 /* Avoid arithmetic overflow. Really, we could
1844                  * save rt_expires in seconds, likely valid_lft,
1845                  * but it would require division in fib gc, that it
1846                  * not good.
1847                  */
1848                 if (HZ > USER_HZ)
1849                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1850                 else
1851                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1852
1853                 if (addrconf_finite_timeout(rt_expires))
1854                         rt_expires *= HZ;
1855
1856                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1857                                                pinfo->prefix_len,
1858                                                dev,
1859                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1860                                                RTF_GATEWAY | RTF_DEFAULT);
1861
1862                 if (rt) {
1863                         /* Autoconf prefix route */
1864                         if (valid_lft == 0) {
1865                                 ip6_del_rt(rt);
1866                                 rt = NULL;
1867                         } else if (addrconf_finite_timeout(rt_expires)) {
1868                                 /* not infinity */
1869                                 rt6_set_expires(rt, jiffies + rt_expires);
1870                         } else {
1871                                 rt6_clean_expires(rt);
1872                         }
1873                 } else if (valid_lft) {
1874                         clock_t expires = 0;
1875                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1876                         if (addrconf_finite_timeout(rt_expires)) {
1877                                 /* not infinity */
1878                                 flags |= RTF_EXPIRES;
1879                                 expires = jiffies_to_clock_t(rt_expires);
1880                         }
1881                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1882                                               dev, expires, flags);
1883                 }
1884                 if (rt)
1885                         dst_release(&rt->dst);
1886         }
1887
1888         /* Try to figure out our local address for this prefix */
1889
1890         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1891                 struct inet6_ifaddr *ifp;
1892                 struct in6_addr addr;
1893                 int create = 0, update_lft = 0;
1894
1895                 if (pinfo->prefix_len == 64) {
1896                         memcpy(&addr, &pinfo->prefix, 8);
1897                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1898                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1899                                 in6_dev_put(in6_dev);
1900                                 return;
1901                         }
1902                         goto ok;
1903                 }
1904                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
1905                                     pinfo->prefix_len);
1906                 in6_dev_put(in6_dev);
1907                 return;
1908
1909 ok:
1910
1911                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1912
1913                 if (ifp == NULL && valid_lft) {
1914                         int max_addresses = in6_dev->cnf.max_addresses;
1915                         u32 addr_flags = 0;
1916
1917 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1918                         if (in6_dev->cnf.optimistic_dad &&
1919                             !net->ipv6.devconf_all->forwarding && sllao)
1920                                 addr_flags = IFA_F_OPTIMISTIC;
1921 #endif
1922
1923                         /* Do not allow to create too much of autoconfigured
1924                          * addresses; this would be too easy way to crash kernel.
1925                          */
1926                         if (!max_addresses ||
1927                             ipv6_count_addresses(in6_dev) < max_addresses)
1928                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1929                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1930                                                     addr_flags);
1931
1932                         if (!ifp || IS_ERR(ifp)) {
1933                                 in6_dev_put(in6_dev);
1934                                 return;
1935                         }
1936
1937                         update_lft = create = 1;
1938                         ifp->cstamp = jiffies;
1939                         addrconf_dad_start(ifp);
1940                 }
1941
1942                 if (ifp) {
1943                         int flags;
1944                         unsigned long now;
1945 #ifdef CONFIG_IPV6_PRIVACY
1946                         struct inet6_ifaddr *ift;
1947 #endif
1948                         u32 stored_lft;
1949
1950                         /* update lifetime (RFC2462 5.5.3 e) */
1951                         spin_lock(&ifp->lock);
1952                         now = jiffies;
1953                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1954                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1955                         else
1956                                 stored_lft = 0;
1957                         if (!update_lft && stored_lft) {
1958                                 if (valid_lft > MIN_VALID_LIFETIME ||
1959                                     valid_lft > stored_lft)
1960                                         update_lft = 1;
1961                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1962                                         /* valid_lft <= stored_lft is always true */
1963                                         /*
1964                                          * RFC 4862 Section 5.5.3e:
1965                                          * "Note that the preferred lifetime of
1966                                          *  the corresponding address is always
1967                                          *  reset to the Preferred Lifetime in
1968                                          *  the received Prefix Information
1969                                          *  option, regardless of whether the
1970                                          *  valid lifetime is also reset or
1971                                          *  ignored."
1972                                          *
1973                                          *  So if the preferred lifetime in
1974                                          *  this advertisement is different
1975                                          *  than what we have stored, but the
1976                                          *  valid lifetime is invalid, just
1977                                          *  reset prefered_lft.
1978                                          *
1979                                          *  We must set the valid lifetime
1980                                          *  to the stored lifetime since we'll
1981                                          *  be updating the timestamp below,
1982                                          *  else we'll set it back to the
1983                                          *  minimum.
1984                                          */
1985                                         if (prefered_lft != ifp->prefered_lft) {
1986                                                 valid_lft = stored_lft;
1987                                                 update_lft = 1;
1988                                         }
1989                                 } else {
1990                                         valid_lft = MIN_VALID_LIFETIME;
1991                                         if (valid_lft < prefered_lft)
1992                                                 prefered_lft = valid_lft;
1993                                         update_lft = 1;
1994                                 }
1995                         }
1996
1997                         if (update_lft) {
1998                                 ifp->valid_lft = valid_lft;
1999                                 ifp->prefered_lft = prefered_lft;
2000                                 ifp->tstamp = now;
2001                                 flags = ifp->flags;
2002                                 ifp->flags &= ~IFA_F_DEPRECATED;
2003                                 spin_unlock(&ifp->lock);
2004
2005                                 if (!(flags&IFA_F_TENTATIVE))
2006                                         ipv6_ifa_notify(0, ifp);
2007                         } else
2008                                 spin_unlock(&ifp->lock);
2009
2010 #ifdef CONFIG_IPV6_PRIVACY
2011                         read_lock_bh(&in6_dev->lock);
2012                         /* update all temporary addresses in the list */
2013                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2014                                             tmp_list) {
2015                                 int age, max_valid, max_prefered;
2016
2017                                 if (ifp != ift->ifpub)
2018                                         continue;
2019
2020                                 /*
2021                                  * RFC 4941 section 3.3:
2022                                  * If a received option will extend the lifetime
2023                                  * of a public address, the lifetimes of
2024                                  * temporary addresses should be extended,
2025                                  * subject to the overall constraint that no
2026                                  * temporary addresses should ever remain
2027                                  * "valid" or "preferred" for a time longer than
2028                                  * (TEMP_VALID_LIFETIME) or
2029                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2030                                  * respectively.
2031                                  */
2032                                 age = (now - ift->cstamp) / HZ;
2033                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2034                                 if (max_valid < 0)
2035                                         max_valid = 0;
2036
2037                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2038                                                in6_dev->cnf.max_desync_factor -
2039                                                age;
2040                                 if (max_prefered < 0)
2041                                         max_prefered = 0;
2042
2043                                 if (valid_lft > max_valid)
2044                                         valid_lft = max_valid;
2045
2046                                 if (prefered_lft > max_prefered)
2047                                         prefered_lft = max_prefered;
2048
2049                                 spin_lock(&ift->lock);
2050                                 flags = ift->flags;
2051                                 ift->valid_lft = valid_lft;
2052                                 ift->prefered_lft = prefered_lft;
2053                                 ift->tstamp = now;
2054                                 if (prefered_lft > 0)
2055                                         ift->flags &= ~IFA_F_DEPRECATED;
2056
2057                                 spin_unlock(&ift->lock);
2058                                 if (!(flags&IFA_F_TENTATIVE))
2059                                         ipv6_ifa_notify(0, ift);
2060                         }
2061
2062                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2063                                 /*
2064                                  * When a new public address is created as
2065                                  * described in [ADDRCONF], also create a new
2066                                  * temporary address. Also create a temporary
2067                                  * address if it's enabled but no temporary
2068                                  * address currently exists.
2069                                  */
2070                                 read_unlock_bh(&in6_dev->lock);
2071                                 ipv6_create_tempaddr(ifp, NULL);
2072                         } else {
2073                                 read_unlock_bh(&in6_dev->lock);
2074                         }
2075 #endif
2076                         in6_ifa_put(ifp);
2077                         addrconf_verify(0);
2078                 }
2079         }
2080         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2081         in6_dev_put(in6_dev);
2082 }
2083
2084 /*
2085  *      Set destination address.
2086  *      Special case for SIT interfaces where we create a new "virtual"
2087  *      device.
2088  */
2089 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2090 {
2091         struct in6_ifreq ireq;
2092         struct net_device *dev;
2093         int err = -EINVAL;
2094
2095         rtnl_lock();
2096
2097         err = -EFAULT;
2098         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2099                 goto err_exit;
2100
2101         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2102
2103         err = -ENODEV;
2104         if (dev == NULL)
2105                 goto err_exit;
2106
2107 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2108         if (dev->type == ARPHRD_SIT) {
2109                 const struct net_device_ops *ops = dev->netdev_ops;
2110                 struct ifreq ifr;
2111                 struct ip_tunnel_parm p;
2112
2113                 err = -EADDRNOTAVAIL;
2114                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2115                         goto err_exit;
2116
2117                 memset(&p, 0, sizeof(p));
2118                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2119                 p.iph.saddr = 0;
2120                 p.iph.version = 4;
2121                 p.iph.ihl = 5;
2122                 p.iph.protocol = IPPROTO_IPV6;
2123                 p.iph.ttl = 64;
2124                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2125
2126                 if (ops->ndo_do_ioctl) {
2127                         mm_segment_t oldfs = get_fs();
2128
2129                         set_fs(KERNEL_DS);
2130                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2131                         set_fs(oldfs);
2132                 } else
2133                         err = -EOPNOTSUPP;
2134
2135                 if (err == 0) {
2136                         err = -ENOBUFS;
2137                         dev = __dev_get_by_name(net, p.name);
2138                         if (!dev)
2139                                 goto err_exit;
2140                         err = dev_open(dev);
2141                 }
2142         }
2143 #endif
2144
2145 err_exit:
2146         rtnl_unlock();
2147         return err;
2148 }
2149
2150 /*
2151  *      Manual configuration of address on an interface
2152  */
2153 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2154                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2155                           __u32 valid_lft)
2156 {
2157         struct inet6_ifaddr *ifp;
2158         struct inet6_dev *idev;
2159         struct net_device *dev;
2160         int scope;
2161         u32 flags;
2162         clock_t expires;
2163         unsigned long timeout;
2164
2165         ASSERT_RTNL();
2166
2167         if (plen > 128)
2168                 return -EINVAL;
2169
2170         /* check the lifetime */
2171         if (!valid_lft || prefered_lft > valid_lft)
2172                 return -EINVAL;
2173
2174         dev = __dev_get_by_index(net, ifindex);
2175         if (!dev)
2176                 return -ENODEV;
2177
2178         idev = addrconf_add_dev(dev);
2179         if (IS_ERR(idev))
2180                 return PTR_ERR(idev);
2181
2182         scope = ipv6_addr_scope(pfx);
2183
2184         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2185         if (addrconf_finite_timeout(timeout)) {
2186                 expires = jiffies_to_clock_t(timeout * HZ);
2187                 valid_lft = timeout;
2188                 flags = RTF_EXPIRES;
2189         } else {
2190                 expires = 0;
2191                 flags = 0;
2192                 ifa_flags |= IFA_F_PERMANENT;
2193         }
2194
2195         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2196         if (addrconf_finite_timeout(timeout)) {
2197                 if (timeout == 0)
2198                         ifa_flags |= IFA_F_DEPRECATED;
2199                 prefered_lft = timeout;
2200         }
2201
2202         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2203
2204         if (!IS_ERR(ifp)) {
2205                 spin_lock_bh(&ifp->lock);
2206                 ifp->valid_lft = valid_lft;
2207                 ifp->prefered_lft = prefered_lft;
2208                 ifp->tstamp = jiffies;
2209                 spin_unlock_bh(&ifp->lock);
2210
2211                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2212                                       expires, flags);
2213                 /*
2214                  * Note that section 3.1 of RFC 4429 indicates
2215                  * that the Optimistic flag should not be set for
2216                  * manually configured addresses
2217                  */
2218                 addrconf_dad_start(ifp);
2219                 in6_ifa_put(ifp);
2220                 addrconf_verify(0);
2221                 return 0;
2222         }
2223
2224         return PTR_ERR(ifp);
2225 }
2226
2227 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2228                           unsigned int plen)
2229 {
2230         struct inet6_ifaddr *ifp;
2231         struct inet6_dev *idev;
2232         struct net_device *dev;
2233
2234         if (plen > 128)
2235                 return -EINVAL;
2236
2237         dev = __dev_get_by_index(net, ifindex);
2238         if (!dev)
2239                 return -ENODEV;
2240
2241         if ((idev = __in6_dev_get(dev)) == NULL)
2242                 return -ENXIO;
2243
2244         read_lock_bh(&idev->lock);
2245         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2246                 if (ifp->prefix_len == plen &&
2247                     ipv6_addr_equal(pfx, &ifp->addr)) {
2248                         in6_ifa_hold(ifp);
2249                         read_unlock_bh(&idev->lock);
2250
2251                         ipv6_del_addr(ifp);
2252
2253                         /* If the last address is deleted administratively,
2254                            disable IPv6 on this interface.
