Pileus Git - ~andy/linux/blob - net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #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 int __ipv6_regen_rndid(struct inet6_dev *idev);
 131 static int __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         rcu_read_lock();
 498         for_each_netdev_rcu(net, dev) {
 499                 idev = __in6_dev_get(dev);
 500                 if (idev) {
 501                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 502                         idev->cnf.forwarding = newf;
 503                         if (changed)
 504                                 dev_forward_change(idev);
 505                 }
 506         }
 507         rcu_read_unlock();
 508 }
 509
 510 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 511 {
 512         struct net *net;
 513         int old;
 514
 515         if (!rtnl_trylock())
 516                 return restart_syscall();
 517
 518         net = (struct net *)table->extra2;
 519         old = *p;
 520         *p = newf;
 521
 522         if (p == &net->ipv6.devconf_dflt->forwarding) {
 523                 rtnl_unlock();
 524                 return 0;
 525         }
 526
 527         if (p == &net->ipv6.devconf_all->forwarding) {
 528                 net->ipv6.devconf_dflt->forwarding = newf;
 529                 addrconf_forward_change(net, newf);
 530         } else if ((!newf) ^ (!old))
 531                 dev_forward_change((struct inet6_dev *)table->extra1);
 532         rtnl_unlock();
 533
 534         if (newf)
 535                 rt6_purge_dflt_routers(net);
 536         return 1;
 537 }
 538 #endif
 539
 540 /* Nobody refers to this ifaddr, destroy it */
 541 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 542 {
 543         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 544
 545 #ifdef NET_REFCNT_DEBUG
 546         pr_debug("%s\n", __func__);
 547 #endif
 548
 549         in6_dev_put(ifp->idev);
 550
 551         if (del_timer(&ifp->timer))
 552                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 553
 554         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 555                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 556                 return;
 557         }
 558         dst_release(&ifp->rt->dst);
 559
 560         kfree_rcu(ifp, rcu);
 561 }
 562
 563 static void
 564 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 565 {
 566         struct list_head *p;
 567         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 568
 569         /*
 570          * Each device address list is sorted in order of scope -
 571          * global before linklocal.
 572          */
 573         list_for_each(p, &idev->addr_list) {
 574                 struct inet6_ifaddr *ifa
 575                         = list_entry(p, struct inet6_ifaddr, if_list);
 576                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 577                         break;
 578         }
 579
 580         list_add_tail(&ifp->if_list, p);
 581 }
 582
 583 static u32 inet6_addr_hash(const struct in6_addr *addr)
 584 {
 585         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 586 }
 587
 588 /* On success it returns ifp with increased reference count */
 589
 590 static struct inet6_ifaddr *
 591 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 592               int scope, u32 flags)
 593 {
 594         struct inet6_ifaddr *ifa = NULL;
 595         struct rt6_info *rt;
 596         unsigned int hash;
 597         int err = 0;
 598         int addr_type = ipv6_addr_type(addr);
 599
 600         if (addr_type == IPV6_ADDR_ANY ||
 601             addr_type & IPV6_ADDR_MULTICAST ||
 602             (!(idev->dev->flags & IFF_LOOPBACK) &&
 603              addr_type & IPV6_ADDR_LOOPBACK))
 604                 return ERR_PTR(-EADDRNOTAVAIL);
 605
 606         rcu_read_lock_bh();
 607         if (idev->dead) {
 608                 err = -ENODEV;                  /*XXX*/
 609                 goto out2;
 610         }
 611
 612         if (idev->cnf.disable_ipv6) {
 613                 err = -EACCES;
 614                 goto out2;
 615         }
 616
 617         spin_lock(&addrconf_hash_lock);
 618
 619         /* Ignore adding duplicate addresses on an interface */
 620         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 621                 ADBG(("ipv6_add_addr: already assigned\n"));
 622                 err = -EEXIST;
 623                 goto out;
 624         }
 625
 626         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 627
 628         if (ifa == NULL) {
 629                 ADBG(("ipv6_add_addr: malloc failed\n"));
 630                 err = -ENOBUFS;
 631                 goto out;
 632         }
 633
 634         rt = addrconf_dst_alloc(idev, addr, false);
 635         if (IS_ERR(rt)) {
 636                 err = PTR_ERR(rt);
 637                 goto out;
 638         }
 639
 640         ifa->addr = *addr;
 641
 642         spin_lock_init(&ifa->lock);
 643         spin_lock_init(&ifa->state_lock);
 644         init_timer(&ifa->timer);
 645         INIT_HLIST_NODE(&ifa->addr_lst);
 646         ifa->timer.data = (unsigned long) ifa;
 647         ifa->scope = scope;
 648         ifa->prefix_len = pfxlen;
 649         ifa->flags = flags | IFA_F_TENTATIVE;
 650         ifa->cstamp = ifa->tstamp = jiffies;
 651
 652         ifa->rt = rt;
 653
 654         ifa->idev = idev;
 655         in6_dev_hold(idev);
 656         /* For caller */
 657         in6_ifa_hold(ifa);
 658
 659         /* Add to big hash table */
 660         hash = inet6_addr_hash(addr);
 661
 662         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 663         spin_unlock(&addrconf_hash_lock);
 664
 665         write_lock(&idev->lock);
 666         /* Add to inet6_dev unicast addr list. */
 667         ipv6_link_dev_addr(idev, ifa);
 668
 669 #ifdef CONFIG_IPV6_PRIVACY
 670         if (ifa->flags&IFA_F_TEMPORARY) {
 671                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 672                 in6_ifa_hold(ifa);
 673         }
 674 #endif
 675
 676         in6_ifa_hold(ifa);
 677         write_unlock(&idev->lock);
 678 out2:
 679         rcu_read_unlock_bh();
 680
 681         if (likely(err == 0))
 682                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 683         else {
 684                 kfree(ifa);
 685                 ifa = ERR_PTR(err);
 686         }
 687
 688         return ifa;
 689 out:
 690         spin_unlock(&addrconf_hash_lock);
 691         goto out2;
 692 }
 693
 694 /* This function wants to get referenced ifp and releases it before return */
 695
 696 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 697 {
 698         struct inet6_ifaddr *ifa, *ifn;
 699         struct inet6_dev *idev = ifp->idev;
 700         int state;
 701         int deleted = 0, onlink = 0;
 702         unsigned long expires = jiffies;
 703
 704         spin_lock_bh(&ifp->state_lock);
 705         state = ifp->state;
 706         ifp->state = INET6_IFADDR_STATE_DEAD;
 707         spin_unlock_bh(&ifp->state_lock);
 708
 709         if (state == INET6_IFADDR_STATE_DEAD)
 710                 goto out;
 711
 712         spin_lock_bh(&addrconf_hash_lock);
 713         hlist_del_init_rcu(&ifp->addr_lst);
 714         spin_unlock_bh(&addrconf_hash_lock);
 715
 716         write_lock_bh(&idev->lock);
 717 #ifdef CONFIG_IPV6_PRIVACY
 718         if (ifp->flags&IFA_F_TEMPORARY) {
 719                 list_del(&ifp->tmp_list);
 720                 if (ifp->ifpub) {
 721                         in6_ifa_put(ifp->ifpub);
 722                         ifp->ifpub = NULL;
 723                 }
 724                 __in6_ifa_put(ifp);
 725         }
 726 #endif
 727
 728         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 729                 if (ifa == ifp) {
 730                         list_del_init(&ifp->if_list);
 731                         __in6_ifa_put(ifp);
 732
 733                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 734                                 break;
 735                         deleted = 1;
 736                         continue;
 737                 } else if (ifp->flags & IFA_F_PERMANENT) {
 738                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 739                                               ifp->prefix_len)) {
 740                                 if (ifa->flags & IFA_F_PERMANENT) {
 741                                         onlink = 1;
 742                                         if (deleted)
 743                                                 break;
 744                                 } else {
 745                                         unsigned long lifetime;
 746
 747                                         if (!onlink)
 748                                                 onlink = -1;
 749
 750                                         spin_lock(&ifa->lock);
 751
 752                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 753                                         /*
 754                                          * Note: Because this address is
 755                                          * not permanent, lifetime <
 756                                          * LONG_MAX / HZ here.
 757                                          */
 758                                         if (time_before(expires,
 759                                                         ifa->tstamp + lifetime * HZ))
 760                                                 expires = ifa->tstamp + lifetime * HZ;
 761                                         spin_unlock(&ifa->lock);
 762                                 }
 763                         }
 764                 }
 765         }
 766         write_unlock_bh(&idev->lock);
 767
 768         addrconf_del_timer(ifp);
 769
 770         ipv6_ifa_notify(RTM_DELADDR, ifp);
 771
 772         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 773
 774         /*
 775          * Purge or update corresponding prefix
 776          *
 777          * 1) we don't purge prefix here if address was not permanent.
 778          *    prefix is managed by its own lifetime.
 779          * 2) if there're no addresses, delete prefix.
 780          * 3) if there're still other permanent address(es),
 781          *    corresponding prefix is still permanent.
 782          * 4) otherwise, update prefix lifetime to the
 783          *    longest valid lifetime among the corresponding
 784          *    addresses on the device.
 785          *    Note: subsequent RA will update lifetime.
 786          *
 787          * --yoshfuji
 788          */
 789         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 790                 struct in6_addr prefix;
 791                 struct rt6_info *rt;
 792                 struct net *net = dev_net(ifp->idev->dev);
 793                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 794                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 795
 796                 if (rt && addrconf_is_prefix_route(rt)) {
 797                         if (onlink == 0) {
 798                                 ip6_del_rt(rt);
 799                                 rt = NULL;
 800                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 801                                 rt6_set_expires(rt, expires);
 802                         }
 803                 }
 804                 dst_release(&rt->dst);
 805         }
 806
 807         /* clean up prefsrc entries */
 808         rt6_remove_prefsrc(ifp);
 809 out:
 810         in6_ifa_put(ifp);
 811 }
 812
 813 #ifdef CONFIG_IPV6_PRIVACY
 814 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 815 {
 816         struct inet6_dev *idev = ifp->idev;
 817         struct in6_addr addr, *tmpaddr;
 818         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 819         unsigned long regen_advance;
 820         int tmp_plen;
 821         int ret = 0;
 822         int max_addresses;
 823         u32 addr_flags;
 824         unsigned long now = jiffies;
 825
 826         write_lock(&idev->lock);
 827         if (ift) {
 828                 spin_lock_bh(&ift->lock);
 829                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 830                 spin_unlock_bh(&ift->lock);
 831                 tmpaddr = &addr;
 832         } else {
 833                 tmpaddr = NULL;
 834         }
 835 retry:
 836         in6_dev_hold(idev);
 837         if (idev->cnf.use_tempaddr <= 0) {
 838                 write_unlock(&idev->lock);
 839                 pr_info("%s: use_tempaddr is disabled\n", __func__);
 840                 in6_dev_put(idev);
 841                 ret = -1;
 842                 goto out;
 843         }
 844         spin_lock_bh(&ifp->lock);
 845         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 846                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 847                 spin_unlock_bh(&ifp->lock);
 848                 write_unlock(&idev->lock);
 849                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
 850                         __func__);
 851                 in6_dev_put(idev);
 852                 ret = -1;
 853                 goto out;
 854         }
 855         in6_ifa_hold(ifp);
 856         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 857         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 858                 spin_unlock_bh(&ifp->lock);
 859                 write_unlock(&idev->lock);
 860                 pr_warn("%s: regeneration of randomized interface id failed\n",
 861                         __func__);
 862                 in6_ifa_put(ifp);
 863                 in6_dev_put(idev);
 864                 ret = -1;
 865                 goto out;
 866         }
 867         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 868         age = (now - ifp->tstamp) / HZ;
 869         tmp_valid_lft = min_t(__u32,
 870                               ifp->valid_lft,
 871                               idev->cnf.temp_valid_lft + age);
 872         tmp_prefered_lft = min_t(__u32,
 873                                  ifp->prefered_lft,
 874                                  idev->cnf.temp_prefered_lft + age -
 875                                  idev->cnf.max_desync_factor);
 876         tmp_plen = ifp->prefix_len;
 877         max_addresses = idev->cnf.max_addresses;
 878         tmp_tstamp = ifp->tstamp;
 879         spin_unlock_bh(&ifp->lock);
 880
 881         regen_advance = idev->cnf.regen_max_retry *
 882                         idev->cnf.dad_transmits *
 883                         idev->nd_parms->retrans_time / HZ;
 884         write_unlock(&idev->lock);
 885
 886         /* A temporary address is created only if this calculated Preferred
 887          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 888          * an implementation must not create a temporary address with a zero
 889          * Preferred Lifetime.
 890          */
 891         if (tmp_prefered_lft <= regen_advance) {
 892                 in6_ifa_put(ifp);
 893                 in6_dev_put(idev);
 894                 ret = -1;
 895                 goto out;
 896         }
 897
 898         addr_flags = IFA_F_TEMPORARY;
 899         /* set in addrconf_prefix_rcv() */
 900         if (ifp->flags & IFA_F_OPTIMISTIC)
 901                 addr_flags |= IFA_F_OPTIMISTIC;
 902
 903         ift = !max_addresses ||
 904               ipv6_count_addresses(idev) < max_addresses ?
 905                 ipv6_add_addr(idev, &addr, tmp_plen,
 906                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 907                               addr_flags) : NULL;
 908         if (!ift || IS_ERR(ift)) {
 909                 in6_ifa_put(ifp);
 910                 in6_dev_put(idev);
 911                 pr_info("%s: retry temporary address regeneration\n", __func__);
 912                 tmpaddr = &addr;
 913                 write_lock(&idev->lock);
 914                 goto retry;
 915         }
 916
 917         spin_lock_bh(&ift->lock);
 918         ift->ifpub = ifp;
 919         ift->valid_lft = tmp_valid_lft;
 920         ift->prefered_lft = tmp_prefered_lft;
 921         ift->cstamp = now;
 922         ift->tstamp = tmp_tstamp;
 923         spin_unlock_bh(&ift->lock);
 924
 925         addrconf_dad_start(ift);
 926         in6_ifa_put(ift);
 927         in6_dev_put(idev);
 928 out:
 929         return ret;
 930 }
 931 #endif
 932
 933 /*
 934  *      Choose an appropriate source address (RFC3484)
 935  */
 936 enum {
 937         IPV6_SADDR_RULE_INIT = 0,
 938         IPV6_SADDR_RULE_LOCAL,
 939         IPV6_SADDR_RULE_SCOPE,
 940         IPV6_SADDR_RULE_PREFERRED,
 941 #ifdef CONFIG_IPV6_MIP6
 942         IPV6_SADDR_RULE_HOA,
 943 #endif
 944         IPV6_SADDR_RULE_OIF,
 945         IPV6_SADDR_RULE_LABEL,
 946 #ifdef CONFIG_IPV6_PRIVACY
 947         IPV6_SADDR_RULE_PRIVACY,
 948 #endif
 949         IPV6_SADDR_RULE_ORCHID,
 950         IPV6_SADDR_RULE_PREFIX,
 951         IPV6_SADDR_RULE_MAX
 952 };
 953
 954 struct ipv6_saddr_score {
 955         int                     rule;
 956         int                     addr_type;
 957         struct inet6_ifaddr     *ifa;
 958         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 959         int                     scopedist;
 960         int                     matchlen;
 961 };
 962
 963 struct ipv6_saddr_dst {
 964         const struct in6_addr *addr;
 965         int ifindex;
 966         int scope;
 967         int label;
 968         unsigned int prefs;
 969 };
 970
 971 static inline int ipv6_saddr_preferred(int type)
 972 {
 973         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 974                 return 1;
 975         return 0;
 976 }
 977
 978 static int ipv6_get_saddr_eval(struct net *net,
 979                                struct ipv6_saddr_score *score,
 980                                struct ipv6_saddr_dst *dst,
 981                                int i)
 982 {
 983         int ret;
 984
 985         if (i <= score->rule) {
 986                 switch (i) {
 987                 case IPV6_SADDR_RULE_SCOPE:
 988                         ret = score->scopedist;
 989                         break;
 990                 case IPV6_SADDR_RULE_PREFIX:
 991                         ret = score->matchlen;
 992                         break;
 993                 default:
 994                         ret = !!test_bit(i, score->scorebits);
 995                 }
 996                 goto out;
 997         }
 998
 999         switch (i) {
1000         case IPV6_SADDR_RULE_INIT:
1001                 /* Rule 0: remember if hiscore is not ready yet */
1002                 ret = !!score->ifa;
1003                 break;
1004         case IPV6_SADDR_RULE_LOCAL:
1005                 /* Rule 1: Prefer same address */
1006                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1007                 break;
1008         case IPV6_SADDR_RULE_SCOPE:
1009                 /* Rule 2: Prefer appropriate scope
1010                  *
1011                  *      ret
1012                  *       ^
1013                  *    -1 |  d 15
1014                  *    ---+--+-+---> scope
1015                  *       |
1016                  *       |             d is scope of the destination.
