net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90 #include <linux/export.h>
  91
  92 /* Set to 3 to get tracing... */
  93 #define ACONF_DEBUG 2
  94
  95 #if ACONF_DEBUG >= 3
  96 #define ADBG(x) printk x
  97 #else
  98 #define ADBG(x)
  99 #endif
 100
 101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 102
 103 static inline u32 cstamp_delta(unsigned long cstamp)
 104 {
 105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 106 }
 107
 108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 111
 112 #ifdef CONFIG_SYSCTL
 113 static void addrconf_sysctl_register(struct inet6_dev *idev);
 114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 115 #else
 116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 117 {
 118 }
 119
 120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 121 {
 122 }
 123 #endif
 124
 125 #ifdef CONFIG_IPV6_PRIVACY
 126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 128 static void ipv6_regen_rndid(unsigned long data);
 129 #endif
 130
 131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 132 static int ipv6_count_addresses(struct inet6_dev *idev);
 133
 134 /*
 135  *      Configured unicast address hash table
 136  */
 137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 138 static DEFINE_SPINLOCK(addrconf_hash_lock);
 139
 140 static void addrconf_verify(unsigned long);
 141
 142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 143 static DEFINE_SPINLOCK(addrconf_verify_lock);
 144
 145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 147
 148 static void addrconf_type_change(struct net_device *dev,
 149                                  unsigned long event);
 150 static int addrconf_ifdown(struct net_device *dev, int how);
 151
 152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 153 static void addrconf_dad_timer(unsigned long data);
 154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 155 static void addrconf_dad_run(struct inet6_dev *idev);
 156 static void addrconf_rs_timer(unsigned long data);
 157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159
 160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 161                                 struct prefix_info *pinfo);
 162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 163                                struct net_device *dev);
 164
 165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 166
 167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 168         .forwarding             = 0,
 169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 170         .mtu6                   = IPV6_MIN_MTU,
 171         .accept_ra              = 1,
 172         .accept_redirects       = 1,
 173         .autoconf               = 1,
 174         .force_mld_version      = 0,
 175         .dad_transmits          = 1,
 176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 179 #ifdef CONFIG_IPV6_PRIVACY
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185 #endif
 186         .max_addresses          = IPV6_MAX_ADDRESSES,
 187         .accept_ra_defrtr       = 1,
 188         .accept_ra_pinfo        = 1,
 189 #ifdef CONFIG_IPV6_ROUTER_PREF
 190         .accept_ra_rtr_pref     = 1,
 191         .rtr_probe_interval     = 60 * HZ,
 192 #ifdef CONFIG_IPV6_ROUTE_INFO
 193         .accept_ra_rt_info_max_plen = 0,
 194 #endif
 195 #endif
 196         .proxy_ndp              = 0,
 197         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 198         .disable_ipv6           = 0,
 199         .accept_dad             = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .dad_transmits          = 1,
 210         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 211         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 212         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 213 #ifdef CONFIG_IPV6_PRIVACY
 214         .use_tempaddr           = 0,
 215         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 216         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 217         .regen_max_retry        = REGEN_MAX_RETRY,
 218         .max_desync_factor      = MAX_DESYNC_FACTOR,
 219 #endif
 220         .max_addresses          = IPV6_MAX_ADDRESSES,
 221         .accept_ra_defrtr       = 1,
 222         .accept_ra_pinfo        = 1,
 223 #ifdef CONFIG_IPV6_ROUTER_PREF
 224         .accept_ra_rtr_pref     = 1,
 225         .rtr_probe_interval     = 60 * HZ,
 226 #ifdef CONFIG_IPV6_ROUTE_INFO
 227         .accept_ra_rt_info_max_plen = 0,
 228 #endif
 229 #endif
 230         .proxy_ndp              = 0,
 231         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 232         .disable_ipv6           = 0,
 233         .accept_dad             = 1,
 234 };
 235
 236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 241
 242 /* Check if a valid qdisc is available */
 243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 244 {
 245         return !qdisc_tx_is_noop(dev);
 246 }
 247
 248 /* Check if a route is valid prefix route */
 249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 250 {
 251         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 252 }
 253
 254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 255 {
 256         if (del_timer(&ifp->timer))
 257                 __in6_ifa_put(ifp);
 258 }
 259
 260 enum addrconf_timer_t {
 261         AC_NONE,
 262         AC_DAD,
 263         AC_RS,
 264 };