2255                          */
2256                         if (list_empty(&idev->addr_list))
2257                                 addrconf_ifdown(idev->dev, 1);
2258                         return 0;
2259                 }
2260         }
2261         read_unlock_bh(&idev->lock);
2262         return -EADDRNOTAVAIL;
2263 }
2264
2265
2266 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2267 {
2268         struct in6_ifreq ireq;
2269         int err;
2270
2271         if (!capable(CAP_NET_ADMIN))
2272                 return -EPERM;
2273
2274         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2275                 return -EFAULT;
2276
2277         rtnl_lock();
2278         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2279                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2280                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2281         rtnl_unlock();
2282         return err;
2283 }
2284
2285 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2286 {
2287         struct in6_ifreq ireq;
2288         int err;
2289
2290         if (!capable(CAP_NET_ADMIN))
2291                 return -EPERM;
2292
2293         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2294                 return -EFAULT;
2295
2296         rtnl_lock();
2297         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2298                              ireq.ifr6_prefixlen);
2299         rtnl_unlock();
2300         return err;
2301 }
2302
2303 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2304                      int plen, int scope)
2305 {
2306         struct inet6_ifaddr *ifp;
2307
2308         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2309         if (!IS_ERR(ifp)) {
2310                 spin_lock_bh(&ifp->lock);
2311                 ifp->flags &= ~IFA_F_TENTATIVE;
2312                 spin_unlock_bh(&ifp->lock);
2313                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2314                 in6_ifa_put(ifp);
2315         }
2316 }
2317
2318 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2319 static void sit_add_v4_addrs(struct inet6_dev *idev)
2320 {
2321         struct in6_addr addr;
2322         struct net_device *dev;
2323         struct net *net = dev_net(idev->dev);
2324         int scope;
2325
2326         ASSERT_RTNL();
2327
2328         memset(&addr, 0, sizeof(struct in6_addr));
2329         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2330
2331         if (idev->dev->flags&IFF_POINTOPOINT) {
2332                 addr.s6_addr32[0] = htonl(0xfe800000);
2333                 scope = IFA_LINK;
2334         } else {
2335                 scope = IPV6_ADDR_COMPATv4;
2336         }
2337
2338         if (addr.s6_addr32[3]) {
2339                 add_addr(idev, &addr, 128, scope);
2340                 return;
2341         }
2342
2343         for_each_netdev(net, dev) {
2344                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2345                 if (in_dev && (dev->flags & IFF_UP)) {
2346                         struct in_ifaddr *ifa;
2347
2348                         int flag = scope;
2349
2350                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2351                                 int plen;
2352
2353                                 addr.s6_addr32[3] = ifa->ifa_local;
2354
2355                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2356                                         continue;
2357                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2358                                         if (idev->dev->flags&IFF_POINTOPOINT)
2359                                                 continue;
2360                                         flag |= IFA_HOST;
2361                                 }
2362                                 if (idev->dev->flags&IFF_POINTOPOINT)
2363                                         plen = 64;
2364                                 else
2365                                         plen = 96;
2366
2367                                 add_addr(idev, &addr, plen, flag);
2368                         }
2369                 }
2370         }
2371 }
2372 #endif
2373
2374 static void init_loopback(struct net_device *dev)
2375 {
2376         struct inet6_dev  *idev;
2377
2378         /* ::1 */
2379
2380         ASSERT_RTNL();
2381
2382         if ((idev = ipv6_find_idev(dev)) == NULL) {
2383                 pr_debug("%s: add_dev failed\n", __func__);
2384                 return;
2385         }
2386
2387         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2388 }
2389
2390 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2391 {
2392         struct inet6_ifaddr *ifp;
2393         u32 addr_flags = IFA_F_PERMANENT;
2394
2395 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2396         if (idev->cnf.optimistic_dad &&
2397             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2398                 addr_flags |= IFA_F_OPTIMISTIC;
2399 #endif
2400
2401
2402         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2403         if (!IS_ERR(ifp)) {
2404                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2405                 addrconf_dad_start(ifp);
2406                 in6_ifa_put(ifp);
2407         }
2408 }
2409
2410 static void addrconf_dev_config(struct net_device *dev)
2411 {
2412         struct in6_addr addr;
2413         struct inet6_dev *idev;
2414
2415         ASSERT_RTNL();
2416
2417         if ((dev->type != ARPHRD_ETHER) &&
2418             (dev->type != ARPHRD_FDDI) &&
2419             (dev->type != ARPHRD_ARCNET) &&
2420             (dev->type != ARPHRD_INFINIBAND) &&
2421             (dev->type != ARPHRD_IEEE802154)) {
2422                 /* Alas, we support only Ethernet autoconfiguration. */
2423                 return;
2424         }
2425
2426         idev = addrconf_add_dev(dev);
2427         if (IS_ERR(idev))
2428                 return;
2429
2430         memset(&addr, 0, sizeof(struct in6_addr));
2431         addr.s6_addr32[0] = htonl(0xFE800000);
2432
2433         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2434                 addrconf_add_linklocal(idev, &addr);
2435 }
2436
2437 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2438 static void addrconf_sit_config(struct net_device *dev)
2439 {
2440         struct inet6_dev *idev;
2441
2442         ASSERT_RTNL();
2443
2444         /*
2445          * Configure the tunnel with one of our IPv4
2446          * addresses... we should configure all of
2447          * our v4 addrs in the tunnel
2448          */
2449
2450         if ((idev = ipv6_find_idev(dev)) == NULL) {
2451                 pr_debug("%s: add_dev failed\n", __func__);
2452                 return;
2453         }
2454
2455         if (dev->priv_flags & IFF_ISATAP) {
2456                 struct in6_addr addr;
2457
2458                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2459                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2460                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2461                         addrconf_add_linklocal(idev, &addr);
2462                 return;
2463         }
2464
2465         sit_add_v4_addrs(idev);
2466
2467         if (dev->flags&IFF_POINTOPOINT)
2468                 addrconf_add_mroute(dev);
2469         else
2470                 sit_route_add(dev);
2471 }
2472 #endif
2473
2474 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2475 static void addrconf_gre_config(struct net_device *dev)
2476 {
2477         struct inet6_dev *idev;
2478         struct in6_addr addr;
2479
2480         pr_info("%s(%s)\n", __func__, dev->name);
2481
2482         ASSERT_RTNL();
2483
2484         if ((idev = ipv6_find_idev(dev)) == NULL) {
2485                 pr_debug("%s: add_dev failed\n", __func__);
2486                 return;
2487         }
2488
2489         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2490         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2491
2492         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2493                 addrconf_add_linklocal(idev, &addr);
2494 }
2495 #endif
2496
2497 static inline int
2498 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2499 {
2500         struct in6_addr lladdr;
2501
2502         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2503                 addrconf_add_linklocal(idev, &lladdr);
2504                 return 0;
2505         }
2506         return -1;
2507 }
2508
2509 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2510 {
2511         struct net_device *link_dev;
2512         struct net *net = dev_net(idev->dev);
2513
2514         /* first try to inherit the link-local address from the link device */
2515         if (idev->dev->iflink &&
2516             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2517                 if (!ipv6_inherit_linklocal(idev, link_dev))
2518                         return;
2519         }
2520         /* then try to inherit it from any device */
2521         for_each_netdev(net, link_dev) {
2522                 if (!ipv6_inherit_linklocal(idev, link_dev))
2523                         return;
2524         }
2525         pr_debug("init ip6-ip6: add_linklocal failed\n");
2526 }
2527
2528 /*
2529  * Autoconfigure tunnel with a link-local address so routing protocols,
2530  * DHCPv6, MLD etc. can be run over the virtual link
2531  */
2532
2533 static void addrconf_ip6_tnl_config(struct net_device *dev)
2534 {
2535         struct inet6_dev *idev;
2536
2537         ASSERT_RTNL();
2538
2539         idev = addrconf_add_dev(dev);
2540         if (IS_ERR(idev)) {
2541                 pr_debug("init ip6-ip6: add_dev failed\n");
2542                 return;
2543         }
2544         ip6_tnl_add_linklocal(idev);
2545 }
2546
2547 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2548                            void *data)
2549 {
2550         struct net_device *dev = (struct net_device *) data;
2551         struct inet6_dev *idev = __in6_dev_get(dev);
2552         int run_pending = 0;
2553         int err;
2554
2555         switch (event) {
2556         case NETDEV_REGISTER:
2557                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2558                         idev = ipv6_add_dev(dev);
2559                         if (!idev)
2560                                 return notifier_from_errno(-ENOMEM);
2561                 }
2562                 break;
2563
2564         case NETDEV_UP:
2565         case NETDEV_CHANGE:
2566                 if (dev->flags & IFF_SLAVE)
2567                         break;
2568
2569                 if (event == NETDEV_UP) {
2570                         if (!addrconf_qdisc_ok(dev)) {
2571                                 /* device is not ready yet. */
2572                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2573                                         dev->name);
2574                                 break;
2575                         }
2576
2577                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2578                                 idev = ipv6_add_dev(dev);
2579
2580                         if (idev) {
2581                                 idev->if_flags |= IF_READY;
2582                                 run_pending = 1;
2583                         }
2584                 } else {
2585                         if (!addrconf_qdisc_ok(dev)) {
2586                                 /* device is still not ready. */
2587                                 break;
2588                         }
2589
2590                         if (idev) {
2591                                 if (idev->if_flags & IF_READY)
2592                                         /* device is already configured. */
2593                                         break;
2594                                 idev->if_flags |= IF_READY;
2595                         }
2596
2597                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2598                                 dev->name);
2599
2600                         run_pending = 1;
2601                 }
2602
2603                 switch (dev->type) {
2604 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2605                 case ARPHRD_SIT:
2606                         addrconf_sit_config(dev);
2607                         break;
2608 #endif
2609 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2610                 case ARPHRD_IPGRE:
2611                         addrconf_gre_config(dev);
2612                         break;
2613 #endif
2614                 case ARPHRD_TUNNEL6:
2615                         addrconf_ip6_tnl_config(dev);
2616                         break;
2617                 case ARPHRD_LOOPBACK:
2618                         init_loopback(dev);
2619                         break;
2620
2621                 default:
2622                         addrconf_dev_config(dev);
2623                         break;
2624                 }
2625
2626                 if (idev) {
2627                         if (run_pending)
2628                                 addrconf_dad_run(idev);
2629
2630                         /*
2631                          * If the MTU changed during the interface down,
2632                          * when the interface up, the changed MTU must be
2633                          * reflected in the idev as well as routers.
