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         for_each_netdev(net, dev) {
 498                 idev = __in6_dev_get(dev);
 499                 if (idev) {
 500                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 501                         idev->cnf.forwarding = newf;
 502                         if (changed)
 503                                 dev_forward_change(idev);
 504                 }
 505         }
 506 }
 507
 508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 509 {
 510         struct net *net;
 511         int old;
 512
 513         if (!rtnl_trylock())
 514                 return restart_syscall();
 515
 516         net = (struct net *)table->extra2;
 517         old = *p;
 518         *p = newf;
 519
 520         if (p == &net->ipv6.devconf_dflt->forwarding) {
 521                 rtnl_unlock();
 522                 return 0;
 523         }
 524
 525         if (p == &net->ipv6.devconf_all->forwarding) {
 526                 net->ipv6.devconf_dflt->forwarding = newf;
 527                 addrconf_forward_change(net, newf);
 528         } else if ((!newf) ^ (!old))
 529                 dev_forward_change((struct inet6_dev *)table->extra1);
 530         rtnl_unlock();
 531
 532         if (newf)
 533                 rt6_purge_dflt_routers(net);
 534         return 1;
 535 }
 536 #endif
 537
 538 /* Nobody refers to this ifaddr, destroy it */
 539 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 540 {
 541         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 542
 543 #ifdef NET_REFCNT_DEBUG
 544         pr_debug("%s\n", __func__);
 545 #endif
 546
 547         in6_dev_put(ifp->idev);
 548
 549         if (del_timer(&ifp->timer))
 550                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 551
 552         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 553                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 554                 return;
 555         }
 556         dst_release(&ifp->rt->dst);
 557
 558         kfree_rcu(ifp, rcu);
 559 }
 560
 561 static void
 562 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 563 {
 564         struct list_head *p;
 565         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 566
 567         /*
 568          * Each device address list is sorted in order of scope -
 569          * global before linklocal.
 570          */
 571         list_for_each(p, &idev->addr_list) {
 572                 struct inet6_ifaddr *ifa
 573                         = list_entry(p, struct inet6_ifaddr, if_list);
 574                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 575                         break;
 576         }
 577
 578         list_add_tail(&ifp->if_list, p);
 579 }
 580
 581 static u32 inet6_addr_hash(const struct in6_addr *addr)
 582 {
 583         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, false);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ifa->addr = *addr;
 639
 640         spin_lock_init(&ifa->lock);
 641         spin_lock_init(&ifa->state_lock);
 642         init_timer(&ifa->timer);
 643         INIT_HLIST_NODE(&ifa->addr_lst);
 644         ifa->timer.data = (unsigned long) ifa;
 645         ifa->scope = scope;
 646         ifa->prefix_len = pfxlen;
 647         ifa->flags = flags | IFA_F_TENTATIVE;
 648         ifa->cstamp = ifa->tstamp = jiffies;
 649
 650         ifa->rt = rt;
 651
 652         ifa->idev = idev;
 653         in6_dev_hold(idev);
 654         /* For caller */
 655         in6_ifa_hold(ifa);
 656
 657         /* Add to big hash table */
 658         hash = inet6_addr_hash(addr);
 659
 660         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 661         spin_unlock(&addrconf_hash_lock);
 662
 663         write_lock(&idev->lock);
 664         /* Add to inet6_dev unicast addr list. */
 665         ipv6_link_dev_addr(idev, ifa);
 666
 667 #ifdef CONFIG_IPV6_PRIVACY
 668         if (ifa->flags&IFA_F_TEMPORARY) {
 669                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 670                 in6_ifa_hold(ifa);
 671         }
 672 #endif
 673
 674         in6_ifa_hold(ifa);
 675         write_unlock(&idev->lock);
 676 out2:
 677         rcu_read_unlock_bh();
 678
 679         if (likely(err == 0))
 680                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 681         else {
 682                 kfree(ifa);
 683                 ifa = ERR_PTR(err);
 684         }
 685
 686         return ifa;
 687 out:
 688         spin_unlock(&addrconf_hash_lock);
 689         goto out2;
 690 }
 691
 692 /* This function wants to get referenced ifp and releases it before return */
 693
 694 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 695 {
 696         struct inet6_ifaddr *ifa, *ifn;
 697         struct inet6_dev *idev = ifp->idev;
 698         int state;
 699         int deleted = 0, onlink = 0;
 700         unsigned long expires = jiffies;
 701
 702         spin_lock_bh(&ifp->state_lock);
 703         state = ifp->state;
 704         ifp->state = INET6_IFADDR_STATE_DEAD;
 705         spin_unlock_bh(&ifp->state_lock);
 706
 707         if (state == INET6_IFADDR_STATE_DEAD)
 708                 goto out;
 709
 710         spin_lock_bh(&addrconf_hash_lock);
 711         hlist_del_init_rcu(&ifp->addr_lst);
 712         spin_unlock_bh(&addrconf_hash_lock);
 713
 714         write_lock_bh(&idev->lock);
 715 #ifdef CONFIG_IPV6_PRIVACY
 716         if (ifp->flags&IFA_F_TEMPORARY) {
 717                 list_del(&ifp->tmp_list);
 718                 if (ifp->ifpub) {
 719                         in6_ifa_put(ifp->ifpub);
 720                         ifp->ifpub = NULL;
 721                 }
 722                 __in6_ifa_put(ifp);
 723         }
 724 #endif
 725
 726         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 727                 if (ifa == ifp) {
 728                         list_del_init(&ifp->if_list);
 729                         __in6_ifa_put(ifp);
 730
 731                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 732                                 break;
 733                         deleted = 1;
 734                         continue;
 735                 } else if (ifp->flags & IFA_F_PERMANENT) {
 736                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 737                                               ifp->prefix_len)) {
 738                                 if (ifa->flags & IFA_F_PERMANENT) {
 739                                         onlink = 1;
 740                                         if (deleted)
 741                                                 break;
 742                                 } else {
 743                                         unsigned long lifetime;
 744
 745                                         if (!onlink)
 746                                                 onlink = -1;
 747
 748                                         spin_lock(&ifa->lock);
 749
 750                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 751                                         /*
 752                                          * Note: Because this address is
 753                                          * not permanent, lifetime <
 754                                          * LONG_MAX / HZ here.
 755                                          */
 756                                         if (time_before(expires,
 757                                                         ifa->tstamp + lifetime * HZ))
 758                                                 expires = ifa->tstamp + lifetime * HZ;
 759                                         spin_unlock(&ifa->lock);
 760                                 }
 761                         }
 762                 }
 763         }
 764         write_unlock_bh(&idev->lock);
 765
 766         addrconf_del_timer(ifp);
 767
 768         ipv6_ifa_notify(RTM_DELADDR, ifp);
 769
 770         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 771
 772         /*
 773          * Purge or update corresponding prefix
 774          *
 775          * 1) we don't purge prefix here if address was not permanent.
 776          *    prefix is managed by its own lifetime.
 777          * 2) if there're no addresses, delete prefix.
 778          * 3) if there're still other permanent address(es),
 779          *    corresponding prefix is still permanent.
 780          * 4) otherwise, update prefix lifetime to the
 781          *    longest valid lifetime among the corresponding
 782          *    addresses on the device.
 783          *    Note: subsequent RA will update lifetime.
 784          *
 785          * --yoshfuji
 786          */
 787         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 788                 struct in6_addr prefix;
 789                 struct rt6_info *rt;
 790                 struct net *net = dev_net(ifp->idev->dev);
 791                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 792                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 793
 794                 if (rt && addrconf_is_prefix_route(rt)) {
 795                         if (onlink == 0) {
 796                                 ip6_del_rt(rt);
 797                                 rt = NULL;
 798                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 799                                 rt6_set_expires(rt, expires);
 800                         }
 801                 }
 802                 dst_release(&rt->dst);
 803         }
 804
 805         /* clean up prefsrc entries */
 806         rt6_remove_prefsrc(ifp);
 807 out:
 808         in6_ifa_put(ifp);
 809 }
 810
 811 #ifdef CONFIG_IPV6_PRIVACY
 812 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 813 {
 814         struct inet6_dev *idev = ifp->idev;
 815         struct in6_addr addr, *tmpaddr;
 816         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 817         unsigned long regen_advance;
 818         int tmp_plen;
 819         int ret = 0;
 820         int max_addresses;
 821         u32 addr_flags;
 822         unsigned long now = jiffies;
 823
 824         write_lock(&idev->lock);
 825         if (ift) {
 826                 spin_lock_bh(&ift->lock);
 827                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 828                 spin_unlock_bh(&ift->lock);
 829                 tmpaddr = &addr;
 830         } else {
 831                 tmpaddr = NULL;
 832         }
 833 retry:
 834         in6_dev_hold(idev);
 835         if (idev->cnf.use_tempaddr <= 0) {
 836                 write_unlock(&idev->lock);
 837                 pr_info("%s: use_tempaddr is disabled\n", __func__);
 838                 in6_dev_put(idev);
 839                 ret = -1;
 840                 goto out;
 841         }
 842         spin_lock_bh(&ifp->lock);
 843         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 844                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 845                 spin_unlock_bh(&ifp->lock);
 846                 write_unlock(&idev->lock);
 847                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
 848                         __func__);
 849                 in6_dev_put(idev);
 850                 ret = -1;
 851                 goto out;
 852         }
 853         in6_ifa_hold(ifp);
 854         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 855         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 856                 spin_unlock_bh(&ifp->lock);
 857                 write_unlock(&idev->lock);
 858                 pr_warn("%s: regeneration of randomized interface id failed\n",
 859                         __func__);
 860                 in6_ifa_put(ifp);
 861                 in6_dev_put(idev);
 862                 ret = -1;
 863                 goto out;
 864         }
 865         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 866         age = (now - ifp->tstamp) / HZ;
 867         tmp_valid_lft = min_t(__u32,
 868                               ifp->valid_lft,
 869                               idev->cnf.temp_valid_lft + age);
 870         tmp_prefered_lft = min_t(__u32,
 871                                  ifp->prefered_lft,
 872                                  idev->cnf.temp_prefered_lft + age -
 873                                  idev->cnf.max_desync_factor);
 874         tmp_plen = ifp->prefix_len;
 875         max_addresses = idev->cnf.max_addresses;
 876         tmp_tstamp = ifp->tstamp;
 877         spin_unlock_bh(&ifp->lock);
 878
 879         regen_advance = idev->cnf.regen_max_retry *
 880                         idev->cnf.dad_transmits *
 881                         idev->nd_parms->retrans_time / HZ;
 882         write_unlock(&idev->lock);
 883
 884         /* A temporary address is created only if this calculated Preferred
 885          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 886          * an implementation must not create a temporary address with a zero
 887          * Preferred Lifetime.
 888          */
 889         if (tmp_prefered_lft <= regen_advance) {
 890                 in6_ifa_put(ifp);
 891                 in6_dev_put(idev);
 892                 ret = -1;
 893                 goto out;
 894         }
 895
 896         addr_flags = IFA_F_TEMPORARY;
 897         /* set in addrconf_prefix_rcv() */
 898         if (ifp->flags & IFA_F_OPTIMISTIC)
 899                 addr_flags |= IFA_F_OPTIMISTIC;
 900
 901         ift = !max_addresses ||
 902               ipv6_count_addresses(idev) < max_addresses ?
 903                 ipv6_add_addr(idev, &addr, tmp_plen,
 904                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 905                               addr_flags) : NULL;
 906         if (!ift || IS_ERR(ift)) {
 907                 in6_ifa_put(ifp);
 908                 in6_dev_put(idev);
 909                 pr_info("%s: retry temporary address regeneration\n", __func__);
 910                 tmpaddr = &addr;
 911                 write_lock(&idev->lock);
 912                 goto retry;
 913         }
 914
 915         spin_lock_bh(&ift->lock);
 916         ift->ifpub = ifp;
 917         ift->valid_lft = tmp_valid_lft;
 918         ift->prefered_lft = tmp_prefered_lft;
 919         ift->cstamp = now;
 920         ift->tstamp = tmp_tstamp;
 921         spin_unlock_bh(&ift->lock);
 922
 923         addrconf_dad_start(ift);
 924         in6_ifa_put(ift);
 925         in6_dev_put(idev);
 926 out:
 927         return ret;
 928 }
 929 #endif
 930
 931 /*
 932  *      Choose an appropriate source address (RFC3484)
 933  */
 934 enum {
 935         IPV6_SADDR_RULE_INIT = 0,
 936         IPV6_SADDR_RULE_LOCAL,
 937         IPV6_SADDR_RULE_SCOPE,
 938         IPV6_SADDR_RULE_PREFERRED,
 939 #ifdef CONFIG_IPV6_MIP6
 940         IPV6_SADDR_RULE_HOA,
 941 #endif
 942         IPV6_SADDR_RULE_OIF,
 943         IPV6_SADDR_RULE_LABEL,
 944 #ifdef CONFIG_IPV6_PRIVACY
 945         IPV6_SADDR_RULE_PRIVACY,
 946 #endif
 947         IPV6_SADDR_RULE_ORCHID,
 948         IPV6_SADDR_RULE_PREFIX,
 949         IPV6_SADDR_RULE_MAX
 950 };
 951
 952 struct ipv6_saddr_score {
 953         int                     rule;
 954         int                     addr_type;
 955         struct inet6_ifaddr     *ifa;
 956         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 957         int                     scopedist;
 958         int                     matchlen;
 959 };
 960
 961 struct ipv6_saddr_dst {
 962         const struct in6_addr *addr;
 963         int ifindex;
 964         int scope;
 965         int label;
 966         unsigned int prefs;
 967 };
 968
 969 static inline int ipv6_saddr_preferred(int type)
 970 {
 971         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 972                 return 1;
 973         return 0;
 974 }
 975
 976 static int ipv6_get_saddr_eval(struct net *net,
 977                                struct ipv6_saddr_score *score,
 978                                struct ipv6_saddr_dst *dst,
 979                                int i)
 980 {
 981         int ret;
 982
 983         if (i <= score->rule) {
 984                 switch (i) {
 985                 case IPV6_SADDR_RULE_SCOPE:
 986                         ret = score->scopedist;
 987                         break;
 988                 case IPV6_SADDR_RULE_PREFIX:
 989                         ret = score->matchlen;
 990                         break;
 991                 default:
 992                         ret = !!test_bit(i, score->scorebits);
 993                 }
 994                 goto out;
 995         }
 996
 997         switch (i) {
 998         case IPV6_SADDR_RULE_INIT:
 999                 /* Rule 0: remember if hiscore is not ready yet */
1000                 ret = !!score->ifa;
1001                 break;
1002         case IPV6_SADDR_RULE_LOCAL:
1003                 /* Rule 1: Prefer same address */
1004                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1005                 break;
1006         case IPV6_SADDR_RULE_SCOPE:
1007                 /* Rule 2: Prefer appropriate scope
1008                  *
1009                  *      ret
1010                  *       ^
1011                  *    -1 |  d 15
1012                  *    ---+--+-+---> scope
1013                  *       |
1014                  *       |             d is scope of the destination.
1015                  *  B-d  |  \
1016                  *       |   \      <- smaller scope is better if
1017                  *  B-15 |    \        if scope is enough for destinaion.
1018                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1019                  * d-C-1 | /
1020                  *       |/         <- greater is better
1021                  *   -C  /             if scope is not enough for destination.
1022                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1023                  *
1024                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1025                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1026                  * Assume B = 0 and we get C > 29.