 265
 266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 267                                enum addrconf_timer_t what,
 268                                unsigned long when)
 269 {
 270         if (!del_timer(&ifp->timer))
 271                 in6_ifa_hold(ifp);
 272
 273         switch (what) {
 274         case AC_DAD:
 275                 ifp->timer.function = addrconf_dad_timer;
 276                 break;
 277         case AC_RS:
 278                 ifp->timer.function = addrconf_rs_timer;
 279                 break;
 280         default:
 281                 break;
 282         }
 283         ifp->timer.expires = jiffies + when;
 284         add_timer(&ifp->timer);
 285 }
 286
 287 static int snmp6_alloc_dev(struct inet6_dev *idev)
 288 {
 289         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 290                           sizeof(struct ipstats_mib),
 291                           __alignof__(struct ipstats_mib)) < 0)
 292                 goto err_ip;
 293         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 294                                         GFP_KERNEL);
 295         if (!idev->stats.icmpv6dev)
 296                 goto err_icmp;
 297         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 298                                            GFP_KERNEL);
 299         if (!idev->stats.icmpv6msgdev)
 300                 goto err_icmpmsg;
 301
 302         return 0;
 303
 304 err_icmpmsg:
 305         kfree(idev->stats.icmpv6dev);
 306 err_icmp:
 307         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 308 err_ip:
 309         return -ENOMEM;
 310 }
 311
 312 static void snmp6_free_dev(struct inet6_dev *idev)
 313 {
 314         kfree(idev->stats.icmpv6msgdev);
 315         kfree(idev->stats.icmpv6dev);
 316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 317 }
 318
 319 /* Nobody refers to this device, we may destroy it. */
 320
 321 void in6_dev_finish_destroy(struct inet6_dev *idev)
 322 {
 323         struct net_device *dev = idev->dev;
 324
 325         WARN_ON(!list_empty(&idev->addr_list));
 326         WARN_ON(idev->mc_list != NULL);
 327
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 pr_warning("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         kfree_rcu(idev, rcu);
 338 }
 339
 340 EXPORT_SYMBOL(in6_dev_finish_destroy);
 341
 342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 343 {
 344         struct inet6_dev *ndev;
 345
 346         ASSERT_RTNL();
 347
 348         if (dev->mtu < IPV6_MIN_MTU)
 349                 return NULL;
 350
 351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 352
 353         if (ndev == NULL)
 354                 return NULL;
 355
 356         rwlock_init(&ndev->lock);
 357         ndev->dev = dev;
 358         INIT_LIST_HEAD(&ndev->addr_list);
 359
 360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (ndev->nd_parms == NULL) {
 365                 kfree(ndev);
 366                 return NULL;
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 dev_put(dev);
 379                 kfree(ndev);
 380                 return NULL;
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG((KERN_WARNING
 385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name));
 387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 388                 ndev->dead = 1;
 389                 in6_dev_finish_destroy(ndev);
 390                 return NULL;
 391         }
 392
 393         /* One reference from device.  We must do this before
 394          * we invoke __ipv6_regen_rndid().
 395          */
 396         in6_dev_hold(ndev);
 397
 398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 399                 ndev->cnf.accept_dad = -1;
 400
 401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Multicast RS\n",
 405                        dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 ndev->cnf.use_tempaddr = -1;
 419         } else {
 420                 in6_dev_hold(ndev);
 421                 ipv6_regen_rndid((unsigned long) ndev);
 422         }
 423 #endif
 424
 425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 426                 ndev->if_flags |= IF_READY;
 427
 428         ipv6_mc_init_dev(ndev);
 429         ndev->tstamp = jiffies;
 430         addrconf_sysctl_register(ndev);
 431         /* protected by rtnl_lock */
 432         rcu_assign_pointer(dev->ip6_ptr, ndev);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         /* Join all-router multicast group if forwarding is set */
 438         if (ndev->cnf.forwarding && dev && (dev->flags & IFF_MULTICAST))
 439                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 #ifdef CONFIG_SYSCTL
 463 static void dev_forward_change(struct inet6_dev *idev)
 464 {
 465         struct net_device *dev;
 466         struct inet6_ifaddr *ifa;
 467
 468         if (!idev)
 469                 return;
 470         dev = idev->dev;
 471         if (idev->cnf.forwarding)
 472                 dev_disable_lro(dev);
 473         if (dev && (dev->flags & IFF_MULTICAST)) {
 474                 if (idev->cnf.forwarding)
 475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 476                 else
 477                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 478         }
 479
 480         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 481                 if (ifa->flags&IFA_F_TENTATIVE)