2634                          */
2635                         if (idev->cnf.mtu6 != dev->mtu &&
2636                             dev->mtu >= IPV6_MIN_MTU) {
2637                                 rt6_mtu_change(dev, dev->mtu);
2638                                 idev->cnf.mtu6 = dev->mtu;
2639                         }
2640                         idev->tstamp = jiffies;
2641                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2642
2643                         /*
2644                          * If the changed mtu during down is lower than
2645                          * IPV6_MIN_MTU stop IPv6 on this interface.
2646                          */
2647                         if (dev->mtu < IPV6_MIN_MTU)
2648                                 addrconf_ifdown(dev, 1);
2649                 }
2650                 break;
2651
2652         case NETDEV_CHANGEMTU:
2653                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2654                         rt6_mtu_change(dev, dev->mtu);
2655                         idev->cnf.mtu6 = dev->mtu;
2656                         break;
2657                 }
2658
2659                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2660                         idev = ipv6_add_dev(dev);
2661                         if (idev)
2662                                 break;
2663                 }
2664
2665                 /*
2666                  * MTU falled under IPV6_MIN_MTU.
2667                  * Stop IPv6 on this interface.
2668                  */
2669
2670         case NETDEV_DOWN:
2671         case NETDEV_UNREGISTER:
2672                 /*
2673                  *      Remove all addresses from this interface.
2674                  */
2675                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2676                 break;
2677
2678         case NETDEV_CHANGENAME:
2679                 if (idev) {
2680                         snmp6_unregister_dev(idev);
2681                         addrconf_sysctl_unregister(idev);
2682                         addrconf_sysctl_register(idev);
2683                         err = snmp6_register_dev(idev);
2684                         if (err)
2685                                 return notifier_from_errno(err);
2686                 }
2687                 break;
2688
2689         case NETDEV_PRE_TYPE_CHANGE:
2690         case NETDEV_POST_TYPE_CHANGE:
2691                 addrconf_type_change(dev, event);
2692                 break;
2693         }
2694
2695         return NOTIFY_OK;
2696 }
2697
2698 /*
2699  *      addrconf module should be notified of a device going up
2700  */
2701 static struct notifier_block ipv6_dev_notf = {
2702         .notifier_call = addrconf_notify,
2703 };
2704
2705 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2706 {
2707         struct inet6_dev *idev;
2708         ASSERT_RTNL();
2709
2710         idev = __in6_dev_get(dev);
2711
2712         if (event == NETDEV_POST_TYPE_CHANGE)
2713                 ipv6_mc_remap(idev);
2714         else if (event == NETDEV_PRE_TYPE_CHANGE)
2715                 ipv6_mc_unmap(idev);
2716 }
2717
2718 static int addrconf_ifdown(struct net_device *dev, int how)
2719 {
2720         struct net *net = dev_net(dev);
2721         struct inet6_dev *idev;
2722         struct inet6_ifaddr *ifa;
2723         int state, i;
2724
2725         ASSERT_RTNL();
2726
2727         rt6_ifdown(net, dev);
2728         neigh_ifdown(&nd_tbl, dev);
2729
2730         idev = __in6_dev_get(dev);
2731         if (idev == NULL)
2732                 return -ENODEV;
2733
2734         /*
2735          * Step 1: remove reference to ipv6 device from parent device.
2736          *         Do not dev_put!
2737          */
2738         if (how) {
2739                 idev->dead = 1;
2740
2741                 /* protected by rtnl_lock */
2742                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2743
2744                 /* Step 1.5: remove snmp6 entry */
2745                 snmp6_unregister_dev(idev);
2746
2747         }
2748
2749         /* Step 2: clear hash table */
2750         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2751                 struct hlist_head *h = &inet6_addr_lst[i];
2752                 struct hlist_node *n;
2753
2754                 spin_lock_bh(&addrconf_hash_lock);
2755         restart:
2756                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2757                         if (ifa->idev == idev) {
2758                                 hlist_del_init_rcu(&ifa->addr_lst);
2759                                 addrconf_del_timer(ifa);
2760                                 goto restart;
2761                         }
2762                 }
2763                 spin_unlock_bh(&addrconf_hash_lock);
2764         }
2765
2766         write_lock_bh(&idev->lock);
2767
2768         /* Step 2: clear flags for stateless addrconf */
2769         if (!how)
2770                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2771
2772 #ifdef CONFIG_IPV6_PRIVACY
2773         if (how && del_timer(&idev->regen_timer))
2774                 in6_dev_put(idev);
2775
2776         /* Step 3: clear tempaddr list */
2777         while (!list_empty(&idev->tempaddr_list)) {
2778                 ifa = list_first_entry(&idev->tempaddr_list,
2779                                        struct inet6_ifaddr, tmp_list);
2780                 list_del(&ifa->tmp_list);
2781                 write_unlock_bh(&idev->lock);
2782                 spin_lock_bh(&ifa->lock);
2783
2784                 if (ifa->ifpub) {
2785                         in6_ifa_put(ifa->ifpub);
2786                         ifa->ifpub = NULL;
2787                 }
2788                 spin_unlock_bh(&ifa->lock);
2789                 in6_ifa_put(ifa);
2790                 write_lock_bh(&idev->lock);
2791         }
2792 #endif
2793
2794         while (!list_empty(&idev->addr_list)) {
2795                 ifa = list_first_entry(&idev->addr_list,
2796                                        struct inet6_ifaddr, if_list);
2797                 addrconf_del_timer(ifa);
2798
2799                 list_del(&ifa->if_list);
2800
2801                 write_unlock_bh(&idev->lock);
2802
2803                 spin_lock_bh(&ifa->state_lock);
2804                 state = ifa->state;
2805                 ifa->state = INET6_IFADDR_STATE_DEAD;
2806                 spin_unlock_bh(&ifa->state_lock);
2807
2808                 if (state != INET6_IFADDR_STATE_DEAD) {
2809                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2810                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2811                 }
2812                 in6_ifa_put(ifa);
2813
2814                 write_lock_bh(&idev->lock);
2815         }
2816
2817         write_unlock_bh(&idev->lock);
2818
2819         /* Step 5: Discard multicast list */
2820         if (how)
2821                 ipv6_mc_destroy_dev(idev);
2822         else
2823                 ipv6_mc_down(idev);
2824
2825         idev->tstamp = jiffies;
2826
2827         /* Last: Shot the device (if unregistered) */
2828         if (how) {
2829                 addrconf_sysctl_unregister(idev);
2830                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2831                 neigh_ifdown(&nd_tbl, dev);
2832                 in6_dev_put(idev);
2833         }
2834         return 0;
2835 }
2836
2837 static void addrconf_rs_timer(unsigned long data)
2838 {
2839         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2840         struct inet6_dev *idev = ifp->idev;
2841
2842         read_lock(&idev->lock);
2843         if (idev->dead || !(idev->if_flags & IF_READY))
2844                 goto out;
2845
2846         if (idev->cnf.forwarding)
2847                 goto out;
2848
2849         /* Announcement received after solicitation was sent */
2850         if (idev->if_flags & IF_RA_RCVD)
2851                 goto out;
2852
2853         spin_lock(&ifp->lock);
2854         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2855                 /* The wait after the last probe can be shorter */
2856                 addrconf_mod_timer(ifp, AC_RS,
2857                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2858                                    idev->cnf.rtr_solicit_delay :
2859                                    idev->cnf.rtr_solicit_interval);
2860                 spin_unlock(&ifp->lock);
2861
2862                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2863         } else {
2864                 spin_unlock(&ifp->lock);
2865                 /*
2866                  * Note: we do not support deprecated "all on-link"
2867                  * assumption any longer.
2868                  */
2869                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
2870         }
2871
2872 out:
2873         read_unlock(&idev->lock);
2874         in6_ifa_put(ifp);
2875 }
2876
2877 /*
2878  *      Duplicate Address Detection
2879  */
2880 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2881 {
2882         unsigned long rand_num;
2883         struct inet6_dev *idev = ifp->idev;
2884
2885         if (ifp->flags & IFA_F_OPTIMISTIC)
2886                 rand_num = 0;
2887         else
2888                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2889
2890         ifp->probes = idev->cnf.dad_transmits;
2891         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2892 }
2893
2894 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
2895 {
2896         struct inet6_dev *idev = ifp->idev;
2897         struct net_device *dev = idev->dev;
2898
2899         addrconf_join_solict(dev, &ifp->addr);
2900
2901         net_srandom(ifp->addr.s6_addr32[3]);
2902
2903         read_lock_bh(&idev->lock);
2904         spin_lock(&ifp->lock);
2905         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2906                 goto out;
2907
2908         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2909             idev->cnf.accept_dad < 1 ||
2910             !(ifp->flags&IFA_F_TENTATIVE) ||
2911             ifp->flags & IFA_F_NODAD) {
2912                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2913                 spin_unlock(&ifp->lock);
2914                 read_unlock_bh(&idev->lock);
2915
2916                 addrconf_dad_completed(ifp);
2917                 return;
2918         }
2919
2920         if (!(idev->if_flags & IF_READY)) {
2921                 spin_unlock(&ifp->lock);
2922                 read_unlock_bh(&idev->lock);
2923                 /*
2924                  * If the device is not ready:
2925                  * - keep it tentative if it is a permanent address.
2926                  * - otherwise, kill it.