1027                  */
1028                 ret = __ipv6_addr_src_scope(score->addr_type);
1029                 if (ret >= dst->scope)
1030                         ret = -ret;
1031                 else
1032                         ret -= 128;     /* 30 is enough */
1033                 score->scopedist = ret;
1034                 break;
1035         case IPV6_SADDR_RULE_PREFERRED:
1036                 /* Rule 3: Avoid deprecated and optimistic addresses */
1037                 ret = ipv6_saddr_preferred(score->addr_type) ||
1038                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1039                 break;
1040 #ifdef CONFIG_IPV6_MIP6
1041         case IPV6_SADDR_RULE_HOA:
1042             {
1043                 /* Rule 4: Prefer home address */
1044                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1045                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1046                 break;
1047             }
1048 #endif
1049         case IPV6_SADDR_RULE_OIF:
1050                 /* Rule 5: Prefer outgoing interface */
1051                 ret = (!dst->ifindex ||
1052                        dst->ifindex == score->ifa->idev->dev->ifindex);
1053                 break;
1054         case IPV6_SADDR_RULE_LABEL:
1055                 /* Rule 6: Prefer matching label */
1056                 ret = ipv6_addr_label(net,
1057                                       &score->ifa->addr, score->addr_type,
1058                                       score->ifa->idev->dev->ifindex) == dst->label;
1059                 break;
1060 #ifdef CONFIG_IPV6_PRIVACY
1061         case IPV6_SADDR_RULE_PRIVACY:
1062             {
1063                 /* Rule 7: Prefer public address
1064                  * Note: prefer temporary address if use_tempaddr >= 2
1065                  */
1066                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1067                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1068                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1069                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1070                 break;
1071             }
1072 #endif
1073         case IPV6_SADDR_RULE_ORCHID:
1074                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1075                  *          non-ORCHID vs non-ORCHID
1076                  */
1077                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1078                         ipv6_addr_orchid(dst->addr));
1079                 break;
1080         case IPV6_SADDR_RULE_PREFIX:
1081                 /* Rule 8: Use longest matching prefix */
1082                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1083                                                        dst->addr);
1084                 break;
1085         default:
1086                 ret = 0;
1087         }
1088
1089         if (ret)
1090                 __set_bit(i, score->scorebits);
1091         score->rule = i;
1092 out:
1093         return ret;
1094 }
1095
1096 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1097                        const struct in6_addr *daddr, unsigned int prefs,
1098                        struct in6_addr *saddr)
1099 {
1100         struct ipv6_saddr_score scores[2],
1101                                 *score = &scores[0], *hiscore = &scores[1];
1102         struct ipv6_saddr_dst dst;
1103         struct net_device *dev;
1104         int dst_type;
1105
1106         dst_type = __ipv6_addr_type(daddr);
1107         dst.addr = daddr;
1108         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1109         dst.scope = __ipv6_addr_src_scope(dst_type);
1110         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1111         dst.prefs = prefs;
1112
1113         hiscore->rule = -1;
1114         hiscore->ifa = NULL;
1115
1116         rcu_read_lock();
1117
1118         for_each_netdev_rcu(net, dev) {
1119                 struct inet6_dev *idev;
1120
1121                 /* Candidate Source Address (section 4)
1122                  *  - multicast and link-local destination address,
1123                  *    the set of candidate source address MUST only
1124                  *    include addresses assigned to interfaces
1125                  *    belonging to the same link as the outgoing
1126                  *    interface.
1127                  * (- For site-local destination addresses, the
1128                  *    set of candidate source addresses MUST only
1129                  *    include addresses assigned to interfaces
1130                  *    belonging to the same site as the outgoing
1131                  *    interface.)
1132                  */
1133                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1134                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1135                     dst.ifindex && dev->ifindex != dst.ifindex)
1136                         continue;
1137
1138                 idev = __in6_dev_get(dev);
1139                 if (!idev)
1140                         continue;
1141
1142                 read_lock_bh(&idev->lock);
1143                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1144                         int i;
1145
1146                         /*
1147                          * - Tentative Address (RFC2462 section 5.4)
1148                          *  - A tentative address is not considered
1149                          *    "assigned to an interface" in the traditional
1150                          *    sense, unless it is also flagged as optimistic.
1151                          * - Candidate Source Address (section 4)
1152                          *  - In any case, anycast addresses, multicast
1153                          *    addresses, and the unspecified address MUST
1154                          *    NOT be included in a candidate set.
1155                          */
1156                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1157                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1158                                 continue;
1159
1160                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1161
1162                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1163                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1164                                 LIMIT_NETDEBUG(KERN_DEBUG
1165                                                "ADDRCONF: unspecified / multicast address "
1166                                                "assigned as unicast address on %s",
1167                                                dev->name);
1168                                 continue;
1169                         }
1170
1171                         score->rule = -1;
1172                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1173
1174                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1175                                 int minihiscore, miniscore;
1176
1177                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1178                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1179
1180                                 if (minihiscore > miniscore) {
1181                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1182                                             score->scopedist > 0) {
1183                                                 /*
1184                                                  * special case:
1185                                                  * each remaining entry
1186                                                  * has too small (not enough)
1187                                                  * scope, because ifa entries
1188                                                  * are sorted by their scope
1189                                                  * values.
1190                                                  */
1191                                                 goto try_nextdev;
1192                                         }
1193                                         break;
1194                                 } else if (minihiscore < miniscore) {
1195                                         if (hiscore->ifa)
1196                                                 in6_ifa_put(hiscore->ifa);
1197
1198                                         in6_ifa_hold(score->ifa);
1199
1200                                         swap(hiscore, score);
1201
1202                                         /* restore our iterator */
1203                                         score->ifa = hiscore->ifa;
1204
1205                                         break;
1206                                 }
1207                         }
1208                 }
1209 try_nextdev:
1210                 read_unlock_bh(&idev->lock);
1211         }
1212         rcu_read_unlock();
1213
1214         if (!hiscore->ifa)
1215                 return -EADDRNOTAVAIL;
1216
1217         *saddr = hiscore->ifa->addr;
1218         in6_ifa_put(hiscore->ifa);
1219         return 0;
1220 }
1221 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1222
1223 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1224                     unsigned char banned_flags)
1225 {
1226         struct inet6_dev *idev;
1227         int err = -EADDRNOTAVAIL;
1228
1229         rcu_read_lock();
1230         idev = __in6_dev_get(dev);
1231         if (idev) {
1232                 struct inet6_ifaddr *ifp;
1233
1234                 read_lock_bh(&idev->lock);
1235                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1236                         if (ifp->scope == IFA_LINK &&
1237                             !(ifp->flags & banned_flags)) {
1238                                 *addr = ifp->addr;
1239                                 err = 0;
1240                                 break;
1241                         }
1242                 }
1243                 read_unlock_bh(&idev->lock);
1244         }
1245         rcu_read_unlock();
1246         return err;
1247 }
1248
1249 static int ipv6_count_addresses(struct inet6_dev *idev)
1250 {
1251         int cnt = 0;
1252         struct inet6_ifaddr *ifp;
1253
1254         read_lock_bh(&idev->lock);
1255         list_for_each_entry(ifp, &idev->addr_list, if_list)
1256                 cnt++;
1257         read_unlock_bh(&idev->lock);
1258         return cnt;
1259 }
1260
1261 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1262                   struct net_device *dev, int strict)
1263 {
1264         struct inet6_ifaddr *ifp;
1265         struct hlist_node *node;
1266         unsigned int hash = inet6_addr_hash(addr);
1267
1268         rcu_read_lock_bh();
1269         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1270                 if (!net_eq(dev_net(ifp->idev->dev), net))
1271                         continue;
1272                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1273                     !(ifp->flags&IFA_F_TENTATIVE) &&
1274                     (dev == NULL || ifp->idev->dev == dev ||
1275                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1276                         rcu_read_unlock_bh();
1277                         return 1;
1278                 }
1279         }
1280
1281         rcu_read_unlock_bh();
1282         return 0;
1283 }
1284 EXPORT_SYMBOL(ipv6_chk_addr);
1285
1286 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1287                                struct net_device *dev)
1288 {
1289         unsigned int hash = inet6_addr_hash(addr);
1290         struct inet6_ifaddr *ifp;
1291         struct hlist_node *node;
1292
1293         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1294                 if (!net_eq(dev_net(ifp->idev->dev), net))
1295                         continue;
1296                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1297                         if (dev == NULL || ifp->idev->dev == dev)
1298                                 return true;
1299                 }
1300         }
1301         return false;
1302 }
1303
1304 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1305 {
1306         struct inet6_dev *idev;
1307         struct inet6_ifaddr *ifa;
1308         int     onlink;
1309
1310         onlink = 0;
1311         rcu_read_lock();
1312         idev = __in6_dev_get(dev);
1313         if (idev) {
1314                 read_lock_bh(&idev->lock);
1315                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1316                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1317                                                    ifa->prefix_len);
1318                         if (onlink)
1319                                 break;
1320                 }
1321                 read_unlock_bh(&idev->lock);
1322         }
1323         rcu_read_unlock();
1324         return onlink;
1325 }
1326 EXPORT_SYMBOL(ipv6_chk_prefix);
1327
1328 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1329                                      struct net_device *dev, int strict)
1330 {
1331         struct inet6_ifaddr *ifp, *result = NULL;
1332         unsigned int hash = inet6_addr_hash(addr);
1333         struct hlist_node *node;
1334
1335         rcu_read_lock_bh();
1336         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1337                 if (!net_eq(dev_net(ifp->idev->dev), net))
1338                         continue;
1339                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1340                         if (dev == NULL || ifp->idev->dev == dev ||
1341                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1342                                 result = ifp;
1343                                 in6_ifa_hold(ifp);
1344                                 break;
1345                         }
1346                 }
1347         }
1348         rcu_read_unlock_bh();
1349
1350         return result;
1351 }
1352
1353 /* Gets referenced address, destroys ifaddr */
1354
1355 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1356 {
1357         if (ifp->flags&IFA_F_PERMANENT) {
1358                 spin_lock_bh(&ifp->lock);
1359                 addrconf_del_timer(ifp);
1360                 ifp->flags |= IFA_F_TENTATIVE;
1361                 if (dad_failed)
1362                         ifp->flags |= IFA_F_DADFAILED;
1363                 spin_unlock_bh(&ifp->lock);
1364                 if (dad_failed)
1365                         ipv6_ifa_notify(0, ifp);
1366                 in6_ifa_put(ifp);
1367 #ifdef CONFIG_IPV6_PRIVACY
1368         } else if (ifp->flags&IFA_F_TEMPORARY) {
1369                 struct inet6_ifaddr *ifpub;
1370                 spin_lock_bh(&ifp->lock);
1371                 ifpub = ifp->ifpub;
1372                 if (ifpub) {
1373                         in6_ifa_hold(ifpub);
1374                         spin_unlock_bh(&ifp->lock);
1375                         ipv6_create_tempaddr(ifpub, ifp);
1376                         in6_ifa_put(ifpub);
1377                 } else {
1378                         spin_unlock_bh(&ifp->lock);
1379                 }
1380                 ipv6_del_addr(ifp);
1381 #endif
1382         } else
1383                 ipv6_del_addr(ifp);
1384 }
1385
1386 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1387 {
1388         int err = -ENOENT;
1389
1390         spin_lock(&ifp->state_lock);
1391         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1392                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1393                 err = 0;
1394         }
1395         spin_unlock(&ifp->state_lock);
1396
1397         return err;
1398 }
1399
1400 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1401 {
1402         struct inet6_dev *idev = ifp->idev;
1403
1404         if (addrconf_dad_end(ifp)) {
1405                 in6_ifa_put(ifp);
1406                 return;
1407         }
1408
1409         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1410                              ifp->idev->dev->name, &ifp->addr);
1411
1412         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1413                 struct in6_addr addr;
1414
1415                 addr.s6_addr32[0] = htonl(0xfe800000);
1416                 addr.s6_addr32[1] = 0;
1417
1418                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1419                     ipv6_addr_equal(&ifp->addr, &addr)) {
1420                         /* DAD failed for link-local based on MAC address */
1421                         idev->cnf.disable_ipv6 = 1;
1422
1423                         pr_info("%s: IPv6 being disabled!\n",
1424                                 ifp->idev->dev->name);
1425                 }
1426         }
1427
1428         addrconf_dad_stop(ifp, 1);
1429 }
1430
1431 /* Join to solicited addr multicast group. */
1432
1433 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1434 {
1435         struct in6_addr maddr;
1436
1437         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1438                 return;
1439
1440         addrconf_addr_solict_mult(addr, &maddr);
1441         ipv6_dev_mc_inc(dev, &maddr);
1442 }
1443
1444 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1445 {
1446         struct in6_addr maddr;
1447
1448         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1449                 return;
1450
1451         addrconf_addr_solict_mult(addr, &maddr);
1452         __ipv6_dev_mc_dec(idev, &maddr);
1453 }
1454
1455 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1456 {
1457         struct in6_addr addr;
1458         if (ifp->prefix_len == 127) /* RFC 6164 */
1459                 return;
1460         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1461         if (ipv6_addr_any(&addr))
1462                 return;
1463         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1464 }
1465
1466 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1467 {
1468         struct in6_addr addr;
1469         if (ifp->prefix_len == 127) /* RFC 6164 */
1470                 return;
1471         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1472         if (ipv6_addr_any(&addr))
1473                 return;
1474         __ipv6_dev_ac_dec(ifp->idev, &addr);
1475 }
1476
1477 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1478 {
1479         if (dev->addr_len != ETH_ALEN)
1480                 return -1;
1481         memcpy(eui, dev->dev_addr, 3);
1482         memcpy(eui + 5, dev->dev_addr + 3, 3);
1483
1484         /*
1485          * The zSeries OSA network cards can be shared among various
1486          * OS instances, but the OSA cards have only one MAC address.
1487          * This leads to duplicate address conflicts in conjunction
1488          * with IPv6 if more than one instance uses the same card.
1489          *
1490          * The driver for these cards can deliver a unique 16-bit
1491          * identifier for each instance sharing the same card.  It is
1492          * placed instead of 0xFFFE in the interface identifier.  The
1493          * "u" bit of the interface identifier is not inverted in this
1494          * case.  Hence the resulting interface identifier has local
1495          * scope according to RFC2373.