 482                         continue;
 483                 if (idev->cnf.forwarding)
 484                         addrconf_join_anycast(ifa);
 485                 else
 486                         addrconf_leave_anycast(ifa);
 487         }
 488 }
 489
 490
 491 static void addrconf_forward_change(struct net *net, __s32 newf)
 492 {
 493         struct net_device *dev;
 494         struct inet6_dev *idev;
 495
 496         rcu_read_lock();
 497         for_each_netdev_rcu(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         rcu_read_unlock();
 507 }
 508
 509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 510 {
 511         struct net *net;
 512
 513         net = (struct net *)table->extra2;
 514         if (p == &net->ipv6.devconf_dflt->forwarding)
 515                 return 0;
 516
 517         if (!rtnl_trylock()) {
 518                 /* Restore the original values before restarting */
 519                 *p = old;
 520                 return restart_syscall();
 521         }
 522
 523         if (p == &net->ipv6.devconf_all->forwarding) {
 524                 __s32 newf = net->ipv6.devconf_all->forwarding;
 525                 net->ipv6.devconf_dflt->forwarding = newf;
 526                 addrconf_forward_change(net, newf);
 527         } else if ((!*p) ^ (!old))
 528                 dev_forward_change((struct inet6_dev *)table->extra1);
 529         rtnl_unlock();
 530
 531         if (*p)
 532                 rt6_purge_dflt_routers(net);
 533         return 1;
 534 }
 535 #endif
 536
 537 /* Nobody refers to this ifaddr, destroy it */
 538 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 539 {
 540         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 541
 542 #ifdef NET_REFCNT_DEBUG
 543         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 544 #endif
 545
 546         in6_dev_put(ifp->idev);
 547
 548         if (del_timer(&ifp->timer))
 549                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 550
 551         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 552                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 553                 return;
 554         }
 555         dst_release(&ifp->rt->dst);
 556
 557         kfree_rcu(ifp, rcu);
 558 }
 559
 560 static void
 561 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 562 {
 563         struct list_head *p;
 564         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 565
 566         /*
 567          * Each device address list is sorted in order of scope -
 568          * global before linklocal.
 569          */
 570         list_for_each(p, &idev->addr_list) {
 571                 struct inet6_ifaddr *ifa
 572                         = list_entry(p, struct inet6_ifaddr, if_list);
 573                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 574                         break;
 575         }
 576
 577         list_add_tail(&ifp->if_list, p);
 578 }
 579
 580 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 581 {
 582         /*
 583          * We perform the hash function over the last 64 bits of the address
 584          * This will include the IEEE address token on links that support it.
 585          */
 586         return jhash_2words((__force u32)addr->s6_addr32[2],
 587                             (__force u32)addr->s6_addr32[3], 0)
 588                 & (IN6_ADDR_HSIZE - 1);
 589 }
 590
 591 /* On success it returns ifp with increased reference count */
 592
 593 static struct inet6_ifaddr *
 594 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 595               int scope, u32 flags)
 596 {
 597         struct inet6_ifaddr *ifa = NULL;
 598         struct rt6_info *rt;
 599         unsigned int hash;
 600         int err = 0;
 601         int addr_type = ipv6_addr_type(addr);
 602
 603         if (addr_type == IPV6_ADDR_ANY ||
 604             addr_type & IPV6_ADDR_MULTICAST ||
 605             (!(idev->dev->flags & IFF_LOOPBACK) &&
 606              addr_type & IPV6_ADDR_LOOPBACK))
 607                 return ERR_PTR(-EADDRNOTAVAIL);
 608
 609         rcu_read_lock_bh();
 610         if (idev->dead) {
 611                 err = -ENODEV;                  /*XXX*/
 612                 goto out2;
 613         }
 614
 615         if (idev->cnf.disable_ipv6) {
 616                 err = -EACCES;
 617                 goto out2;
 618         }
 619
 620         spin_lock(&addrconf_hash_lock);
 621
 622         /* Ignore adding duplicate addresses on an interface */
 623         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 624                 ADBG(("ipv6_add_addr: already assigned\n"));
 625                 err = -EEXIST;
 626                 goto out;
 627         }
 628
 629         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 630
 631         if (ifa == NULL) {
 632                 ADBG(("ipv6_add_addr: malloc failed\n"));
 633                 err = -ENOBUFS;
 634                 goto out;
 635         }
 636
 637         rt = addrconf_dst_alloc(idev, addr, 0);
 638         if (IS_ERR(rt)) {
 639                 err = PTR_ERR(rt);
 640                 goto out;
 641         }
 642
 643         ipv6_addr_copy(&ifa->addr, addr);
 644
 645         spin_lock_init(&ifa->lock);
 646         spin_lock_init(&ifa->state_lock);
 647         init_timer(&ifa->timer);
 648         INIT_HLIST_NODE(&ifa->addr_lst);