2927                  */
2928                 in6_ifa_hold(ifp);
2929                 addrconf_dad_stop(ifp, 0);
2930                 return;
2931         }
2932
2933         /*
2934          * Optimistic nodes can start receiving
2935          * Frames right away
2936          */
2937         if (ifp->flags & IFA_F_OPTIMISTIC)
2938                 ip6_ins_rt(ifp->rt);
2939
2940         addrconf_dad_kick(ifp);
2941 out:
2942         spin_unlock(&ifp->lock);
2943         read_unlock_bh(&idev->lock);
2944 }
2945
2946 static void addrconf_dad_timer(unsigned long data)
2947 {
2948         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2949         struct inet6_dev *idev = ifp->idev;
2950         struct in6_addr mcaddr;
2951
2952         if (!ifp->probes && addrconf_dad_end(ifp))
2953                 goto out;
2954
2955         read_lock(&idev->lock);
2956         if (idev->dead || !(idev->if_flags & IF_READY)) {
2957                 read_unlock(&idev->lock);
2958                 goto out;
2959         }
2960
2961         spin_lock(&ifp->lock);
2962         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2963                 spin_unlock(&ifp->lock);
2964                 read_unlock(&idev->lock);
2965                 goto out;
2966         }
2967
2968         if (ifp->probes == 0) {
2969                 /*
2970                  * DAD was successful
2971                  */
2972
2973                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2974                 spin_unlock(&ifp->lock);
2975                 read_unlock(&idev->lock);
2976
2977                 addrconf_dad_completed(ifp);
2978
2979                 goto out;
2980         }
2981
2982         ifp->probes--;
2983         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2984         spin_unlock(&ifp->lock);
2985         read_unlock(&idev->lock);
2986
2987         /* send a neighbour solicitation for our addr */
2988         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2989         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2990 out:
2991         in6_ifa_put(ifp);
2992 }
2993
2994 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2995 {
2996         struct net_device *dev = ifp->idev->dev;
2997
2998         /*
2999          *      Configure the address for reception. Now it is valid.
3000          */
3001
3002         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3003
3004         /* If added prefix is link local and we are prepared to process
3005            router advertisements, start sending router solicitations.
3006          */
3007
3008         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3009              ifp->idev->cnf.accept_ra == 2) &&
3010             ifp->idev->cnf.rtr_solicits > 0 &&
3011             (dev->flags&IFF_LOOPBACK) == 0 &&
3012             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3013                 /*
3014                  *      If a host as already performed a random delay
3015                  *      [...] as part of DAD [...] there is no need
3016                  *      to delay again before sending the first RS
3017                  */
3018                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3019
3020                 spin_lock_bh(&ifp->lock);
3021                 ifp->probes = 1;
3022                 ifp->idev->if_flags |= IF_RS_SENT;
3023                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3024                 spin_unlock_bh(&ifp->lock);
3025         }
3026 }
3027
3028 static void addrconf_dad_run(struct inet6_dev *idev)
3029 {
3030         struct inet6_ifaddr *ifp;
3031
3032         read_lock_bh(&idev->lock);
3033         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3034                 spin_lock(&ifp->lock);
3035                 if (ifp->flags & IFA_F_TENTATIVE &&
3036                     ifp->state == INET6_IFADDR_STATE_DAD)
3037                         addrconf_dad_kick(ifp);
3038                 spin_unlock(&ifp->lock);
3039         }
3040         read_unlock_bh(&idev->lock);
3041 }
3042
3043 #ifdef CONFIG_PROC_FS
3044 struct if6_iter_state {
3045         struct seq_net_private p;
3046         int bucket;
3047         int offset;
3048 };
3049
3050 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3051 {
3052         struct inet6_ifaddr *ifa = NULL;
3053         struct if6_iter_state *state = seq->private;
3054         struct net *net = seq_file_net(seq);
3055         int p = 0;
3056
3057         /* initial bucket if pos is 0 */
3058         if (pos == 0) {
3059                 state->bucket = 0;
3060                 state->offset = 0;
3061         }
3062
3063         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3064                 struct hlist_node *n;
3065                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3066                                          addr_lst) {
3067                         /* sync with offset */
3068                         if (p < state->offset) {
3069                                 p++;
3070                                 continue;
3071                         }
3072                         state->offset++;
3073                         if (net_eq(dev_net(ifa->idev->dev), net))
3074                                 return ifa;
3075                 }
3076
3077                 /* prepare for next bucket */
3078                 state->offset = 0;
3079                 p = 0;
3080         }
3081         return NULL;
3082 }
3083
3084 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3085                                          struct inet6_ifaddr *ifa)
3086 {
3087         struct if6_iter_state *state = seq->private;
3088         struct net *net = seq_file_net(seq);
3089         struct hlist_node *n = &ifa->addr_lst;
3090
3091         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3092                 state->offset++;
3093                 if (net_eq(dev_net(ifa->idev->dev), net))
3094                         return ifa;
3095         }
3096
3097         while (++state->bucket < IN6_ADDR_HSIZE) {
3098                 state->offset = 0;
3099                 hlist_for_each_entry_rcu_bh(ifa, n,
3100                                      &inet6_addr_lst[state->bucket], addr_lst) {
3101                         state->offset++;
3102                         if (net_eq(dev_net(ifa->idev->dev), net))
3103                                 return ifa;
3104                 }
3105         }
3106
3107         return NULL;
3108 }
3109
3110 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3111         __acquires(rcu_bh)
3112 {
3113         rcu_read_lock_bh();
3114         return if6_get_first(seq, *pos);
3115 }
3116
3117 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3118 {
3119         struct inet6_ifaddr *ifa;
3120
3121         ifa = if6_get_next(seq, v);
3122         ++*pos;
3123         return ifa;
3124 }
3125
3126 static void if6_seq_stop(struct seq_file *seq, void *v)
3127         __releases(rcu_bh)
3128 {
3129         rcu_read_unlock_bh();
3130 }
3131
3132 static int if6_seq_show(struct seq_file *seq, void *v)
3133 {
3134         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3135         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3136                    &ifp->addr,
3137                    ifp->idev->dev->ifindex,
3138                    ifp->prefix_len,
3139                    ifp->scope,
3140                    ifp->flags,
3141                    ifp->idev->dev->name);
3142         return 0;
3143 }
3144
3145 static const struct seq_operations if6_seq_ops = {
3146         .start  = if6_seq_start,
3147         .next   = if6_seq_next,
3148         .show   = if6_seq_show,
3149         .stop   = if6_seq_stop,
3150 };
3151
3152 static int if6_seq_open(struct inode *inode, struct file *file)
3153 {
3154         return seq_open_net(inode, file, &if6_seq_ops,
3155                             sizeof(struct if6_iter_state));
3156 }
3157
3158 static const struct file_operations if6_fops = {
3159         .owner          = THIS_MODULE,
3160         .open           = if6_seq_open,
3161         .read           = seq_read,
3162         .llseek         = seq_lseek,
3163         .release        = seq_release_net,
3164 };
3165
3166 static int __net_init if6_proc_net_init(struct net *net)
3167 {
3168         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3169                 return -ENOMEM;
3170         return 0;
3171 }
3172
3173 static void __net_exit if6_proc_net_exit(struct net *net)
3174 {
3175        proc_net_remove(net, "if_inet6");
3176 }
3177
3178 static struct pernet_operations if6_proc_net_ops = {
3179        .init = if6_proc_net_init,
3180        .exit = if6_proc_net_exit,
3181 };
3182
3183 int __init if6_proc_init(void)
3184 {
3185         return register_pernet_subsys(&if6_proc_net_ops);
3186 }
3187
3188 void if6_proc_exit(void)
3189 {
3190         unregister_pernet_subsys(&if6_proc_net_ops);
3191 }
3192 #endif  /* CONFIG_PROC_FS */
3193
3194 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3195 /* Check if address is a home address configured on any interface. */
3196 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3197 {
3198         int ret = 0;
3199         struct inet6_ifaddr *ifp = NULL;
3200         struct hlist_node *n;
3201         unsigned int hash = inet6_addr_hash(addr);
3202
3203         rcu_read_lock_bh();
3204         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3205                 if (!net_eq(dev_net(ifp->idev->dev), net))
3206                         continue;
3207                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3208                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3209                         ret = 1;
3210                         break;
3211                 }
3212         }
3213         rcu_read_unlock_bh();
3214         return ret;
3215 }
3216 #endif
3217
3218 /*
3219  *      Periodic address status verification
3220  */
3221
3222 static void addrconf_verify(unsigned long foo)
3223 {
3224         unsigned long now, next, next_sec, next_sched;
3225         struct inet6_ifaddr *ifp;
3226         struct hlist_node *node;
3227         int i;
3228
3229         rcu_read_lock_bh();
3230         spin_lock(&addrconf_verify_lock);
3231         now = jiffies;
3232         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3233
3234         del_timer(&addr_chk_timer);
3235
3236         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3237 restart:
3238                 hlist_for_each_entry_rcu_bh(ifp, node,
3239                                          &inet6_addr_lst[i], addr_lst) {
3240                         unsigned long age;
3241
3242                         if (ifp->flags & IFA_F_PERMANENT)
3243                                 continue;
3244
3245                         spin_lock(&ifp->lock);
3246                         /* We try to batch several events at once. */
3247                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3248
3249                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3250                             age >= ifp->valid_lft) {
3251                                 spin_unlock(&ifp->lock);
3252                                 in6_ifa_hold(ifp);
3253                                 ipv6_del_addr(ifp);
3254                                 goto restart;
3255                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3256                                 spin_unlock(&ifp->lock);
3257                                 continue;
3258                         } else if (age >= ifp->prefered_lft) {
3259                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3260                                 int deprecate = 0;
3261
3262                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3263                                         deprecate = 1;
3264                                         ifp->flags |= IFA_F_DEPRECATED;
3265                                 }
3266
3267                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3268                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3269
3270                                 spin_unlock(&ifp->lock);
3271
3272                                 if (deprecate) {
3273                                         in6_ifa_hold(ifp);
3274
3275                                         ipv6_ifa_notify(0, ifp);
3276                                         in6_ifa_put(ifp);
3277                                         goto restart;
3278                                 }
3279 #ifdef CONFIG_IPV6_PRIVACY
3280                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3281                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3282                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3283                                         ifp->idev->cnf.dad_transmits *
3284                                         ifp->idev->nd_parms->retrans_time / HZ;
3285
3286                                 if (age >= ifp->prefered_lft - regen_advance) {
3287                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3288                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3289                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3290                                         if (!ifp->regen_count && ifpub) {
3291                                                 ifp->regen_count++;
3292                                                 in6_ifa_hold(ifp);
3293                                                 in6_ifa_hold(ifpub);
3294                                                 spin_unlock(&ifp->lock);
3295
3296                                                 spin_lock(&ifpub->lock);
3297                                                 ifpub->regen_count = 0;
3298                                                 spin_unlock(&ifpub->lock);
3299                                                 ipv6_create_tempaddr(ifpub, ifp);
3300                                                 in6_ifa_put(ifpub);
3301                                                 in6_ifa_put(ifp);
3302                                                 goto restart;
3303                                         }
3304                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3305                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3306                                 spin_unlock(&ifp->lock);
3307 #endif
3308                         } else {