1496          */
1497         if (dev->dev_id) {
1498                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1499                 eui[4] = dev->dev_id & 0xFF;
1500         } else {
1501                 eui[3] = 0xFF;
1502                 eui[4] = 0xFE;
1503                 eui[0] ^= 2;
1504         }
1505         return 0;
1506 }
1507
1508 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1509 {
1510         if (dev->addr_len != IEEE802154_ADDR_LEN)
1511                 return -1;
1512         memcpy(eui, dev->dev_addr, 8);
1513         return 0;
1514 }
1515
1516 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1517 {
1518         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1519         if (dev->addr_len != ARCNET_ALEN)
1520                 return -1;
1521         memset(eui, 0, 7);
1522         eui[7] = *(u8 *)dev->dev_addr;
1523         return 0;
1524 }
1525
1526 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1527 {
1528         if (dev->addr_len != INFINIBAND_ALEN)
1529                 return -1;
1530         memcpy(eui, dev->dev_addr + 12, 8);
1531         eui[0] |= 2;
1532         return 0;
1533 }
1534
1535 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1536 {
1537         if (addr == 0)
1538                 return -1;
1539         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1540                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1541                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1542                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1543                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1544                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1545         eui[1] = 0;
1546         eui[2] = 0x5E;
1547         eui[3] = 0xFE;
1548         memcpy(eui + 4, &addr, 4);
1549         return 0;
1550 }
1551
1552 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1553 {
1554         if (dev->priv_flags & IFF_ISATAP)
1555                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1556         return -1;
1557 }
1558
1559 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1560 {
1561         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1562 }
1563
1564 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1565 {
1566         switch (dev->type) {
1567         case ARPHRD_ETHER:
1568         case ARPHRD_FDDI:
1569                 return addrconf_ifid_eui48(eui, dev);
1570         case ARPHRD_ARCNET:
1571                 return addrconf_ifid_arcnet(eui, dev);
1572         case ARPHRD_INFINIBAND:
1573                 return addrconf_ifid_infiniband(eui, dev);
1574         case ARPHRD_SIT:
1575                 return addrconf_ifid_sit(eui, dev);
1576         case ARPHRD_IPGRE:
1577                 return addrconf_ifid_gre(eui, dev);
1578         case ARPHRD_IEEE802154:
1579                 return addrconf_ifid_eui64(eui, dev);
1580         }
1581         return -1;
1582 }
1583
1584 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1585 {
1586         int err = -1;
1587         struct inet6_ifaddr *ifp;
1588
1589         read_lock_bh(&idev->lock);
1590         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1591                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1592                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1593                         err = 0;
1594                         break;
1595                 }
1596         }
1597         read_unlock_bh(&idev->lock);
1598         return err;
1599 }
1600
1601 #ifdef CONFIG_IPV6_PRIVACY
1602 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1603 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1604 {
1605 regen:
1606         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1607         idev->rndid[0] &= ~0x02;
1608
1609         /*
1610          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1611          * check if generated address is not inappropriate
1612          *
1613          *  - Reserved subnet anycast (RFC 2526)
1614          *      11111101 11....11 1xxxxxxx
1615          *  - ISATAP (RFC4214) 6.1
1616          *      00-00-5E-FE-xx-xx-xx-xx
1617          *  - value 0
1618          *  - XXX: already assigned to an address on the device
1619          */
1620         if (idev->rndid[0] == 0xfd &&
1621             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1622             (idev->rndid[7]&0x80))
1623                 goto regen;
1624         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1625                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1626                         goto regen;
1627                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1628                         goto regen;
1629         }
1630
1631         return 0;
1632 }
1633
1634 static void ipv6_regen_rndid(unsigned long data)
1635 {
1636         struct inet6_dev *idev = (struct inet6_dev *) data;
1637         unsigned long expires;
1638
1639         rcu_read_lock_bh();
1640         write_lock_bh(&idev->lock);
1641
1642         if (idev->dead)
1643                 goto out;
1644
1645         if (__ipv6_regen_rndid(idev) < 0)
1646                 goto out;
1647
1648         expires = jiffies +
1649                 idev->cnf.temp_prefered_lft * HZ -
1650                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1651                 idev->cnf.max_desync_factor * HZ;
1652         if (time_before(expires, jiffies)) {
1653                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1654                         __func__, idev->dev->name);
1655                 goto out;
1656         }
1657
1658         if (!mod_timer(&idev->regen_timer, expires))
1659                 in6_dev_hold(idev);
1660
1661 out:
1662         write_unlock_bh(&idev->lock);
1663         rcu_read_unlock_bh();
1664         in6_dev_put(idev);
1665 }
1666
1667 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1668 {
1669         int ret = 0;
1670
1671         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1672                 ret = __ipv6_regen_rndid(idev);
1673         return ret;
1674 }
1675 #endif
1676
1677 /*
1678  *      Add prefix route.
1679  */
1680
1681 static void
1682 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1683                       unsigned long expires, u32 flags)
1684 {
1685         struct fib6_config cfg = {
1686                 .fc_table = RT6_TABLE_PREFIX,
1687                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1688                 .fc_ifindex = dev->ifindex,
1689                 .fc_expires = expires,
1690                 .fc_dst_len = plen,
1691                 .fc_flags = RTF_UP | flags,
1692                 .fc_nlinfo.nl_net = dev_net(dev),
1693                 .fc_protocol = RTPROT_KERNEL,
1694         };
1695
1696         cfg.fc_dst = *pfx;
1697
1698         /* Prevent useless cloning on PtP SIT.
1699            This thing is done here expecting that the whole
1700            class of non-broadcast devices need not cloning.
1701          */
1702 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1703         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1704                 cfg.fc_flags |= RTF_NONEXTHOP;
1705 #endif
1706
1707         ip6_route_add(&cfg);
1708 }
1709
1710
1711 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1712                                                   int plen,
1713                                                   const struct net_device *dev,
1714                                                   u32 flags, u32 noflags)
1715 {
1716         struct fib6_node *fn;
1717         struct rt6_info *rt = NULL;
1718         struct fib6_table *table;
1719
1720         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1721         if (table == NULL)
1722                 return NULL;
1723
1724         write_lock_bh(&table->tb6_lock);
1725         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1726         if (!fn)
1727                 goto out;
1728         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1729                 if (rt->dst.dev->ifindex != dev->ifindex)
1730                         continue;
1731                 if ((rt->rt6i_flags & flags) != flags)
1732                         continue;
1733                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1734                         continue;
1735                 dst_hold(&rt->dst);
1736                 break;
1737         }
1738 out:
1739         write_unlock_bh(&table->tb6_lock);
1740         return rt;
1741 }
1742
1743
1744 /* Create "default" multicast route to the interface */
1745
1746 static void addrconf_add_mroute(struct net_device *dev)
1747 {
1748         struct fib6_config cfg = {
1749                 .fc_table = RT6_TABLE_LOCAL,
1750                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1751                 .fc_ifindex = dev->ifindex,
1752                 .fc_dst_len = 8,
1753                 .fc_flags = RTF_UP,
1754                 .fc_nlinfo.nl_net = dev_net(dev),
1755         };
1756
1757         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1758
1759         ip6_route_add(&cfg);
1760 }
1761
1762 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1763 static void sit_route_add(struct net_device *dev)
1764 {
1765         struct fib6_config cfg = {
1766                 .fc_table = RT6_TABLE_MAIN,
1767                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1768                 .fc_ifindex = dev->ifindex,
1769                 .fc_dst_len = 96,
1770                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1771                 .fc_nlinfo.nl_net = dev_net(dev),
1772         };
1773
1774         /* prefix length - 96 bits "::d.d.d.d" */
1775         ip6_route_add(&cfg);
1776 }
1777 #endif
1778
1779 static void addrconf_add_lroute(struct net_device *dev)
1780 {
1781         struct in6_addr addr;
1782
1783         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1784         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1785 }
1786
1787 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1788 {
1789         struct inet6_dev *idev;
1790
1791         ASSERT_RTNL();
1792
1793         idev = ipv6_find_idev(dev);
1794         if (!idev)
1795                 return ERR_PTR(-ENOBUFS);
1796
1797         if (idev->cnf.disable_ipv6)
1798                 return ERR_PTR(-EACCES);
1799
1800         /* Add default multicast route */
1801         if (!(dev->flags & IFF_LOOPBACK))
1802                 addrconf_add_mroute(dev);
1803
1804         /* Add link local route */
1805         addrconf_add_lroute(dev);
1806         return idev;
1807 }
1808
1809 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1810 {
1811         struct prefix_info *pinfo;
1812         __u32 valid_lft;
1813         __u32 prefered_lft;
1814         int addr_type;
1815         struct inet6_dev *in6_dev;
1816         struct net *net = dev_net(dev);
1817
1818         pinfo = (struct prefix_info *) opt;
1819
1820         if (len < sizeof(struct prefix_info)) {
1821                 ADBG(("addrconf: prefix option too short\n"));
1822                 return;
1823         }
1824
1825         /*
1826          *      Validation checks ([ADDRCONF], page 19)
1827          */
1828
1829         addr_type = ipv6_addr_type(&pinfo->prefix);
1830
1831         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1832                 return;
1833
1834         valid_lft = ntohl(pinfo->valid);
1835         prefered_lft = ntohl(pinfo->prefered);
1836
1837         if (prefered_lft > valid_lft) {
1838                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1839                 return;
1840         }
1841
1842         in6_dev = in6_dev_get(dev);
1843
1844         if (in6_dev == NULL) {
1845                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1846                                     dev->name);
1847                 return;
1848         }
1849
1850         /*
1851          *      Two things going on here:
1852          *      1) Add routes for on-link prefixes
1853          *      2) Configure prefixes with the auto flag set
1854          */
1855
1856         if (pinfo->onlink) {
1857                 struct rt6_info *rt;
1858                 unsigned long rt_expires;
1859
1860                 /* Avoid arithmetic overflow. Really, we could
1861                  * save rt_expires in seconds, likely valid_lft,
1862                  * but it would require division in fib gc, that it
1863                  * not good.
1864                  */
1865                 if (HZ > USER_HZ)
1866                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1867                 else
1868                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1869
1870                 if (addrconf_finite_timeout(rt_expires))
1871                         rt_expires *= HZ;
1872
1873                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1874                                                pinfo->prefix_len,
1875                                                dev,
1876                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1877                                                RTF_GATEWAY | RTF_DEFAULT);
1878
1879                 if (rt) {
1880                         /* Autoconf prefix route */
1881                         if (valid_lft == 0) {
1882                                 ip6_del_rt(rt);
1883                                 rt = NULL;
1884                         } else if (addrconf_finite_timeout(rt_expires)) {
1885                                 /* not infinity */
1886                                 rt6_set_expires(rt, jiffies + rt_expires);
1887                         } else {
1888                                 rt6_clean_expires(rt);
1889                         }
1890                 } else if (valid_lft) {
1891                         clock_t expires = 0;
1892                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1893                         if (addrconf_finite_timeout(rt_expires)) {
1894                                 /* not infinity */
1895                                 flags |= RTF_EXPIRES;
1896                                 expires = jiffies_to_clock_t(rt_expires);
1897                         }
1898                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1899                                               dev, expires, flags);
1900                 }
1901                 if (rt)
1902                         dst_release(&rt->dst);
1903         }
1904
1905         /* Try to figure out our local address for this prefix */
1906
1907         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1908                 struct inet6_ifaddr *ifp;
1909                 struct in6_addr addr;
1910                 int create = 0, update_lft = 0;
1911
1912                 if (pinfo->prefix_len == 64) {
1913                         memcpy(&addr, &pinfo->prefix, 8);
1914                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1915                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1916                                 in6_dev_put(in6_dev);
1917                                 return;
1918                         }
1919                         goto ok;
1920                 }
1921                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
1922                                     pinfo->prefix_len);
1923                 in6_dev_put(in6_dev);
1924                 return;
1925
1926 ok:
1927
1928                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1929
1930                 if (ifp == NULL && valid_lft) {
1931                         int max_addresses = in6_dev->cnf.max_addresses;
1932                         u32 addr_flags = 0;
1933
1934 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1935                         if (in6_dev->cnf.optimistic_dad &&
1936                             !net->ipv6.devconf_all->forwarding && sllao)
1937                                 addr_flags = IFA_F_OPTIMISTIC;
1938 #endif
1939
1940                         /* Do not allow to create too much of autoconfigured
1941                          * addresses; this would be too easy way to crash kernel.
1942                          */
1943                         if (!max_addresses ||
1944                             ipv6_count_addresses(in6_dev) < max_addresses)
1945                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1946                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1947                                                     addr_flags);
1948
1949                         if (!ifp || IS_ERR(ifp)) {
1950                                 in6_dev_put(in6_dev);
1951                                 return;
1952                         }
1953
1954                         update_lft = create = 1;
1955                         ifp->cstamp = jiffies;
1956                         addrconf_dad_start(ifp);
1957                 }
1958
1959                 if (ifp) {
1960                         int flags;
1961                         unsigned long now;
1962 #ifdef CONFIG_IPV6_PRIVACY
1963                         struct inet6_ifaddr *ift;
1964 #endif
1965                         u32 stored_lft;
1966
1967                         /* update lifetime (RFC2462 5.5.3 e) */
1968                         spin_lock(&ifp->lock);
1969                         now = jiffies;
1970                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1971                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1972                         else
1973                                 stored_lft = 0;
1974                         if (!update_lft && stored_lft) {
1975                                 if (valid_lft > MIN_VALID_LIFETIME ||
1976                                     valid_lft > stored_lft)
1977                                         update_lft = 1;
1978                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1979                                         /* valid_lft <= stored_lft is always true */
1980                                         /*
1981                                          * RFC 4862 Section 5.5.3e:
1982                                          * "Note that the preferred lifetime of
1983                                          *  the corresponding address is always
1984                                          *  reset to the Preferred Lifetime in
1985                                          *  the received Prefix Information
1986                                          *  option, regardless of whether the
1987                                          *  valid lifetime is also reset or
1988                                          *  ignored."
1989                                          *
1990                                          *  So if the preferred lifetime in
1991                                          *  this advertisement is different
1992                                          *  than what we have stored, but the
1993                                          *  valid lifetime is invalid, just
1994                                          *  reset prefered_lft.
1995                                          *
1996                                          *  We must set the valid lifetime
1997                                          *  to the stored lifetime since we'll
1998                                          *  be updating the timestamp below,
1999                                          *  else we'll set it back to the
2000                                          *  minimum.
2001                                          */
2002                                         if (prefered_lft != ifp->prefered_lft) {
2003                                                 valid_lft = stored_lft;
2004                                                 update_lft = 1;
2005                                         }
2006                                 } else {
2007                                         valid_lft = MIN_VALID_LIFETIME;
2008                                         if (valid_lft < prefered_lft)
2009                                                 prefered_lft = valid_lft;
2010                                         update_lft = 1;
2011                                 }
2012                         }
2013
2014                         if (update_lft) {
2015                                 ifp->valid_lft = valid_lft;
2016                                 ifp->prefered_lft = prefered_lft;
2017                                 ifp->tstamp = now;
2018                                 flags = ifp->flags;
2019                                 ifp->flags &= ~IFA_F_DEPRECATED;
2020                                 spin_unlock(&ifp->lock);
2021
2022                                 if (!(flags&IFA_F_TENTATIVE))
2023                                         ipv6_ifa_notify(0, ifp);
2024                         } else
2025                                 spin_unlock(&ifp->lock);
2026
2027 #ifdef CONFIG_IPV6_PRIVACY
2028                         read_lock_bh(&in6_dev->lock);
2029                         /* update all temporary addresses in the list */
2030                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2031                                             tmp_list) {
2032                                 int age, max_valid, max_prefered;
2033
2034                                 if (ifp != ift->ifpub)
2035                                         continue;
2036
2037                                 /*
2038                                  * RFC 4941 section 3.3:
2039                                  * If a received option will extend the lifetime
2040                                  * of a public address, the lifetimes of
2041                                  * temporary addresses should be extended,
2042                                  * subject to the overall constraint that no
2043                                  * temporary addresses should ever remain
2044                                  * "valid" or "preferred" for a time longer than
2045                                  * (TEMP_VALID_LIFETIME) or
2046                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2047                                  * respectively.
2048                                  */
2049                                 age = (now - ift->cstamp) / HZ;
2050                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2051                                 if (max_valid < 0)
2052                                         max_valid = 0;
2053
2054                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2055                                                in6_dev->cnf.max_desync_factor -
2056                                                age;
2057                                 if (max_prefered < 0)
2058                                         max_prefered = 0;
2059
2060                                 if (valid_lft > max_valid)
2061                                         valid_lft = max_valid;
2062
2063                                 if (prefered_lft > max_prefered)
2064                                         prefered_lft = max_prefered;
2065
2066                                 spin_lock(&ift->lock);
2067                                 flags = ift->flags;
2068                                 ift->valid_lft = valid_lft;
2069                                 ift->prefered_lft = prefered_lft;
2070                                 ift->tstamp = now;
2071                                 if (prefered_lft > 0)
2072                                         ift->flags &= ~IFA_F_DEPRECATED;
2073
2074                                 spin_unlock(&ift->lock);
2075                                 if (!(flags&IFA_F_TENTATIVE))
2076                                         ipv6_ifa_notify(0, ift);
2077                         }
2078
2079                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2080                                 /*
2081                                  * When a new public address is created as
2082                                  * described in [ADDRCONF], also create a new
2083                                  * temporary address. Also create a temporary
2084                                  * address if it's enabled but no temporary
2085                                  * address currently exists.
2086                                  */
2087                                 read_unlock_bh(&in6_dev->lock);
2088                                 ipv6_create_tempaddr(ifp, NULL);
2089                         } else {
2090                                 read_unlock_bh(&in6_dev->lock);
2091                         }
2092 #endif
2093                         in6_ifa_put(ifp);
2094                         addrconf_verify(0);
2095                 }
2096         }
2097         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2098         in6_dev_put(in6_dev);
2099 }
2100
2101 /*
2102  *      Set destination address.
2103  *      Special case for SIT interfaces where we create a new "virtual"
2104  *      device.