 649         ifa->timer.data = (unsigned long) ifa;
 650         ifa->scope = scope;
 651         ifa->prefix_len = pfxlen;
 652         ifa->flags = flags | IFA_F_TENTATIVE;
 653         ifa->cstamp = ifa->tstamp = jiffies;
 654
 655         ifa->rt = rt;
 656
 657         /*
 658          * part one of RFC 4429, section 3.3
 659          * We should not configure an address as
 660          * optimistic if we do not yet know the link
 661          * layer address of our nexhop router
 662          */
 663
 664         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 665                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 666
 667         ifa->idev = idev;
 668         in6_dev_hold(idev);
 669         /* For caller */
 670         in6_ifa_hold(ifa);
 671
 672         /* Add to big hash table */
 673         hash = ipv6_addr_hash(addr);
 674
 675         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 676         spin_unlock(&addrconf_hash_lock);
 677
 678         write_lock(&idev->lock);
 679         /* Add to inet6_dev unicast addr list. */
 680         ipv6_link_dev_addr(idev, ifa);
 681
 682 #ifdef CONFIG_IPV6_PRIVACY
 683         if (ifa->flags&IFA_F_TEMPORARY) {
 684                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 685                 in6_ifa_hold(ifa);
 686         }
 687 #endif
 688
 689         in6_ifa_hold(ifa);
 690         write_unlock(&idev->lock);
 691 out2:
 692         rcu_read_unlock_bh();
 693
 694         if (likely(err == 0))
 695                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 696         else {
 697                 kfree(ifa);
 698                 ifa = ERR_PTR(err);
 699         }
 700
 701         return ifa;
 702 out:
 703         spin_unlock(&addrconf_hash_lock);
 704         goto out2;
 705 }
 706
 707 /* This function wants to get referenced ifp and releases it before return */
 708
 709 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 710 {
 711         struct inet6_ifaddr *ifa, *ifn;
 712         struct inet6_dev *idev = ifp->idev;
 713         int state;
 714         int deleted = 0, onlink = 0;
 715         unsigned long expires = jiffies;
 716
 717         spin_lock_bh(&ifp->state_lock);
 718         state = ifp->state;
 719         ifp->state = INET6_IFADDR_STATE_DEAD;
 720         spin_unlock_bh(&ifp->state_lock);
 721
 722         if (state == INET6_IFADDR_STATE_DEAD)
 723                 goto out;
 724
 725         spin_lock_bh(&addrconf_hash_lock);
 726         hlist_del_init_rcu(&ifp->addr_lst);
 727         spin_unlock_bh(&addrconf_hash_lock);
 728
 729         write_lock_bh(&idev->lock);
 730 #ifdef CONFIG_IPV6_PRIVACY
 731         if (ifp->flags&IFA_F_TEMPORARY) {
 732                 list_del(&ifp->tmp_list);
 733                 if (ifp->ifpub) {
 734                         in6_ifa_put(ifp->ifpub);
 735                         ifp->ifpub = NULL;
 736                 }
 737                 __in6_ifa_put(ifp);
 738         }
 739 #endif
 740
 741         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 742                 if (ifa == ifp) {
 743                         list_del_init(&ifp->if_list);
 744                         __in6_ifa_put(ifp);
 745
 746                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 747                                 break;
 748                         deleted = 1;
 749                         continue;
 750                 } else if (ifp->flags & IFA_F_PERMANENT) {
 751                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 752                                               ifp->prefix_len)) {
 753                                 if (ifa->flags & IFA_F_PERMANENT) {
 754                                         onlink = 1;
 755                                         if (deleted)
 756                                                 break;
 757                                 } else {
 758                                         unsigned long lifetime;
 759
 760                                         if (!onlink)
 761                                                 onlink = -1;
 762
 763                                         spin_lock(&ifa->lock);
 764
 765                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 766                                         /*
 767                                          * Note: Because this address is
 768                                          * not permanent, lifetime <
 769                                          * LONG_MAX / HZ here.
 770                                          */
 771                                         if (time_before(expires,
 772                                                         ifa->tstamp + lifetime * HZ))
 773                                                 expires = ifa->tstamp + lifetime * HZ;
 774                                         spin_unlock(&ifa->lock);
 775                                 }
 776                         }
 777                 }
 778         }
 779         write_unlock_bh(&idev->lock);
 780
 781         addrconf_del_timer(ifp);
 782
 783         ipv6_ifa_notify(RTM_DELADDR, ifp);
 784
 785         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 786
 787         /*
 788          * Purge or update corresponding prefix
 789          *
 790          * 1) we don't purge prefix here if address was not permanent.