3309                                 /* ifp->prefered_lft <= ifp->valid_lft */
3310                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3311                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3312                                 spin_unlock(&ifp->lock);
3313                         }
3314                 }
3315         }
3316
3317         next_sec = round_jiffies_up(next);
3318         next_sched = next;
3319
3320         /* If rounded timeout is accurate enough, accept it. */
3321         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3322                 next_sched = next_sec;
3323
3324         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3325         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3326                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3327
3328         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3329               now, next, next_sec, next_sched));
3330
3331         addr_chk_timer.expires = next_sched;
3332         add_timer(&addr_chk_timer);
3333         spin_unlock(&addrconf_verify_lock);
3334         rcu_read_unlock_bh();
3335 }
3336
3337 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3338 {
3339         struct in6_addr *pfx = NULL;
3340
3341         if (addr)
3342                 pfx = nla_data(addr);
3343
3344         if (local) {
3345                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3346                         pfx = NULL;
3347                 else
3348                         pfx = nla_data(local);
3349         }
3350
3351         return pfx;
3352 }
3353
3354 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3355         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3356         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3357         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3358 };
3359
3360 static int
3361 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3362 {
3363         struct net *net = sock_net(skb->sk);
3364         struct ifaddrmsg *ifm;
3365         struct nlattr *tb[IFA_MAX+1];
3366         struct in6_addr *pfx;
3367         int err;
3368
3369         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3370         if (err < 0)
3371                 return err;
3372
3373         ifm = nlmsg_data(nlh);
3374         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3375         if (pfx == NULL)
3376                 return -EINVAL;
3377
3378         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3379 }
3380
3381 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3382                              u32 prefered_lft, u32 valid_lft)
3383 {
3384         u32 flags;
3385         clock_t expires;
3386         unsigned long timeout;
3387
3388         if (!valid_lft || (prefered_lft > valid_lft))
3389                 return -EINVAL;
3390
3391         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3392         if (addrconf_finite_timeout(timeout)) {
3393                 expires = jiffies_to_clock_t(timeout * HZ);
3394                 valid_lft = timeout;
3395                 flags = RTF_EXPIRES;
3396         } else {
3397                 expires = 0;
3398                 flags = 0;
3399                 ifa_flags |= IFA_F_PERMANENT;
3400         }
3401
3402         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3403         if (addrconf_finite_timeout(timeout)) {
3404                 if (timeout == 0)
3405                         ifa_flags |= IFA_F_DEPRECATED;
3406                 prefered_lft = timeout;
3407         }
3408
3409         spin_lock_bh(&ifp->lock);
3410         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3411         ifp->tstamp = jiffies;
3412         ifp->valid_lft = valid_lft;
3413         ifp->prefered_lft = prefered_lft;
3414
3415         spin_unlock_bh(&ifp->lock);
3416         if (!(ifp->flags&IFA_F_TENTATIVE))
3417                 ipv6_ifa_notify(0, ifp);
3418
3419         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3420                               expires, flags);
3421         addrconf_verify(0);
3422
3423         return 0;
3424 }
3425
3426 static int
3427 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3428 {
3429         struct net *net = sock_net(skb->sk);
3430         struct ifaddrmsg *ifm;
3431         struct nlattr *tb[IFA_MAX+1];
3432         struct in6_addr *pfx;
3433         struct inet6_ifaddr *ifa;
3434         struct net_device *dev;
3435         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3436         u8 ifa_flags;
3437         int err;
3438
3439         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3440         if (err < 0)
3441                 return err;
3442
3443         ifm = nlmsg_data(nlh);
3444         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3445         if (pfx == NULL)
3446                 return -EINVAL;
3447
3448         if (tb[IFA_CACHEINFO]) {
3449                 struct ifa_cacheinfo *ci;
3450
3451                 ci = nla_data(tb[IFA_CACHEINFO]);
3452                 valid_lft = ci->ifa_valid;
3453                 preferred_lft = ci->ifa_prefered;
3454         } else {
3455                 preferred_lft = INFINITY_LIFE_TIME;
3456                 valid_lft = INFINITY_LIFE_TIME;
3457         }
3458
3459         dev =  __dev_get_by_index(net, ifm->ifa_index);
3460         if (dev == NULL)
3461                 return -ENODEV;
3462
3463         /* We ignore other flags so far. */
3464         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3465
3466         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3467         if (ifa == NULL) {
3468                 /*
3469                  * It would be best to check for !NLM_F_CREATE here but
3470                  * userspace alreay relies on not having to provide this.
3471                  */
3472                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3473                                       ifm->ifa_prefixlen, ifa_flags,
3474                                       preferred_lft, valid_lft);
3475         }
3476
3477         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3478             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3479                 err = -EEXIST;
3480         else
3481                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3482
3483         in6_ifa_put(ifa);
3484
3485         return err;
3486 }
3487
3488 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3489                           u8 scope, int ifindex)
3490 {
3491         struct ifaddrmsg *ifm;
3492
3493         ifm = nlmsg_data(nlh);
3494         ifm->ifa_family = AF_INET6;
3495         ifm->ifa_prefixlen = prefixlen;
3496         ifm->ifa_flags = flags;
3497         ifm->ifa_scope = scope;
3498         ifm->ifa_index = ifindex;
3499 }
3500
3501 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3502                          unsigned long tstamp, u32 preferred, u32 valid)
3503 {
3504         struct ifa_cacheinfo ci;
3505
3506         ci.cstamp = cstamp_delta(cstamp);
3507         ci.tstamp = cstamp_delta(tstamp);
3508         ci.ifa_prefered = preferred;
3509         ci.ifa_valid = valid;
3510
3511         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3512 }
3513
3514 static inline int rt_scope(int ifa_scope)
3515 {
3516         if (ifa_scope & IFA_HOST)
3517                 return RT_SCOPE_HOST;
3518         else if (ifa_scope & IFA_LINK)
3519                 return RT_SCOPE_LINK;
3520         else if (ifa_scope & IFA_SITE)
3521                 return RT_SCOPE_SITE;
3522         else
3523                 return RT_SCOPE_UNIVERSE;
3524 }
3525
3526 static inline int inet6_ifaddr_msgsize(void)
3527 {
3528         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3529                + nla_total_size(16) /* IFA_ADDRESS */
3530                + nla_total_size(sizeof(struct ifa_cacheinfo));
3531 }
3532
3533 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3534                              u32 portid, u32 seq, int event, unsigned int flags)
3535 {
3536         struct nlmsghdr  *nlh;
3537         u32 preferred, valid;
3538
3539         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3540         if (nlh == NULL)
3541                 return -EMSGSIZE;
3542
3543         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3544                       ifa->idev->dev->ifindex);
3545
3546         if (!(ifa->flags&IFA_F_PERMANENT)) {
3547                 preferred = ifa->prefered_lft;
3548                 valid = ifa->valid_lft;
3549                 if (preferred != INFINITY_LIFE_TIME) {
3550                         long tval = (jiffies - ifa->tstamp)/HZ;
3551                         if (preferred > tval)
3552                                 preferred -= tval;
3553                         else
3554                                 preferred = 0;
3555                         if (valid != INFINITY_LIFE_TIME) {
3556                                 if (valid > tval)
3557                                         valid -= tval;
3558                                 else
3559                                         valid = 0;
3560                         }
3561                 }
3562         } else {
3563                 preferred = INFINITY_LIFE_TIME;
3564                 valid = INFINITY_LIFE_TIME;
3565         }
3566
3567         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3568             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3569                 nlmsg_cancel(skb, nlh);
3570                 return -EMSGSIZE;
3571         }
3572
3573         return nlmsg_end(skb, nlh);
3574 }
3575
3576 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3577                                 u32 portid, u32 seq, int event, u16 flags)
3578 {
3579         struct nlmsghdr  *nlh;
3580         u8 scope = RT_SCOPE_UNIVERSE;
3581         int ifindex = ifmca->idev->dev->ifindex;
3582
3583         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3584                 scope = RT_SCOPE_SITE;
3585
3586         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3587         if (nlh == NULL)
3588                 return -EMSGSIZE;
3589
3590         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3591         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3592             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3593                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3594                 nlmsg_cancel(skb, nlh);
3595                 return -EMSGSIZE;
3596         }
3597
3598         return nlmsg_end(skb, nlh);
3599 }
3600
3601 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3602                                 u32 portid, u32 seq, int event, unsigned int flags)
3603 {
3604         struct nlmsghdr  *nlh;
3605         u8 scope = RT_SCOPE_UNIVERSE;
3606         int ifindex = ifaca->aca_idev->dev->ifindex;
3607
3608         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3609                 scope = RT_SCOPE_SITE;
3610
3611         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3612         if (nlh == NULL)
3613                 return -EMSGSIZE;
3614
3615         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3616         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3617             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3618                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3619                 nlmsg_cancel(skb, nlh);
3620                 return -EMSGSIZE;
3621         }
3622
3623         return nlmsg_end(skb, nlh);
3624 }
3625
3626 enum addr_type_t {
3627         UNICAST_ADDR,
3628         MULTICAST_ADDR,
3629         ANYCAST_ADDR,
3630 };
3631
3632 /* called with rcu_read_lock() */
3633 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3634                           struct netlink_callback *cb, enum addr_type_t type,
3635                           int s_ip_idx, int *p_ip_idx)
3636 {
3637         struct ifmcaddr6 *ifmca;
3638         struct ifacaddr6 *ifaca;
3639         int err = 1;
3640         int ip_idx = *p_ip_idx;
3641
3642         read_lock_bh(&idev->lock);
3643         switch (type) {
3644         case UNICAST_ADDR: {
3645                 struct inet6_ifaddr *ifa;
3646
3647                 /* unicast address incl. temp addr */
3648                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3649                         if (++ip_idx < s_ip_idx)
3650                                 continue;
3651                         err = inet6_fill_ifaddr(skb, ifa,
3652                                                 NETLINK_CB(cb->skb).portid,
3653                                                 cb->nlh->nlmsg_seq,
3654                                                 RTM_NEWADDR,
3655                                                 NLM_F_MULTI);
3656                         if (err <= 0)
3657                                 break;
3658                 }
3659                 break;
3660         }
3661         case MULTICAST_ADDR:
3662                 /* multicast address */
3663                 for (ifmca = idev->mc_list; ifmca;
3664                      ifmca = ifmca->next, ip_idx++) {
3665                         if (ip_idx < s_ip_idx)
3666                                 continue;
3667                         err = inet6_fill_ifmcaddr(skb, ifmca,
3668                                                   NETLINK_CB(cb->skb).portid,
3669                                                   cb->nlh->nlmsg_seq,
3670                                                   RTM_GETMULTICAST,
3671                                                   NLM_F_MULTI);
3672                         if (err <= 0)
3673                                 break;
3674                 }
3675                 break;
3676         case ANYCAST_ADDR:
3677                 /* anycast address */
3678                 for (ifaca = idev->ac_list; ifaca;
3679                      ifaca = ifaca->aca_next, ip_idx++) {
3680                         if (ip_idx < s_ip_idx)
3681                                 continue;
3682                         err = inet6_fill_ifacaddr(skb, ifaca,
3683                                                   NETLINK_CB(cb->skb).portid,
3684                                                   cb->nlh->nlmsg_seq,
3685                                                   RTM_GETANYCAST,
3686                                                   NLM_F_MULTI);
3687                         if (err <= 0)
3688                                 break;
3689                 }
3690                 break;