1017                  *  B-d  |  \
1018                  *       |   \      <- smaller scope is better if
1019                  *  B-15 |    \        if scope is enough for destinaion.
1020                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1021                  * d-C-1 | /
1022                  *       |/         <- greater is better
1023                  *   -C  /             if scope is not enough for destination.
1024                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1025                  *
1026                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1027                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1028                  * Assume B = 0 and we get C > 29.
1029                  */
1030                 ret = __ipv6_addr_src_scope(score->addr_type);
1031                 if (ret >= dst->scope)
1032                         ret = -ret;
1033                 else
1034                         ret -= 128;     /* 30 is enough */
1035                 score->scopedist = ret;
1036                 break;
1037         case IPV6_SADDR_RULE_PREFERRED:
1038                 /* Rule 3: Avoid deprecated and optimistic addresses */
1039                 ret = ipv6_saddr_preferred(score->addr_type) ||
1040                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1041                 break;
1042 #ifdef CONFIG_IPV6_MIP6
1043         case IPV6_SADDR_RULE_HOA:
1044             {
1045                 /* Rule 4: Prefer home address */
1046                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1047                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1048                 break;
1049             }
1050 #endif
1051         case IPV6_SADDR_RULE_OIF:
1052                 /* Rule 5: Prefer outgoing interface */
1053                 ret = (!dst->ifindex ||
1054                        dst->ifindex == score->ifa->idev->dev->ifindex);
1055                 break;
1056         case IPV6_SADDR_RULE_LABEL:
1057                 /* Rule 6: Prefer matching label */
1058                 ret = ipv6_addr_label(net,
1059                                       &score->ifa->addr, score->addr_type,
1060                                       score->ifa->idev->dev->ifindex) == dst->label;
1061                 break;
1062 #ifdef CONFIG_IPV6_PRIVACY
1063         case IPV6_SADDR_RULE_PRIVACY:
1064             {
1065                 /* Rule 7: Prefer public address
1066                  * Note: prefer temporary address if use_tempaddr >= 2
1067                  */
1068                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1069                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1070                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1071                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1072                 break;
1073             }
1074 #endif
1075         case IPV6_SADDR_RULE_ORCHID:
1076                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1077                  *          non-ORCHID vs non-ORCHID
1078                  */
1079                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1080                         ipv6_addr_orchid(dst->addr));
1081                 break;
1082         case IPV6_SADDR_RULE_PREFIX:
1083                 /* Rule 8: Use longest matching prefix */
1084                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1085                                                        dst->addr);
1086                 break;
1087         default:
1088                 ret = 0;
1089         }
1090
1091         if (ret)
1092                 __set_bit(i, score->scorebits);
1093         score->rule = i;
1094 out:
1095         return ret;
1096 }
1097
1098 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1099                        const struct in6_addr *daddr, unsigned int prefs,
1100                        struct in6_addr *saddr)
1101 {
1102         struct ipv6_saddr_score scores[2],
1103                                 *score = &scores[0], *hiscore = &scores[1];
1104         struct ipv6_saddr_dst dst;
1105         struct net_device *dev;
1106         int dst_type;
1107
1108         dst_type = __ipv6_addr_type(daddr);
1109         dst.addr = daddr;
1110         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1111         dst.scope = __ipv6_addr_src_scope(dst_type);
1112         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1113         dst.prefs = prefs;
1114
1115         hiscore->rule = -1;
1116         hiscore->ifa = NULL;
1117
1118         rcu_read_lock();
1119
1120         for_each_netdev_rcu(net, dev) {
1121                 struct inet6_dev *idev;
1122
1123                 /* Candidate Source Address (section 4)
1124                  *  - multicast and link-local destination address,
1125                  *    the set of candidate source address MUST only
1126                  *    include addresses assigned to interfaces
1127                  *    belonging to the same link as the outgoing
1128                  *    interface.
1129                  * (- For site-local destination addresses, the
1130                  *    set of candidate source addresses MUST only
1131                  *    include addresses assigned to interfaces
1132                  *    belonging to the same site as the outgoing
1133                  *    interface.)
1134                  */
1135                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1136                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1137                     dst.ifindex && dev->ifindex != dst.ifindex)
1138                         continue;
1139
1140                 idev = __in6_dev_get(dev);
1141                 if (!idev)
1142                         continue;
1143
1144                 read_lock_bh(&idev->lock);
1145                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1146                         int i;
1147
1148                         /*
1149                          * - Tentative Address (RFC2462 section 5.4)
1150                          *  - A tentative address is not considered
1151                          *    "assigned to an interface" in the traditional
1152                          *    sense, unless it is also flagged as optimistic.
1153                          * - Candidate Source Address (section 4)
1154                          *  - In any case, anycast addresses, multicast
1155                          *    addresses, and the unspecified address MUST
1156                          *    NOT be included in a candidate set.
1157                          */
1158                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1159                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1160                                 continue;
1161
1162                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1163
1164                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1165                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1166                                 LIMIT_NETDEBUG(KERN_DEBUG
1167                                                "ADDRCONF: unspecified / multicast address "
1168                                                "assigned as unicast address on %s",
1169                                                dev->name);
1170                                 continue;
1171                         }
1172
1173                         score->rule = -1;
1174                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1175
1176                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1177                                 int minihiscore, miniscore;
1178
1179                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1180                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1181
1182                                 if (minihiscore > miniscore) {
1183                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1184                                             score->scopedist > 0) {
1185                                                 /*
1186                                                  * special case:
1187                                                  * each remaining entry
1188                                                  * has too small (not enough)
1189                                                  * scope, because ifa entries
1190                                                  * are sorted by their scope
1191                                                  * values.
1192                                                  */
1193                                                 goto try_nextdev;
1194                                         }
1195                                         break;
1196                                 } else if (minihiscore < miniscore) {
1197                                         if (hiscore->ifa)
1198                                                 in6_ifa_put(hiscore->ifa);
1199
1200                                         in6_ifa_hold(score->ifa);
1201
1202                                         swap(hiscore, score);
1203
1204                                         /* restore our iterator */
1205                                         score->ifa = hiscore->ifa;
1206
1207                                         break;
1208                                 }
1209                         }
1210                 }
1211 try_nextdev:
1212                 read_unlock_bh(&idev->lock);
1213         }
1214         rcu_read_unlock();
1215
1216         if (!hiscore->ifa)
1217                 return -EADDRNOTAVAIL;
1218
1219         *saddr = hiscore->ifa->addr;
1220         in6_ifa_put(hiscore->ifa);
1221         return 0;
1222 }
1223 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1224
1225 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1226                     unsigned char banned_flags)
1227 {
1228         struct inet6_dev *idev;
1229         int err = -EADDRNOTAVAIL;
1230
1231         rcu_read_lock();
1232         idev = __in6_dev_get(dev);
1233         if (idev) {
1234                 struct inet6_ifaddr *ifp;
1235
1236                 read_lock_bh(&idev->lock);
1237                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1238                         if (ifp->scope == IFA_LINK &&
1239                             !(ifp->flags & banned_flags)) {
1240                                 *addr = ifp->addr;
1241                                 err = 0;
1242                                 break;
1243                         }
1244                 }
1245                 read_unlock_bh(&idev->lock);
1246         }
1247         rcu_read_unlock();
1248         return err;
1249 }
1250
1251 static int ipv6_count_addresses(struct inet6_dev *idev)
1252 {
1253         int cnt = 0;
1254         struct inet6_ifaddr *ifp;
1255
1256         read_lock_bh(&idev->lock);
1257         list_for_each_entry(ifp, &idev->addr_list, if_list)
1258                 cnt++;
1259         read_unlock_bh(&idev->lock);
1260         return cnt;
1261 }
1262
1263 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1264                   struct net_device *dev, int strict)
1265 {
1266         struct inet6_ifaddr *ifp;
1267         struct hlist_node *node;
1268         unsigned int hash = inet6_addr_hash(addr);
1269
1270         rcu_read_lock_bh();
1271         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1272                 if (!net_eq(dev_net(ifp->idev->dev), net))
1273                         continue;
1274                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1275                     !(ifp->flags&IFA_F_TENTATIVE) &&
1276                     (dev == NULL || ifp->idev->dev == dev ||
1277                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1278                         rcu_read_unlock_bh();
1279                         return 1;
1280                 }
1281         }
1282
1283         rcu_read_unlock_bh();
1284         return 0;
1285 }
1286 EXPORT_SYMBOL(ipv6_chk_addr);
1287
1288 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1289                                struct net_device *dev)
1290 {
1291         unsigned int hash = inet6_addr_hash(addr);
1292         struct inet6_ifaddr *ifp;
1293         struct hlist_node *node;
1294
1295         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1296                 if (!net_eq(dev_net(ifp->idev->dev), net))
1297                         continue;
1298                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1299                         if (dev == NULL || ifp->idev->dev == dev)
1300                                 return true;
1301                 }
1302         }
1303         return false;
1304 }
1305
1306 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1307 {
1308         struct inet6_dev *idev;
1309         struct inet6_ifaddr *ifa;
1310         int     onlink;
1311
1312         onlink = 0;
1313         rcu_read_lock();
1314         idev = __in6_dev_get(dev);
1315         if (idev) {
1316                 read_lock_bh(&idev->lock);
1317                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1318                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1319                                                    ifa->prefix_len);
1320                         if (onlink)
1321                                 break;
1322                 }
1323                 read_unlock_bh(&idev->lock);
1324         }
1325         rcu_read_unlock();
1326         return onlink;
1327 }
1328 EXPORT_SYMBOL(ipv6_chk_prefix);
1329
1330 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1331                                      struct net_device *dev, int strict)
1332 {
1333         struct inet6_ifaddr *ifp, *result = NULL;
1334         unsigned int hash = inet6_addr_hash(addr);
1335         struct hlist_node *node;
1336
1337         rcu_read_lock_bh();
1338         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1339                 if (!net_eq(dev_net(ifp->idev->dev), net))
1340                         continue;
1341                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1342                         if (dev == NULL || ifp->idev->dev == dev ||
1343                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1344                                 result = ifp;
1345                                 in6_ifa_hold(ifp);
1346                                 break;
1347                         }
1348                 }
1349         }
1350         rcu_read_unlock_bh();
1351
1352         return result;
1353 }
1354
1355 /* Gets referenced address, destroys ifaddr */
1356
1357 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1358 {
1359         if (ifp->flags&IFA_F_PERMANENT) {
1360                 spin_lock_bh(&ifp->lock);
1361                 addrconf_del_timer(ifp);
1362                 ifp->flags |= IFA_F_TENTATIVE;
1363                 if (dad_failed)
1364                         ifp->flags |= IFA_F_DADFAILED;
1365                 spin_unlock_bh(&ifp->lock);
1366                 if (dad_failed)
1367                         ipv6_ifa_notify(0, ifp);
1368                 in6_ifa_put(ifp);
1369 #ifdef CONFIG_IPV6_PRIVACY
1370         } else if (ifp->flags&IFA_F_TEMPORARY) {
1371                 struct inet6_ifaddr *ifpub;
1372                 spin_lock_bh(&ifp->lock);
1373                 ifpub = ifp->ifpub;
1374                 if (ifpub) {
1375                         in6_ifa_hold(ifpub);
1376                         spin_unlock_bh(&ifp->lock);
1377                         ipv6_create_tempaddr(ifpub, ifp);
1378                         in6_ifa_put(ifpub);
1379                 } else {
1380                         spin_unlock_bh(&ifp->lock);
1381                 }
1382                 ipv6_del_addr(ifp);
1383 #endif
1384         } else
1385                 ipv6_del_addr(ifp);
1386 }
1387
1388 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1389 {
1390         int err = -ENOENT;
1391
1392         spin_lock(&ifp->state_lock);
1393         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1394                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1395                 err = 0;
1396         }
1397         spin_unlock(&ifp->state_lock);
1398
1399         return err;
1400 }
1401
1402 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1403 {
1404         struct inet6_dev *idev = ifp->idev;
1405
1406         if (addrconf_dad_end(ifp)) {
1407                 in6_ifa_put(ifp);
1408                 return;
1409         }
1410
1411         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1412                              ifp->idev->dev->name, &ifp->addr);
1413
1414         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1415                 struct in6_addr addr;
1416
1417                 addr.s6_addr32[0] = htonl(0xfe800000);
1418                 addr.s6_addr32[1] = 0;
1419
1420                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1421                     ipv6_addr_equal(&ifp->addr, &addr)) {
1422                         /* DAD failed for link-local based on MAC address */
1423                         idev->cnf.disable_ipv6 = 1;
1424
1425                         pr_info("%s: IPv6 being disabled!\n",
1426                                 ifp->idev->dev->name);
1427                 }
1428         }
1429
1430         addrconf_dad_stop(ifp, 1);
1431 }
1432
1433 /* Join to solicited addr multicast group. */
1434
1435 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1436 {
1437         struct in6_addr maddr;
1438
1439         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1440                 return;
1441
1442         addrconf_addr_solict_mult(addr, &maddr);
1443         ipv6_dev_mc_inc(dev, &maddr);
1444 }
1445
1446 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1447 {
1448         struct in6_addr maddr;
1449
1450         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1451                 return;
1452
1453         addrconf_addr_solict_mult(addr, &maddr);
1454         __ipv6_dev_mc_dec(idev, &maddr);
1455 }
1456
1457 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1458 {
1459         struct in6_addr addr;
1460         if (ifp->prefix_len == 127) /* RFC 6164 */
1461                 return;
1462         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1463         if (ipv6_addr_any(&addr))
1464                 return;
1465         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1466 }
1467
1468 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1469 {
1470         struct in6_addr addr;
1471         if (ifp->prefix_len == 127) /* RFC 6164 */
1472                 return;
1473         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1474         if (ipv6_addr_any(&addr))
1475                 return;
1476         __ipv6_dev_ac_dec(ifp->idev, &addr);
1477 }
1478
1479 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1480 {
1481         if (dev->addr_len != ETH_ALEN)
1482                 return -1;
1483         memcpy(eui, dev->dev_addr, 3);
1484         memcpy(eui + 5, dev->dev_addr + 3, 3);
1485
1486         /*
1487          * The zSeries OSA network cards can be shared among various
1488          * OS instances, but the OSA cards have only one MAC address.
1489          * This leads to duplicate address conflicts in conjunction
1490          * with IPv6 if more than one instance uses the same card.
1491          *
1492          * The driver for these cards can deliver a unique 16-bit
1493          * identifier for each instance sharing the same card.  It is
1494          * placed instead of 0xFFFE in the interface identifier.  The
1495          * "u" bit of the interface identifier is not inverted in this
1496          * case.  Hence the resulting interface identifier has local
1497          * scope according to RFC2373.