 791          *    prefix is managed by its own lifetime.
 792          * 2) if there're no addresses, delete prefix.
 793          * 3) if there're still other permanent address(es),
 794          *    corresponding prefix is still permanent.
 795          * 4) otherwise, update prefix lifetime to the
 796          *    longest valid lifetime among the corresponding
 797          *    addresses on the device.
 798          *    Note: subsequent RA will update lifetime.
 799          *
 800          * --yoshfuji
 801          */
 802         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 803                 struct in6_addr prefix;
 804                 struct rt6_info *rt;
 805                 struct net *net = dev_net(ifp->idev->dev);
 806                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 807                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 808
 809                 if (rt && addrconf_is_prefix_route(rt)) {
 810                         if (onlink == 0) {
 811                                 ip6_del_rt(rt);
 812                                 rt = NULL;
 813                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 814                                 rt->rt6i_expires = expires;
 815                                 rt->rt6i_flags |= RTF_EXPIRES;
 816                         }
 817                 }
 818                 dst_release(&rt->dst);
 819         }
 820
 821         /* clean up prefsrc entries */
 822         rt6_remove_prefsrc(ifp);
 823 out:
 824         in6_ifa_put(ifp);
 825 }
 826
 827 #ifdef CONFIG_IPV6_PRIVACY
 828 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 829 {
 830         struct inet6_dev *idev = ifp->idev;
 831         struct in6_addr addr, *tmpaddr;
 832         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 833         unsigned long regen_advance;
 834         int tmp_plen;
 835         int ret = 0;
 836         int max_addresses;
 837         u32 addr_flags;
 838         unsigned long now = jiffies;
 839
 840         write_lock(&idev->lock);
 841         if (ift) {
 842                 spin_lock_bh(&ift->lock);
 843                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 844                 spin_unlock_bh(&ift->lock);
 845                 tmpaddr = &addr;
 846         } else {
 847                 tmpaddr = NULL;
 848         }
 849 retry:
 850         in6_dev_hold(idev);
 851         if (idev->cnf.use_tempaddr <= 0) {
 852                 write_unlock(&idev->lock);
 853                 printk(KERN_INFO
 854                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 855                 in6_dev_put(idev);
 856                 ret = -1;
 857                 goto out;
 858         }
 859         spin_lock_bh(&ifp->lock);
 860         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 861                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 862                 spin_unlock_bh(&ifp->lock);
 863                 write_unlock(&idev->lock);
 864                 printk(KERN_WARNING
 865                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 866                 in6_dev_put(idev);
 867                 ret = -1;
 868                 goto out;
 869         }
 870         in6_ifa_hold(ifp);
 871         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 872         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 873                 spin_unlock_bh(&ifp->lock);
 874                 write_unlock(&idev->lock);
 875                 printk(KERN_WARNING
 876                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 877                 in6_ifa_put(ifp);
 878                 in6_dev_put(idev);
 879                 ret = -1;
 880                 goto out;
 881         }
 882         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 883         age = (now - ifp->tstamp) / HZ;
 884         tmp_valid_lft = min_t(__u32,
 885                               ifp->valid_lft,
 886                               idev->cnf.temp_valid_lft + age);
 887         tmp_prefered_lft = min_t(__u32,
 888                                  ifp->prefered_lft,
 889                                  idev->cnf.temp_prefered_lft + age -
 890                                  idev->cnf.max_desync_factor);
 891         tmp_plen = ifp->prefix_len;
 892         max_addresses = idev->cnf.max_addresses;
 893         tmp_tstamp = ifp->tstamp;
 894         spin_unlock_bh(&ifp->lock);
 895
 896         regen_advance = idev->cnf.regen_max_retry *
 897                         idev->cnf.dad_transmits *
 898                         idev->nd_parms->retrans_time / HZ;
 899         write_unlock(&idev->lock);
 900
 901         /* A temporary address is created only if this calculated Preferred
 902          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 903          * an implementation must not create a temporary address with a zero
 904          * Preferred Lifetime.