2105  */
2106 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2107 {
2108         struct in6_ifreq ireq;
2109         struct net_device *dev;
2110         int err = -EINVAL;
2111
2112         rtnl_lock();
2113
2114         err = -EFAULT;
2115         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2116                 goto err_exit;
2117
2118         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2119
2120         err = -ENODEV;
2121         if (dev == NULL)
2122                 goto err_exit;
2123
2124 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2125         if (dev->type == ARPHRD_SIT) {
2126                 const struct net_device_ops *ops = dev->netdev_ops;
2127                 struct ifreq ifr;
2128                 struct ip_tunnel_parm p;
2129
2130                 err = -EADDRNOTAVAIL;
2131                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2132                         goto err_exit;
2133
2134                 memset(&p, 0, sizeof(p));
2135                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2136                 p.iph.saddr = 0;
2137                 p.iph.version = 4;
2138                 p.iph.ihl = 5;
2139                 p.iph.protocol = IPPROTO_IPV6;
2140                 p.iph.ttl = 64;
2141                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2142
2143                 if (ops->ndo_do_ioctl) {
2144                         mm_segment_t oldfs = get_fs();
2145
2146                         set_fs(KERNEL_DS);
2147                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2148                         set_fs(oldfs);
2149                 } else
2150                         err = -EOPNOTSUPP;
2151
2152                 if (err == 0) {
2153                         err = -ENOBUFS;
2154                         dev = __dev_get_by_name(net, p.name);
2155                         if (!dev)
2156                                 goto err_exit;
2157                         err = dev_open(dev);
2158                 }
2159         }
2160 #endif
2161
2162 err_exit:
2163         rtnl_unlock();
2164         return err;
2165 }
2166
2167 /*
2168  *      Manual configuration of address on an interface
2169  */
2170 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2171                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2172                           __u32 valid_lft)
2173 {
2174         struct inet6_ifaddr *ifp;
2175         struct inet6_dev *idev;
2176         struct net_device *dev;
2177         int scope;
2178         u32 flags;
2179         clock_t expires;
2180         unsigned long timeout;
2181
2182         ASSERT_RTNL();
2183
2184         if (plen > 128)
2185                 return -EINVAL;
2186
2187         /* check the lifetime */
2188         if (!valid_lft || prefered_lft > valid_lft)
2189                 return -EINVAL;
2190
2191         dev = __dev_get_by_index(net, ifindex);
2192         if (!dev)
2193                 return -ENODEV;
2194
2195         idev = addrconf_add_dev(dev);
2196         if (IS_ERR(idev))
2197                 return PTR_ERR(idev);
2198
2199         scope = ipv6_addr_scope(pfx);
2200
2201         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2202         if (addrconf_finite_timeout(timeout)) {
2203                 expires = jiffies_to_clock_t(timeout * HZ);
2204                 valid_lft = timeout;
2205                 flags = RTF_EXPIRES;
2206         } else {
2207                 expires = 0;
2208                 flags = 0;
2209                 ifa_flags |= IFA_F_PERMANENT;
2210         }
2211
2212         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2213         if (addrconf_finite_timeout(timeout)) {
2214                 if (timeout == 0)
2215                         ifa_flags |= IFA_F_DEPRECATED;
2216                 prefered_lft = timeout;
2217         }
2218
2219         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2220
2221         if (!IS_ERR(ifp)) {
2222                 spin_lock_bh(&ifp->lock);
2223                 ifp->valid_lft = valid_lft;
2224                 ifp->prefered_lft = prefered_lft;
2225                 ifp->tstamp = jiffies;
2226                 spin_unlock_bh(&ifp->lock);
2227
2228                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2229                                       expires, flags);
2230                 /*
2231                  * Note that section 3.1 of RFC 4429 indicates
2232                  * that the Optimistic flag should not be set for
2233                  * manually configured addresses
2234                  */
2235                 addrconf_dad_start(ifp);
2236                 in6_ifa_put(ifp);
2237                 addrconf_verify(0);
2238                 return 0;
2239         }
2240
2241         return PTR_ERR(ifp);
2242 }
2243
2244 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2245                           unsigned int plen)
2246 {
2247         struct inet6_ifaddr *ifp;
2248         struct inet6_dev *idev;
2249         struct net_device *dev;
2250
2251         if (plen > 128)
2252                 return -EINVAL;
2253
2254         dev = __dev_get_by_index(net, ifindex);
2255         if (!dev)
2256                 return -ENODEV;
2257
2258         if ((idev = __in6_dev_get(dev)) == NULL)
2259                 return -ENXIO;
2260
2261         read_lock_bh(&idev->lock);
2262         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2263                 if (ifp->prefix_len == plen &&
2264                     ipv6_addr_equal(pfx, &ifp->addr)) {
2265                         in6_ifa_hold(ifp);
2266                         read_unlock_bh(&idev->lock);
2267
2268                         ipv6_del_addr(ifp);
2269
2270                         /* If the last address is deleted administratively,
2271                            disable IPv6 on this interface.
2272                          */
2273                         if (list_empty(&idev->addr_list))
2274                                 addrconf_ifdown(idev->dev, 1);
2275                         return 0;
2276                 }
2277         }
2278         read_unlock_bh(&idev->lock);
2279         return -EADDRNOTAVAIL;
2280 }
2281
2282
2283 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2284 {
2285         struct in6_ifreq ireq;
2286         int err;
2287
2288         if (!capable(CAP_NET_ADMIN))
2289                 return -EPERM;
2290
2291         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2292                 return -EFAULT;
2293
2294         rtnl_lock();
2295         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2296                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2297                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2298         rtnl_unlock();
2299         return err;
2300 }
2301
2302 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2303 {
2304         struct in6_ifreq ireq;
2305         int err;
2306
2307         if (!capable(CAP_NET_ADMIN))
2308                 return -EPERM;
2309
2310         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2311                 return -EFAULT;
2312
2313         rtnl_lock();
2314         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2315                              ireq.ifr6_prefixlen);
2316         rtnl_unlock();
2317         return err;
2318 }
2319
2320 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2321                      int plen, int scope)
2322 {
2323         struct inet6_ifaddr *ifp;
2324
2325         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2326         if (!IS_ERR(ifp)) {
2327                 spin_lock_bh(&ifp->lock);
2328                 ifp->flags &= ~IFA_F_TENTATIVE;
2329                 spin_unlock_bh(&ifp->lock);
2330                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2331                 in6_ifa_put(ifp);
2332         }
2333 }
2334
2335 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2336 static void sit_add_v4_addrs(struct inet6_dev *idev)
2337 {
2338         struct in6_addr addr;
2339         struct net_device *dev;
2340         struct net *net = dev_net(idev->dev);
2341         int scope;
2342
2343         ASSERT_RTNL();
2344
2345         memset(&addr, 0, sizeof(struct in6_addr));
2346         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2347
2348         if (idev->dev->flags&IFF_POINTOPOINT) {
2349                 addr.s6_addr32[0] = htonl(0xfe800000);
2350                 scope = IFA_LINK;
2351         } else {
2352                 scope = IPV6_ADDR_COMPATv4;
2353         }
2354
2355         if (addr.s6_addr32[3]) {
2356                 add_addr(idev, &addr, 128, scope);
2357                 return;
2358         }
2359
2360         for_each_netdev(net, dev) {
2361                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2362                 if (in_dev && (dev->flags & IFF_UP)) {
2363                         struct in_ifaddr *ifa;
2364
2365                         int flag = scope;
2366
2367                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2368                                 int plen;
2369
2370                                 addr.s6_addr32[3] = ifa->ifa_local;
2371
2372                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2373                                         continue;
2374                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2375                                         if (idev->dev->flags&IFF_POINTOPOINT)
2376                                                 continue;
2377                                         flag |= IFA_HOST;
2378                                 }
2379                                 if (idev->dev->flags&IFF_POINTOPOINT)
2380                                         plen = 64;
2381                                 else
2382                                         plen = 96;
2383
2384                                 add_addr(idev, &addr, plen, flag);
2385                         }
2386                 }
2387         }
2388 }
2389 #endif
2390
2391 static void init_loopback(struct net_device *dev)
2392 {
2393         struct inet6_dev  *idev;
2394
2395         /* ::1 */
2396
2397         ASSERT_RTNL();
2398
2399         if ((idev = ipv6_find_idev(dev)) == NULL) {
2400                 pr_debug("%s: add_dev failed\n", __func__);
2401                 return;
2402         }
2403
2404         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2405 }
2406
2407 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2408 {
2409         struct inet6_ifaddr *ifp;
2410         u32 addr_flags = IFA_F_PERMANENT;
2411
2412 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2413         if (idev->cnf.optimistic_dad &&
2414             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2415                 addr_flags |= IFA_F_OPTIMISTIC;
2416 #endif
2417
2418
2419         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2420         if (!IS_ERR(ifp)) {
2421                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2422                 addrconf_dad_start(ifp);
2423                 in6_ifa_put(ifp);
2424         }
2425 }
2426
2427 static void addrconf_dev_config(struct net_device *dev)
2428 {
2429         struct in6_addr addr;
2430         struct inet6_dev *idev;
2431
2432         ASSERT_RTNL();
2433
2434         if ((dev->type != ARPHRD_ETHER) &&
2435             (dev->type != ARPHRD_FDDI) &&
2436             (dev->type != ARPHRD_ARCNET) &&
2437             (dev->type != ARPHRD_INFINIBAND) &&
2438             (dev->type != ARPHRD_IEEE802154)) {
2439                 /* Alas, we support only Ethernet autoconfiguration. */
2440                 return;
2441         }
2442
2443         idev = addrconf_add_dev(dev);
2444         if (IS_ERR(idev))
2445                 return;
2446
2447         memset(&addr, 0, sizeof(struct in6_addr));
2448         addr.s6_addr32[0] = htonl(0xFE800000);
2449
2450         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2451                 addrconf_add_linklocal(idev, &addr);
2452 }
2453
2454 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2455 static void addrconf_sit_config(struct net_device *dev)
2456 {
2457         struct inet6_dev *idev;
2458
2459         ASSERT_RTNL();
2460
2461         /*
2462          * Configure the tunnel with one of our IPv4
2463          * addresses... we should configure all of
2464          * our v4 addrs in the tunnel
2465          */
2466
2467         if ((idev = ipv6_find_idev(dev)) == NULL) {
2468                 pr_debug("%s: add_dev failed\n", __func__);
2469                 return;
2470         }
2471
2472         if (dev->priv_flags & IFF_ISATAP) {
2473                 struct in6_addr addr;
2474
2475                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2476                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2477                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2478                         addrconf_add_linklocal(idev, &addr);
2479                 return;
2480         }
2481
2482         sit_add_v4_addrs(idev);
2483
2484         if (dev->flags&IFF_POINTOPOINT) {
2485                 addrconf_add_mroute(dev);
2486                 addrconf_add_lroute(dev);
2487         } else
2488                 sit_route_add(dev);
2489 }
2490 #endif
2491
2492 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2493 static void addrconf_gre_config(struct net_device *dev)
2494 {
2495         struct inet6_dev *idev;
2496         struct in6_addr addr;
2497
2498         pr_info("%s(%s)\n", __func__, dev->name);
2499
2500         ASSERT_RTNL();
2501
2502         if ((idev = ipv6_find_idev(dev)) == NULL) {
2503                 pr_debug("%s: add_dev failed\n", __func__);
2504                 return;
2505         }
2506
2507         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2508         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2509
2510         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2511                 addrconf_add_linklocal(idev, &addr);
2512 }
2513 #endif
2514
2515 static inline int
2516 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2517 {
2518         struct in6_addr lladdr;
2519
2520         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2521                 addrconf_add_linklocal(idev, &lladdr);
2522                 return 0;
2523         }
2524         return -1;
2525 }
2526
2527 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2528 {
2529         struct net_device *link_dev;
2530         struct net *net = dev_net(idev->dev);
2531
2532         /* first try to inherit the link-local address from the link device */
2533         if (idev->dev->iflink &&
2534             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2535                 if (!ipv6_inherit_linklocal(idev, link_dev))
2536                         return;
2537         }
2538         /* then try to inherit it from any device */
2539         for_each_netdev(net, link_dev) {
2540                 if (!ipv6_inherit_linklocal(idev, link_dev))
2541                         return;
2542         }
2543         pr_debug("init ip6-ip6: add_linklocal failed\n");
2544 }
2545
2546 /*
2547  * Autoconfigure tunnel with a link-local address so routing protocols,
2548  * DHCPv6, MLD etc. can be run over the virtual link
2549  */
2550
2551 static void addrconf_ip6_tnl_config(struct net_device *dev)
2552 {
2553         struct inet6_dev *idev;
2554
2555         ASSERT_RTNL();
2556
2557         idev = addrconf_add_dev(dev);
2558         if (IS_ERR(idev)) {
2559                 pr_debug("init ip6-ip6: add_dev failed\n");
2560                 return;
2561         }
2562         ip6_tnl_add_linklocal(idev);
2563 }
2564
2565 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2566                            void *data)
2567 {
2568         struct net_device *dev = (struct net_device *) data;
2569         struct inet6_dev *idev = __in6_dev_get(dev);
2570         int run_pending = 0;
2571         int err;
2572
2573         switch (event) {
2574         case NETDEV_REGISTER:
2575                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2576                         idev = ipv6_add_dev(dev);
2577                         if (!idev)
2578                                 return notifier_from_errno(-ENOMEM);
2579                 }
2580                 break;
2581
2582         case NETDEV_UP:
2583         case NETDEV_CHANGE:
2584                 if (dev->flags & IFF_SLAVE)
2585                         break;
2586
2587                 if (event == NETDEV_UP) {
2588                         if (!addrconf_qdisc_ok(dev)) {
2589                                 /* device is not ready yet. */
2590                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2591                                         dev->name);
2592                                 break;
2593                         }
2594
2595                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2596                                 idev = ipv6_add_dev(dev);
2597
2598                         if (idev) {
2599                                 idev->if_flags |= IF_READY;
2600                                 run_pending = 1;
2601                         }
2602                 } else {
2603                         if (!addrconf_qdisc_ok(dev)) {
2604                                 /* device is still not ready. */
2605                                 break;
2606                         }
2607
2608                         if (idev) {
2609                                 if (idev->if_flags & IF_READY)
2610                                         /* device is already configured. */
2611                                         break;
2612                                 idev->if_flags |= IF_READY;
2613                         }
2614
2615                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2616                                 dev->name);
2617
2618                         run_pending = 1;
2619                 }
2620
2621                 switch (dev->type) {
2622 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2623                 case ARPHRD_SIT:
2624                         addrconf_sit_config(dev);
2625                         break;
2626 #endif
2627 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2628                 case ARPHRD_IPGRE:
2629                         addrconf_gre_config(dev);
2630                         break;
2631 #endif
2632                 case ARPHRD_TUNNEL6:
2633                         addrconf_ip6_tnl_config(dev);
2634                         break;
2635                 case ARPHRD_LOOPBACK:
2636                         init_loopback(dev);
2637                         break;
2638
2639                 default:
2640                         addrconf_dev_config(dev);
2641                         break;
2642                 }
2643
2644                 if (idev) {
2645                         if (run_pending)
2646                                 addrconf_dad_run(idev);
2647
2648                         /*
2649                          * If the MTU changed during the interface down,
2650                          * when the interface up, the changed MTU must be
2651                          * reflected in the idev as well as routers.
2652                          */
2653                         if (idev->cnf.mtu6 != dev->mtu &&
2654                             dev->mtu >= IPV6_MIN_MTU) {
2655                                 rt6_mtu_change(dev, dev->mtu);
2656                                 idev->cnf.mtu6 = dev->mtu;
2657                         }
2658                         idev->tstamp = jiffies;
2659                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2660
2661                         /*
2662                          * If the changed mtu during down is lower than
2663                          * IPV6_MIN_MTU stop IPv6 on this interface.
2664                          */
2665                         if (dev->mtu < IPV6_MIN_MTU)
2666                                 addrconf_ifdown(dev, 1);
2667                 }
2668                 break;
2669
2670         case NETDEV_CHANGEMTU:
2671                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2672                         rt6_mtu_change(dev, dev->mtu);
2673                         idev->cnf.mtu6 = dev->mtu;
2674                         break;
2675                 }
2676
2677                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2678                         idev = ipv6_add_dev(dev);
2679                         if (idev)
2680                                 break;
2681                 }
2682
2683                 /*
2684                  * MTU falled under IPV6_MIN_MTU.