1498          */
1499         if (dev->dev_id) {
1500                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1501                 eui[4] = dev->dev_id & 0xFF;
1502         } else {
1503                 eui[3] = 0xFF;
1504                 eui[4] = 0xFE;
1505                 eui[0] ^= 2;
1506         }
1507         return 0;
1508 }
1509
1510 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1511 {
1512         if (dev->addr_len != IEEE802154_ADDR_LEN)
1513                 return -1;
1514         memcpy(eui, dev->dev_addr, 8);
1515         return 0;
1516 }
1517
1518 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1519 {
1520         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1521         if (dev->addr_len != ARCNET_ALEN)
1522                 return -1;
1523         memset(eui, 0, 7);
1524         eui[7] = *(u8 *)dev->dev_addr;
1525         return 0;
1526 }
1527
1528 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1529 {
1530         if (dev->addr_len != INFINIBAND_ALEN)
1531                 return -1;
1532         memcpy(eui, dev->dev_addr + 12, 8);
1533         eui[0] |= 2;
1534         return 0;
1535 }
1536
1537 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1538 {
1539         if (addr == 0)
1540                 return -1;
1541         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1542                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1543                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1544                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1545                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1546                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1547         eui[1] = 0;
1548         eui[2] = 0x5E;
1549         eui[3] = 0xFE;
1550         memcpy(eui + 4, &addr, 4);
1551         return 0;
1552 }
1553
1554 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1555 {
1556         if (dev->priv_flags & IFF_ISATAP)
1557                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1558         return -1;
1559 }
1560
1561 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1562 {
1563         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1564 }
1565
1566 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1567 {
1568         switch (dev->type) {
1569         case ARPHRD_ETHER:
1570         case ARPHRD_FDDI:
1571                 return addrconf_ifid_eui48(eui, dev);
1572         case ARPHRD_ARCNET:
1573                 return addrconf_ifid_arcnet(eui, dev);
1574         case ARPHRD_INFINIBAND:
1575                 return addrconf_ifid_infiniband(eui, dev);
1576         case ARPHRD_SIT:
1577                 return addrconf_ifid_sit(eui, dev);
1578         case ARPHRD_IPGRE:
1579                 return addrconf_ifid_gre(eui, dev);
1580         case ARPHRD_IEEE802154:
1581                 return addrconf_ifid_eui64(eui, dev);
1582         }
1583         return -1;
1584 }
1585
1586 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1587 {
1588         int err = -1;
1589         struct inet6_ifaddr *ifp;
1590
1591         read_lock_bh(&idev->lock);
1592         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1593                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1594                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1595                         err = 0;
1596                         break;
1597                 }
1598         }
1599         read_unlock_bh(&idev->lock);
1600         return err;
1601 }
1602
1603 #ifdef CONFIG_IPV6_PRIVACY
1604 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1605 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1606 {
1607 regen:
1608         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1609         idev->rndid[0] &= ~0x02;
1610
1611         /*
1612          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1613          * check if generated address is not inappropriate
1614          *
1615          *  - Reserved subnet anycast (RFC 2526)
1616          *      11111101 11....11 1xxxxxxx
1617          *  - ISATAP (RFC4214) 6.1
1618          *      00-00-5E-FE-xx-xx-xx-xx
1619          *  - value 0
1620          *  - XXX: already assigned to an address on the device
1621          */
1622         if (idev->rndid[0] == 0xfd &&
1623             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1624             (idev->rndid[7]&0x80))
1625                 goto regen;
1626         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1627                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1628                         goto regen;
1629                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1630                         goto regen;
1631         }
1632
1633         return 0;
1634 }
1635
1636 static void ipv6_regen_rndid(unsigned long data)
1637 {
1638         struct inet6_dev *idev = (struct inet6_dev *) data;
1639         unsigned long expires;
1640
1641         rcu_read_lock_bh();
1642         write_lock_bh(&idev->lock);
1643
1644         if (idev->dead)
1645                 goto out;
1646
1647         if (__ipv6_regen_rndid(idev) < 0)
1648                 goto out;
1649
1650         expires = jiffies +
1651                 idev->cnf.temp_prefered_lft * HZ -
1652                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1653                 idev->cnf.max_desync_factor * HZ;
1654         if (time_before(expires, jiffies)) {
1655                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1656                         __func__, idev->dev->name);
1657                 goto out;
1658         }
1659
1660         if (!mod_timer(&idev->regen_timer, expires))
1661                 in6_dev_hold(idev);
1662
1663 out:
1664         write_unlock_bh(&idev->lock);
1665         rcu_read_unlock_bh();
1666         in6_dev_put(idev);
1667 }
1668
1669 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1670 {
1671         int ret = 0;
1672
1673         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1674                 ret = __ipv6_regen_rndid(idev);
1675         return ret;
1676 }
1677 #endif
1678
1679 /*
1680  *      Add prefix route.
1681  */
1682
1683 static void
1684 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1685                       unsigned long expires, u32 flags)
1686 {
1687         struct fib6_config cfg = {
1688                 .fc_table = RT6_TABLE_PREFIX,
1689                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1690                 .fc_ifindex = dev->ifindex,
1691                 .fc_expires = expires,
1692                 .fc_dst_len = plen,
1693                 .fc_flags = RTF_UP | flags,
1694                 .fc_nlinfo.nl_net = dev_net(dev),
1695                 .fc_protocol = RTPROT_KERNEL,
1696         };
1697
1698         cfg.fc_dst = *pfx;
1699
1700         /* Prevent useless cloning on PtP SIT.
1701            This thing is done here expecting that the whole
1702            class of non-broadcast devices need not cloning.
1703          */
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1706                 cfg.fc_flags |= RTF_NONEXTHOP;
1707 #endif
1708
1709         ip6_route_add(&cfg);
1710 }
1711
1712
1713 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1714                                                   int plen,
1715                                                   const struct net_device *dev,
1716                                                   u32 flags, u32 noflags)
1717 {
1718         struct fib6_node *fn;
1719         struct rt6_info *rt = NULL;
1720         struct fib6_table *table;
1721
1722         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1723         if (table == NULL)
1724                 return NULL;
1725
1726         write_lock_bh(&table->tb6_lock);
1727         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1728         if (!fn)
1729                 goto out;
1730         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1731                 if (rt->dst.dev->ifindex != dev->ifindex)
1732                         continue;
1733                 if ((rt->rt6i_flags & flags) != flags)
1734                         continue;
1735                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1736                         continue;
1737                 dst_hold(&rt->dst);
1738                 break;
1739         }
1740 out:
1741         write_unlock_bh(&table->tb6_lock);
1742         return rt;
1743 }
1744
1745
1746 /* Create "default" multicast route to the interface */
1747
1748 static void addrconf_add_mroute(struct net_device *dev)
1749 {
1750         struct fib6_config cfg = {
1751                 .fc_table = RT6_TABLE_LOCAL,
1752                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1753                 .fc_ifindex = dev->ifindex,
1754                 .fc_dst_len = 8,
1755                 .fc_flags = RTF_UP,
1756                 .fc_nlinfo.nl_net = dev_net(dev),
1757         };
1758
1759         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1760
1761         ip6_route_add(&cfg);
1762 }
1763
1764 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1765 static void sit_route_add(struct net_device *dev)
1766 {
1767         struct fib6_config cfg = {
1768                 .fc_table = RT6_TABLE_MAIN,
1769                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1770                 .fc_ifindex = dev->ifindex,
1771                 .fc_dst_len = 96,
1772                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1773                 .fc_nlinfo.nl_net = dev_net(dev),
1774         };
1775
1776         /* prefix length - 96 bits "::d.d.d.d" */
1777         ip6_route_add(&cfg);
1778 }
1779 #endif
1780
1781 static void addrconf_add_lroute(struct net_device *dev)
1782 {
1783         struct in6_addr addr;
1784
1785         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1786         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1787 }
1788
1789 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1790 {
1791         struct inet6_dev *idev;
1792
1793         ASSERT_RTNL();
1794
1795         idev = ipv6_find_idev(dev);
1796         if (!idev)
1797                 return ERR_PTR(-ENOBUFS);
1798
1799         if (idev->cnf.disable_ipv6)
1800                 return ERR_PTR(-EACCES);
1801
1802         /* Add default multicast route */
1803         if (!(dev->flags & IFF_LOOPBACK))
1804                 addrconf_add_mroute(dev);
1805
1806         /* Add link local route */
1807         addrconf_add_lroute(dev);
1808         return idev;
1809 }
1810
1811 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1812 {
1813         struct prefix_info *pinfo;
1814         __u32 valid_lft;
1815         __u32 prefered_lft;
1816         int addr_type;
1817         struct inet6_dev *in6_dev;
1818         struct net *net = dev_net(dev);
1819
1820         pinfo = (struct prefix_info *) opt;
1821
1822         if (len < sizeof(struct prefix_info)) {
1823                 ADBG(("addrconf: prefix option too short\n"));
1824                 return;
1825         }
1826
1827         /*
1828          *      Validation checks ([ADDRCONF], page 19)
1829          */
1830
1831         addr_type = ipv6_addr_type(&pinfo->prefix);
1832
1833         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1834                 return;
1835
1836         valid_lft = ntohl(pinfo->valid);
1837         prefered_lft = ntohl(pinfo->prefered);
1838
1839         if (prefered_lft > valid_lft) {
1840                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1841                 return;
1842         }
1843
1844         in6_dev = in6_dev_get(dev);
1845
1846         if (in6_dev == NULL) {
1847                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1848                                     dev->name);
1849                 return;
1850         }
1851
1852         /*
1853          *      Two things going on here:
1854          *      1) Add routes for on-link prefixes
1855          *      2) Configure prefixes with the auto flag set
1856          */
1857
1858         if (pinfo->onlink) {
1859                 struct rt6_info *rt;
1860                 unsigned long rt_expires;
1861
1862                 /* Avoid arithmetic overflow. Really, we could
1863                  * save rt_expires in seconds, likely valid_lft,
1864                  * but it would require division in fib gc, that it
1865                  * not good.
1866                  */
1867                 if (HZ > USER_HZ)
1868                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1869                 else
1870                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1871
1872                 if (addrconf_finite_timeout(rt_expires))
1873                         rt_expires *= HZ;
1874
1875                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1876                                                pinfo->prefix_len,
1877                                                dev,
1878                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1879                                                RTF_GATEWAY | RTF_DEFAULT);
1880
1881                 if (rt) {
1882                         /* Autoconf prefix route */
1883                         if (valid_lft == 0) {
1884                                 ip6_del_rt(rt);
1885                                 rt = NULL;
1886                         } else if (addrconf_finite_timeout(rt_expires)) {
1887                                 /* not infinity */
1888                                 rt6_set_expires(rt, jiffies + rt_expires);
1889                         } else {
1890                                 rt6_clean_expires(rt);
1891                         }
1892                 } else if (valid_lft) {
1893                         clock_t expires = 0;
1894                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1895                         if (addrconf_finite_timeout(rt_expires)) {
1896                                 /* not infinity */
1897                                 flags |= RTF_EXPIRES;
1898                                 expires = jiffies_to_clock_t(rt_expires);
1899                         }
1900                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1901                                               dev, expires, flags);
1902                 }
1903                 if (rt)
1904                         dst_release(&rt->dst);
1905         }
1906
1907         /* Try to figure out our local address for this prefix */
1908
1909         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1910                 struct inet6_ifaddr *ifp;
1911                 struct in6_addr addr;
1912                 int create = 0, update_lft = 0;
1913
1914                 if (pinfo->prefix_len == 64) {
1915                         memcpy(&addr, &pinfo->prefix, 8);
1916                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1917                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1918                                 in6_dev_put(in6_dev);
1919                                 return;
1920                         }
1921                         goto ok;
1922                 }
1923                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
1924                                     pinfo->prefix_len);
1925                 in6_dev_put(in6_dev);
1926                 return;
1927
1928 ok:
1929
1930                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1931
1932                 if (ifp == NULL && valid_lft) {
1933                         int max_addresses = in6_dev->cnf.max_addresses;
1934                         u32 addr_flags = 0;
1935
1936 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1937                         if (in6_dev->cnf.optimistic_dad &&
1938                             !net->ipv6.devconf_all->forwarding && sllao)
1939                                 addr_flags = IFA_F_OPTIMISTIC;
1940 #endif
1941
1942                         /* Do not allow to create too much of autoconfigured
1943                          * addresses; this would be too easy way to crash kernel.
1944                          */
1945                         if (!max_addresses ||
1946                             ipv6_count_addresses(in6_dev) < max_addresses)
1947                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1948                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1949                                                     addr_flags);
1950
1951                         if (!ifp || IS_ERR(ifp)) {
1952                                 in6_dev_put(in6_dev);
1953                                 return;
1954                         }
1955
1956                         update_lft = create = 1;
1957                         ifp->cstamp = jiffies;
1958                         addrconf_dad_start(ifp);
1959                 }
1960
1961                 if (ifp) {
1962                         int flags;
1963                         unsigned long now;
1964 #ifdef CONFIG_IPV6_PRIVACY
1965                         struct inet6_ifaddr *ift;
1966 #endif
1967                         u32 stored_lft;
1968
1969                         /* update lifetime (RFC2462 5.5.3 e) */
1970                         spin_lock(&ifp->lock);
1971                         now = jiffies;
1972                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1973                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1974                         else
1975                                 stored_lft = 0;
1976                         if (!update_lft && stored_lft) {
1977                                 if (valid_lft > MIN_VALID_LIFETIME ||
1978                                     valid_lft > stored_lft)
1979                                         update_lft = 1;
1980                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1981                                         /* valid_lft <= stored_lft is always true */
1982                                         /*
1983                                          * RFC 4862 Section 5.5.3e:
1984                                          * "Note that the preferred lifetime of
1985                                          *  the corresponding address is always
1986                                          *  reset to the Preferred Lifetime in
1987                                          *  the received Prefix Information
1988                                          *  option, regardless of whether the
1989                                          *  valid lifetime is also reset or
1990                                          *  ignored."
1991                                          *
1992                                          *  So if the preferred lifetime in
1993                                          *  this advertisement is different
1994                                          *  than what we have stored, but the
1995                                          *  valid lifetime is invalid, just
1996                                          *  reset prefered_lft.
1997                                          *
1998                                          *  We must set the valid lifetime
1999                                          *  to the stored lifetime since we'll
2000                                          *  be updating the timestamp below,
2001                                          *  else we'll set it back to the
2002                                          *  minimum.
2003                                          */
2004                                         if (prefered_lft != ifp->prefered_lft) {
2005                                                 valid_lft = stored_lft;
2006                                                 update_lft = 1;
2007                                         }
2008                                 } else {
2009                                         valid_lft = MIN_VALID_LIFETIME;
2010                                         if (valid_lft < prefered_lft)
2011                                                 prefered_lft = valid_lft;
2012                                         update_lft = 1;
2013                                 }
2014                         }
2015
2016                         if (update_lft) {
2017                                 ifp->valid_lft = valid_lft;
2018                                 ifp->prefered_lft = prefered_lft;
2019                                 ifp->tstamp = now;
2020                                 flags = ifp->flags;
2021                                 ifp->flags &= ~IFA_F_DEPRECATED;
2022                                 spin_unlock(&ifp->lock);
2023
2024                                 if (!(flags&IFA_F_TENTATIVE))
2025                                         ipv6_ifa_notify(0, ifp);
2026                         } else
2027                                 spin_unlock(&ifp->lock);
2028
2029 #ifdef CONFIG_IPV6_PRIVACY
2030                         read_lock_bh(&in6_dev->lock);
2031                         /* update all temporary addresses in the list */
2032                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2033                                             tmp_list) {
2034                                 int age, max_valid, max_prefered;
2035
2036                                 if (ifp != ift->ifpub)
2037                                         continue;
2038
2039                                 /*
2040                                  * RFC 4941 section 3.3:
2041                                  * If a received option will extend the lifetime
2042                                  * of a public address, the lifetimes of
2043                                  * temporary addresses should be extended,
2044                                  * subject to the overall constraint that no
2045                                  * temporary addresses should ever remain
2046                                  * "valid" or "preferred" for a time longer than
2047                                  * (TEMP_VALID_LIFETIME) or
2048                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2049                                  * respectively.
2050                                  */
2051                                 age = (now - ift->cstamp) / HZ;
2052                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2053                                 if (max_valid < 0)
2054                                         max_valid = 0;
2055
2056                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2057                                                in6_dev->cnf.max_desync_factor -
2058                                                age;
2059                                 if (max_prefered < 0)
2060                                         max_prefered = 0;
2061
2062                                 if (valid_lft > max_valid)
2063                                         valid_lft = max_valid;
2064
2065                                 if (prefered_lft > max_prefered)
2066                                         prefered_lft = max_prefered;
2067
2068                                 spin_lock(&ift->lock);
2069                                 flags = ift->flags;
2070                                 ift->valid_lft = valid_lft;
2071                                 ift->prefered_lft = prefered_lft;
2072                                 ift->tstamp = now;
2073                                 if (prefered_lft > 0)
2074                                         ift->flags &= ~IFA_F_DEPRECATED;
2075
2076                                 spin_unlock(&ift->lock);
2077                                 if (!(flags&IFA_F_TENTATIVE))
2078                                         ipv6_ifa_notify(0, ift);
2079                         }
2080
2081                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2082                                 /*
2083                                  * When a new public address is created as
2084                                  * described in [ADDRCONF], also create a new
2085                                  * temporary address. Also create a temporary
2086                                  * address if it's enabled but no temporary
2087                                  * address currently exists.