 905          */
 906         if (tmp_prefered_lft <= regen_advance) {
 907                 in6_ifa_put(ifp);
 908                 in6_dev_put(idev);
 909                 ret = -1;
 910                 goto out;
 911         }
 912
 913         addr_flags = IFA_F_TEMPORARY;
 914         /* set in addrconf_prefix_rcv() */
 915         if (ifp->flags & IFA_F_OPTIMISTIC)
 916                 addr_flags |= IFA_F_OPTIMISTIC;
 917
 918         ift = !max_addresses ||
 919               ipv6_count_addresses(idev) < max_addresses ?
 920                 ipv6_add_addr(idev, &addr, tmp_plen,
 921                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 922                               addr_flags) : NULL;
 923         if (!ift || IS_ERR(ift)) {
 924                 in6_ifa_put(ifp);
 925                 in6_dev_put(idev);
 926                 printk(KERN_INFO
 927                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 928                 tmpaddr = &addr;
 929                 write_lock(&idev->lock);
 930                 goto retry;
 931         }
 932
 933         spin_lock_bh(&ift->lock);
 934         ift->ifpub = ifp;
 935         ift->valid_lft = tmp_valid_lft;
 936         ift->prefered_lft = tmp_prefered_lft;
 937         ift->cstamp = now;
 938         ift->tstamp = tmp_tstamp;
 939         spin_unlock_bh(&ift->lock);
 940
 941         addrconf_dad_start(ift, 0);
 942         in6_ifa_put(ift);
 943         in6_dev_put(idev);
 944 out:
 945         return ret;
 946 }
 947 #endif
 948
 949 /*
 950  *      Choose an appropriate source address (RFC3484)
 951  */
 952 enum {
 953         IPV6_SADDR_RULE_INIT = 0,
 954         IPV6_SADDR_RULE_LOCAL,
 955         IPV6_SADDR_RULE_SCOPE,
 956         IPV6_SADDR_RULE_PREFERRED,
 957 #ifdef CONFIG_IPV6_MIP6
 958         IPV6_SADDR_RULE_HOA,
 959 #endif
 960         IPV6_SADDR_RULE_OIF,
 961         IPV6_SADDR_RULE_LABEL,
 962 #ifdef CONFIG_IPV6_PRIVACY
 963         IPV6_SADDR_RULE_PRIVACY,
 964 #endif
 965         IPV6_SADDR_RULE_ORCHID,
 966         IPV6_SADDR_RULE_PREFIX,
 967         IPV6_SADDR_RULE_MAX
 968 };
 969
 970 struct ipv6_saddr_score {
 971         int                     rule;
 972         int                     addr_type;
 973         struct inet6_ifaddr     *ifa;
 974         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 975         int                     scopedist;
 976         int                     matchlen;
 977 };
 978
 979 struct ipv6_saddr_dst {
 980         const struct in6_addr *addr;
 981         int ifindex;
 982         int scope;
 983         int label;
 984         unsigned int prefs;
 985 };
 986
 987 static inline int ipv6_saddr_preferred(int type)
 988 {
 989         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 990                 return 1;
 991         return 0;
 992 }
 993
 994 static int ipv6_get_saddr_eval(struct net *net,
 995                                struct ipv6_saddr_score *score,
 996                                struct ipv6_saddr_dst *dst,
 997                                int i)
 998 {
 999         int ret;
1000
1001         if (i <= score->rule) {
1002                 switch (i) {
1003                 case IPV6_SADDR_RULE_SCOPE:
1004                         ret = score->scopedist;
1005                         break;
1006                 case IPV6_SADDR_RULE_PREFIX:
1007                         ret = score->matchlen;
1008                         break;
1009                 default:
1010                         ret = !!test_bit(i, score->scorebits);
1011                 }
1012                 goto out;
1013         }
1014
1015         switch (i) {
1016         case IPV6_SADDR_RULE_INIT:
1017                 /* Rule 0: remember if hiscore is not ready yet */
1018                 ret = !!score->ifa;
1019                 break;
1020         case IPV6_SADDR_RULE_LOCAL:
1021                 /* Rule 1: Prefer same address */
1022                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1023                 break;
1024         case IPV6_SADDR_RULE_SCOPE:
1025                 /* Rule 2: Prefer appropriate scope
1026                  *
1027                  *      ret
1028                  *       ^
1029                  *    -1 |  d 15
1030                  *    ---+--+-+---> scope
1031                  *       |
1032                  *       |             d is scope of the destination.
1033                  *  B-d  |  \
1034                  *       |   \      <- smaller scope is better if
1035                  *  B-15 |    \        if scope is enough for destinaion.
1036                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1037                  * d-C-1 | /
1038                  *       |/         <- greater is better
1039                  *   -C  /             if scope is not enough for destination.
1040                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1041                  *
1042                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1043                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1044                  * Assume B = 0 and we get C > 29.