2685                  * Stop IPv6 on this interface.
2686                  */
2687
2688         case NETDEV_DOWN:
2689         case NETDEV_UNREGISTER:
2690                 /*
2691                  *      Remove all addresses from this interface.
2692                  */
2693                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2694                 break;
2695
2696         case NETDEV_CHANGENAME:
2697                 if (idev) {
2698                         snmp6_unregister_dev(idev);
2699                         addrconf_sysctl_unregister(idev);
2700                         addrconf_sysctl_register(idev);
2701                         err = snmp6_register_dev(idev);
2702                         if (err)
2703                                 return notifier_from_errno(err);
2704                 }
2705                 break;
2706
2707         case NETDEV_PRE_TYPE_CHANGE:
2708         case NETDEV_POST_TYPE_CHANGE:
2709                 addrconf_type_change(dev, event);
2710                 break;
2711         }
2712
2713         return NOTIFY_OK;
2714 }
2715
2716 /*
2717  *      addrconf module should be notified of a device going up
2718  */
2719 static struct notifier_block ipv6_dev_notf = {
2720         .notifier_call = addrconf_notify,
2721 };
2722
2723 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2724 {
2725         struct inet6_dev *idev;
2726         ASSERT_RTNL();
2727
2728         idev = __in6_dev_get(dev);
2729
2730         if (event == NETDEV_POST_TYPE_CHANGE)
2731                 ipv6_mc_remap(idev);
2732         else if (event == NETDEV_PRE_TYPE_CHANGE)
2733                 ipv6_mc_unmap(idev);
2734 }
2735
2736 static int addrconf_ifdown(struct net_device *dev, int how)
2737 {
2738         struct net *net = dev_net(dev);
2739         struct inet6_dev *idev;
2740         struct inet6_ifaddr *ifa;
2741         int state, i;
2742
2743         ASSERT_RTNL();
2744
2745         rt6_ifdown(net, dev);
2746         neigh_ifdown(&nd_tbl, dev);
2747
2748         idev = __in6_dev_get(dev);
2749         if (idev == NULL)
2750                 return -ENODEV;
2751
2752         /*
2753          * Step 1: remove reference to ipv6 device from parent device.
2754          *         Do not dev_put!
2755          */
2756         if (how) {
2757                 idev->dead = 1;
2758
2759                 /* protected by rtnl_lock */
2760                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2761
2762                 /* Step 1.5: remove snmp6 entry */
2763                 snmp6_unregister_dev(idev);
2764
2765         }
2766
2767         /* Step 2: clear hash table */
2768         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2769                 struct hlist_head *h = &inet6_addr_lst[i];
2770                 struct hlist_node *n;
2771
2772                 spin_lock_bh(&addrconf_hash_lock);
2773         restart:
2774                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2775                         if (ifa->idev == idev) {
2776                                 hlist_del_init_rcu(&ifa->addr_lst);
2777                                 addrconf_del_timer(ifa);
2778                                 goto restart;
2779                         }
2780                 }
2781                 spin_unlock_bh(&addrconf_hash_lock);
2782         }
2783
2784         write_lock_bh(&idev->lock);
2785
2786         /* Step 2: clear flags for stateless addrconf */
2787         if (!how)
2788                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2789
2790 #ifdef CONFIG_IPV6_PRIVACY
2791         if (how && del_timer(&idev->regen_timer))
2792                 in6_dev_put(idev);
2793
2794         /* Step 3: clear tempaddr list */
2795         while (!list_empty(&idev->tempaddr_list)) {
2796                 ifa = list_first_entry(&idev->tempaddr_list,
2797                                        struct inet6_ifaddr, tmp_list);
2798                 list_del(&ifa->tmp_list);
2799                 write_unlock_bh(&idev->lock);
2800                 spin_lock_bh(&ifa->lock);
2801
2802                 if (ifa->ifpub) {
2803                         in6_ifa_put(ifa->ifpub);
2804                         ifa->ifpub = NULL;
2805                 }
2806                 spin_unlock_bh(&ifa->lock);
2807                 in6_ifa_put(ifa);
2808                 write_lock_bh(&idev->lock);
2809         }
2810 #endif
2811
2812         while (!list_empty(&idev->addr_list)) {
2813                 ifa = list_first_entry(&idev->addr_list,
2814                                        struct inet6_ifaddr, if_list);
2815                 addrconf_del_timer(ifa);
2816
2817                 list_del(&ifa->if_list);
2818
2819                 write_unlock_bh(&idev->lock);
2820
2821                 spin_lock_bh(&ifa->state_lock);
2822                 state = ifa->state;
2823                 ifa->state = INET6_IFADDR_STATE_DEAD;
2824                 spin_unlock_bh(&ifa->state_lock);
2825
2826                 if (state != INET6_IFADDR_STATE_DEAD) {
2827                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2828                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2829                 }
2830                 in6_ifa_put(ifa);
2831
2832                 write_lock_bh(&idev->lock);
2833         }
2834
2835         write_unlock_bh(&idev->lock);
2836
2837         /* Step 5: Discard multicast list */
2838         if (how)
2839                 ipv6_mc_destroy_dev(idev);
2840         else
2841                 ipv6_mc_down(idev);
2842
2843         idev->tstamp = jiffies;
2844
2845         /* Last: Shot the device (if unregistered) */
2846         if (how) {
2847                 addrconf_sysctl_unregister(idev);
2848                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2849                 neigh_ifdown(&nd_tbl, dev);
2850                 in6_dev_put(idev);
2851         }
2852         return 0;
2853 }
2854
2855 static void addrconf_rs_timer(unsigned long data)
2856 {
2857         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2858         struct inet6_dev *idev = ifp->idev;
2859
2860         read_lock(&idev->lock);
2861         if (idev->dead || !(idev->if_flags & IF_READY))
2862                 goto out;
2863
2864         if (idev->cnf.forwarding)
2865                 goto out;
2866
2867         /* Announcement received after solicitation was sent */
2868         if (idev->if_flags & IF_RA_RCVD)
2869                 goto out;
2870
2871         spin_lock(&ifp->lock);
2872         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2873                 /* The wait after the last probe can be shorter */
2874                 addrconf_mod_timer(ifp, AC_RS,
2875                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2876                                    idev->cnf.rtr_solicit_delay :
2877                                    idev->cnf.rtr_solicit_interval);
2878                 spin_unlock(&ifp->lock);
2879
2880                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2881         } else {
2882                 spin_unlock(&ifp->lock);
2883                 /*
2884                  * Note: we do not support deprecated "all on-link"
2885                  * assumption any longer.
2886                  */
2887                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
2888         }
2889
2890 out:
2891         read_unlock(&idev->lock);
2892         in6_ifa_put(ifp);
2893 }
2894
2895 /*
2896  *      Duplicate Address Detection
2897  */
2898 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2899 {
2900         unsigned long rand_num;
2901         struct inet6_dev *idev = ifp->idev;
2902
2903         if (ifp->flags & IFA_F_OPTIMISTIC)
2904                 rand_num = 0;
2905         else
2906                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2907
2908         ifp->probes = idev->cnf.dad_transmits;
2909         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2910 }
2911
2912 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
2913 {
2914         struct inet6_dev *idev = ifp->idev;
2915         struct net_device *dev = idev->dev;
2916
2917         addrconf_join_solict(dev, &ifp->addr);
2918
2919         net_srandom(ifp->addr.s6_addr32[3]);
2920
2921         read_lock_bh(&idev->lock);
2922         spin_lock(&ifp->lock);
2923         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2924                 goto out;
2925
2926         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2927             idev->cnf.accept_dad < 1 ||
2928             !(ifp->flags&IFA_F_TENTATIVE) ||
2929             ifp->flags & IFA_F_NODAD) {
2930                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2931                 spin_unlock(&ifp->lock);
2932                 read_unlock_bh(&idev->lock);
2933
2934                 addrconf_dad_completed(ifp);
2935                 return;
2936         }
2937
2938         if (!(idev->if_flags & IF_READY)) {
2939                 spin_unlock(&ifp->lock);
2940                 read_unlock_bh(&idev->lock);
2941                 /*
2942                  * If the device is not ready:
2943                  * - keep it tentative if it is a permanent address.
2944                  * - otherwise, kill it.
2945                  */
2946                 in6_ifa_hold(ifp);
2947                 addrconf_dad_stop(ifp, 0);
2948                 return;
2949         }
2950
2951         /*
2952          * Optimistic nodes can start receiving
2953          * Frames right away
2954          */
2955         if (ifp->flags & IFA_F_OPTIMISTIC)
2956                 ip6_ins_rt(ifp->rt);
2957
2958         addrconf_dad_kick(ifp);
2959 out:
2960         spin_unlock(&ifp->lock);
2961         read_unlock_bh(&idev->lock);
2962 }
2963
2964 static void addrconf_dad_timer(unsigned long data)
2965 {
2966         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2967         struct inet6_dev *idev = ifp->idev;
2968         struct in6_addr mcaddr;
2969
2970         if (!ifp->probes && addrconf_dad_end(ifp))
2971                 goto out;
2972
2973         read_lock(&idev->lock);
2974         if (idev->dead || !(idev->if_flags & IF_READY)) {
2975                 read_unlock(&idev->lock);
2976                 goto out;
2977         }
2978
2979         spin_lock(&ifp->lock);
2980         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2981                 spin_unlock(&ifp->lock);
2982                 read_unlock(&idev->lock);
2983                 goto out;
2984         }
2985
2986         if (ifp->probes == 0) {
2987                 /*
2988                  * DAD was successful
2989                  */
2990
2991                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2992                 spin_unlock(&ifp->lock);
2993                 read_unlock(&idev->lock);
2994
2995                 addrconf_dad_completed(ifp);
2996
2997                 goto out;
2998         }
2999
3000         ifp->probes--;
3001         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3002         spin_unlock(&ifp->lock);
3003         read_unlock(&idev->lock);
3004
3005         /* send a neighbour solicitation for our addr */
3006         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3007         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3008 out:
3009         in6_ifa_put(ifp);
3010 }
3011
3012 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3013 {
3014         struct net_device *dev = ifp->idev->dev;
3015
3016         /*
3017          *      Configure the address for reception. Now it is valid.
3018          */
3019
3020         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3021
3022         /* If added prefix is link local and we are prepared to process
3023            router advertisements, start sending router solicitations.
3024          */
3025
3026         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3027              ifp->idev->cnf.accept_ra == 2) &&
3028             ifp->idev->cnf.rtr_solicits > 0 &&
3029             (dev->flags&IFF_LOOPBACK) == 0 &&
3030             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3031                 /*
3032                  *      If a host as already performed a random delay
3033                  *      [...] as part of DAD [...] there is no need
3034                  *      to delay again before sending the first RS
3035                  */
3036                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3037
3038                 spin_lock_bh(&ifp->lock);
3039                 ifp->probes = 1;
3040                 ifp->idev->if_flags |= IF_RS_SENT;
3041                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3042                 spin_unlock_bh(&ifp->lock);
3043         }
3044 }
3045
3046 static void addrconf_dad_run(struct inet6_dev *idev)
3047 {
3048         struct inet6_ifaddr *ifp;
3049
3050         read_lock_bh(&idev->lock);
3051         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3052                 spin_lock(&ifp->lock);
3053                 if (ifp->flags & IFA_F_TENTATIVE &&
3054                     ifp->state == INET6_IFADDR_STATE_DAD)
3055                         addrconf_dad_kick(ifp);
3056                 spin_unlock(&ifp->lock);
3057         }
3058         read_unlock_bh(&idev->lock);
3059 }
3060
3061 #ifdef CONFIG_PROC_FS
3062 struct if6_iter_state {
3063         struct seq_net_private p;
3064         int bucket;
3065         int offset;
3066 };
3067
3068 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3069 {
3070         struct inet6_ifaddr *ifa = NULL;
3071         struct if6_iter_state *state = seq->private;
3072         struct net *net = seq_file_net(seq);
3073         int p = 0;
3074
3075         /* initial bucket if pos is 0 */
3076         if (pos == 0) {
3077                 state->bucket = 0;
3078                 state->offset = 0;
3079         }
3080
3081         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3082                 struct hlist_node *n;
3083                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3084                                          addr_lst) {
3085                         /* sync with offset */
3086                         if (p < state->offset) {
3087                                 p++;
3088                                 continue;
3089                         }
3090                         state->offset++;
3091                         if (net_eq(dev_net(ifa->idev->dev), net))
3092                                 return ifa;
3093                 }
3094
3095                 /* prepare for next bucket */
3096                 state->offset = 0;
3097                 p = 0;
3098         }
3099         return NULL;
3100 }
3101
3102 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3103                                          struct inet6_ifaddr *ifa)
3104 {
3105         struct if6_iter_state *state = seq->private;
3106         struct net *net = seq_file_net(seq);
3107         struct hlist_node *n = &ifa->addr_lst;
3108
3109         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3110                 state->offset++;
3111                 if (net_eq(dev_net(ifa->idev->dev), net))
3112                         return ifa;
3113         }
3114
3115         while (++state->bucket < IN6_ADDR_HSIZE) {
3116                 state->offset = 0;
3117                 hlist_for_each_entry_rcu_bh(ifa, n,
3118                                      &inet6_addr_lst[state->bucket], addr_lst) {
3119                         state->offset++;
3120                         if (net_eq(dev_net(ifa->idev->dev), net))
3121                                 return ifa;
3122                 }
3123         }
3124
3125         return NULL;
3126 }
3127
3128 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3129         __acquires(rcu_bh)
3130 {
3131         rcu_read_lock_bh();
3132         return if6_get_first(seq, *pos);
3133 }
3134
3135 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3136 {
3137         struct inet6_ifaddr *ifa;
3138
3139         ifa = if6_get_next(seq, v);
3140         ++*pos;
3141         return ifa;
3142 }
3143
3144 static void if6_seq_stop(struct seq_file *seq, void *v)
3145         __releases(rcu_bh)
3146 {
3147         rcu_read_unlock_bh();
3148 }
3149
3150 static int if6_seq_show(struct seq_file *seq, void *v)
3151 {
3152         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3153         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3154                    &ifp->addr,
3155                    ifp->idev->dev->ifindex,
3156                    ifp->prefix_len,
3157                    ifp->scope,
3158                    ifp->flags,
3159                    ifp->idev->dev->name);
3160         return 0;
3161 }
3162
3163 static const struct seq_operations if6_seq_ops = {
3164         .start  = if6_seq_start,
3165         .next   = if6_seq_next,
3166         .show   = if6_seq_show,
3167         .stop   = if6_seq_stop,
3168 };
3169
3170 static int if6_seq_open(struct inode *inode, struct file *file)
3171 {
3172         return seq_open_net(inode, file, &if6_seq_ops,
3173                             sizeof(struct if6_iter_state));
3174 }
3175
3176 static const struct file_operations if6_fops = {
3177         .owner          = THIS_MODULE,
3178         .open           = if6_seq_open,
3179         .read           = seq_read,
3180         .llseek         = seq_lseek,
3181         .release        = seq_release_net,
3182 };
3183
3184 static int __net_init if6_proc_net_init(struct net *net)
3185 {
3186         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3187                 return -ENOMEM;
3188         return 0;
3189 }
3190
3191 static void __net_exit if6_proc_net_exit(struct net *net)
3192 {
3193        proc_net_remove(net, "if_inet6");
3194 }
3195
3196 static struct pernet_operations if6_proc_net_ops = {
3197        .init = if6_proc_net_init,
3198        .exit = if6_proc_net_exit,
3199 };
3200
3201 int __init if6_proc_init(void)
3202 {
3203         return register_pernet_subsys(&if6_proc_net_ops);
3204 }
3205
3206 void if6_proc_exit(void)
3207 {
3208         unregister_pernet_subsys(&if6_proc_net_ops);
3209 }
3210 #endif  /* CONFIG_PROC_FS */
3211
3212 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3213 /* Check if address is a home address configured on any interface. */
3214 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3215 {
3216         int ret = 0;
3217         struct inet6_ifaddr *ifp = NULL;
3218         struct hlist_node *n;
3219         unsigned int hash = inet6_addr_hash(addr);
3220
3221         rcu_read_lock_bh();
3222         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3223                 if (!net_eq(dev_net(ifp->idev->dev), net))
3224                         continue;
3225                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3226                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3227                         ret = 1;
3228                         break;
3229                 }
3230         }
3231         rcu_read_unlock_bh();
3232         return ret;
3233 }
3234 #endif
3235
3236 /*
3237  *      Periodic address status verification
3238  */
3239
3240 static void addrconf_verify(unsigned long foo)
3241 {
3242         unsigned long now, next, next_sec, next_sched;
3243         struct inet6_ifaddr *ifp;
3244         struct hlist_node *node;
3245         int i;
3246
3247         rcu_read_lock_bh();
3248         spin_lock(&addrconf_verify_lock);
3249         now = jiffies;
3250         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3251
3252         del_timer(&addr_chk_timer);
3253
3254         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3255 restart:
3256                 hlist_for_each_entry_rcu_bh(ifp, node,
3257                                          &inet6_addr_lst[i], addr_lst) {
3258                         unsigned long age;
3259
3260                         if (ifp->flags & IFA_F_PERMANENT)
3261                                 continue;
3262
3263                         spin_lock(&ifp->lock);
3264                         /* We try to batch several events at once. */
3265                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3266
3267                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3268                             age >= ifp->valid_lft) {
3269                                 spin_unlock(&ifp->lock);