2088                                  */
2089                                 read_unlock_bh(&in6_dev->lock);
2090                                 ipv6_create_tempaddr(ifp, NULL);
2091                         } else {
2092                                 read_unlock_bh(&in6_dev->lock);
2093                         }
2094 #endif
2095                         in6_ifa_put(ifp);
2096                         addrconf_verify(0);
2097                 }
2098         }
2099         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2100         in6_dev_put(in6_dev);
2101 }
2102
2103 /*
2104  *      Set destination address.
2105  *      Special case for SIT interfaces where we create a new "virtual"
2106  *      device.
2107  */
2108 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2109 {
2110         struct in6_ifreq ireq;
2111         struct net_device *dev;
2112         int err = -EINVAL;
2113
2114         rtnl_lock();
2115
2116         err = -EFAULT;
2117         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2118                 goto err_exit;
2119
2120         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2121
2122         err = -ENODEV;
2123         if (dev == NULL)
2124                 goto err_exit;
2125
2126 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2127         if (dev->type == ARPHRD_SIT) {
2128                 const struct net_device_ops *ops = dev->netdev_ops;
2129                 struct ifreq ifr;
2130                 struct ip_tunnel_parm p;
2131
2132                 err = -EADDRNOTAVAIL;
2133                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2134                         goto err_exit;
2135
2136                 memset(&p, 0, sizeof(p));
2137                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2138                 p.iph.saddr = 0;
2139                 p.iph.version = 4;
2140                 p.iph.ihl = 5;
2141                 p.iph.protocol = IPPROTO_IPV6;
2142                 p.iph.ttl = 64;
2143                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2144
2145                 if (ops->ndo_do_ioctl) {
2146                         mm_segment_t oldfs = get_fs();
2147
2148                         set_fs(KERNEL_DS);
2149                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2150                         set_fs(oldfs);
2151                 } else
2152                         err = -EOPNOTSUPP;
2153
2154                 if (err == 0) {
2155                         err = -ENOBUFS;
2156                         dev = __dev_get_by_name(net, p.name);
2157                         if (!dev)
2158                                 goto err_exit;
2159                         err = dev_open(dev);
2160                 }
2161         }
2162 #endif
2163
2164 err_exit:
2165         rtnl_unlock();
2166         return err;
2167 }
2168
2169 /*
2170  *      Manual configuration of address on an interface
2171  */
2172 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2173                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2174                           __u32 valid_lft)
2175 {
2176         struct inet6_ifaddr *ifp;
2177         struct inet6_dev *idev;
2178         struct net_device *dev;
2179         int scope;
2180         u32 flags;
2181         clock_t expires;
2182         unsigned long timeout;
2183
2184         ASSERT_RTNL();
2185
2186         if (plen > 128)
2187                 return -EINVAL;
2188
2189         /* check the lifetime */
2190         if (!valid_lft || prefered_lft > valid_lft)
2191                 return -EINVAL;
2192
2193         dev = __dev_get_by_index(net, ifindex);
2194         if (!dev)
2195                 return -ENODEV;
2196
2197         idev = addrconf_add_dev(dev);
2198         if (IS_ERR(idev))
2199                 return PTR_ERR(idev);
2200
2201         scope = ipv6_addr_scope(pfx);
2202
2203         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2204         if (addrconf_finite_timeout(timeout)) {
2205                 expires = jiffies_to_clock_t(timeout * HZ);
2206                 valid_lft = timeout;
2207                 flags = RTF_EXPIRES;
2208         } else {
2209                 expires = 0;
2210                 flags = 0;
2211                 ifa_flags |= IFA_F_PERMANENT;
2212         }
2213
2214         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2215         if (addrconf_finite_timeout(timeout)) {
2216                 if (timeout == 0)
2217                         ifa_flags |= IFA_F_DEPRECATED;
2218                 prefered_lft = timeout;
2219         }
2220
2221         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2222
2223         if (!IS_ERR(ifp)) {
2224                 spin_lock_bh(&ifp->lock);
2225                 ifp->valid_lft = valid_lft;
2226                 ifp->prefered_lft = prefered_lft;
2227                 ifp->tstamp = jiffies;
2228                 spin_unlock_bh(&ifp->lock);
2229
2230                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2231                                       expires, flags);
2232                 /*
2233                  * Note that section 3.1 of RFC 4429 indicates
2234                  * that the Optimistic flag should not be set for
2235                  * manually configured addresses
2236                  */
2237                 addrconf_dad_start(ifp);
2238                 in6_ifa_put(ifp);
2239                 addrconf_verify(0);
2240                 return 0;
2241         }
2242
2243         return PTR_ERR(ifp);
2244 }
2245
2246 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2247                           unsigned int plen)
2248 {
2249         struct inet6_ifaddr *ifp;
2250         struct inet6_dev *idev;
2251         struct net_device *dev;
2252
2253         if (plen > 128)
2254                 return -EINVAL;
2255
2256         dev = __dev_get_by_index(net, ifindex);
2257         if (!dev)
2258                 return -ENODEV;
2259
2260         if ((idev = __in6_dev_get(dev)) == NULL)
2261                 return -ENXIO;
2262
2263         read_lock_bh(&idev->lock);
2264         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2265                 if (ifp->prefix_len == plen &&
2266                     ipv6_addr_equal(pfx, &ifp->addr)) {
2267                         in6_ifa_hold(ifp);
2268                         read_unlock_bh(&idev->lock);
2269
2270                         ipv6_del_addr(ifp);
2271
2272                         /* If the last address is deleted administratively,
2273                            disable IPv6 on this interface.
2274                          */
2275                         if (list_empty(&idev->addr_list))
2276                                 addrconf_ifdown(idev->dev, 1);
2277                         return 0;
2278                 }
2279         }
2280         read_unlock_bh(&idev->lock);
2281         return -EADDRNOTAVAIL;
2282 }
2283
2284
2285 int addrconf_add_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_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2298                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2299                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2300         rtnl_unlock();
2301         return err;
2302 }
2303
2304 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2305 {
2306         struct in6_ifreq ireq;
2307         int err;
2308
2309         if (!capable(CAP_NET_ADMIN))
2310                 return -EPERM;
2311
2312         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2313                 return -EFAULT;
2314
2315         rtnl_lock();
2316         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2317                              ireq.ifr6_prefixlen);
2318         rtnl_unlock();
2319         return err;
2320 }
2321
2322 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2323                      int plen, int scope)
2324 {
2325         struct inet6_ifaddr *ifp;
2326
2327         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2328         if (!IS_ERR(ifp)) {
2329                 spin_lock_bh(&ifp->lock);
2330                 ifp->flags &= ~IFA_F_TENTATIVE;
2331                 spin_unlock_bh(&ifp->lock);
2332                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2333                 in6_ifa_put(ifp);
2334         }
2335 }
2336
2337 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2338 static void sit_add_v4_addrs(struct inet6_dev *idev)
2339 {
2340         struct in6_addr addr;
2341         struct net_device *dev;
2342         struct net *net = dev_net(idev->dev);
2343         int scope;
2344
2345         ASSERT_RTNL();
2346
2347         memset(&addr, 0, sizeof(struct in6_addr));
2348         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2349
2350         if (idev->dev->flags&IFF_POINTOPOINT) {
2351                 addr.s6_addr32[0] = htonl(0xfe800000);
2352                 scope = IFA_LINK;
2353         } else {
2354                 scope = IPV6_ADDR_COMPATv4;
2355         }
2356
2357         if (addr.s6_addr32[3]) {
2358                 add_addr(idev, &addr, 128, scope);
2359                 return;
2360         }
2361
2362         for_each_netdev(net, dev) {
2363                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2364                 if (in_dev && (dev->flags & IFF_UP)) {
2365                         struct in_ifaddr *ifa;
2366
2367                         int flag = scope;
2368
2369                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2370                                 int plen;
2371
2372                                 addr.s6_addr32[3] = ifa->ifa_local;
2373
2374                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2375                                         continue;
2376                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2377                                         if (idev->dev->flags&IFF_POINTOPOINT)
2378                                                 continue;
2379                                         flag |= IFA_HOST;
2380                                 }
2381                                 if (idev->dev->flags&IFF_POINTOPOINT)
2382                                         plen = 64;
2383                                 else
2384                                         plen = 96;
2385
2386                                 add_addr(idev, &addr, plen, flag);
2387                         }
2388                 }
2389         }
2390 }
2391 #endif
2392
2393 static void init_loopback(struct net_device *dev)
2394 {
2395         struct inet6_dev  *idev;
2396
2397         /* ::1 */
2398
2399         ASSERT_RTNL();
2400
2401         if ((idev = ipv6_find_idev(dev)) == NULL) {
2402                 pr_debug("%s: add_dev failed\n", __func__);
2403                 return;
2404         }
2405
2406         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2407 }
2408
2409 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2410 {
2411         struct inet6_ifaddr *ifp;
2412         u32 addr_flags = IFA_F_PERMANENT;
2413
2414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2415         if (idev->cnf.optimistic_dad &&
2416             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2417                 addr_flags |= IFA_F_OPTIMISTIC;
2418 #endif
2419
2420
2421         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2422         if (!IS_ERR(ifp)) {
2423                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2424                 addrconf_dad_start(ifp);
2425                 in6_ifa_put(ifp);
2426         }
2427 }
2428
2429 static void addrconf_dev_config(struct net_device *dev)
2430 {
2431         struct in6_addr addr;
2432         struct inet6_dev *idev;
2433
2434         ASSERT_RTNL();
2435
2436         if ((dev->type != ARPHRD_ETHER) &&
2437             (dev->type != ARPHRD_FDDI) &&
2438             (dev->type != ARPHRD_ARCNET) &&
2439             (dev->type != ARPHRD_INFINIBAND) &&
2440             (dev->type != ARPHRD_IEEE802154)) {
2441                 /* Alas, we support only Ethernet autoconfiguration. */
2442                 return;
2443         }
2444
2445         idev = addrconf_add_dev(dev);
2446         if (IS_ERR(idev))
2447                 return;
2448
2449         memset(&addr, 0, sizeof(struct in6_addr));
2450         addr.s6_addr32[0] = htonl(0xFE800000);
2451
2452         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2453                 addrconf_add_linklocal(idev, &addr);
2454 }
2455
2456 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2457 static void addrconf_sit_config(struct net_device *dev)
2458 {
2459         struct inet6_dev *idev;
2460
2461         ASSERT_RTNL();
2462
2463         /*
2464          * Configure the tunnel with one of our IPv4
2465          * addresses... we should configure all of
2466          * our v4 addrs in the tunnel
2467          */
2468
2469         if ((idev = ipv6_find_idev(dev)) == NULL) {
2470                 pr_debug("%s: add_dev failed\n", __func__);
2471                 return;
2472         }
2473
2474         if (dev->priv_flags & IFF_ISATAP) {
2475                 struct in6_addr addr;
2476
2477                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2478                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2479                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2480                         addrconf_add_linklocal(idev, &addr);
2481                 return;
2482         }
2483
2484         sit_add_v4_addrs(idev);
2485
2486         if (dev->flags&IFF_POINTOPOINT) {
2487                 addrconf_add_mroute(dev);
2488                 addrconf_add_lroute(dev);
2489         } else
2490                 sit_route_add(dev);
2491 }
2492 #endif
2493
2494 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2495 static void addrconf_gre_config(struct net_device *dev)
2496 {
2497         struct inet6_dev *idev;
2498         struct in6_addr addr;
2499
2500         pr_info("%s(%s)\n", __func__, dev->name);
2501
2502         ASSERT_RTNL();
2503
2504         if ((idev = ipv6_find_idev(dev)) == NULL) {
2505                 pr_debug("%s: add_dev failed\n", __func__);
2506                 return;
2507         }
2508
2509         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2510         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2511
2512         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2513                 addrconf_add_linklocal(idev, &addr);
2514 }
2515 #endif
2516
2517 static inline int
2518 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2519 {
2520         struct in6_addr lladdr;
2521
2522         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2523                 addrconf_add_linklocal(idev, &lladdr);
2524                 return 0;
2525         }
2526         return -1;
2527 }
2528
2529 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2530 {
2531         struct net_device *link_dev;
2532         struct net *net = dev_net(idev->dev);
2533
2534         /* first try to inherit the link-local address from the link device */
2535         if (idev->dev->iflink &&
2536             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2537                 if (!ipv6_inherit_linklocal(idev, link_dev))
2538                         return;
2539         }
2540         /* then try to inherit it from any device */
2541         for_each_netdev(net, link_dev) {
2542                 if (!ipv6_inherit_linklocal(idev, link_dev))
2543                         return;
2544         }
2545         pr_debug("init ip6-ip6: add_linklocal failed\n");
2546 }
2547
2548 /*
2549  * Autoconfigure tunnel with a link-local address so routing protocols,
2550  * DHCPv6, MLD etc. can be run over the virtual link
2551  */
2552
2553 static void addrconf_ip6_tnl_config(struct net_device *dev)
2554 {
2555         struct inet6_dev *idev;
2556
2557         ASSERT_RTNL();
2558
2559         idev = addrconf_add_dev(dev);
2560         if (IS_ERR(idev)) {
2561                 pr_debug("init ip6-ip6: add_dev failed\n");
2562                 return;
2563         }
2564         ip6_tnl_add_linklocal(idev);
2565 }
2566
2567 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2568                            void *data)
2569 {
2570         struct net_device *dev = (struct net_device *) data;
2571         struct inet6_dev *idev = __in6_dev_get(dev);
2572         int run_pending = 0;
2573         int err;
2574
2575         switch (event) {
2576         case NETDEV_REGISTER:
2577                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2578                         idev = ipv6_add_dev(dev);
2579                         if (!idev)
2580                                 return notifier_from_errno(-ENOMEM);
2581                 }
2582                 break;
2583
2584         case NETDEV_UP:
2585         case NETDEV_CHANGE:
2586                 if (dev->flags & IFF_SLAVE)
2587                         break;
2588
2589                 if (event == NETDEV_UP) {
2590                         if (!addrconf_qdisc_ok(dev)) {
2591                                 /* device is not ready yet. */
2592                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2593                                         dev->name);
2594                                 break;
2595                         }
2596
2597                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2598                                 idev = ipv6_add_dev(dev);
2599
2600                         if (idev) {
2601                                 idev->if_flags |= IF_READY;
2602                                 run_pending = 1;
2603                         }
2604                 } else {
2605                         if (!addrconf_qdisc_ok(dev)) {
2606                                 /* device is still not ready. */
2607                                 break;
2608                         }
2609
2610                         if (idev) {
2611                                 if (idev->if_flags & IF_READY)
2612                                         /* device is already configured. */
2613                                         break;
2614                                 idev->if_flags |= IF_READY;
2615                         }
2616
2617                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2618                                 dev->name);
2619
2620                         run_pending = 1;
2621                 }
2622
2623                 switch (dev->type) {
2624 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2625                 case ARPHRD_SIT:
2626                         addrconf_sit_config(dev);
2627                         break;
2628 #endif
2629 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2630                 case ARPHRD_IPGRE:
2631                         addrconf_gre_config(dev);
2632                         break;
2633 #endif
2634                 case ARPHRD_TUNNEL6:
2635                         addrconf_ip6_tnl_config(dev);
2636                         break;
2637                 case ARPHRD_LOOPBACK:
2638                         init_loopback(dev);
2639                         break;
2640
2641                 default:
2642                         addrconf_dev_config(dev);
2643                         break;
2644                 }
2645
2646                 if (idev) {
2647                         if (run_pending)
2648                                 addrconf_dad_run(idev);
2649
2650                         /*
2651                          * If the MTU changed during the interface down,
2652                          * when the interface up, the changed MTU must be
2653                          * reflected in the idev as well as routers.