1045                  */
1046                 ret = __ipv6_addr_src_scope(score->addr_type);
1047                 if (ret >= dst->scope)
1048                         ret = -ret;
1049                 else
1050                         ret -= 128;     /* 30 is enough */
1051                 score->scopedist = ret;
1052                 break;
1053         case IPV6_SADDR_RULE_PREFERRED:
1054                 /* Rule 3: Avoid deprecated and optimistic addresses */
1055                 ret = ipv6_saddr_preferred(score->addr_type) ||
1056                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1057                 break;
1058 #ifdef CONFIG_IPV6_MIP6
1059         case IPV6_SADDR_RULE_HOA:
1060             {
1061                 /* Rule 4: Prefer home address */
1062                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1063                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1064                 break;
1065             }
1066 #endif
1067         case IPV6_SADDR_RULE_OIF:
1068                 /* Rule 5: Prefer outgoing interface */
1069                 ret = (!dst->ifindex ||
1070                        dst->ifindex == score->ifa->idev->dev->ifindex);
1071                 break;
1072         case IPV6_SADDR_RULE_LABEL:
1073                 /* Rule 6: Prefer matching label */
1074                 ret = ipv6_addr_label(net,
1075                                       &score->ifa->addr, score->addr_type,
1076                                       score->ifa->idev->dev->ifindex) == dst->label;
1077                 break;
1078 #ifdef CONFIG_IPV6_PRIVACY
1079         case IPV6_SADDR_RULE_PRIVACY:
1080             {
1081                 /* Rule 7: Prefer public address
1082                  * Note: prefer temporary address if use_tempaddr >= 2
1083                  */
1084                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1085                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1086                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1087                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1088                 break;
1089             }
1090 #endif
1091         case IPV6_SADDR_RULE_ORCHID:
1092                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1093                  *          non-ORCHID vs non-ORCHID
1094                  */
1095                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1096                         ipv6_addr_orchid(dst->addr));
1097                 break;
1098         case IPV6_SADDR_RULE_PREFIX:
1099                 /* Rule 8: Use longest matching prefix */
1100                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1101                                                        dst->addr);
1102                 break;
1103         default:
1104                 ret = 0;
1105         }
1106
1107         if (ret)
1108                 __set_bit(i, score->scorebits);
1109         score->rule = i;
1110 out:
1111         return ret;
1112 }
1113
1114 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1115                        const struct in6_addr *daddr, unsigned int prefs,
1116                        struct in6_addr *saddr)
1117 {
1118         struct ipv6_saddr_score scores[2],
1119                                 *score = &scores[0], *hiscore = &scores[1];
1120         struct ipv6_saddr_dst dst;
1121         struct net_device *dev;
1122         int dst_type;
1123
1124         dst_type = __ipv6_addr_type(daddr);
1125         dst.addr = daddr;
1126         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1127         dst.scope = __ipv6_addr_src_scope(dst_type);
1128         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1129         dst.prefs = prefs;
1130
1131         hiscore->rule = -1;
1132         hiscore->ifa = NULL;
1133
1134         rcu_read_lock();
1135
1136         for_each_netdev_rcu(net, dev) {
1137                 struct inet6_dev *idev;
1138
1139                 /* Candidate Source Address (section 4)
1140                  *  - multicast and link-local destination address,
1141                  *    the set of candidate source address MUST only
1142                  *    include addresses assigned to interfaces
1143                  *    belonging to the same link as the outgoing
1144                  *    interface.
1145                  * (- For site-local destination addresses, the
1146                  *    set of candidate source addresses MUST only
1147                  *    include addresses assigned to interfaces
1148                  *    belonging to the same site as the outgoing
1149                  *    interface.)
1150                  */
1151                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1152                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1153                     dst.ifindex && dev->ifindex != dst.ifindex)
1154                         continue;
1155
1156                 idev = __in6_dev_get(dev);
1157                 if (!idev)
1158                         continue;
1159
1160                 read_lock_bh(&idev->lock);
1161                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1162                         int i;
1163
1164                         /*
1165                          * - Tentative Address (RFC2462 section 5.4)
1166                          *  - A tentative address is not considered
1167                          *    "assigned to an interface" in the traditional
1168                          *    sense, unless it is also flagged as optimistic.
1169                          * - Candidate Source Address (section 4)
1170                          *  - In any case, anycast addresses, multicast
1171                          *    addresses, and the unspecified address MUST
1172                          *    NOT be included in a candidate set.