3691         default:
3692                 break;
3693         }
3694         read_unlock_bh(&idev->lock);
3695         *p_ip_idx = ip_idx;
3696         return err;
3697 }
3698
3699 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3700                            enum addr_type_t type)
3701 {
3702         struct net *net = sock_net(skb->sk);
3703         int h, s_h;
3704         int idx, ip_idx;
3705         int s_idx, s_ip_idx;
3706         struct net_device *dev;
3707         struct inet6_dev *idev;
3708         struct hlist_head *head;
3709         struct hlist_node *node;
3710
3711         s_h = cb->args[0];
3712         s_idx = idx = cb->args[1];
3713         s_ip_idx = ip_idx = cb->args[2];
3714
3715         rcu_read_lock();
3716         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3717                 idx = 0;
3718                 head = &net->dev_index_head[h];
3719                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3720                         if (idx < s_idx)
3721                                 goto cont;
3722                         if (h > s_h || idx > s_idx)
3723                                 s_ip_idx = 0;
3724                         ip_idx = 0;
3725                         idev = __in6_dev_get(dev);
3726                         if (!idev)
3727                                 goto cont;
3728
3729                         if (in6_dump_addrs(idev, skb, cb, type,
3730                                            s_ip_idx, &ip_idx) <= 0)
3731                                 goto done;
3732 cont:
3733                         idx++;
3734                 }
3735         }
3736 done:
3737         rcu_read_unlock();
3738         cb->args[0] = h;
3739         cb->args[1] = idx;
3740         cb->args[2] = ip_idx;
3741
3742         return skb->len;
3743 }
3744
3745 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3746 {
3747         enum addr_type_t type = UNICAST_ADDR;
3748
3749         return inet6_dump_addr(skb, cb, type);
3750 }
3751
3752 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3753 {
3754         enum addr_type_t type = MULTICAST_ADDR;
3755
3756         return inet6_dump_addr(skb, cb, type);
3757 }
3758
3759
3760 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3761 {
3762         enum addr_type_t type = ANYCAST_ADDR;
3763
3764         return inet6_dump_addr(skb, cb, type);
3765 }
3766
3767 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3768                              void *arg)
3769 {
3770         struct net *net = sock_net(in_skb->sk);
3771         struct ifaddrmsg *ifm;
3772         struct nlattr *tb[IFA_MAX+1];
3773         struct in6_addr *addr = NULL;
3774         struct net_device *dev = NULL;
3775         struct inet6_ifaddr *ifa;
3776         struct sk_buff *skb;
3777         int err;
3778
3779         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3780         if (err < 0)
3781                 goto errout;
3782
3783         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3784         if (addr == NULL) {
3785                 err = -EINVAL;
3786                 goto errout;
3787         }
3788
3789         ifm = nlmsg_data(nlh);
3790         if (ifm->ifa_index)
3791                 dev = __dev_get_by_index(net, ifm->ifa_index);
3792
3793         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3794         if (!ifa) {
3795                 err = -EADDRNOTAVAIL;
3796                 goto errout;
3797         }
3798
3799         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3800         if (!skb) {
3801                 err = -ENOBUFS;
3802                 goto errout_ifa;
3803         }
3804
3805         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3806                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3807         if (err < 0) {
3808                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3809                 WARN_ON(err == -EMSGSIZE);
3810                 kfree_skb(skb);
3811                 goto errout_ifa;
3812         }
3813         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3814 errout_ifa:
3815         in6_ifa_put(ifa);
3816 errout:
3817         return err;
3818 }
3819
3820 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3821 {
3822         struct sk_buff *skb;
3823         struct net *net = dev_net(ifa->idev->dev);
3824         int err = -ENOBUFS;
3825
3826         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3827         if (skb == NULL)
3828                 goto errout;
3829
3830         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3831         if (err < 0) {
3832                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3833                 WARN_ON(err == -EMSGSIZE);
3834                 kfree_skb(skb);
3835                 goto errout;
3836         }
3837         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3838         return;
3839 errout:
3840         if (err < 0)
3841                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3842 }
3843
3844 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3845                                 __s32 *array, int bytes)
3846 {
3847         BUG_ON(bytes < (DEVCONF_MAX * 4));
3848
3849         memset(array, 0, bytes);
3850         array[DEVCONF_FORWARDING] = cnf->forwarding;
3851         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3852         array[DEVCONF_MTU6] = cnf->mtu6;
3853         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3854         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3855         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3856         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3857         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3858         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3859                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3860         array[DEVCONF_RTR_SOLICIT_DELAY] =
3861                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3862         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3863 #ifdef CONFIG_IPV6_PRIVACY
3864         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3865         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3866         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3867         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3868         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3869 #endif
3870         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3871         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3872         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3873 #ifdef CONFIG_IPV6_ROUTER_PREF
3874         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3875         array[DEVCONF_RTR_PROBE_INTERVAL] =
3876                 jiffies_to_msecs(cnf->rtr_probe_interval);
3877 #ifdef CONFIG_IPV6_ROUTE_INFO
3878         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3879 #endif
3880 #endif
3881         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3882         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3883 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3884         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3885 #endif
3886 #ifdef CONFIG_IPV6_MROUTE
3887         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3888 #endif
3889         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3890         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3891         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3892 }
3893
3894 static inline size_t inet6_ifla6_size(void)
3895 {
3896         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3897              + nla_total_size(sizeof(struct ifla_cacheinfo))
3898              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3899              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3900              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3901 }
3902
3903 static inline size_t inet6_if_nlmsg_size(void)
3904 {
3905         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3906                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3907                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3908                + nla_total_size(4) /* IFLA_MTU */
3909                + nla_total_size(4) /* IFLA_LINK */
3910                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3911 }
3912
3913 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3914                                       int items, int bytes)
3915 {
3916         int i;
3917         int pad = bytes - sizeof(u64) * items;
3918         BUG_ON(pad < 0);
3919
3920         /* Use put_unaligned() because stats may not be aligned for u64. */
3921         put_unaligned(items, &stats[0]);
3922         for (i = 1; i < items; i++)
3923                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3924
3925         memset(&stats[items], 0, pad);
3926 }
3927
3928 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3929                                       int items, int bytes, size_t syncpoff)
3930 {
3931         int i;
3932         int pad = bytes - sizeof(u64) * items;
3933         BUG_ON(pad < 0);
3934
3935         /* Use put_unaligned() because stats may not be aligned for u64. */
3936         put_unaligned(items, &stats[0]);
3937         for (i = 1; i < items; i++)
3938                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3939
3940         memset(&stats[items], 0, pad);
3941 }
3942
3943 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3944                              int bytes)
3945 {
3946         switch (attrtype) {
3947         case IFLA_INET6_STATS:
3948                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3949                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3950                 break;
3951         case IFLA_INET6_ICMP6STATS:
3952                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3953                 break;
3954         }
3955 }
3956
3957 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3958 {
3959         struct nlattr *nla;
3960         struct ifla_cacheinfo ci;
3961
3962         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
3963                 goto nla_put_failure;
3964         ci.max_reasm_len = IPV6_MAXPLEN;
3965         ci.tstamp = cstamp_delta(idev->tstamp);
3966         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3967         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3968         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
3969                 goto nla_put_failure;
3970         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3971         if (nla == NULL)
3972                 goto nla_put_failure;
3973         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3974
3975         /* XXX - MC not implemented */
3976
3977         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3978         if (nla == NULL)
3979                 goto nla_put_failure;
3980         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3981
3982         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3983         if (nla == NULL)
3984                 goto nla_put_failure;
3985         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3986
3987         return 0;
3988
3989 nla_put_failure:
3990         return -EMSGSIZE;
3991 }
3992
3993 static size_t inet6_get_link_af_size(const struct net_device *dev)
3994 {
3995         if (!__in6_dev_get(dev))
3996                 return 0;
3997
3998         return inet6_ifla6_size();
3999 }
4000
4001 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4002 {
4003         struct inet6_dev *idev = __in6_dev_get(dev);
4004
4005         if (!idev)
4006                 return -ENODATA;
4007
4008         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4009                 return -EMSGSIZE;
4010
4011         return 0;
4012 }
4013
4014 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4015                              u32 portid, u32 seq, int event, unsigned int flags)
4016 {
4017         struct net_device *dev = idev->dev;
4018         struct ifinfomsg *hdr;
4019         struct nlmsghdr *nlh;
4020         void *protoinfo;
4021
4022         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4023         if (nlh == NULL)
4024                 return -EMSGSIZE;
4025
4026         hdr = nlmsg_data(nlh);
4027         hdr->ifi_family = AF_INET6;
4028         hdr->__ifi_pad = 0;
4029         hdr->ifi_type = dev->type;
4030         hdr->ifi_index = dev->ifindex;
4031         hdr->ifi_flags = dev_get_flags(dev);
4032         hdr->ifi_change = 0;
4033
4034         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4035             (dev->addr_len &&
4036              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4037             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4038             (dev->ifindex != dev->iflink &&
4039              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4040                 goto nla_put_failure;
4041         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4042         if (protoinfo == NULL)
4043                 goto nla_put_failure;
4044
4045         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4046                 goto nla_put_failure;
4047
4048         nla_nest_end(skb, protoinfo);
4049         return nlmsg_end(skb, nlh);
4050
4051 nla_put_failure:
4052         nlmsg_cancel(skb, nlh);
4053         return -EMSGSIZE;
4054 }
4055
4056 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4057 {
4058         struct net *net = sock_net(skb->sk);
4059         int h, s_h;
4060         int idx = 0, s_idx;
4061         struct net_device *dev;
4062         struct inet6_dev *idev;
4063         struct hlist_head *head;
4064         struct hlist_node *node;
4065
4066         s_h = cb->args[0];
4067         s_idx = cb->args[1];
4068
4069         rcu_read_lock();
4070         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4071                 idx = 0;
4072                 head = &net->dev_index_head[h];
4073                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4074                         if (idx < s_idx)
4075                                 goto cont;
4076                         idev = __in6_dev_get(dev);
4077                         if (!idev)
4078                                 goto cont;
4079                         if (inet6_fill_ifinfo(skb, idev,