2654                          */
2655                         if (idev->cnf.mtu6 != dev->mtu &&
2656                             dev->mtu >= IPV6_MIN_MTU) {
2657                                 rt6_mtu_change(dev, dev->mtu);
2658                                 idev->cnf.mtu6 = dev->mtu;
2659                         }
2660                         idev->tstamp = jiffies;
2661                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2662
2663                         /*
2664                          * If the changed mtu during down is lower than
2665                          * IPV6_MIN_MTU stop IPv6 on this interface.
2666                          */
2667                         if (dev->mtu < IPV6_MIN_MTU)
2668                                 addrconf_ifdown(dev, 1);
2669                 }
2670                 break;
2671
2672         case NETDEV_CHANGEMTU:
2673                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2674                         rt6_mtu_change(dev, dev->mtu);
2675                         idev->cnf.mtu6 = dev->mtu;
2676                         break;
2677                 }
2678
2679                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2680                         idev = ipv6_add_dev(dev);
2681                         if (idev)
2682                                 break;
2683                 }
2684
2685                 /*
2686                  * MTU falled under IPV6_MIN_MTU.
2687                  * Stop IPv6 on this interface.
2688                  */
2689
2690         case NETDEV_DOWN:
2691         case NETDEV_UNREGISTER:
2692                 /*
2693                  *      Remove all addresses from this interface.
2694                  */
2695                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2696                 break;
2697
2698         case NETDEV_CHANGENAME:
2699                 if (idev) {
2700                         snmp6_unregister_dev(idev);
2701                         addrconf_sysctl_unregister(idev);
2702                         addrconf_sysctl_register(idev);
2703                         err = snmp6_register_dev(idev);
2704                         if (err)
2705                                 return notifier_from_errno(err);
2706                 }
2707                 break;
2708
2709         case NETDEV_PRE_TYPE_CHANGE:
2710         case NETDEV_POST_TYPE_CHANGE:
2711                 addrconf_type_change(dev, event);
2712                 break;
2713         }
2714
2715         return NOTIFY_OK;
2716 }
2717
2718 /*
2719  *      addrconf module should be notified of a device going up
2720  */
2721 static struct notifier_block ipv6_dev_notf = {
2722         .notifier_call = addrconf_notify,
2723 };
2724
2725 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2726 {
2727         struct inet6_dev *idev;
2728         ASSERT_RTNL();
2729
2730         idev = __in6_dev_get(dev);
2731
2732         if (event == NETDEV_POST_TYPE_CHANGE)
2733                 ipv6_mc_remap(idev);
2734         else if (event == NETDEV_PRE_TYPE_CHANGE)
2735                 ipv6_mc_unmap(idev);
2736 }
2737
2738 static int addrconf_ifdown(struct net_device *dev, int how)
2739 {
2740         struct net *net = dev_net(dev);
2741         struct inet6_dev *idev;
2742         struct inet6_ifaddr *ifa;
2743         int state, i;
2744
2745         ASSERT_RTNL();
2746
2747         rt6_ifdown(net, dev);
2748         neigh_ifdown(&nd_tbl, dev);
2749
2750         idev = __in6_dev_get(dev);
2751         if (idev == NULL)
2752                 return -ENODEV;
2753
2754         /*
2755          * Step 1: remove reference to ipv6 device from parent device.
2756          *         Do not dev_put!
2757          */
2758         if (how) {
2759                 idev->dead = 1;
2760
2761                 /* protected by rtnl_lock */
2762                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2763
2764                 /* Step 1.5: remove snmp6 entry */
2765                 snmp6_unregister_dev(idev);
2766
2767         }
2768
2769         /* Step 2: clear hash table */
2770         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2771                 struct hlist_head *h = &inet6_addr_lst[i];
2772                 struct hlist_node *n;
2773
2774                 spin_lock_bh(&addrconf_hash_lock);
2775         restart:
2776                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2777                         if (ifa->idev == idev) {
2778                                 hlist_del_init_rcu(&ifa->addr_lst);
2779                                 addrconf_del_timer(ifa);
2780                                 goto restart;
2781                         }
2782                 }
2783                 spin_unlock_bh(&addrconf_hash_lock);
2784         }
2785
2786         write_lock_bh(&idev->lock);
2787
2788         /* Step 2: clear flags for stateless addrconf */
2789         if (!how)
2790                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2791
2792 #ifdef CONFIG_IPV6_PRIVACY
2793         if (how && del_timer(&idev->regen_timer))
2794                 in6_dev_put(idev);
2795
2796         /* Step 3: clear tempaddr list */
2797         while (!list_empty(&idev->tempaddr_list)) {
2798                 ifa = list_first_entry(&idev->tempaddr_list,
2799                                        struct inet6_ifaddr, tmp_list);
2800                 list_del(&ifa->tmp_list);
2801                 write_unlock_bh(&idev->lock);
2802                 spin_lock_bh(&ifa->lock);
2803
2804                 if (ifa->ifpub) {
2805                         in6_ifa_put(ifa->ifpub);
2806                         ifa->ifpub = NULL;
2807                 }
2808                 spin_unlock_bh(&ifa->lock);
2809                 in6_ifa_put(ifa);
2810                 write_lock_bh(&idev->lock);
2811         }
2812 #endif
2813
2814         while (!list_empty(&idev->addr_list)) {
2815                 ifa = list_first_entry(&idev->addr_list,
2816                                        struct inet6_ifaddr, if_list);
2817                 addrconf_del_timer(ifa);
2818
2819                 list_del(&ifa->if_list);
2820
2821                 write_unlock_bh(&idev->lock);
2822
2823                 spin_lock_bh(&ifa->state_lock);
2824                 state = ifa->state;
2825                 ifa->state = INET6_IFADDR_STATE_DEAD;
2826                 spin_unlock_bh(&ifa->state_lock);
2827
2828                 if (state != INET6_IFADDR_STATE_DEAD) {
2829                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2830                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2831                 }
2832                 in6_ifa_put(ifa);
2833
2834                 write_lock_bh(&idev->lock);
2835         }
2836
2837         write_unlock_bh(&idev->lock);
2838
2839         /* Step 5: Discard multicast list */
2840         if (how)
2841                 ipv6_mc_destroy_dev(idev);
2842         else
2843                 ipv6_mc_down(idev);
2844
2845         idev->tstamp = jiffies;
2846
2847         /* Last: Shot the device (if unregistered) */
2848         if (how) {
2849                 addrconf_sysctl_unregister(idev);
2850                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2851                 neigh_ifdown(&nd_tbl, dev);
2852                 in6_dev_put(idev);
2853         }
2854         return 0;
2855 }
2856
2857 static void addrconf_rs_timer(unsigned long data)
2858 {
2859         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2860         struct inet6_dev *idev = ifp->idev;
2861
2862         read_lock(&idev->lock);
2863         if (idev->dead || !(idev->if_flags & IF_READY))
2864                 goto out;
2865
2866         if (idev->cnf.forwarding)
2867                 goto out;
2868
2869         /* Announcement received after solicitation was sent */
2870         if (idev->if_flags & IF_RA_RCVD)
2871                 goto out;
2872
2873         spin_lock(&ifp->lock);
2874         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2875                 /* The wait after the last probe can be shorter */
2876                 addrconf_mod_timer(ifp, AC_RS,
2877                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2878                                    idev->cnf.rtr_solicit_delay :
2879                                    idev->cnf.rtr_solicit_interval);
2880                 spin_unlock(&ifp->lock);
2881
2882                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2883         } else {
2884                 spin_unlock(&ifp->lock);
2885                 /*
2886                  * Note: we do not support deprecated "all on-link"
2887                  * assumption any longer.
2888                  */
2889                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
2890         }
2891
2892 out:
2893         read_unlock(&idev->lock);
2894         in6_ifa_put(ifp);
2895 }
2896
2897 /*
2898  *      Duplicate Address Detection
2899  */
2900 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2901 {
2902         unsigned long rand_num;
2903         struct inet6_dev *idev = ifp->idev;
2904
2905         if (ifp->flags & IFA_F_OPTIMISTIC)
2906                 rand_num = 0;
2907         else
2908                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2909
2910         ifp->probes = idev->cnf.dad_transmits;
2911         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2912 }
2913
2914 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
2915 {
2916         struct inet6_dev *idev = ifp->idev;
2917         struct net_device *dev = idev->dev;
2918
2919         addrconf_join_solict(dev, &ifp->addr);
2920
2921         net_srandom(ifp->addr.s6_addr32[3]);
2922
2923         read_lock_bh(&idev->lock);
2924         spin_lock(&ifp->lock);
2925         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2926                 goto out;
2927
2928         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2929             idev->cnf.accept_dad < 1 ||
2930             !(ifp->flags&IFA_F_TENTATIVE) ||
2931             ifp->flags & IFA_F_NODAD) {
2932                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2933                 spin_unlock(&ifp->lock);
2934                 read_unlock_bh(&idev->lock);
2935
2936                 addrconf_dad_completed(ifp);
2937                 return;
2938         }
2939
2940         if (!(idev->if_flags & IF_READY)) {
2941                 spin_unlock(&ifp->lock);
2942                 read_unlock_bh(&idev->lock);
2943                 /*
2944                  * If the device is not ready:
2945                  * - keep it tentative if it is a permanent address.
2946                  * - otherwise, kill it.
2947                  */
2948                 in6_ifa_hold(ifp);
2949                 addrconf_dad_stop(ifp, 0);
2950                 return;
2951         }
2952
2953         /*
2954          * Optimistic nodes can start receiving
2955          * Frames right away
2956          */
2957         if (ifp->flags & IFA_F_OPTIMISTIC)
2958                 ip6_ins_rt(ifp->rt);
2959
2960         addrconf_dad_kick(ifp);
2961 out:
2962         spin_unlock(&ifp->lock);
2963         read_unlock_bh(&idev->lock);
2964 }
2965
2966 static void addrconf_dad_timer(unsigned long data)
2967 {
2968         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2969         struct inet6_dev *idev = ifp->idev;
2970         struct in6_addr mcaddr;
2971
2972         if (!ifp->probes && addrconf_dad_end(ifp))
2973                 goto out;
2974
2975         read_lock(&idev->lock);
2976         if (idev->dead || !(idev->if_flags & IF_READY)) {
2977                 read_unlock(&idev->lock);
2978                 goto out;
2979         }
2980
2981         spin_lock(&ifp->lock);
2982         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2983                 spin_unlock(&ifp->lock);
2984                 read_unlock(&idev->lock);
2985                 goto out;
2986         }
2987
2988         if (ifp->probes == 0) {
2989                 /*
2990                  * DAD was successful
2991                  */
2992
2993                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2994                 spin_unlock(&ifp->lock);
2995                 read_unlock(&idev->lock);
2996
2997                 addrconf_dad_completed(ifp);
2998
2999                 goto out;
3000         }
3001
3002         ifp->probes--;
3003         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3004         spin_unlock(&ifp->lock);
3005         read_unlock(&idev->lock);
3006
3007         /* send a neighbour solicitation for our addr */
3008         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3009         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3010 out:
3011         in6_ifa_put(ifp);
3012 }
3013
3014 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3015 {
3016         struct net_device *dev = ifp->idev->dev;
3017
3018         /*
3019          *      Configure the address for reception. Now it is valid.
3020          */
3021
3022         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3023
3024         /* If added prefix is link local and we are prepared to process
3025            router advertisements, start sending router solicitations.
3026          */
3027
3028         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3029              ifp->idev->cnf.accept_ra == 2) &&
3030             ifp->idev->cnf.rtr_solicits > 0 &&
3031             (dev->flags&IFF_LOOPBACK) == 0 &&
3032             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3033                 /*
3034                  *      If a host as already performed a random delay
3035                  *      [...] as part of DAD [...] there is no need
3036                  *      to delay again before sending the first RS
3037                  */
3038                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3039
3040                 spin_lock_bh(&ifp->lock);
3041                 ifp->probes = 1;
3042                 ifp->idev->if_flags |= IF_RS_SENT;
3043                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3044                 spin_unlock_bh(&ifp->lock);
3045         }
3046 }
3047
3048 static void addrconf_dad_run(struct inet6_dev *idev)
3049 {
3050         struct inet6_ifaddr *ifp;
3051
3052         read_lock_bh(&idev->lock);
3053         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3054                 spin_lock(&ifp->lock);
3055                 if (ifp->flags & IFA_F_TENTATIVE &&
3056                     ifp->state == INET6_IFADDR_STATE_DAD)
3057                         addrconf_dad_kick(ifp);
3058                 spin_unlock(&ifp->lock);
3059         }
3060         read_unlock_bh(&idev->lock);
3061 }
3062
3063 #ifdef CONFIG_PROC_FS
3064 struct if6_iter_state {
3065         struct seq_net_private p;
3066         int bucket;
3067         int offset;
3068 };
3069
3070 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3071 {
3072         struct inet6_ifaddr *ifa = NULL;
3073         struct if6_iter_state *state = seq->private;
3074         struct net *net = seq_file_net(seq);
3075         int p = 0;
3076
3077         /* initial bucket if pos is 0 */
3078         if (pos == 0) {
3079                 state->bucket = 0;
3080                 state->offset = 0;
3081         }
3082
3083         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3084                 struct hlist_node *n;
3085                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3086                                          addr_lst) {
3087                         /* sync with offset */
3088                         if (p < state->offset) {
3089                                 p++;
3090                                 continue;
3091                         }
3092                         state->offset++;
3093                         if (net_eq(dev_net(ifa->idev->dev), net))
3094                                 return ifa;
3095                 }
3096
3097                 /* prepare for next bucket */
3098                 state->offset = 0;
3099                 p = 0;
3100         }
3101         return NULL;
3102 }
3103
3104 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3105                                          struct inet6_ifaddr *ifa)
3106 {
3107         struct if6_iter_state *state = seq->private;
3108         struct net *net = seq_file_net(seq);
3109         struct hlist_node *n = &ifa->addr_lst;
3110
3111         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3112                 state->offset++;
3113                 if (net_eq(dev_net(ifa->idev->dev), net))
3114                         return ifa;
3115         }
3116
3117         while (++state->bucket < IN6_ADDR_HSIZE) {
3118                 state->offset = 0;
3119                 hlist_for_each_entry_rcu_bh(ifa, n,
3120                                      &inet6_addr_lst[state->bucket], addr_lst) {
3121                         state->offset++;
3122                         if (net_eq(dev_net(ifa->idev->dev), net))
3123                                 return ifa;
3124                 }
3125         }
3126
3127         return NULL;
3128 }
3129
3130 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3131         __acquires(rcu_bh)
3132 {
3133         rcu_read_lock_bh();
3134         return if6_get_first(seq, *pos);
3135 }
3136
3137 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3138 {
3139         struct inet6_ifaddr *ifa;
3140
3141         ifa = if6_get_next(seq, v);
3142         ++*pos;
3143         return ifa;
3144 }
3145
3146 static void if6_seq_stop(struct seq_file *seq, void *v)
3147         __releases(rcu_bh)
3148 {
3149         rcu_read_unlock_bh();
3150 }
3151
3152 static int if6_seq_show(struct seq_file *seq, void *v)
3153 {
3154         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3155         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3156                    &ifp->addr,
3157                    ifp->idev->dev->ifindex,
3158                    ifp->prefix_len,
3159                    ifp->scope,
3160                    ifp->flags,
3161                    ifp->idev->dev->name);
3162         return 0;
3163 }
3164
3165 static const struct seq_operations if6_seq_ops = {
3166         .start  = if6_seq_start,
3167         .next   = if6_seq_next,
3168         .show   = if6_seq_show,
3169         .stop   = if6_seq_stop,
3170 };
3171
3172 static int if6_seq_open(struct inode *inode, struct file *file)
3173 {
3174         return seq_open_net(inode, file, &if6_seq_ops,
3175                             sizeof(struct if6_iter_state));
3176 }
3177
3178 static const struct file_operations if6_fops = {
3179         .owner          = THIS_MODULE,
3180         .open           = if6_seq_open,
3181         .read           = seq_read,
3182         .llseek         = seq_lseek,
3183         .release        = seq_release_net,
3184 };
3185
3186 static int __net_init if6_proc_net_init(struct net *net)
3187 {
3188         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3189                 return -ENOMEM;
3190         return 0;
3191 }
3192
3193 static void __net_exit if6_proc_net_exit(struct net *net)
3194 {
3195        proc_net_remove(net, "if_inet6");
3196 }
3197
3198 static struct pernet_operations if6_proc_net_ops = {
3199        .init = if6_proc_net_init,
3200        .exit = if6_proc_net_exit,
3201 };
3202
3203 int __init if6_proc_init(void)
3204 {
3205         return register_pernet_subsys(&if6_proc_net_ops);
3206 }
3207
3208 void if6_proc_exit(void)
3209 {
3210         unregister_pernet_subsys(&if6_proc_net_ops);
3211 }
3212 #endif  /* CONFIG_PROC_FS */
3213
3214 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3215 /* Check if address is a home address configured on any interface. */
3216 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3217 {
3218         int ret = 0;
3219         struct inet6_ifaddr *ifp = NULL;
3220         struct hlist_node *n;
3221         unsigned int hash = inet6_addr_hash(addr);
3222
3223         rcu_read_lock_bh();
3224         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3225                 if (!net_eq(dev_net(ifp->idev->dev), net))
3226                         continue;
3227                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3228                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3229                         ret = 1;
3230                         break;
3231                 }
3232         }
3233         rcu_read_unlock_bh();
3234         return ret;
3235 }
3236 #endif
3237
3238 /*
3239  *      Periodic address status verification
3240  */
3241
3242 static void addrconf_verify(unsigned long foo)
3243 {
3244         unsigned long now, next, next_sec, next_sched;
3245         struct inet6_ifaddr *ifp;
3246         struct hlist_node *node;
3247         int i;
3248
3249         rcu_read_lock_bh();