3270                                 in6_ifa_hold(ifp);
3271                                 ipv6_del_addr(ifp);
3272                                 goto restart;
3273                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3274                                 spin_unlock(&ifp->lock);
3275                                 continue;
3276                         } else if (age >= ifp->prefered_lft) {
3277                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3278                                 int deprecate = 0;
3279
3280                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3281                                         deprecate = 1;
3282                                         ifp->flags |= IFA_F_DEPRECATED;
3283                                 }
3284
3285                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3286                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3287
3288                                 spin_unlock(&ifp->lock);
3289
3290                                 if (deprecate) {
3291                                         in6_ifa_hold(ifp);
3292
3293                                         ipv6_ifa_notify(0, ifp);
3294                                         in6_ifa_put(ifp);
3295                                         goto restart;
3296                                 }
3297 #ifdef CONFIG_IPV6_PRIVACY
3298                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3299                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3300                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3301                                         ifp->idev->cnf.dad_transmits *
3302                                         ifp->idev->nd_parms->retrans_time / HZ;
3303
3304                                 if (age >= ifp->prefered_lft - regen_advance) {
3305                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3306                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3307                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3308                                         if (!ifp->regen_count && ifpub) {
3309                                                 ifp->regen_count++;
3310                                                 in6_ifa_hold(ifp);
3311                                                 in6_ifa_hold(ifpub);
3312                                                 spin_unlock(&ifp->lock);
3313
3314                                                 spin_lock(&ifpub->lock);
3315                                                 ifpub->regen_count = 0;
3316                                                 spin_unlock(&ifpub->lock);
3317                                                 ipv6_create_tempaddr(ifpub, ifp);
3318                                                 in6_ifa_put(ifpub);
3319                                                 in6_ifa_put(ifp);
3320                                                 goto restart;
3321                                         }
3322                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3323                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3324                                 spin_unlock(&ifp->lock);
3325 #endif
3326                         } else {
3327                                 /* ifp->prefered_lft <= ifp->valid_lft */
3328                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3329                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3330                                 spin_unlock(&ifp->lock);
3331                         }
3332                 }
3333         }
3334
3335         next_sec = round_jiffies_up(next);
3336         next_sched = next;
3337
3338         /* If rounded timeout is accurate enough, accept it. */
3339         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3340                 next_sched = next_sec;
3341
3342         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3343         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3344                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3345
3346         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3347               now, next, next_sec, next_sched));
3348
3349         addr_chk_timer.expires = next_sched;
3350         add_timer(&addr_chk_timer);
3351         spin_unlock(&addrconf_verify_lock);
3352         rcu_read_unlock_bh();
3353 }
3354
3355 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3356 {
3357         struct in6_addr *pfx = NULL;
3358
3359         if (addr)
3360                 pfx = nla_data(addr);
3361
3362         if (local) {
3363                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3364                         pfx = NULL;
3365                 else
3366                         pfx = nla_data(local);
3367         }
3368
3369         return pfx;
3370 }
3371
3372 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3373         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3374         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3375         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3376 };
3377
3378 static int
3379 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3380 {
3381         struct net *net = sock_net(skb->sk);
3382         struct ifaddrmsg *ifm;
3383         struct nlattr *tb[IFA_MAX+1];
3384         struct in6_addr *pfx;
3385         int err;
3386
3387         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3388         if (err < 0)
3389                 return err;
3390
3391         ifm = nlmsg_data(nlh);
3392         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3393         if (pfx == NULL)
3394                 return -EINVAL;
3395
3396         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3397 }
3398
3399 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3400                              u32 prefered_lft, u32 valid_lft)
3401 {
3402         u32 flags;
3403         clock_t expires;
3404         unsigned long timeout;
3405
3406         if (!valid_lft || (prefered_lft > valid_lft))
3407                 return -EINVAL;
3408
3409         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3410         if (addrconf_finite_timeout(timeout)) {
3411                 expires = jiffies_to_clock_t(timeout * HZ);
3412                 valid_lft = timeout;
3413                 flags = RTF_EXPIRES;
3414         } else {
3415                 expires = 0;
3416                 flags = 0;
3417                 ifa_flags |= IFA_F_PERMANENT;
3418         }
3419
3420         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3421         if (addrconf_finite_timeout(timeout)) {
3422                 if (timeout == 0)
3423                         ifa_flags |= IFA_F_DEPRECATED;
3424                 prefered_lft = timeout;
3425         }
3426
3427         spin_lock_bh(&ifp->lock);
3428         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3429         ifp->tstamp = jiffies;
3430         ifp->valid_lft = valid_lft;
3431         ifp->prefered_lft = prefered_lft;
3432
3433         spin_unlock_bh(&ifp->lock);
3434         if (!(ifp->flags&IFA_F_TENTATIVE))
3435                 ipv6_ifa_notify(0, ifp);
3436
3437         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3438                               expires, flags);
3439         addrconf_verify(0);
3440
3441         return 0;
3442 }
3443
3444 static int
3445 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3446 {
3447         struct net *net = sock_net(skb->sk);
3448         struct ifaddrmsg *ifm;
3449         struct nlattr *tb[IFA_MAX+1];
3450         struct in6_addr *pfx;
3451         struct inet6_ifaddr *ifa;
3452         struct net_device *dev;
3453         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3454         u8 ifa_flags;
3455         int err;
3456
3457         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3458         if (err < 0)
3459                 return err;
3460
3461         ifm = nlmsg_data(nlh);
3462         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3463         if (pfx == NULL)
3464                 return -EINVAL;
3465
3466         if (tb[IFA_CACHEINFO]) {
3467                 struct ifa_cacheinfo *ci;
3468
3469                 ci = nla_data(tb[IFA_CACHEINFO]);
3470                 valid_lft = ci->ifa_valid;
3471                 preferred_lft = ci->ifa_prefered;
3472         } else {
3473                 preferred_lft = INFINITY_LIFE_TIME;
3474                 valid_lft = INFINITY_LIFE_TIME;
3475         }
3476
3477         dev =  __dev_get_by_index(net, ifm->ifa_index);
3478         if (dev == NULL)
3479                 return -ENODEV;
3480
3481         /* We ignore other flags so far. */
3482         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3483
3484         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3485         if (ifa == NULL) {
3486                 /*
3487                  * It would be best to check for !NLM_F_CREATE here but
3488                  * userspace alreay relies on not having to provide this.
3489                  */
3490                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3491                                       ifm->ifa_prefixlen, ifa_flags,
3492                                       preferred_lft, valid_lft);
3493         }
3494
3495         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3496             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3497                 err = -EEXIST;
3498         else
3499                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3500
3501         in6_ifa_put(ifa);
3502
3503         return err;
3504 }
3505
3506 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3507                           u8 scope, int ifindex)
3508 {
3509         struct ifaddrmsg *ifm;
3510
3511         ifm = nlmsg_data(nlh);
3512         ifm->ifa_family = AF_INET6;
3513         ifm->ifa_prefixlen = prefixlen;
3514         ifm->ifa_flags = flags;
3515         ifm->ifa_scope = scope;
3516         ifm->ifa_index = ifindex;
3517 }
3518
3519 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3520                          unsigned long tstamp, u32 preferred, u32 valid)
3521 {
3522         struct ifa_cacheinfo ci;
3523
3524         ci.cstamp = cstamp_delta(cstamp);
3525         ci.tstamp = cstamp_delta(tstamp);
3526         ci.ifa_prefered = preferred;
3527         ci.ifa_valid = valid;
3528
3529         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3530 }
3531
3532 static inline int rt_scope(int ifa_scope)
3533 {
3534         if (ifa_scope & IFA_HOST)
3535                 return RT_SCOPE_HOST;
3536         else if (ifa_scope & IFA_LINK)
3537                 return RT_SCOPE_LINK;
3538         else if (ifa_scope & IFA_SITE)
3539                 return RT_SCOPE_SITE;
3540         else
3541                 return RT_SCOPE_UNIVERSE;
3542 }
3543
3544 static inline int inet6_ifaddr_msgsize(void)
3545 {
3546         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3547                + nla_total_size(16) /* IFA_ADDRESS */
3548                + nla_total_size(sizeof(struct ifa_cacheinfo));
3549 }
3550
3551 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3552                              u32 pid, u32 seq, int event, unsigned int flags)
3553 {
3554         struct nlmsghdr  *nlh;
3555         u32 preferred, valid;
3556
3557         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3558         if (nlh == NULL)
3559                 return -EMSGSIZE;
3560
3561         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3562                       ifa->idev->dev->ifindex);
3563
3564         if (!(ifa->flags&IFA_F_PERMANENT)) {
3565                 preferred = ifa->prefered_lft;
3566                 valid = ifa->valid_lft;
3567                 if (preferred != INFINITY_LIFE_TIME) {
3568                         long tval = (jiffies - ifa->tstamp)/HZ;
3569                         if (preferred > tval)
3570                                 preferred -= tval;
3571                         else
3572                                 preferred = 0;
3573                         if (valid != INFINITY_LIFE_TIME) {
3574                                 if (valid > tval)
3575                                         valid -= tval;
3576                                 else
3577                                         valid = 0;
3578                         }
3579                 }
3580         } else {
3581                 preferred = INFINITY_LIFE_TIME;
3582                 valid = INFINITY_LIFE_TIME;
3583         }
3584
3585         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3586             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3587                 nlmsg_cancel(skb, nlh);
3588                 return -EMSGSIZE;
3589         }
3590
3591         return nlmsg_end(skb, nlh);
3592 }
3593
3594 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3595                                 u32 pid, u32 seq, int event, u16 flags)
3596 {
3597         struct nlmsghdr  *nlh;
3598         u8 scope = RT_SCOPE_UNIVERSE;
3599         int ifindex = ifmca->idev->dev->ifindex;
3600
3601         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3602                 scope = RT_SCOPE_SITE;
3603
3604         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3605         if (nlh == NULL)
3606                 return -EMSGSIZE;
3607
3608         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3609         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3610             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3611                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3612                 nlmsg_cancel(skb, nlh);
3613                 return -EMSGSIZE;
3614         }
3615
3616         return nlmsg_end(skb, nlh);
3617 }
3618
3619 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3620                                 u32 pid, u32 seq, int event, unsigned int flags)
3621 {
3622         struct nlmsghdr  *nlh;
3623         u8 scope = RT_SCOPE_UNIVERSE;
3624         int ifindex = ifaca->aca_idev->dev->ifindex;
3625
3626         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3627                 scope = RT_SCOPE_SITE;
3628
3629         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3630         if (nlh == NULL)
3631                 return -EMSGSIZE;
3632
3633         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3634         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3635             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3636                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3637                 nlmsg_cancel(skb, nlh);
3638                 return -EMSGSIZE;
3639         }
3640
3641         return nlmsg_end(skb, nlh);
3642 }
3643
3644 enum addr_type_t {
3645         UNICAST_ADDR,
3646         MULTICAST_ADDR,
3647         ANYCAST_ADDR,
3648 };
3649
3650 /* called with rcu_read_lock() */
3651 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3652                           struct netlink_callback *cb, enum addr_type_t type,
3653                           int s_ip_idx, int *p_ip_idx)
3654 {
3655         struct ifmcaddr6 *ifmca;
3656         struct ifacaddr6 *ifaca;
3657         int err = 1;
3658         int ip_idx = *p_ip_idx;
3659
3660         read_lock_bh(&idev->lock);
3661         switch (type) {
3662         case UNICAST_ADDR: {
3663                 struct inet6_ifaddr *ifa;
3664
3665                 /* unicast address incl. temp addr */
3666                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3667                         if (++ip_idx < s_ip_idx)
3668                                 continue;
3669                         err = inet6_fill_ifaddr(skb, ifa,
3670                                                 NETLINK_CB(cb->skb).pid,
3671                                                 cb->nlh->nlmsg_seq,
3672                                                 RTM_NEWADDR,
3673                                                 NLM_F_MULTI);
3674                         if (err <= 0)
3675                                 break;
3676                 }
3677                 break;
3678         }
3679         case MULTICAST_ADDR:
3680                 /* multicast address */
3681                 for (ifmca = idev->mc_list; ifmca;
3682                      ifmca = ifmca->next, ip_idx++) {
3683                         if (ip_idx < s_ip_idx)
3684                                 continue;
3685                         err = inet6_fill_ifmcaddr(skb, ifmca,
3686                                                   NETLINK_CB(cb->skb).pid,
3687                                                   cb->nlh->nlmsg_seq,
3688                                                   RTM_GETMULTICAST,
3689                                                   NLM_F_MULTI);
3690                         if (err <= 0)
3691                                 break;
3692                 }
3693                 break;
3694         case ANYCAST_ADDR:
3695                 /* anycast address */
3696                 for (ifaca = idev->ac_list; ifaca;
3697                      ifaca = ifaca->aca_next, ip_idx++) {
3698                         if (ip_idx < s_ip_idx)
3699                                 continue;
3700                         err = inet6_fill_ifacaddr(skb, ifaca,
3701                                                   NETLINK_CB(cb->skb).pid,
3702                                                   cb->nlh->nlmsg_seq,
3703                                                   RTM_GETANYCAST,
3704                                                   NLM_F_MULTI);
3705                         if (err <= 0)
3706                                 break;
3707                 }
3708                 break;
3709         default:
3710                 break;
3711         }
3712         read_unlock_bh(&idev->lock);
3713         *p_ip_idx = ip_idx;
3714         return err;
3715 }
3716
3717 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3718                            enum addr_type_t type)
3719 {
3720         struct net *net = sock_net(skb->sk);
3721         int h, s_h;
3722         int idx, ip_idx;
3723         int s_idx, s_ip_idx;
3724         struct net_device *dev;
3725         struct inet6_dev *idev;
3726         struct hlist_head *head;
3727         struct hlist_node *node;
3728
3729         s_h = cb->args[0];
3730         s_idx = idx = cb->args[1];
3731         s_ip_idx = ip_idx = cb->args[2];
3732
3733         rcu_read_lock();
3734         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3735                 idx = 0;
3736                 head = &net->dev_index_head[h];
3737                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3738                         if (idx < s_idx)
3739                                 goto cont;
3740                         if (h > s_h || idx > s_idx)
3741                                 s_ip_idx = 0;
3742                         ip_idx = 0;
3743                         idev = __in6_dev_get(dev);
3744                         if (!idev)
3745                                 goto cont;
3746
3747                         if (in6_dump_addrs(idev, skb, cb, type,
3748                                            s_ip_idx, &ip_idx) <= 0)
3749                                 goto done;
3750 cont:
3751                         idx++;
3752                 }
3753         }
3754 done:
3755         rcu_read_unlock();
3756         cb->args[0] = h;
3757         cb->args[1] = idx;
3758         cb->args[2] = ip_idx;
3759
3760         return skb->len;
3761 }
3762
3763 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3764 {
3765         enum addr_type_t type = UNICAST_ADDR;
3766
3767         return inet6_dump_addr(skb, cb, type);
3768 }
3769
3770 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3771 {
3772         enum addr_type_t type = MULTICAST_ADDR;
3773
3774         return inet6_dump_addr(skb, cb, type);
3775 }
3776
3777
3778 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3779 {
3780         enum addr_type_t type = ANYCAST_ADDR;
3781
3782         return inet6_dump_addr(skb, cb, type);
3783 }
3784
3785 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3786                              void *arg)
3787 {
3788         struct net *net = sock_net(in_skb->sk);
3789         struct ifaddrmsg *ifm;