4080                                               NETLINK_CB(cb->skb).portid,
4081                                               cb->nlh->nlmsg_seq,
4082                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4083                                 goto out;
4084 cont:
4085                         idx++;
4086                 }
4087         }
4088 out:
4089         rcu_read_unlock();
4090         cb->args[1] = idx;
4091         cb->args[0] = h;
4092
4093         return skb->len;
4094 }
4095
4096 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4097 {
4098         struct sk_buff *skb;
4099         struct net *net = dev_net(idev->dev);
4100         int err = -ENOBUFS;
4101
4102         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4103         if (skb == NULL)
4104                 goto errout;
4105
4106         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4107         if (err < 0) {
4108                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4109                 WARN_ON(err == -EMSGSIZE);
4110                 kfree_skb(skb);
4111                 goto errout;
4112         }
4113         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4114         return;
4115 errout:
4116         if (err < 0)
4117                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4118 }
4119
4120 static inline size_t inet6_prefix_nlmsg_size(void)
4121 {
4122         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4123                + nla_total_size(sizeof(struct in6_addr))
4124                + nla_total_size(sizeof(struct prefix_cacheinfo));
4125 }
4126
4127 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4128                              struct prefix_info *pinfo, u32 portid, u32 seq,
4129                              int event, unsigned int flags)
4130 {
4131         struct prefixmsg *pmsg;
4132         struct nlmsghdr *nlh;
4133         struct prefix_cacheinfo ci;
4134
4135         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4136         if (nlh == NULL)
4137                 return -EMSGSIZE;
4138
4139         pmsg = nlmsg_data(nlh);
4140         pmsg->prefix_family = AF_INET6;
4141         pmsg->prefix_pad1 = 0;
4142         pmsg->prefix_pad2 = 0;
4143         pmsg->prefix_ifindex = idev->dev->ifindex;
4144         pmsg->prefix_len = pinfo->prefix_len;
4145         pmsg->prefix_type = pinfo->type;
4146         pmsg->prefix_pad3 = 0;
4147         pmsg->prefix_flags = 0;
4148         if (pinfo->onlink)
4149                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4150         if (pinfo->autoconf)
4151                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4152
4153         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4154                 goto nla_put_failure;
4155         ci.preferred_time = ntohl(pinfo->prefered);
4156         ci.valid_time = ntohl(pinfo->valid);
4157         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4158                 goto nla_put_failure;
4159         return nlmsg_end(skb, nlh);
4160
4161 nla_put_failure:
4162         nlmsg_cancel(skb, nlh);
4163         return -EMSGSIZE;
4164 }
4165
4166 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4167                          struct prefix_info *pinfo)
4168 {
4169         struct sk_buff *skb;
4170         struct net *net = dev_net(idev->dev);
4171         int err = -ENOBUFS;
4172
4173         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4174         if (skb == NULL)
4175                 goto errout;
4176
4177         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4178         if (err < 0) {
4179                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4180                 WARN_ON(err == -EMSGSIZE);
4181                 kfree_skb(skb);
4182                 goto errout;
4183         }
4184         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4185         return;
4186 errout:
4187         if (err < 0)
4188                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4189 }
4190
4191 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4192 {
4193         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4194
4195         switch (event) {
4196         case RTM_NEWADDR:
4197                 /*
4198                  * If the address was optimistic
4199                  * we inserted the route at the start of
4200                  * our DAD process, so we don't need
4201                  * to do it again
4202                  */
4203                 if (!(ifp->rt->rt6i_node))
4204                         ip6_ins_rt(ifp->rt);
4205                 if (ifp->idev->cnf.forwarding)
4206                         addrconf_join_anycast(ifp);
4207                 break;
4208         case RTM_DELADDR:
4209                 if (ifp->idev->cnf.forwarding)
4210                         addrconf_leave_anycast(ifp);
4211                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4212                 dst_hold(&ifp->rt->dst);
4213
4214                 if (ip6_del_rt(ifp->rt))
4215                         dst_free(&ifp->rt->dst);
4216                 break;
4217         }
4218 }
4219
4220 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4221 {
4222         rcu_read_lock_bh();
4223         if (likely(ifp->idev->dead == 0))
4224                 __ipv6_ifa_notify(event, ifp);
4225         rcu_read_unlock_bh();
4226 }
4227
4228 #ifdef CONFIG_SYSCTL
4229
4230 static
4231 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4232                            void __user *buffer, size_t *lenp, loff_t *ppos)
4233 {
4234         int *valp = ctl->data;
4235         int val = *valp;
4236         loff_t pos = *ppos;
4237         ctl_table lctl;
4238         int ret;
4239
4240         /*
4241          * ctl->data points to idev->cnf.forwarding, we should
4242          * not modify it until we get the rtnl lock.
4243          */
4244         lctl = *ctl;
4245         lctl.data = &val;
4246
4247         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4248
4249         if (write)
4250                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4251         if (ret)
4252                 *ppos = pos;
4253         return ret;
4254 }
4255
4256 static void dev_disable_change(struct inet6_dev *idev)
4257 {
4258         if (!idev || !idev->dev)
4259                 return;
4260
4261         if (idev->cnf.disable_ipv6)
4262                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4263         else
4264                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4265 }
4266
4267 static void addrconf_disable_change(struct net *net, __s32 newf)
4268 {
4269         struct net_device *dev;
4270         struct inet6_dev *idev;
4271
4272         rcu_read_lock();
4273         for_each_netdev_rcu(net, dev) {
4274                 idev = __in6_dev_get(dev);
4275                 if (idev) {
4276                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4277                         idev->cnf.disable_ipv6 = newf;
4278                         if (changed)
4279                                 dev_disable_change(idev);
4280                 }
4281         }
4282         rcu_read_unlock();
4283 }
4284
4285 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4286 {
4287         struct net *net;
4288         int old;
4289
4290         if (!rtnl_trylock())
4291                 return restart_syscall();
4292
4293         net = (struct net *)table->extra2;
4294         old = *p;
4295         *p = newf;
4296
4297         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4298                 rtnl_unlock();
4299                 return 0;
4300         }
4301
4302         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4303                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4304                 addrconf_disable_change(net, newf);
4305         } else if ((!newf) ^ (!old))
4306                 dev_disable_change((struct inet6_dev *)table->extra1);
4307
4308         rtnl_unlock();
4309         return 0;
4310 }
4311
4312 static
4313 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4314                             void __user *buffer, size_t *lenp, loff_t *ppos)
4315 {
4316         int *valp = ctl->data;
4317         int val = *valp;
4318         loff_t pos = *ppos;
4319         ctl_table lctl;
4320         int ret;
4321
4322         /*
4323          * ctl->data points to idev->cnf.disable_ipv6, we should
4324          * not modify it until we get the rtnl lock.
4325          */
4326         lctl = *ctl;
4327         lctl.data = &val;
4328
4329         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4330
4331         if (write)
4332                 ret = addrconf_disable_ipv6(ctl, valp, val);
4333         if (ret)
4334                 *ppos = pos;
4335         return ret;
4336 }
4337
4338 static struct addrconf_sysctl_table
4339 {
4340         struct ctl_table_header *sysctl_header;
4341         ctl_table addrconf_vars[DEVCONF_MAX+1];
4342 } addrconf_sysctl __read_mostly = {
4343         .sysctl_header = NULL,
4344         .addrconf_vars = {
4345                 {
4346                         .procname       = "forwarding",
4347                         .data           = &ipv6_devconf.forwarding,
4348                         .maxlen         = sizeof(int),
4349                         .mode           = 0644,
4350                         .proc_handler   = addrconf_sysctl_forward,
4351                 },
4352                 {
4353                         .procname       = "hop_limit",
4354                         .data           = &ipv6_devconf.hop_limit,
4355                         .maxlen         = sizeof(int),
4356                         .mode           = 0644,
4357                         .proc_handler   = proc_dointvec,
4358                 },
4359                 {
4360                         .procname       = "mtu",
4361                         .data           = &ipv6_devconf.mtu6,
4362                         .maxlen         = sizeof(int),
4363                         .mode           = 0644,
4364                         .proc_handler   = proc_dointvec,
4365                 },
4366                 {
4367                         .procname       = "accept_ra",
4368                         .data           = &ipv6_devconf.accept_ra,
4369                         .maxlen         = sizeof(int),
4370                         .mode           = 0644,
4371                         .proc_handler   = proc_dointvec,
4372                 },
4373                 {
4374                         .procname       = "accept_redirects",
4375                         .data           = &ipv6_devconf.accept_redirects,
4376                         .maxlen         = sizeof(int),
4377                         .mode           = 0644,
4378                         .proc_handler   = proc_dointvec,
4379                 },
4380                 {
4381                         .procname       = "autoconf",
4382                         .data           = &ipv6_devconf.autoconf,
4383                         .maxlen         = sizeof(int),
4384                         .mode           = 0644,
4385                         .proc_handler   = proc_dointvec,
4386                 },
4387                 {
4388                         .procname       = "dad_transmits",
4389                         .data           = &ipv6_devconf.dad_transmits,
4390                         .maxlen         = sizeof(int),
4391                         .mode           = 0644,
4392                         .proc_handler   = proc_dointvec,
4393                 },
4394                 {
4395                         .procname       = "router_solicitations",
4396                         .data           = &ipv6_devconf.rtr_solicits,
4397                         .maxlen         = sizeof(int),
4398                         .mode           = 0644,
4399                         .proc_handler   = proc_dointvec,
4400                 },
4401                 {
4402                         .procname       = "router_solicitation_interval",
4403                         .data           = &ipv6_devconf.rtr_solicit_interval,
4404                         .maxlen         = sizeof(int),
4405                         .mode           = 0644,
4406                         .proc_handler   = proc_dointvec_jiffies,
4407                 },
4408                 {
4409                         .procname       = "router_solicitation_delay",
4410                         .data           = &ipv6_devconf.rtr_solicit_delay,
4411                         .maxlen         = sizeof(int),
4412                         .mode           = 0644,
4413                         .proc_handler   = proc_dointvec_jiffies,
4414                 },
4415                 {
4416                         .procname       = "force_mld_version",
4417                         .data           = &ipv6_devconf.force_mld_version,
4418                         .maxlen         = sizeof(int),
4419                         .mode           = 0644,
4420                         .proc_handler   = proc_dointvec,
4421                 },
4422 #ifdef CONFIG_IPV6_PRIVACY
4423                 {
4424                         .procname       = "use_tempaddr",
4425                         .data           = &ipv6_devconf.use_tempaddr,
4426                         .maxlen         = sizeof(int),
4427                         .mode           = 0644,
4428                         .proc_handler   = proc_dointvec,
4429                 },
4430                 {
4431                         .procname       = "temp_valid_lft",
4432                         .data           = &ipv6_devconf.temp_valid_lft,
4433                         .maxlen         = sizeof(int),
4434                         .mode           = 0644,
4435                         .proc_handler   = proc_dointvec,
4436                 },
4437                 {
4438                         .procname       = "temp_prefered_lft",
4439                         .data           = &ipv6_devconf.temp_prefered_lft,
4440                         .maxlen         = sizeof(int),
4441                         .mode           = 0644,
4442                         .proc_handler   = proc_dointvec,
4443                 },
4444                 {
4445                         .procname       = "regen_max_retry",
4446                         .data           = &ipv6_devconf.regen_max_retry,
4447                         .maxlen         = sizeof(int),
4448                         .mode           = 0644,