3250         spin_lock(&addrconf_verify_lock);
3251         now = jiffies;
3252         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3253
3254         del_timer(&addr_chk_timer);
3255
3256         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3257 restart:
3258                 hlist_for_each_entry_rcu_bh(ifp, node,
3259                                          &inet6_addr_lst[i], addr_lst) {
3260                         unsigned long age;
3261
3262                         if (ifp->flags & IFA_F_PERMANENT)
3263                                 continue;
3264
3265                         spin_lock(&ifp->lock);
3266                         /* We try to batch several events at once. */
3267                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3268
3269                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3270                             age >= ifp->valid_lft) {
3271                                 spin_unlock(&ifp->lock);
3272                                 in6_ifa_hold(ifp);
3273                                 ipv6_del_addr(ifp);
3274                                 goto restart;
3275                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3276                                 spin_unlock(&ifp->lock);
3277                                 continue;
3278                         } else if (age >= ifp->prefered_lft) {
3279                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3280                                 int deprecate = 0;
3281
3282                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3283                                         deprecate = 1;
3284                                         ifp->flags |= IFA_F_DEPRECATED;
3285                                 }
3286
3287                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3288                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3289
3290                                 spin_unlock(&ifp->lock);
3291
3292                                 if (deprecate) {
3293                                         in6_ifa_hold(ifp);
3294
3295                                         ipv6_ifa_notify(0, ifp);
3296                                         in6_ifa_put(ifp);
3297                                         goto restart;
3298                                 }
3299 #ifdef CONFIG_IPV6_PRIVACY
3300                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3301                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3302                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3303                                         ifp->idev->cnf.dad_transmits *
3304                                         ifp->idev->nd_parms->retrans_time / HZ;
3305
3306                                 if (age >= ifp->prefered_lft - regen_advance) {
3307                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3308                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3309                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3310                                         if (!ifp->regen_count && ifpub) {
3311                                                 ifp->regen_count++;
3312                                                 in6_ifa_hold(ifp);
3313                                                 in6_ifa_hold(ifpub);
3314                                                 spin_unlock(&ifp->lock);
3315
3316                                                 spin_lock(&ifpub->lock);
3317                                                 ifpub->regen_count = 0;
3318                                                 spin_unlock(&ifpub->lock);
3319                                                 ipv6_create_tempaddr(ifpub, ifp);
3320                                                 in6_ifa_put(ifpub);
3321                                                 in6_ifa_put(ifp);
3322                                                 goto restart;
3323                                         }
3324                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3325                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3326                                 spin_unlock(&ifp->lock);
3327 #endif
3328                         } else {
3329                                 /* ifp->prefered_lft <= ifp->valid_lft */
3330                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3331                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3332                                 spin_unlock(&ifp->lock);
3333                         }
3334                 }
3335         }
3336
3337         next_sec = round_jiffies_up(next);
3338         next_sched = next;
3339
3340         /* If rounded timeout is accurate enough, accept it. */
3341         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3342                 next_sched = next_sec;
3343
3344         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3345         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3346                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3347
3348         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3349               now, next, next_sec, next_sched));
3350
3351         addr_chk_timer.expires = next_sched;
3352         add_timer(&addr_chk_timer);
3353         spin_unlock(&addrconf_verify_lock);
3354         rcu_read_unlock_bh();
3355 }
3356
3357 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3358 {
3359         struct in6_addr *pfx = NULL;
3360
3361         if (addr)
3362                 pfx = nla_data(addr);
3363
3364         if (local) {
3365                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3366                         pfx = NULL;
3367                 else
3368                         pfx = nla_data(local);
3369         }
3370
3371         return pfx;
3372 }
3373
3374 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3375         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3376         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3377         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3378 };
3379
3380 static int
3381 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3382 {
3383         struct net *net = sock_net(skb->sk);
3384         struct ifaddrmsg *ifm;
3385         struct nlattr *tb[IFA_MAX+1];
3386         struct in6_addr *pfx;
3387         int err;
3388
3389         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3390         if (err < 0)
3391                 return err;
3392
3393         ifm = nlmsg_data(nlh);
3394         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3395         if (pfx == NULL)
3396                 return -EINVAL;
3397
3398         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3399 }
3400
3401 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3402                              u32 prefered_lft, u32 valid_lft)
3403 {
3404         u32 flags;
3405         clock_t expires;
3406         unsigned long timeout;
3407
3408         if (!valid_lft || (prefered_lft > valid_lft))
3409                 return -EINVAL;
3410
3411         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3412         if (addrconf_finite_timeout(timeout)) {
3413                 expires = jiffies_to_clock_t(timeout * HZ);
3414                 valid_lft = timeout;
3415                 flags = RTF_EXPIRES;
3416         } else {
3417                 expires = 0;
3418                 flags = 0;
3419                 ifa_flags |= IFA_F_PERMANENT;
3420         }
3421
3422         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3423         if (addrconf_finite_timeout(timeout)) {
3424                 if (timeout == 0)
3425                         ifa_flags |= IFA_F_DEPRECATED;
3426                 prefered_lft = timeout;
3427         }
3428
3429         spin_lock_bh(&ifp->lock);
3430         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3431         ifp->tstamp = jiffies;
3432         ifp->valid_lft = valid_lft;
3433         ifp->prefered_lft = prefered_lft;
3434
3435         spin_unlock_bh(&ifp->lock);
3436         if (!(ifp->flags&IFA_F_TENTATIVE))
3437                 ipv6_ifa_notify(0, ifp);
3438
3439         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3440                               expires, flags);
3441         addrconf_verify(0);
3442
3443         return 0;
3444 }
3445
3446 static int
3447 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3448 {
3449         struct net *net = sock_net(skb->sk);
3450         struct ifaddrmsg *ifm;
3451         struct nlattr *tb[IFA_MAX+1];
3452         struct in6_addr *pfx;
3453         struct inet6_ifaddr *ifa;
3454         struct net_device *dev;
3455         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3456         u8 ifa_flags;
3457         int err;
3458
3459         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3460         if (err < 0)
3461                 return err;
3462
3463         ifm = nlmsg_data(nlh);
3464         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3465         if (pfx == NULL)
3466                 return -EINVAL;
3467
3468         if (tb[IFA_CACHEINFO]) {
3469                 struct ifa_cacheinfo *ci;
3470
3471                 ci = nla_data(tb[IFA_CACHEINFO]);
3472                 valid_lft = ci->ifa_valid;
3473                 preferred_lft = ci->ifa_prefered;
3474         } else {
3475                 preferred_lft = INFINITY_LIFE_TIME;
3476                 valid_lft = INFINITY_LIFE_TIME;
3477         }
3478
3479         dev =  __dev_get_by_index(net, ifm->ifa_index);
3480         if (dev == NULL)
3481                 return -ENODEV;
3482
3483         /* We ignore other flags so far. */
3484         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3485
3486         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3487         if (ifa == NULL) {
3488                 /*
3489                  * It would be best to check for !NLM_F_CREATE here but
3490                  * userspace alreay relies on not having to provide this.
3491                  */
3492                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3493                                       ifm->ifa_prefixlen, ifa_flags,
3494                                       preferred_lft, valid_lft);
3495         }
3496
3497         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3498             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3499                 err = -EEXIST;
3500         else
3501                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3502
3503         in6_ifa_put(ifa);
3504
3505         return err;
3506 }
3507
3508 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3509                           u8 scope, int ifindex)
3510 {
3511         struct ifaddrmsg *ifm;
3512
3513         ifm = nlmsg_data(nlh);
3514         ifm->ifa_family = AF_INET6;
3515         ifm->ifa_prefixlen = prefixlen;
3516         ifm->ifa_flags = flags;
3517         ifm->ifa_scope = scope;
3518         ifm->ifa_index = ifindex;
3519 }
3520
3521 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3522                          unsigned long tstamp, u32 preferred, u32 valid)
3523 {
3524         struct ifa_cacheinfo ci;
3525
3526         ci.cstamp = cstamp_delta(cstamp);
3527         ci.tstamp = cstamp_delta(tstamp);
3528         ci.ifa_prefered = preferred;
3529         ci.ifa_valid = valid;
3530
3531         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3532 }
3533
3534 static inline int rt_scope(int ifa_scope)
3535 {
3536         if (ifa_scope & IFA_HOST)
3537                 return RT_SCOPE_HOST;
3538         else if (ifa_scope & IFA_LINK)
3539                 return RT_SCOPE_LINK;
3540         else if (ifa_scope & IFA_SITE)
3541                 return RT_SCOPE_SITE;
3542         else
3543                 return RT_SCOPE_UNIVERSE;
3544 }
3545
3546 static inline int inet6_ifaddr_msgsize(void)
3547 {
3548         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3549                + nla_total_size(16) /* IFA_ADDRESS */
3550                + nla_total_size(sizeof(struct ifa_cacheinfo));
3551 }
3552
3553 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3554                              u32 pid, u32 seq, int event, unsigned int flags)
3555 {
3556         struct nlmsghdr  *nlh;
3557         u32 preferred, valid;
3558
3559         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3560         if (nlh == NULL)
3561                 return -EMSGSIZE;
3562
3563         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3564                       ifa->idev->dev->ifindex);
3565
3566         if (!(ifa->flags&IFA_F_PERMANENT)) {
3567                 preferred = ifa->prefered_lft;
3568                 valid = ifa->valid_lft;
3569                 if (preferred != INFINITY_LIFE_TIME) {
3570                         long tval = (jiffies - ifa->tstamp)/HZ;
3571                         if (preferred > tval)
3572                                 preferred -= tval;
3573                         else
3574                                 preferred = 0;
3575                         if (valid != INFINITY_LIFE_TIME) {
3576                                 if (valid > tval)
3577                                         valid -= tval;
3578                                 else
3579                                         valid = 0;
3580                         }
3581                 }
3582         } else {
3583                 preferred = INFINITY_LIFE_TIME;
3584                 valid = INFINITY_LIFE_TIME;
3585         }
3586
3587         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3588             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3589                 nlmsg_cancel(skb, nlh);
3590                 return -EMSGSIZE;
3591         }
3592
3593         return nlmsg_end(skb, nlh);
3594 }
3595
3596 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3597                                 u32 pid, u32 seq, int event, u16 flags)
3598 {
3599         struct nlmsghdr  *nlh;
3600         u8 scope = RT_SCOPE_UNIVERSE;
3601         int ifindex = ifmca->idev->dev->ifindex;
3602
3603         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3604                 scope = RT_SCOPE_SITE;
3605
3606         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3607         if (nlh == NULL)
3608                 return -EMSGSIZE;
3609
3610         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3611         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3612             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3613                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3614                 nlmsg_cancel(skb, nlh);
3615                 return -EMSGSIZE;
3616         }
3617
3618         return nlmsg_end(skb, nlh);
3619 }
3620
3621 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3622                                 u32 pid, u32 seq, int event, unsigned int flags)
3623 {
3624         struct nlmsghdr  *nlh;
3625         u8 scope = RT_SCOPE_UNIVERSE;
3626         int ifindex = ifaca->aca_idev->dev->ifindex;
3627
3628         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3629                 scope = RT_SCOPE_SITE;
3630
3631         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3632         if (nlh == NULL)
3633                 return -EMSGSIZE;
3634
3635         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3636         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3637             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3638                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3639                 nlmsg_cancel(skb, nlh);
3640                 return -EMSGSIZE;
3641         }
3642
3643         return nlmsg_end(skb, nlh);
3644 }
3645
3646 enum addr_type_t {
3647         UNICAST_ADDR,
3648         MULTICAST_ADDR,
3649         ANYCAST_ADDR,
3650 };
3651
3652 /* called with rcu_read_lock() */
3653 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3654                           struct netlink_callback *cb, enum addr_type_t type,
3655                           int s_ip_idx, int *p_ip_idx)
3656 {
3657         struct ifmcaddr6 *ifmca;
3658         struct ifacaddr6 *ifaca;
3659         int err = 1;
3660         int ip_idx = *p_ip_idx;
3661
3662         read_lock_bh(&idev->lock);
3663         switch (type) {
3664         case UNICAST_ADDR: {
3665                 struct inet6_ifaddr *ifa;
3666
3667                 /* unicast address incl. temp addr */
3668                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3669                         if (++ip_idx < s_ip_idx)
3670                                 continue;
3671                         err = inet6_fill_ifaddr(skb, ifa,
3672                                                 NETLINK_CB(cb->skb).pid,
3673                                                 cb->nlh->nlmsg_seq,
3674                                                 RTM_NEWADDR,
3675                                                 NLM_F_MULTI);
3676                         if (err <= 0)
3677                                 break;
3678                 }
3679                 break;
3680         }
3681         case MULTICAST_ADDR:
3682                 /* multicast address */
3683                 for (ifmca = idev->mc_list; ifmca;
3684                      ifmca = ifmca->next, ip_idx++) {
3685                         if (ip_idx < s_ip_idx)
3686                                 continue;
3687                         err = inet6_fill_ifmcaddr(skb, ifmca,
3688                                                   NETLINK_CB(cb->skb).pid,
3689                                                   cb->nlh->nlmsg_seq,
3690                                                   RTM_GETMULTICAST,
3691                                                   NLM_F_MULTI);
3692                         if (err <= 0)
3693                                 break;
3694                 }
3695                 break;
3696         case ANYCAST_ADDR:
3697                 /* anycast address */
3698                 for (ifaca = idev->ac_list; ifaca;
3699                      ifaca = ifaca->aca_next, ip_idx++) {
3700                         if (ip_idx < s_ip_idx)
3701                                 continue;
3702                         err = inet6_fill_ifacaddr(skb, ifaca,
3703                                                   NETLINK_CB(cb->skb).pid,
3704                                                   cb->nlh->nlmsg_seq,
3705                                                   RTM_GETANYCAST,
3706                                                   NLM_F_MULTI);
3707                         if (err <= 0)
3708                                 break;
3709                 }
3710                 break;
3711         default:
3712                 break;
3713         }
3714         read_unlock_bh(&idev->lock);
3715         *p_ip_idx = ip_idx;
3716         return err;
3717 }
3718
3719 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3720                            enum addr_type_t type)
3721 {
3722         struct net *net = sock_net(skb->sk);
3723         int h, s_h;
3724         int idx, ip_idx;
3725         int s_idx, s_ip_idx;
3726         struct net_device *dev;
3727         struct inet6_dev *idev;
3728         struct hlist_head *head;
3729         struct hlist_node *node;
3730
3731         s_h = cb->args[0];
3732         s_idx = idx = cb->args[1];
3733         s_ip_idx = ip_idx = cb->args[2];
3734
3735         rcu_read_lock();
3736         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3737                 idx = 0;
3738                 head = &net->dev_index_head[h];
3739                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3740                         if (idx < s_idx)
3741                                 goto cont;
3742                         if (h > s_h || idx > s_idx)
3743                                 s_ip_idx = 0;
3744                         ip_idx = 0;
3745                         idev = __in6_dev_get(dev);
3746                         if (!idev)
3747                                 goto cont;
3748
3749                         if (in6_dump_addrs(idev, skb, cb, type,
3750                                            s_ip_idx, &ip_idx) <= 0)
3751                                 goto done;
3752 cont:
3753                         idx++;
3754                 }
3755         }
3756 done:
3757         rcu_read_unlock();
3758         cb->args[0] = h;
3759         cb->args[1] = idx;
3760         cb->args[2] = ip_idx;
3761
3762         return skb->len;
3763 }
3764
3765 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3766 {
3767         enum addr_type_t type = UNICAST_ADDR;
3768
3769         return inet6_dump_addr(skb, cb, type);
3770 }
3771
3772 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3773 {