3790         struct nlattr *tb[IFA_MAX+1];
3791         struct in6_addr *addr = NULL;
3792         struct net_device *dev = NULL;
3793         struct inet6_ifaddr *ifa;
3794         struct sk_buff *skb;
3795         int err;
3796
3797         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3798         if (err < 0)
3799                 goto errout;
3800
3801         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3802         if (addr == NULL) {
3803                 err = -EINVAL;
3804                 goto errout;
3805         }
3806
3807         ifm = nlmsg_data(nlh);
3808         if (ifm->ifa_index)
3809                 dev = __dev_get_by_index(net, ifm->ifa_index);
3810
3811         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3812         if (!ifa) {
3813                 err = -EADDRNOTAVAIL;
3814                 goto errout;
3815         }
3816
3817         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3818         if (!skb) {
3819                 err = -ENOBUFS;
3820                 goto errout_ifa;
3821         }
3822
3823         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3824                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3825         if (err < 0) {
3826                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3827                 WARN_ON(err == -EMSGSIZE);
3828                 kfree_skb(skb);
3829                 goto errout_ifa;
3830         }
3831         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3832 errout_ifa:
3833         in6_ifa_put(ifa);
3834 errout:
3835         return err;
3836 }
3837
3838 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3839 {
3840         struct sk_buff *skb;
3841         struct net *net = dev_net(ifa->idev->dev);
3842         int err = -ENOBUFS;
3843
3844         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3845         if (skb == NULL)
3846                 goto errout;
3847
3848         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3849         if (err < 0) {
3850                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3851                 WARN_ON(err == -EMSGSIZE);
3852                 kfree_skb(skb);
3853                 goto errout;
3854         }
3855         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3856         return;
3857 errout:
3858         if (err < 0)
3859                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3860 }
3861
3862 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3863                                 __s32 *array, int bytes)
3864 {
3865         BUG_ON(bytes < (DEVCONF_MAX * 4));
3866
3867         memset(array, 0, bytes);
3868         array[DEVCONF_FORWARDING] = cnf->forwarding;
3869         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3870         array[DEVCONF_MTU6] = cnf->mtu6;
3871         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3872         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3873         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3874         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3875         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3876         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3877                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3878         array[DEVCONF_RTR_SOLICIT_DELAY] =
3879                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3880         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3881 #ifdef CONFIG_IPV6_PRIVACY
3882         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3883         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3884         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3885         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3886         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3887 #endif
3888         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3889         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3890         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3891 #ifdef CONFIG_IPV6_ROUTER_PREF
3892         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3893         array[DEVCONF_RTR_PROBE_INTERVAL] =
3894                 jiffies_to_msecs(cnf->rtr_probe_interval);
3895 #ifdef CONFIG_IPV6_ROUTE_INFO
3896         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3897 #endif
3898 #endif
3899         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3900         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3901 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3902         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3903 #endif
3904 #ifdef CONFIG_IPV6_MROUTE
3905         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3906 #endif
3907         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3908         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3909         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3910 }
3911
3912 static inline size_t inet6_ifla6_size(void)
3913 {
3914         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3915              + nla_total_size(sizeof(struct ifla_cacheinfo))
3916              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3917              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3918              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3919 }
3920
3921 static inline size_t inet6_if_nlmsg_size(void)
3922 {
3923         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3924                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3925                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3926                + nla_total_size(4) /* IFLA_MTU */
3927                + nla_total_size(4) /* IFLA_LINK */
3928                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3929 }
3930
3931 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3932                                       int items, int bytes)
3933 {
3934         int i;
3935         int pad = bytes - sizeof(u64) * items;
3936         BUG_ON(pad < 0);
3937
3938         /* Use put_unaligned() because stats may not be aligned for u64. */
3939         put_unaligned(items, &stats[0]);
3940         for (i = 1; i < items; i++)
3941                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3942
3943         memset(&stats[items], 0, pad);
3944 }
3945
3946 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3947                                       int items, int bytes, size_t syncpoff)
3948 {
3949         int i;
3950         int pad = bytes - sizeof(u64) * items;
3951         BUG_ON(pad < 0);
3952
3953         /* Use put_unaligned() because stats may not be aligned for u64. */
3954         put_unaligned(items, &stats[0]);
3955         for (i = 1; i < items; i++)
3956                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3957
3958         memset(&stats[items], 0, pad);
3959 }
3960
3961 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3962                              int bytes)
3963 {
3964         switch (attrtype) {
3965         case IFLA_INET6_STATS:
3966                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3967                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3968                 break;
3969         case IFLA_INET6_ICMP6STATS:
3970                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3971                 break;
3972         }
3973 }
3974
3975 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3976 {
3977         struct nlattr *nla;
3978         struct ifla_cacheinfo ci;
3979
3980         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
3981                 goto nla_put_failure;
3982         ci.max_reasm_len = IPV6_MAXPLEN;
3983         ci.tstamp = cstamp_delta(idev->tstamp);
3984         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3985         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3986         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
3987                 goto nla_put_failure;
3988         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3989         if (nla == NULL)
3990                 goto nla_put_failure;
3991         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3992
3993         /* XXX - MC not implemented */
3994
3995         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3996         if (nla == NULL)
3997                 goto nla_put_failure;
3998         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3999
4000         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4001         if (nla == NULL)
4002                 goto nla_put_failure;
4003         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4004
4005         return 0;
4006
4007 nla_put_failure:
4008         return -EMSGSIZE;
4009 }
4010
4011 static size_t inet6_get_link_af_size(const struct net_device *dev)
4012 {
4013         if (!__in6_dev_get(dev))
4014                 return 0;
4015
4016         return inet6_ifla6_size();
4017 }
4018
4019 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4020 {
4021         struct inet6_dev *idev = __in6_dev_get(dev);
4022
4023         if (!idev)
4024                 return -ENODATA;
4025
4026         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4027                 return -EMSGSIZE;
4028
4029         return 0;
4030 }
4031
4032 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4033                              u32 pid, u32 seq, int event, unsigned int flags)
4034 {
4035         struct net_device *dev = idev->dev;
4036         struct ifinfomsg *hdr;
4037         struct nlmsghdr *nlh;
4038         void *protoinfo;
4039
4040         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4041         if (nlh == NULL)
4042                 return -EMSGSIZE;
4043
4044         hdr = nlmsg_data(nlh);
4045         hdr->ifi_family = AF_INET6;
4046         hdr->__ifi_pad = 0;
4047         hdr->ifi_type = dev->type;
4048         hdr->ifi_index = dev->ifindex;
4049         hdr->ifi_flags = dev_get_flags(dev);
4050         hdr->ifi_change = 0;
4051
4052         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4053             (dev->addr_len &&
4054              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4055             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4056             (dev->ifindex != dev->iflink &&
4057              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4058                 goto nla_put_failure;
4059         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4060         if (protoinfo == NULL)
4061                 goto nla_put_failure;
4062
4063         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4064                 goto nla_put_failure;
4065
4066         nla_nest_end(skb, protoinfo);
4067         return nlmsg_end(skb, nlh);
4068
4069 nla_put_failure:
4070         nlmsg_cancel(skb, nlh);
4071         return -EMSGSIZE;
4072 }
4073
4074 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4075 {
4076         struct net *net = sock_net(skb->sk);
4077         int h, s_h;
4078         int idx = 0, s_idx;
4079         struct net_device *dev;
4080         struct inet6_dev *idev;
4081         struct hlist_head *head;
4082         struct hlist_node *node;
4083
4084         s_h = cb->args[0];
4085         s_idx = cb->args[1];
4086
4087         rcu_read_lock();
4088         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4089                 idx = 0;
4090                 head = &net->dev_index_head[h];
4091                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4092                         if (idx < s_idx)
4093                                 goto cont;
4094                         idev = __in6_dev_get(dev);
4095                         if (!idev)
4096                                 goto cont;
4097                         if (inet6_fill_ifinfo(skb, idev,
4098                                               NETLINK_CB(cb->skb).pid,
4099                                               cb->nlh->nlmsg_seq,
4100                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4101                                 goto out;
4102 cont:
4103                         idx++;
4104                 }
4105         }
4106 out:
4107         rcu_read_unlock();
4108         cb->args[1] = idx;
4109         cb->args[0] = h;
4110
4111         return skb->len;
4112 }
4113
4114 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4115 {
4116         struct sk_buff *skb;
4117         struct net *net = dev_net(idev->dev);
4118         int err = -ENOBUFS;
4119
4120         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4121         if (skb == NULL)
4122                 goto errout;
4123
4124         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4125         if (err < 0) {
4126                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4127                 WARN_ON(err == -EMSGSIZE);
4128                 kfree_skb(skb);
4129                 goto errout;
4130         }
4131         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4132         return;
4133 errout:
4134         if (err < 0)
4135                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4136 }
4137
4138 static inline size_t inet6_prefix_nlmsg_size(void)
4139 {
4140         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4141                + nla_total_size(sizeof(struct in6_addr))
4142                + nla_total_size(sizeof(struct prefix_cacheinfo));
4143 }
4144
4145 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4146                              struct prefix_info *pinfo, u32 pid, u32 seq,
4147                              int event, unsigned int flags)
4148 {
4149         struct prefixmsg *pmsg;
4150         struct nlmsghdr *nlh;
4151         struct prefix_cacheinfo ci;
4152
4153         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4154         if (nlh == NULL)
4155                 return -EMSGSIZE;
4156
4157         pmsg = nlmsg_data(nlh);
4158         pmsg->prefix_family = AF_INET6;
4159         pmsg->prefix_pad1 = 0;
4160         pmsg->prefix_pad2 = 0;
4161         pmsg->prefix_ifindex = idev->dev->ifindex;
4162         pmsg->prefix_len = pinfo->prefix_len;
4163         pmsg->prefix_type = pinfo->type;
4164         pmsg->prefix_pad3 = 0;
4165         pmsg->prefix_flags = 0;
4166         if (pinfo->onlink)
4167                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4168         if (pinfo->autoconf)
4169                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4170
4171         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4172                 goto nla_put_failure;
4173         ci.preferred_time = ntohl(pinfo->prefered);
4174         ci.valid_time = ntohl(pinfo->valid);
4175         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4176                 goto nla_put_failure;
4177         return nlmsg_end(skb, nlh);
4178
4179 nla_put_failure:
4180         nlmsg_cancel(skb, nlh);
4181         return -EMSGSIZE;
4182 }
4183
4184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4185                          struct prefix_info *pinfo)
4186 {
4187         struct sk_buff *skb;
4188         struct net *net = dev_net(idev->dev);
4189         int err = -ENOBUFS;
4190
4191         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4192         if (skb == NULL)
4193                 goto errout;
4194
4195         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4196         if (err < 0) {
4197                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4198                 WARN_ON(err == -EMSGSIZE);
4199                 kfree_skb(skb);
4200                 goto errout;
4201         }
4202         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4203         return;
4204 errout:
4205         if (err < 0)
4206                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4207 }
4208
4209 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4210 {
4211         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4212
4213         switch (event) {
4214         case RTM_NEWADDR:
4215                 /*
4216                  * If the address was optimistic
4217                  * we inserted the route at the start of
4218                  * our DAD process, so we don't need
4219                  * to do it again
4220                  */
4221                 if (!(ifp->rt->rt6i_node))
4222                         ip6_ins_rt(ifp->rt);
4223                 if (ifp->idev->cnf.forwarding)
4224                         addrconf_join_anycast(ifp);
4225                 break;
4226         case RTM_DELADDR:
4227                 if (ifp->idev->cnf.forwarding)
4228                         addrconf_leave_anycast(ifp);
4229                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4230                 dst_hold(&ifp->rt->dst);
4231
4232                 if (ip6_del_rt(ifp->rt))
4233                         dst_free(&ifp->rt->dst);
4234                 break;
4235         }
4236 }
4237
4238 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4239 {
4240         rcu_read_lock_bh();
4241         if (likely(ifp->idev->dead == 0))
4242                 __ipv6_ifa_notify(event, ifp);
4243         rcu_read_unlock_bh();
4244 }
4245
4246 #ifdef CONFIG_SYSCTL
4247
4248 static
4249 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4250                            void __user *buffer, size_t *lenp, loff_t *ppos)
4251 {
4252         int *valp = ctl->data;
4253         int val = *valp;
4254         loff_t pos = *ppos;
4255         ctl_table lctl;
4256         int ret;
4257
4258         /*
4259          * ctl->data points to idev->cnf.forwarding, we should
4260          * not modify it until we get the rtnl lock.
4261          */
4262         lctl = *ctl;
4263         lctl.data = &val;
4264
4265         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4266
4267         if (write)
4268                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4269         if (ret)
4270                 *ppos = pos;
4271         return ret;
4272 }
4273
4274 static void dev_disable_change(struct inet6_dev *idev)
4275 {
4276         if (!idev || !idev->dev)
4277                 return;
4278
4279         if (idev->cnf.disable_ipv6)
4280                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4281         else
4282                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4283 }
4284
4285 static void addrconf_disable_change(struct net *net, __s32 newf)
4286 {
4287         struct net_device *dev;
4288         struct inet6_dev *idev;
4289
4290         rcu_read_lock();
4291         for_each_netdev_rcu(net, dev) {
4292                 idev = __in6_dev_get(dev);
4293                 if (idev) {
4294                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4295                         idev->cnf.disable_ipv6 = newf;
4296                         if (changed)
4297                                 dev_disable_change(idev);
4298                 }
4299         }
4300         rcu_read_unlock();
4301 }
4302
4303 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4304 {
4305         struct net *net;
4306         int old;
4307
4308         if (!rtnl_trylock())
4309                 return restart_syscall();
4310
4311         net = (struct net *)table->extra2;
4312         old = *p;
4313         *p = newf;
4314
4315         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4316                 rtnl_unlock();
4317                 return 0;
4318         }
4319
4320         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4321                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4322                 addrconf_disable_change(net, newf);
4323         } else if ((!newf) ^ (!old))
4324                 dev_disable_change((struct inet6_dev *)table->extra1);
4325
4326         rtnl_unlock();
4327         return 0;
4328 }
4329
4330 static
4331 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4332                             void __user *buffer, size_t *lenp, loff_t *ppos)
4333 {
4334         int *valp = ctl->data;
4335         int val = *valp;
4336         loff_t pos = *ppos;
4337         ctl_table lctl;
4338         int ret;
4339
4340         /*
4341          * ctl->data points to idev->cnf.disable_ipv6, we should
4342          * not modify it until we get the rtnl lock.