4449                         .proc_handler   = proc_dointvec,
4450                 },
4451                 {
4452                         .procname       = "max_desync_factor",
4453                         .data           = &ipv6_devconf.max_desync_factor,
4454                         .maxlen         = sizeof(int),
4455                         .mode           = 0644,
4456                         .proc_handler   = proc_dointvec,
4457                 },
4458 #endif
4459                 {
4460                         .procname       = "max_addresses",
4461                         .data           = &ipv6_devconf.max_addresses,
4462                         .maxlen         = sizeof(int),
4463                         .mode           = 0644,
4464                         .proc_handler   = proc_dointvec,
4465                 },
4466                 {
4467                         .procname       = "accept_ra_defrtr",
4468                         .data           = &ipv6_devconf.accept_ra_defrtr,
4469                         .maxlen         = sizeof(int),
4470                         .mode           = 0644,
4471                         .proc_handler   = proc_dointvec,
4472                 },
4473                 {
4474                         .procname       = "accept_ra_pinfo",
4475                         .data           = &ipv6_devconf.accept_ra_pinfo,
4476                         .maxlen         = sizeof(int),
4477                         .mode           = 0644,
4478                         .proc_handler   = proc_dointvec,
4479                 },
4480 #ifdef CONFIG_IPV6_ROUTER_PREF
4481                 {
4482                         .procname       = "accept_ra_rtr_pref",
4483                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4484                         .maxlen         = sizeof(int),
4485                         .mode           = 0644,
4486                         .proc_handler   = proc_dointvec,
4487                 },
4488                 {
4489                         .procname       = "router_probe_interval",
4490                         .data           = &ipv6_devconf.rtr_probe_interval,
4491                         .maxlen         = sizeof(int),
4492                         .mode           = 0644,
4493                         .proc_handler   = proc_dointvec_jiffies,
4494                 },
4495 #ifdef CONFIG_IPV6_ROUTE_INFO
4496                 {
4497                         .procname       = "accept_ra_rt_info_max_plen",
4498                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4499                         .maxlen         = sizeof(int),
4500                         .mode           = 0644,
4501                         .proc_handler   = proc_dointvec,
4502                 },
4503 #endif
4504 #endif
4505                 {
4506                         .procname       = "proxy_ndp",
4507                         .data           = &ipv6_devconf.proxy_ndp,
4508                         .maxlen         = sizeof(int),
4509                         .mode           = 0644,
4510                         .proc_handler   = proc_dointvec,
4511                 },
4512                 {
4513                         .procname       = "accept_source_route",
4514                         .data           = &ipv6_devconf.accept_source_route,
4515                         .maxlen         = sizeof(int),
4516                         .mode           = 0644,
4517                         .proc_handler   = proc_dointvec,
4518                 },
4519 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4520                 {
4521                         .procname       = "optimistic_dad",
4522                         .data           = &ipv6_devconf.optimistic_dad,
4523                         .maxlen         = sizeof(int),
4524                         .mode           = 0644,
4525                         .proc_handler   = proc_dointvec,
4526
4527                 },
4528 #endif
4529 #ifdef CONFIG_IPV6_MROUTE
4530                 {
4531                         .procname       = "mc_forwarding",
4532                         .data           = &ipv6_devconf.mc_forwarding,
4533                         .maxlen         = sizeof(int),
4534                         .mode           = 0444,
4535                         .proc_handler   = proc_dointvec,
4536                 },
4537 #endif
4538                 {
4539                         .procname       = "disable_ipv6",
4540                         .data           = &ipv6_devconf.disable_ipv6,
4541                         .maxlen         = sizeof(int),
4542                         .mode           = 0644,
4543                         .proc_handler   = addrconf_sysctl_disable,
4544                 },
4545                 {
4546                         .procname       = "accept_dad",
4547                         .data           = &ipv6_devconf.accept_dad,
4548                         .maxlen         = sizeof(int),
4549                         .mode           = 0644,
4550                         .proc_handler   = proc_dointvec,
4551                 },
4552                 {
4553                         .procname       = "force_tllao",
4554                         .data           = &ipv6_devconf.force_tllao,
4555                         .maxlen         = sizeof(int),
4556                         .mode           = 0644,
4557                         .proc_handler   = proc_dointvec
4558                 },
4559                 {
4560                         /* sentinel */
4561                 }
4562         },
4563 };
4564
4565 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4566                 struct inet6_dev *idev, struct ipv6_devconf *p)
4567 {
4568         int i;
4569         struct addrconf_sysctl_table *t;
4570         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4571
4572         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4573         if (t == NULL)
4574                 goto out;
4575
4576         for (i = 0; t->addrconf_vars[i].data; i++) {
4577                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4578                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4579                 t->addrconf_vars[i].extra2 = net;
4580         }
4581
4582         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4583
4584         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4585         if (t->sysctl_header == NULL)
4586                 goto free;
4587
4588         p->sysctl = t;
4589         return 0;
4590
4591 free:
4592         kfree(t);
4593 out:
4594         return -ENOBUFS;
4595 }
4596
4597 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4598 {
4599         struct addrconf_sysctl_table *t;
4600
4601         if (p->sysctl == NULL)
4602                 return;
4603
4604         t = p->sysctl;
4605         p->sysctl = NULL;
4606         unregister_net_sysctl_table(t->sysctl_header);
4607         kfree(t);
4608 }
4609
4610 static void addrconf_sysctl_register(struct inet6_dev *idev)
4611 {
4612         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4613                               &ndisc_ifinfo_sysctl_change);
4614         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4615                                         idev, &idev->cnf);
4616 }
4617
4618 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4619 {
4620         __addrconf_sysctl_unregister(&idev->cnf);
4621         neigh_sysctl_unregister(idev->nd_parms);
4622 }
4623
4624
4625 #endif
4626
4627 static int __net_init addrconf_init_net(struct net *net)
4628 {
4629         int err;
4630         struct ipv6_devconf *all, *dflt;
4631
4632         err = -ENOMEM;
4633         all = &ipv6_devconf;
4634         dflt = &ipv6_devconf_dflt;
4635
4636         if (!net_eq(net, &init_net)) {
4637                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4638                 if (all == NULL)
4639                         goto err_alloc_all;
4640
4641                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4642                 if (dflt == NULL)
4643                         goto err_alloc_dflt;
4644         } else {
4645                 /* these will be inherited by all namespaces */
4646                 dflt->autoconf = ipv6_defaults.autoconf;
4647                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4648         }
4649
4650         net->ipv6.devconf_all = all;
4651         net->ipv6.devconf_dflt = dflt;
4652
4653 #ifdef CONFIG_SYSCTL
4654         err = __addrconf_sysctl_register(net, "all", NULL, all);
4655         if (err < 0)
4656                 goto err_reg_all;
4657
4658         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4659         if (err < 0)
4660                 goto err_reg_dflt;
4661 #endif
4662         return 0;
4663
4664 #ifdef CONFIG_SYSCTL
4665 err_reg_dflt:
4666         __addrconf_sysctl_unregister(all);
4667 err_reg_all:
4668         kfree(dflt);
4669 #endif
4670 err_alloc_dflt:
4671         kfree(all);
4672 err_alloc_all:
4673         return err;
4674 }
4675
4676 static void __net_exit addrconf_exit_net(struct net *net)
4677 {
4678 #ifdef CONFIG_SYSCTL
4679         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4680         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4681 #endif
4682         if (!net_eq(net, &init_net)) {
4683                 kfree(net->ipv6.devconf_dflt);
4684                 kfree(net->ipv6.devconf_all);
4685         }
4686 }
4687
4688 static struct pernet_operations addrconf_ops = {
4689         .init = addrconf_init_net,
4690         .exit = addrconf_exit_net,
4691 };
4692
4693 /*
4694  *      Device notifier
4695  */
4696
4697 int register_inet6addr_notifier(struct notifier_block *nb)
4698 {
4699         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4700 }
4701 EXPORT_SYMBOL(register_inet6addr_notifier);
4702
4703 int unregister_inet6addr_notifier(struct notifier_block *nb)
4704 {
4705         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4706 }
4707 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4708
4709 static struct rtnl_af_ops inet6_ops = {
4710         .family           = AF_INET6,
4711         .fill_link_af     = inet6_fill_link_af,
4712         .get_link_af_size = inet6_get_link_af_size,
4713 };
4714
4715 /*
4716  *      Init / cleanup code
4717  */
4718
4719 int __init addrconf_init(void)
4720 {
4721         int i, err;
4722
4723         err = ipv6_addr_label_init();
4724         if (err < 0) {
4725                 pr_crit("%s: cannot initialize default policy table: %d\n",
4726                         __func__, err);
4727                 goto out;
4728         }
4729
4730         err = register_pernet_subsys(&addrconf_ops);
4731         if (err < 0)
4732                 goto out_addrlabel;
4733
4734         /* The addrconf netdev notifier requires that loopback_dev
4735          * has it's ipv6 private information allocated and setup
4736          * before it can bring up and give link-local addresses
4737          * to other devices which are up.
4738          *
4739          * Unfortunately, loopback_dev is not necessarily the first
4740          * entry in the global dev_base list of net devices.  In fact,
4741          * it is likely to be the very last entry on that list.
4742          * So this causes the notifier registry below to try and
4743          * give link-local addresses to all devices besides loopback_dev
4744          * first, then loopback_dev, which cases all the non-loopback_dev
4745          * devices to fail to get a link-local address.
4746          *
4747          * So, as a temporary fix, allocate the ipv6 structure for
4748          * loopback_dev first by hand.
4749          * Longer term, all of the dependencies ipv6 has upon the loopback
4750          * device and it being up should be removed.
4751          */
4752         rtnl_lock();
4753         if (!ipv6_add_dev(init_net.loopback_dev))
4754                 err = -ENOMEM;
4755         rtnl_unlock();
4756         if (err)
4757                 goto errlo;
4758
4759         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4760                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4761
4762         register_netdevice_notifier(&ipv6_dev_notf);
4763
4764         addrconf_verify(0);
4765
4766         err = rtnl_af_register(&inet6_ops);
4767         if (err < 0)
4768                 goto errout_af;
4769
4770         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4771                               NULL);
4772         if (err < 0)
4773                 goto errout;
4774
4775         /* Only the first call to __rtnl_register can fail */
4776         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4777         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4778         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4779                         inet6_dump_ifaddr, NULL);
4780         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4781                         inet6_dump_ifmcaddr, NULL);
4782         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4783                         inet6_dump_ifacaddr, NULL);
4784
4785         ipv6_addr_label_rtnl_register();
4786
4787         return 0;
4788 errout:
4789         rtnl_af_unregister(&inet6_ops);
4790 errout_af:
4791         unregister_netdevice_notifier(&ipv6_dev_notf);
4792 errlo:
4793         unregister_pernet_subsys(&addrconf_ops);
4794 out_addrlabel:
4795         ipv6_addr_label_cleanup();
4796 out:
4797         return err;
4798 }
4799
4800 void addrconf_cleanup(void)
4801 {
4802         struct net_device *dev;
4803         int i;
4804
4805         unregister_netdevice_notifier(&ipv6_dev_notf);
4806         unregister_pernet_subsys(&addrconf_ops);
4807         ipv6_addr_label_cleanup();
4808
4809         rtnl_lock();
4810
4811         __rtnl_af_unregister(&inet6_ops);
4812
4813         /* clean dev list */
4814         for_each_netdev(&init_net, dev) {
4815                 if (__in6_dev_get(dev) == NULL)
4816                         continue;
4817                 addrconf_ifdown(dev, 1);
4818         }
4819         addrconf_ifdown(init_net.loopback_dev, 2);
4820
4821         /*
4822          *      Check hash table.
4823          */
4824         spin_lock_bh(&addrconf_hash_lock);
4825         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4826                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4827         spin_unlock_bh(&addrconf_hash_lock);
4828
4829         del_timer(&addr_chk_timer);
4830         rtnl_unlock();
4831 }