3774         enum addr_type_t type = MULTICAST_ADDR;
3775
3776         return inet6_dump_addr(skb, cb, type);
3777 }
3778
3779
3780 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3781 {
3782         enum addr_type_t type = ANYCAST_ADDR;
3783
3784         return inet6_dump_addr(skb, cb, type);
3785 }
3786
3787 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3788                              void *arg)
3789 {
3790         struct net *net = sock_net(in_skb->sk);
3791         struct ifaddrmsg *ifm;
3792         struct nlattr *tb[IFA_MAX+1];
3793         struct in6_addr *addr = NULL;
3794         struct net_device *dev = NULL;
3795         struct inet6_ifaddr *ifa;
3796         struct sk_buff *skb;
3797         int err;
3798
3799         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3800         if (err < 0)
3801                 goto errout;
3802
3803         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3804         if (addr == NULL) {
3805                 err = -EINVAL;
3806                 goto errout;
3807         }
3808
3809         ifm = nlmsg_data(nlh);
3810         if (ifm->ifa_index)
3811                 dev = __dev_get_by_index(net, ifm->ifa_index);
3812
3813         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3814         if (!ifa) {
3815                 err = -EADDRNOTAVAIL;
3816                 goto errout;
3817         }
3818
3819         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3820         if (!skb) {
3821                 err = -ENOBUFS;
3822                 goto errout_ifa;
3823         }
3824
3825         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3826                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3827         if (err < 0) {
3828                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3829                 WARN_ON(err == -EMSGSIZE);
3830                 kfree_skb(skb);
3831                 goto errout_ifa;
3832         }
3833         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3834 errout_ifa:
3835         in6_ifa_put(ifa);
3836 errout:
3837         return err;
3838 }
3839
3840 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3841 {
3842         struct sk_buff *skb;
3843         struct net *net = dev_net(ifa->idev->dev);
3844         int err = -ENOBUFS;
3845
3846         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3847         if (skb == NULL)
3848                 goto errout;
3849
3850         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3851         if (err < 0) {
3852                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3853                 WARN_ON(err == -EMSGSIZE);
3854                 kfree_skb(skb);
3855                 goto errout;
3856         }
3857         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3858         return;
3859 errout:
3860         if (err < 0)
3861                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3862 }
3863
3864 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3865                                 __s32 *array, int bytes)
3866 {
3867         BUG_ON(bytes < (DEVCONF_MAX * 4));
3868
3869         memset(array, 0, bytes);
3870         array[DEVCONF_FORWARDING] = cnf->forwarding;
3871         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3872         array[DEVCONF_MTU6] = cnf->mtu6;
3873         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3874         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3875         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3876         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3877         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3878         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3879                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3880         array[DEVCONF_RTR_SOLICIT_DELAY] =
3881                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3882         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3883 #ifdef CONFIG_IPV6_PRIVACY
3884         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3885         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3886         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3887         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3888         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3889 #endif
3890         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3891         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3892         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3893 #ifdef CONFIG_IPV6_ROUTER_PREF
3894         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3895         array[DEVCONF_RTR_PROBE_INTERVAL] =
3896                 jiffies_to_msecs(cnf->rtr_probe_interval);
3897 #ifdef CONFIG_IPV6_ROUTE_INFO
3898         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3899 #endif
3900 #endif
3901         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3902         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3903 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3904         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3905 #endif
3906 #ifdef CONFIG_IPV6_MROUTE
3907         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3908 #endif
3909         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3910         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3911         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3912 }
3913
3914 static inline size_t inet6_ifla6_size(void)
3915 {
3916         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3917              + nla_total_size(sizeof(struct ifla_cacheinfo))
3918              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3919              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3920              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3921 }
3922
3923 static inline size_t inet6_if_nlmsg_size(void)
3924 {
3925         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3926                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3927                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3928                + nla_total_size(4) /* IFLA_MTU */
3929                + nla_total_size(4) /* IFLA_LINK */
3930                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3931 }
3932
3933 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3934                                       int items, int bytes)
3935 {
3936         int i;
3937         int pad = bytes - sizeof(u64) * items;
3938         BUG_ON(pad < 0);
3939
3940         /* Use put_unaligned() because stats may not be aligned for u64. */
3941         put_unaligned(items, &stats[0]);
3942         for (i = 1; i < items; i++)
3943                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3944
3945         memset(&stats[items], 0, pad);
3946 }
3947
3948 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3949                                       int items, int bytes, size_t syncpoff)
3950 {
3951         int i;
3952         int pad = bytes - sizeof(u64) * items;
3953         BUG_ON(pad < 0);
3954
3955         /* Use put_unaligned() because stats may not be aligned for u64. */
3956         put_unaligned(items, &stats[0]);
3957         for (i = 1; i < items; i++)
3958                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3959
3960         memset(&stats[items], 0, pad);
3961 }
3962
3963 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3964                              int bytes)
3965 {
3966         switch (attrtype) {
3967         case IFLA_INET6_STATS:
3968                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3969                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3970                 break;
3971         case IFLA_INET6_ICMP6STATS:
3972                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3973                 break;
3974         }
3975 }
3976
3977 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3978 {
3979         struct nlattr *nla;
3980         struct ifla_cacheinfo ci;
3981
3982         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
3983                 goto nla_put_failure;
3984         ci.max_reasm_len = IPV6_MAXPLEN;
3985         ci.tstamp = cstamp_delta(idev->tstamp);
3986         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3987         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3988         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
3989                 goto nla_put_failure;
3990         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3991         if (nla == NULL)
3992                 goto nla_put_failure;
3993         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3994
3995         /* XXX - MC not implemented */
3996
3997         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3998         if (nla == NULL)
3999                 goto nla_put_failure;
4000         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4001
4002         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4003         if (nla == NULL)
4004                 goto nla_put_failure;
4005         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4006
4007         return 0;
4008
4009 nla_put_failure:
4010         return -EMSGSIZE;
4011 }
4012
4013 static size_t inet6_get_link_af_size(const struct net_device *dev)
4014 {
4015         if (!__in6_dev_get(dev))
4016                 return 0;
4017
4018         return inet6_ifla6_size();
4019 }
4020
4021 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4022 {
4023         struct inet6_dev *idev = __in6_dev_get(dev);
4024
4025         if (!idev)
4026                 return -ENODATA;
4027
4028         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4029                 return -EMSGSIZE;
4030
4031         return 0;
4032 }
4033
4034 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4035                              u32 pid, u32 seq, int event, unsigned int flags)
4036 {
4037         struct net_device *dev = idev->dev;
4038         struct ifinfomsg *hdr;
4039         struct nlmsghdr *nlh;
4040         void *protoinfo;
4041
4042         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4043         if (nlh == NULL)
4044                 return -EMSGSIZE;
4045
4046         hdr = nlmsg_data(nlh);
4047         hdr->ifi_family = AF_INET6;
4048         hdr->__ifi_pad = 0;
4049         hdr->ifi_type = dev->type;
4050         hdr->ifi_index = dev->ifindex;
4051         hdr->ifi_flags = dev_get_flags(dev);
4052         hdr->ifi_change = 0;
4053
4054         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4055             (dev->addr_len &&
4056              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4057             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4058             (dev->ifindex != dev->iflink &&
4059              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4060                 goto nla_put_failure;
4061         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4062         if (protoinfo == NULL)
4063                 goto nla_put_failure;
4064
4065         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4066                 goto nla_put_failure;
4067
4068         nla_nest_end(skb, protoinfo);
4069         return nlmsg_end(skb, nlh);
4070
4071 nla_put_failure:
4072         nlmsg_cancel(skb, nlh);
4073         return -EMSGSIZE;
4074 }
4075
4076 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4077 {
4078         struct net *net = sock_net(skb->sk);
4079         int h, s_h;
4080         int idx = 0, s_idx;
4081         struct net_device *dev;
4082         struct inet6_dev *idev;
4083         struct hlist_head *head;
4084         struct hlist_node *node;
4085
4086         s_h = cb->args[0];
4087         s_idx = cb->args[1];
4088
4089         rcu_read_lock();
4090         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4091                 idx = 0;
4092                 head = &net->dev_index_head[h];
4093                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4094                         if (idx < s_idx)
4095                                 goto cont;
4096                         idev = __in6_dev_get(dev);
4097                         if (!idev)
4098                                 goto cont;
4099                         if (inet6_fill_ifinfo(skb, idev,
4100                                               NETLINK_CB(cb->skb).pid,
4101                                               cb->nlh->nlmsg_seq,
4102                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4103                                 goto out;
4104 cont:
4105                         idx++;
4106                 }
4107         }
4108 out:
4109         rcu_read_unlock();
4110         cb->args[1] = idx;
4111         cb->args[0] = h;
4112
4113         return skb->len;
4114 }
4115
4116 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4117 {
4118         struct sk_buff *skb;
4119         struct net *net = dev_net(idev->dev);
4120         int err = -ENOBUFS;
4121
4122         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4123         if (skb == NULL)
4124                 goto errout;
4125
4126         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4127         if (err < 0) {
4128                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4129                 WARN_ON(err == -EMSGSIZE);
4130                 kfree_skb(skb);
4131                 goto errout;
4132         }
4133         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4134         return;
4135 errout:
4136         if (err < 0)
4137                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4138 }
4139
4140 static inline size_t inet6_prefix_nlmsg_size(void)
4141 {
4142         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4143                + nla_total_size(sizeof(struct in6_addr))
4144                + nla_total_size(sizeof(struct prefix_cacheinfo));
4145 }
4146
4147 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4148                              struct prefix_info *pinfo, u32 pid, u32 seq,
4149                              int event, unsigned int flags)
4150 {
4151         struct prefixmsg *pmsg;
4152         struct nlmsghdr *nlh;
4153         struct prefix_cacheinfo ci;
4154
4155         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4156         if (nlh == NULL)
4157                 return -EMSGSIZE;
4158
4159         pmsg = nlmsg_data(nlh);
4160         pmsg->prefix_family = AF_INET6;
4161         pmsg->prefix_pad1 = 0;
4162         pmsg->prefix_pad2 = 0;
4163         pmsg->prefix_ifindex = idev->dev->ifindex;
4164         pmsg->prefix_len = pinfo->prefix_len;
4165         pmsg->prefix_type = pinfo->type;
4166         pmsg->prefix_pad3 = 0;
4167         pmsg->prefix_flags = 0;
4168         if (pinfo->onlink)
4169                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4170         if (pinfo->autoconf)
4171                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4172
4173         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4174                 goto nla_put_failure;
4175         ci.preferred_time = ntohl(pinfo->prefered);
4176         ci.valid_time = ntohl(pinfo->valid);
4177         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4178                 goto nla_put_failure;
4179         return nlmsg_end(skb, nlh);
4180
4181 nla_put_failure:
4182         nlmsg_cancel(skb, nlh);
4183         return -EMSGSIZE;
4184 }
4185
4186 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4187                          struct prefix_info *pinfo)
4188 {
4189         struct sk_buff *skb;
4190         struct net *net = dev_net(idev->dev);
4191         int err = -ENOBUFS;
4192
4193         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4194         if (skb == NULL)
4195                 goto errout;
4196
4197         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4198         if (err < 0) {
4199                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4200                 WARN_ON(err == -EMSGSIZE);
4201                 kfree_skb(skb);
4202                 goto errout;
4203         }
4204         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4205         return;
4206 errout:
4207         if (err < 0)
4208                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4209 }
4210
4211 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4212 {
4213         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4214
4215         switch (event) {
4216         case RTM_NEWADDR:
4217                 /*
4218                  * If the address was optimistic
4219                  * we inserted the route at the start of
4220                  * our DAD process, so we don't need
4221                  * to do it again
4222                  */
4223                 if (!(ifp->rt->rt6i_node))
4224                         ip6_ins_rt(ifp->rt);
4225                 if (ifp->idev->cnf.forwarding)
4226                         addrconf_join_anycast(ifp);
4227                 break;
4228         case RTM_DELADDR:
4229                 if (ifp->idev->cnf.forwarding)
4230                         addrconf_leave_anycast(ifp);
4231                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4232                 dst_hold(&ifp->rt->dst);
4233
4234                 if (ip6_del_rt(ifp->rt))
4235                         dst_free(&ifp->rt->dst);
4236                 break;
4237         }
4238 }
4239
4240 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4241 {
4242         rcu_read_lock_bh();
4243         if (likely(ifp->idev->dead == 0))
4244                 __ipv6_ifa_notify(event, ifp);
4245         rcu_read_unlock_bh();
4246 }
4247
4248 #ifdef CONFIG_SYSCTL
4249
4250 static
4251 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4252                            void __user *buffer, size_t *lenp, loff_t *ppos)
4253 {
4254         int *valp = ctl->data;
4255         int val = *valp;
4256         loff_t pos = *ppos;
4257         ctl_table lctl;
4258         int ret;
4259
4260         /*
4261          * ctl->data points to idev->cnf.forwarding, we should
4262          * not modify it until we get the rtnl lock.
4263          */
4264         lctl = *ctl;
4265         lctl.data = &val;
4266
4267         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4268
4269         if (write)
4270                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4271         if (ret)
4272                 *ppos = pos;
4273         return ret;
4274 }
4275
4276 static void dev_disable_change(struct inet6_dev *idev)
4277 {
4278         if (!idev || !idev->dev)
4279                 return;
4280
4281         if (idev->cnf.disable_ipv6)
4282                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4283         else
4284                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4285 }
4286
4287 static void addrconf_disable_change(struct net *net, __s32 newf)
4288 {
4289         struct net_device *dev;
4290         struct inet6_dev *idev;
4291
4292         rcu_read_lock();
4293         for_each_netdev_rcu(net, dev) {
4294                 idev = __in6_dev_get(dev);
4295                 if (idev) {
4296                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4297                         idev->cnf.disable_ipv6 = newf;
4298                         if (changed)
4299                                 dev_disable_change(idev);
4300                 }
4301         }
4302         rcu_read_unlock();
4303 }
4304
4305 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4306 {
4307         struct net *net;
4308         int old;
4309
4310         if (!rtnl_trylock())
4311                 return restart_syscall();
4312
4313         net = (struct net *)table->extra2;
4314         old = *p;
4315         *p = newf;
4316
4317         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4318                 rtnl_unlock();
4319                 return 0;
4320         }
4321
4322         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4323                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4324                 addrconf_disable_change(net, newf);
4325         } else if ((!newf) ^ (!old))
4326                 dev_disable_change((struct inet6_dev *)table->extra1);
4327
4328         rtnl_unlock();
4329         return 0;
4330 }
4331
4332 static
4333 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4334                             void __user *buffer, size_t *lenp, loff_t *ppos)
4335 {
4336         int *valp = ctl->data;
4337         int val = *valp;
4338         loff_t pos = *ppos;
4339         ctl_table lctl;
4340         int ret;
4341
4342         /*
4343          * ctl->data points to idev->cnf.disable_ipv6, we should
4344          * not modify it until we get the rtnl lock.