4343          */
4344         lctl = *ctl;
4345         lctl.data = &val;
4346
4347         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4348
4349         if (write)
4350                 ret = addrconf_disable_ipv6(ctl, valp, val);
4351         if (ret)
4352                 *ppos = pos;
4353         return ret;
4354 }
4355
4356 static struct addrconf_sysctl_table
4357 {
4358         struct ctl_table_header *sysctl_header;
4359         ctl_table addrconf_vars[DEVCONF_MAX+1];
4360 } addrconf_sysctl __read_mostly = {
4361         .sysctl_header = NULL,
4362         .addrconf_vars = {
4363                 {
4364                         .procname       = "forwarding",
4365                         .data           = &ipv6_devconf.forwarding,
4366                         .maxlen         = sizeof(int),
4367                         .mode           = 0644,
4368                         .proc_handler   = addrconf_sysctl_forward,
4369                 },
4370                 {
4371                         .procname       = "hop_limit",
4372                         .data           = &ipv6_devconf.hop_limit,
4373                         .maxlen         = sizeof(int),
4374                         .mode           = 0644,
4375                         .proc_handler   = proc_dointvec,
4376                 },
4377                 {
4378                         .procname       = "mtu",
4379                         .data           = &ipv6_devconf.mtu6,
4380                         .maxlen         = sizeof(int),
4381                         .mode           = 0644,
4382                         .proc_handler   = proc_dointvec,
4383                 },
4384                 {
4385                         .procname       = "accept_ra",
4386                         .data           = &ipv6_devconf.accept_ra,
4387                         .maxlen         = sizeof(int),
4388                         .mode           = 0644,
4389                         .proc_handler   = proc_dointvec,
4390                 },
4391                 {
4392                         .procname       = "accept_redirects",
4393                         .data           = &ipv6_devconf.accept_redirects,
4394                         .maxlen         = sizeof(int),
4395                         .mode           = 0644,
4396                         .proc_handler   = proc_dointvec,
4397                 },
4398                 {
4399                         .procname       = "autoconf",
4400                         .data           = &ipv6_devconf.autoconf,
4401                         .maxlen         = sizeof(int),
4402                         .mode           = 0644,
4403                         .proc_handler   = proc_dointvec,
4404                 },
4405                 {
4406                         .procname       = "dad_transmits",
4407                         .data           = &ipv6_devconf.dad_transmits,
4408                         .maxlen         = sizeof(int),
4409                         .mode           = 0644,
4410                         .proc_handler   = proc_dointvec,
4411                 },
4412                 {
4413                         .procname       = "router_solicitations",
4414                         .data           = &ipv6_devconf.rtr_solicits,
4415                         .maxlen         = sizeof(int),
4416                         .mode           = 0644,
4417                         .proc_handler   = proc_dointvec,
4418                 },
4419                 {
4420                         .procname       = "router_solicitation_interval",
4421                         .data           = &ipv6_devconf.rtr_solicit_interval,
4422                         .maxlen         = sizeof(int),
4423                         .mode           = 0644,
4424                         .proc_handler   = proc_dointvec_jiffies,
4425                 },
4426                 {
4427                         .procname       = "router_solicitation_delay",
4428                         .data           = &ipv6_devconf.rtr_solicit_delay,
4429                         .maxlen         = sizeof(int),
4430                         .mode           = 0644,
4431                         .proc_handler   = proc_dointvec_jiffies,
4432                 },
4433                 {
4434                         .procname       = "force_mld_version",
4435                         .data           = &ipv6_devconf.force_mld_version,
4436                         .maxlen         = sizeof(int),
4437                         .mode           = 0644,
4438                         .proc_handler   = proc_dointvec,
4439                 },
4440 #ifdef CONFIG_IPV6_PRIVACY
4441                 {
4442                         .procname       = "use_tempaddr",
4443                         .data           = &ipv6_devconf.use_tempaddr,
4444                         .maxlen         = sizeof(int),
4445                         .mode           = 0644,
4446                         .proc_handler   = proc_dointvec,
4447                 },
4448                 {
4449                         .procname       = "temp_valid_lft",
4450                         .data           = &ipv6_devconf.temp_valid_lft,
4451                         .maxlen         = sizeof(int),
4452                         .mode           = 0644,
4453                         .proc_handler   = proc_dointvec,
4454                 },
4455                 {
4456                         .procname       = "temp_prefered_lft",
4457                         .data           = &ipv6_devconf.temp_prefered_lft,
4458                         .maxlen         = sizeof(int),
4459                         .mode           = 0644,
4460                         .proc_handler   = proc_dointvec,
4461                 },
4462                 {
4463                         .procname       = "regen_max_retry",
4464                         .data           = &ipv6_devconf.regen_max_retry,
4465                         .maxlen         = sizeof(int),
4466                         .mode           = 0644,
4467                         .proc_handler   = proc_dointvec,
4468                 },
4469                 {
4470                         .procname       = "max_desync_factor",
4471                         .data           = &ipv6_devconf.max_desync_factor,
4472                         .maxlen         = sizeof(int),
4473                         .mode           = 0644,
4474                         .proc_handler   = proc_dointvec,
4475                 },
4476 #endif
4477                 {
4478                         .procname       = "max_addresses",
4479                         .data           = &ipv6_devconf.max_addresses,
4480                         .maxlen         = sizeof(int),
4481                         .mode           = 0644,
4482                         .proc_handler   = proc_dointvec,
4483                 },
4484                 {
4485                         .procname       = "accept_ra_defrtr",
4486                         .data           = &ipv6_devconf.accept_ra_defrtr,
4487                         .maxlen         = sizeof(int),
4488                         .mode           = 0644,
4489                         .proc_handler   = proc_dointvec,
4490                 },
4491                 {
4492                         .procname       = "accept_ra_pinfo",
4493                         .data           = &ipv6_devconf.accept_ra_pinfo,
4494                         .maxlen         = sizeof(int),
4495                         .mode           = 0644,
4496                         .proc_handler   = proc_dointvec,
4497                 },
4498 #ifdef CONFIG_IPV6_ROUTER_PREF
4499                 {
4500                         .procname       = "accept_ra_rtr_pref",
4501                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4502                         .maxlen         = sizeof(int),
4503                         .mode           = 0644,
4504                         .proc_handler   = proc_dointvec,
4505                 },
4506                 {
4507                         .procname       = "router_probe_interval",
4508                         .data           = &ipv6_devconf.rtr_probe_interval,
4509                         .maxlen         = sizeof(int),
4510                         .mode           = 0644,
4511                         .proc_handler   = proc_dointvec_jiffies,
4512                 },
4513 #ifdef CONFIG_IPV6_ROUTE_INFO
4514                 {
4515                         .procname       = "accept_ra_rt_info_max_plen",
4516                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4517                         .maxlen         = sizeof(int),
4518                         .mode           = 0644,
4519                         .proc_handler   = proc_dointvec,
4520                 },
4521 #endif
4522 #endif
4523                 {
4524                         .procname       = "proxy_ndp",
4525                         .data           = &ipv6_devconf.proxy_ndp,
4526                         .maxlen         = sizeof(int),
4527                         .mode           = 0644,
4528                         .proc_handler   = proc_dointvec,
4529                 },
4530                 {
4531                         .procname       = "accept_source_route",
4532                         .data           = &ipv6_devconf.accept_source_route,
4533                         .maxlen         = sizeof(int),
4534                         .mode           = 0644,
4535                         .proc_handler   = proc_dointvec,
4536                 },
4537 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4538                 {
4539                         .procname       = "optimistic_dad",
4540                         .data           = &ipv6_devconf.optimistic_dad,
4541                         .maxlen         = sizeof(int),
4542                         .mode           = 0644,
4543                         .proc_handler   = proc_dointvec,
4544
4545                 },
4546 #endif
4547 #ifdef CONFIG_IPV6_MROUTE
4548                 {
4549                         .procname       = "mc_forwarding",
4550                         .data           = &ipv6_devconf.mc_forwarding,
4551                         .maxlen         = sizeof(int),
4552                         .mode           = 0444,
4553                         .proc_handler   = proc_dointvec,
4554                 },
4555 #endif
4556                 {
4557                         .procname       = "disable_ipv6",
4558                         .data           = &ipv6_devconf.disable_ipv6,
4559                         .maxlen         = sizeof(int),
4560                         .mode           = 0644,
4561                         .proc_handler   = addrconf_sysctl_disable,
4562                 },
4563                 {
4564                         .procname       = "accept_dad",
4565                         .data           = &ipv6_devconf.accept_dad,
4566                         .maxlen         = sizeof(int),
4567                         .mode           = 0644,
4568                         .proc_handler   = proc_dointvec,
4569                 },
4570                 {
4571                         .procname       = "force_tllao",
4572                         .data           = &ipv6_devconf.force_tllao,
4573                         .maxlen         = sizeof(int),
4574                         .mode           = 0644,
4575                         .proc_handler   = proc_dointvec
4576                 },
4577                 {
4578                         /* sentinel */
4579                 }
4580         },
4581 };
4582
4583 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4584                 struct inet6_dev *idev, struct ipv6_devconf *p)
4585 {
4586         int i;
4587         struct addrconf_sysctl_table *t;
4588         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4589
4590         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4591         if (t == NULL)
4592                 goto out;
4593
4594         for (i = 0; t->addrconf_vars[i].data; i++) {
4595                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4596                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4597                 t->addrconf_vars[i].extra2 = net;
4598         }
4599
4600         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4601
4602         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4603         if (t->sysctl_header == NULL)
4604                 goto free;
4605
4606         p->sysctl = t;
4607         return 0;
4608
4609 free:
4610         kfree(t);
4611 out:
4612         return -ENOBUFS;
4613 }
4614
4615 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4616 {
4617         struct addrconf_sysctl_table *t;
4618
4619         if (p->sysctl == NULL)
4620                 return;
4621
4622         t = p->sysctl;
4623         p->sysctl = NULL;
4624         unregister_net_sysctl_table(t->sysctl_header);
4625         kfree(t);
4626 }
4627
4628 static void addrconf_sysctl_register(struct inet6_dev *idev)
4629 {
4630         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4631                               &ndisc_ifinfo_sysctl_change);
4632         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4633                                         idev, &idev->cnf);
4634 }
4635
4636 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4637 {
4638         __addrconf_sysctl_unregister(&idev->cnf);
4639         neigh_sysctl_unregister(idev->nd_parms);
4640 }
4641
4642
4643 #endif
4644
4645 static int __net_init addrconf_init_net(struct net *net)
4646 {
4647         int err;
4648         struct ipv6_devconf *all, *dflt;
4649
4650         err = -ENOMEM;
4651         all = &ipv6_devconf;
4652         dflt = &ipv6_devconf_dflt;
4653
4654         if (!net_eq(net, &init_net)) {
4655                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4656                 if (all == NULL)
4657                         goto err_alloc_all;
4658
4659                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4660                 if (dflt == NULL)
4661                         goto err_alloc_dflt;
4662         } else {
4663                 /* these will be inherited by all namespaces */
4664                 dflt->autoconf = ipv6_defaults.autoconf;
4665                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4666         }
4667
4668         net->ipv6.devconf_all = all;
4669         net->ipv6.devconf_dflt = dflt;
4670
4671 #ifdef CONFIG_SYSCTL
4672         err = __addrconf_sysctl_register(net, "all", NULL, all);
4673         if (err < 0)
4674                 goto err_reg_all;
4675
4676         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4677         if (err < 0)
4678                 goto err_reg_dflt;
4679 #endif
4680         return 0;
4681
4682 #ifdef CONFIG_SYSCTL
4683 err_reg_dflt:
4684         __addrconf_sysctl_unregister(all);
4685 err_reg_all:
4686         kfree(dflt);
4687 #endif
4688 err_alloc_dflt:
4689         kfree(all);
4690 err_alloc_all:
4691         return err;
4692 }
4693
4694 static void __net_exit addrconf_exit_net(struct net *net)
4695 {
4696 #ifdef CONFIG_SYSCTL
4697         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4698         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4699 #endif
4700         if (!net_eq(net, &init_net)) {
4701                 kfree(net->ipv6.devconf_dflt);
4702                 kfree(net->ipv6.devconf_all);
4703         }
4704 }
4705
4706 static struct pernet_operations addrconf_ops = {
4707         .init = addrconf_init_net,
4708         .exit = addrconf_exit_net,
4709 };
4710
4711 /*
4712  *      Device notifier
4713  */
4714
4715 int register_inet6addr_notifier(struct notifier_block *nb)
4716 {
4717         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4718 }
4719 EXPORT_SYMBOL(register_inet6addr_notifier);
4720
4721 int unregister_inet6addr_notifier(struct notifier_block *nb)
4722 {
4723         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4724 }
4725 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4726
4727 static struct rtnl_af_ops inet6_ops = {
4728         .family           = AF_INET6,
4729         .fill_link_af     = inet6_fill_link_af,
4730         .get_link_af_size = inet6_get_link_af_size,
4731 };
4732
4733 /*
4734  *      Init / cleanup code
4735  */
4736
4737 int __init addrconf_init(void)
4738 {
4739         int i, err;
4740
4741         err = ipv6_addr_label_init();
4742         if (err < 0) {
4743                 pr_crit("%s: cannot initialize default policy table: %d\n",
4744                         __func__, err);
4745                 goto out;
4746         }
4747
4748         err = register_pernet_subsys(&addrconf_ops);
4749         if (err < 0)
4750                 goto out_addrlabel;
4751
4752         /* The addrconf netdev notifier requires that loopback_dev
4753          * has it's ipv6 private information allocated and setup
4754          * before it can bring up and give link-local addresses
4755          * to other devices which are up.
4756          *
4757          * Unfortunately, loopback_dev is not necessarily the first
4758          * entry in the global dev_base list of net devices.  In fact,
4759          * it is likely to be the very last entry on that list.
4760          * So this causes the notifier registry below to try and
4761          * give link-local addresses to all devices besides loopback_dev
4762          * first, then loopback_dev, which cases all the non-loopback_dev
4763          * devices to fail to get a link-local address.
4764          *
4765          * So, as a temporary fix, allocate the ipv6 structure for
4766          * loopback_dev first by hand.
4767          * Longer term, all of the dependencies ipv6 has upon the loopback
4768          * device and it being up should be removed.
4769          */
4770         rtnl_lock();
4771         if (!ipv6_add_dev(init_net.loopback_dev))
4772                 err = -ENOMEM;
4773         rtnl_unlock();
4774         if (err)
4775                 goto errlo;
4776
4777         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4778                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4779
4780         register_netdevice_notifier(&ipv6_dev_notf);
4781
4782         addrconf_verify(0);
4783
4784         err = rtnl_af_register(&inet6_ops);
4785         if (err < 0)
4786                 goto errout_af;
4787
4788         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4789                               NULL);
4790         if (err < 0)
4791                 goto errout;
4792
4793         /* Only the first call to __rtnl_register can fail */
4794         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4795         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4796         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4797                         inet6_dump_ifaddr, NULL);
4798         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4799                         inet6_dump_ifmcaddr, NULL);
4800         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4801                         inet6_dump_ifacaddr, NULL);
4802
4803         ipv6_addr_label_rtnl_register();
4804
4805         return 0;
4806 errout:
4807         rtnl_af_unregister(&inet6_ops);
4808 errout_af:
4809         unregister_netdevice_notifier(&ipv6_dev_notf);
4810 errlo:
4811         unregister_pernet_subsys(&addrconf_ops);
4812 out_addrlabel:
4813         ipv6_addr_label_cleanup();
4814 out:
4815         return err;
4816 }
4817
4818 void addrconf_cleanup(void)
4819 {
4820         struct net_device *dev;
4821         int i;
4822
4823         unregister_netdevice_notifier(&ipv6_dev_notf);
4824         unregister_pernet_subsys(&addrconf_ops);
4825         ipv6_addr_label_cleanup();
4826
4827         rtnl_lock();
4828
4829         __rtnl_af_unregister(&inet6_ops);
4830
4831         /* clean dev list */
4832         for_each_netdev(&init_net, dev) {
4833                 if (__in6_dev_get(dev) == NULL)
4834                         continue;
4835                 addrconf_ifdown(dev, 1);
4836         }
4837         addrconf_ifdown(init_net.loopback_dev, 2);
4838
4839         /*
4840          *      Check hash table.
4841          */
4842         spin_lock_bh(&addrconf_hash_lock);
4843         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4844                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4845         spin_unlock_bh(&addrconf_hash_lock);
4846
4847         del_timer(&addr_chk_timer);
4848         rtnl_unlock();
4849 }