4345          */
4346         lctl = *ctl;
4347         lctl.data = &val;
4348
4349         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4350
4351         if (write)
4352                 ret = addrconf_disable_ipv6(ctl, valp, val);
4353         if (ret)
4354                 *ppos = pos;
4355         return ret;
4356 }
4357
4358 static struct addrconf_sysctl_table
4359 {
4360         struct ctl_table_header *sysctl_header;
4361         ctl_table addrconf_vars[DEVCONF_MAX+1];
4362 } addrconf_sysctl __read_mostly = {
4363         .sysctl_header = NULL,
4364         .addrconf_vars = {
4365                 {
4366                         .procname       = "forwarding",
4367                         .data           = &ipv6_devconf.forwarding,
4368                         .maxlen         = sizeof(int),
4369                         .mode           = 0644,
4370                         .proc_handler   = addrconf_sysctl_forward,
4371                 },
4372                 {
4373                         .procname       = "hop_limit",
4374                         .data           = &ipv6_devconf.hop_limit,
4375                         .maxlen         = sizeof(int),
4376                         .mode           = 0644,
4377                         .proc_handler   = proc_dointvec,
4378                 },
4379                 {
4380                         .procname       = "mtu",
4381                         .data           = &ipv6_devconf.mtu6,
4382                         .maxlen         = sizeof(int),
4383                         .mode           = 0644,
4384                         .proc_handler   = proc_dointvec,
4385                 },
4386                 {
4387                         .procname       = "accept_ra",
4388                         .data           = &ipv6_devconf.accept_ra,
4389                         .maxlen         = sizeof(int),
4390                         .mode           = 0644,
4391                         .proc_handler   = proc_dointvec,
4392                 },
4393                 {
4394                         .procname       = "accept_redirects",
4395                         .data           = &ipv6_devconf.accept_redirects,
4396                         .maxlen         = sizeof(int),
4397                         .mode           = 0644,
4398                         .proc_handler   = proc_dointvec,
4399                 },
4400                 {
4401                         .procname       = "autoconf",
4402                         .data           = &ipv6_devconf.autoconf,
4403                         .maxlen         = sizeof(int),
4404                         .mode           = 0644,
4405                         .proc_handler   = proc_dointvec,
4406                 },
4407                 {
4408                         .procname       = "dad_transmits",
4409                         .data           = &ipv6_devconf.dad_transmits,
4410                         .maxlen         = sizeof(int),
4411                         .mode           = 0644,
4412                         .proc_handler   = proc_dointvec,
4413                 },
4414                 {
4415                         .procname       = "router_solicitations",
4416                         .data           = &ipv6_devconf.rtr_solicits,
4417                         .maxlen         = sizeof(int),
4418                         .mode           = 0644,
4419                         .proc_handler   = proc_dointvec,
4420                 },
4421                 {
4422                         .procname       = "router_solicitation_interval",
4423                         .data           = &ipv6_devconf.rtr_solicit_interval,
4424                         .maxlen         = sizeof(int),
4425                         .mode           = 0644,
4426                         .proc_handler   = proc_dointvec_jiffies,
4427                 },
4428                 {
4429                         .procname       = "router_solicitation_delay",
4430                         .data           = &ipv6_devconf.rtr_solicit_delay,
4431                         .maxlen         = sizeof(int),
4432                         .mode           = 0644,
4433                         .proc_handler   = proc_dointvec_jiffies,
4434                 },
4435                 {
4436                         .procname       = "force_mld_version",
4437                         .data           = &ipv6_devconf.force_mld_version,
4438                         .maxlen         = sizeof(int),
4439                         .mode           = 0644,
4440                         .proc_handler   = proc_dointvec,
4441                 },
4442 #ifdef CONFIG_IPV6_PRIVACY
4443                 {
4444                         .procname       = "use_tempaddr",
4445                         .data           = &ipv6_devconf.use_tempaddr,
4446                         .maxlen         = sizeof(int),
4447                         .mode           = 0644,
4448                         .proc_handler   = proc_dointvec,
4449                 },
4450                 {
4451                         .procname       = "temp_valid_lft",
4452                         .data           = &ipv6_devconf.temp_valid_lft,
4453                         .maxlen         = sizeof(int),
4454                         .mode           = 0644,
4455                         .proc_handler   = proc_dointvec,
4456                 },
4457                 {
4458                         .procname       = "temp_prefered_lft",
4459                         .data           = &ipv6_devconf.temp_prefered_lft,
4460                         .maxlen         = sizeof(int),
4461                         .mode           = 0644,
4462                         .proc_handler   = proc_dointvec,
4463                 },
4464                 {
4465                         .procname       = "regen_max_retry",
4466                         .data           = &ipv6_devconf.regen_max_retry,
4467                         .maxlen         = sizeof(int),
4468                         .mode           = 0644,
4469                         .proc_handler   = proc_dointvec,
4470                 },
4471                 {
4472                         .procname       = "max_desync_factor",
4473                         .data           = &ipv6_devconf.max_desync_factor,
4474                         .maxlen         = sizeof(int),
4475                         .mode           = 0644,
4476                         .proc_handler   = proc_dointvec,
4477                 },
4478 #endif
4479                 {
4480                         .procname       = "max_addresses",
4481                         .data           = &ipv6_devconf.max_addresses,
4482                         .maxlen         = sizeof(int),
4483                         .mode           = 0644,
4484                         .proc_handler   = proc_dointvec,
4485                 },
4486                 {
4487                         .procname       = "accept_ra_defrtr",
4488                         .data           = &ipv6_devconf.accept_ra_defrtr,
4489                         .maxlen         = sizeof(int),
4490                         .mode           = 0644,
4491                         .proc_handler   = proc_dointvec,
4492                 },
4493                 {
4494                         .procname       = "accept_ra_pinfo",
4495                         .data           = &ipv6_devconf.accept_ra_pinfo,
4496                         .maxlen         = sizeof(int),
4497                         .mode           = 0644,
4498                         .proc_handler   = proc_dointvec,
4499                 },
4500 #ifdef CONFIG_IPV6_ROUTER_PREF
4501                 {
4502                         .procname       = "accept_ra_rtr_pref",
4503                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4504                         .maxlen         = sizeof(int),
4505                         .mode           = 0644,
4506                         .proc_handler   = proc_dointvec,
4507                 },
4508                 {
4509                         .procname       = "router_probe_interval",
4510                         .data           = &ipv6_devconf.rtr_probe_interval,
4511                         .maxlen         = sizeof(int),
4512                         .mode           = 0644,
4513                         .proc_handler   = proc_dointvec_jiffies,
4514                 },
4515 #ifdef CONFIG_IPV6_ROUTE_INFO
4516                 {
4517                         .procname       = "accept_ra_rt_info_max_plen",
4518                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4519                         .maxlen         = sizeof(int),
4520                         .mode           = 0644,
4521                         .proc_handler   = proc_dointvec,
4522                 },
4523 #endif
4524 #endif
4525                 {
4526                         .procname       = "proxy_ndp",
4527                         .data           = &ipv6_devconf.proxy_ndp,
4528                         .maxlen         = sizeof(int),
4529                         .mode           = 0644,
4530                         .proc_handler   = proc_dointvec,
4531                 },
4532                 {
4533                         .procname       = "accept_source_route",
4534                         .data           = &ipv6_devconf.accept_source_route,
4535                         .maxlen         = sizeof(int),
4536                         .mode           = 0644,
4537                         .proc_handler   = proc_dointvec,
4538                 },
4539 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4540                 {
4541                         .procname       = "optimistic_dad",
4542                         .data           = &ipv6_devconf.optimistic_dad,
4543                         .maxlen         = sizeof(int),
4544                         .mode           = 0644,
4545                         .proc_handler   = proc_dointvec,
4546
4547                 },
4548 #endif
4549 #ifdef CONFIG_IPV6_MROUTE
4550                 {
4551                         .procname       = "mc_forwarding",
4552                         .data           = &ipv6_devconf.mc_forwarding,
4553                         .maxlen         = sizeof(int),
4554                         .mode           = 0444,
4555                         .proc_handler   = proc_dointvec,
4556                 },
4557 #endif
4558                 {
4559                         .procname       = "disable_ipv6",
4560                         .data           = &ipv6_devconf.disable_ipv6,
4561                         .maxlen         = sizeof(int),
4562                         .mode           = 0644,
4563                         .proc_handler   = addrconf_sysctl_disable,
4564                 },
4565                 {
4566                         .procname       = "accept_dad",
4567                         .data           = &ipv6_devconf.accept_dad,
4568                         .maxlen         = sizeof(int),
4569                         .mode           = 0644,
4570                         .proc_handler   = proc_dointvec,
4571                 },
4572                 {
4573                         .procname       = "force_tllao",
4574                         .data           = &ipv6_devconf.force_tllao,
4575                         .maxlen         = sizeof(int),
4576                         .mode           = 0644,
4577                         .proc_handler   = proc_dointvec
4578                 },
4579                 {
4580                         /* sentinel */
4581                 }
4582         },
4583 };
4584
4585 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4586                 struct inet6_dev *idev, struct ipv6_devconf *p)
4587 {
4588         int i;
4589         struct addrconf_sysctl_table *t;
4590         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4591
4592         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4593         if (t == NULL)
4594                 goto out;
4595
4596         for (i = 0; t->addrconf_vars[i].data; i++) {
4597                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4598                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4599                 t->addrconf_vars[i].extra2 = net;
4600         }
4601
4602         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4603
4604         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4605         if (t->sysctl_header == NULL)
4606                 goto free;
4607
4608         p->sysctl = t;
4609         return 0;
4610
4611 free:
4612         kfree(t);
4613 out:
4614         return -ENOBUFS;
4615 }
4616
4617 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4618 {
4619         struct addrconf_sysctl_table *t;
4620
4621         if (p->sysctl == NULL)
4622                 return;
4623
4624         t = p->sysctl;
4625         p->sysctl = NULL;
4626         unregister_net_sysctl_table(t->sysctl_header);
4627         kfree(t);
4628 }
4629
4630 static void addrconf_sysctl_register(struct inet6_dev *idev)
4631 {
4632         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4633                               &ndisc_ifinfo_sysctl_change);
4634         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4635                                         idev, &idev->cnf);
4636 }
4637
4638 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4639 {
4640         __addrconf_sysctl_unregister(&idev->cnf);
4641         neigh_sysctl_unregister(idev->nd_parms);
4642 }
4643
4644
4645 #endif
4646
4647 static int __net_init addrconf_init_net(struct net *net)
4648 {
4649         int err;
4650         struct ipv6_devconf *all, *dflt;
4651
4652         err = -ENOMEM;
4653         all = &ipv6_devconf;
4654         dflt = &ipv6_devconf_dflt;
4655
4656         if (!net_eq(net, &init_net)) {
4657                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4658                 if (all == NULL)
4659                         goto err_alloc_all;
4660
4661                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4662                 if (dflt == NULL)
4663                         goto err_alloc_dflt;
4664         } else {
4665                 /* these will be inherited by all namespaces */
4666                 dflt->autoconf = ipv6_defaults.autoconf;
4667                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4668         }
4669
4670         net->ipv6.devconf_all = all;
4671         net->ipv6.devconf_dflt = dflt;
4672
4673 #ifdef CONFIG_SYSCTL
4674         err = __addrconf_sysctl_register(net, "all", NULL, all);
4675         if (err < 0)
4676                 goto err_reg_all;
4677
4678         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4679         if (err < 0)
4680                 goto err_reg_dflt;
4681 #endif
4682         return 0;
4683
4684 #ifdef CONFIG_SYSCTL
4685 err_reg_dflt:
4686         __addrconf_sysctl_unregister(all);
4687 err_reg_all:
4688         kfree(dflt);
4689 #endif
4690 err_alloc_dflt:
4691         kfree(all);
4692 err_alloc_all:
4693         return err;
4694 }
4695
4696 static void __net_exit addrconf_exit_net(struct net *net)
4697 {
4698 #ifdef CONFIG_SYSCTL
4699         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4700         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4701 #endif
4702         if (!net_eq(net, &init_net)) {
4703                 kfree(net->ipv6.devconf_dflt);
4704                 kfree(net->ipv6.devconf_all);
4705         }
4706 }
4707
4708 static struct pernet_operations addrconf_ops = {
4709         .init = addrconf_init_net,
4710         .exit = addrconf_exit_net,
4711 };
4712
4713 /*
4714  *      Device notifier
4715  */
4716
4717 int register_inet6addr_notifier(struct notifier_block *nb)
4718 {
4719         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4720 }
4721 EXPORT_SYMBOL(register_inet6addr_notifier);
4722
4723 int unregister_inet6addr_notifier(struct notifier_block *nb)
4724 {
4725         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4726 }
4727 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4728
4729 static struct rtnl_af_ops inet6_ops = {
4730         .family           = AF_INET6,
4731         .fill_link_af     = inet6_fill_link_af,
4732         .get_link_af_size = inet6_get_link_af_size,
4733 };
4734
4735 /*
4736  *      Init / cleanup code
4737  */
4738
4739 int __init addrconf_init(void)
4740 {
4741         int i, err;
4742
4743         err = ipv6_addr_label_init();
4744         if (err < 0) {
4745                 pr_crit("%s: cannot initialize default policy table: %d\n",
4746                         __func__, err);
4747                 goto out;
4748         }
4749
4750         err = register_pernet_subsys(&addrconf_ops);
4751         if (err < 0)
4752                 goto out_addrlabel;
4753
4754         /* The addrconf netdev notifier requires that loopback_dev
4755          * has it's ipv6 private information allocated and setup
4756          * before it can bring up and give link-local addresses
4757          * to other devices which are up.
4758          *
4759          * Unfortunately, loopback_dev is not necessarily the first
4760          * entry in the global dev_base list of net devices.  In fact,
4761          * it is likely to be the very last entry on that list.
4762          * So this causes the notifier registry below to try and
4763          * give link-local addresses to all devices besides loopback_dev
4764          * first, then loopback_dev, which cases all the non-loopback_dev
4765          * devices to fail to get a link-local address.
4766          *
4767          * So, as a temporary fix, allocate the ipv6 structure for
4768          * loopback_dev first by hand.
4769          * Longer term, all of the dependencies ipv6 has upon the loopback
4770          * device and it being up should be removed.
4771          */
4772         rtnl_lock();
4773         if (!ipv6_add_dev(init_net.loopback_dev))
4774                 err = -ENOMEM;
4775         rtnl_unlock();
4776         if (err)
4777                 goto errlo;
4778
4779         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4780                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4781
4782         register_netdevice_notifier(&ipv6_dev_notf);
4783
4784         addrconf_verify(0);
4785
4786         err = rtnl_af_register(&inet6_ops);
4787         if (err < 0)
4788                 goto errout_af;
4789
4790         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4791                               NULL);
4792         if (err < 0)
4793                 goto errout;
4794
4795         /* Only the first call to __rtnl_register can fail */
4796         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4797         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4798         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4799                         inet6_dump_ifaddr, NULL);
4800         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4801                         inet6_dump_ifmcaddr, NULL);
4802         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4803                         inet6_dump_ifacaddr, NULL);
4804
4805         ipv6_addr_label_rtnl_register();
4806
4807         return 0;
4808 errout:
4809         rtnl_af_unregister(&inet6_ops);
4810 errout_af:
4811         unregister_netdevice_notifier(&ipv6_dev_notf);
4812 errlo:
4813         unregister_pernet_subsys(&addrconf_ops);
4814 out_addrlabel:
4815         ipv6_addr_label_cleanup();
4816 out:
4817         return err;
4818 }
4819
4820 void addrconf_cleanup(void)
4821 {
4822         struct net_device *dev;
4823         int i;
4824
4825         unregister_netdevice_notifier(&ipv6_dev_notf);
4826         unregister_pernet_subsys(&addrconf_ops);
4827         ipv6_addr_label_cleanup();
4828
4829         rtnl_lock();
4830
4831         __rtnl_af_unregister(&inet6_ops);
4832
4833         /* clean dev list */
4834         for_each_netdev(&init_net, dev) {
4835                 if (__in6_dev_get(dev) == NULL)
4836                         continue;
4837                 addrconf_ifdown(dev, 1);
4838         }
4839         addrconf_ifdown(init_net.loopback_dev, 2);
4840
4841         /*
4842          *      Check hash table.
4843          */
4844         spin_lock_bh(&addrconf_hash_lock);
4845         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4846                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4847         spin_unlock_bh(&addrconf_hash_lock);
4848
4849         del_timer(&addr_chk_timer);
4850         rtnl_unlock();
4851 }