1173                          */
1174                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1175                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1176                                 continue;
1177
1178                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1179
1180                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1181                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1182                                 LIMIT_NETDEBUG(KERN_DEBUG
1183                                                "ADDRCONF: unspecified / multicast address "
1184                                                "assigned as unicast address on %s",
1185                                                dev->name);
1186                                 continue;
1187                         }
1188
1189                         score->rule = -1;
1190                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1191
1192                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1193                                 int minihiscore, miniscore;
1194
1195                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1196                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1197
1198                                 if (minihiscore > miniscore) {
1199                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1200                                             score->scopedist > 0) {
1201                                                 /*
1202                                                  * special case:
1203                                                  * each remaining entry
1204                                                  * has too small (not enough)
1205                                                  * scope, because ifa entries
1206                                                  * are sorted by their scope
1207                                                  * values.
1208                                                  */
1209                                                 goto try_nextdev;
1210                                         }
1211                                         break;
1212                                 } else if (minihiscore < miniscore) {
1213                                         if (hiscore->ifa)
1214                                                 in6_ifa_put(hiscore->ifa);
1215
1216                                         in6_ifa_hold(score->ifa);
1217
1218                                         swap(hiscore, score);
1219
1220                                         /* restore our iterator */
1221                                         score->ifa = hiscore->ifa;
1222
1223                                         break;
1224                                 }
1225                         }
1226                 }
1227 try_nextdev:
1228                 read_unlock_bh(&idev->lock);
1229         }
1230         rcu_read_unlock();
1231
1232         if (!hiscore->ifa)
1233                 return -EADDRNOTAVAIL;
1234
1235         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1236         in6_ifa_put(hiscore->ifa);
1237         return 0;
1238 }
1239 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1240
1241 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1242                     unsigned char banned_flags)
1243 {
1244         struct inet6_dev *idev;
1245         int err = -EADDRNOTAVAIL;
1246
1247         rcu_read_lock();
1248         idev = __in6_dev_get(dev);
1249         if (idev) {
1250                 struct inet6_ifaddr *ifp;
1251
1252                 read_lock_bh(&idev->lock);
1253                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1254                         if (ifp->scope == IFA_LINK &&
1255                             !(ifp->flags & banned_flags)) {
1256                                 ipv6_addr_copy(addr, &ifp->addr);
1257                                 err = 0;
1258                                 break;
1259                         }
1260                 }
1261                 read_unlock_bh(&idev->lock);
1262         }
1263         rcu_read_unlock();
1264         return err;
1265 }
1266
1267 static int ipv6_count_addresses(struct inet6_dev *idev)
1268 {
1269         int cnt = 0;
1270         struct inet6_ifaddr *ifp;
1271
1272         read_lock_bh(&idev->lock);
1273         list_for_each_entry(ifp, &idev->addr_list, if_list)
1274                 cnt++;
1275         read_unlock_bh(&idev->lock);
1276         return cnt;
1277 }
1278
1279 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1280                   struct net_device *dev, int strict)
1281 {
1282         struct inet6_ifaddr *ifp;
1283         struct hlist_node *node;
1284         unsigned int hash = ipv6_addr_hash(addr);
1285
1286         rcu_read_lock_bh();
1287         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1288                 if (!net_eq(dev_net(ifp->idev->dev), net))
1289                         continue;
1290                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1291                     !(ifp->flags&IFA_F_TENTATIVE) &&
1292                     (dev == NULL || ifp->idev->dev == dev ||
1293                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1294                         rcu_read_unlock_bh();
1295                         return 1;
1296                 }
1297         }
1298
1299         rcu_read_unlock_bh();
1300         return 0;
1301 }
1302 EXPORT_SYMBOL(ipv6_chk_addr);
1303
1304 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1305                                struct net_device *dev)
1306 {
1307         unsigned int hash = ipv6_addr_hash(addr);
1308         struct inet6_ifaddr *ifp;
1309         struct hlist_node *node;
1310
1311         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1312                 if (!net_eq(dev_net(ifp->idev->dev), net))
1313                         continue;
1314                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1315                         if (dev == NULL || ifp->idev->dev == dev)
1316                                 return true;
1317                 }
1318         }
1319         return false;
1320 }
1321
1322 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1323 {
1324         struct inet6_dev *idev;
1325         struct inet6_ifaddr *ifa;
1326         int     onlink;
1327
1328         onlink = 0;
1329         rcu_read_lock();
1330         idev = __in6_dev_get(dev);
1331         if (idev) {
1332                 read_lock_bh(&idev->lock);
1333                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1334                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1335                                                    ifa->prefix_len);
1336                         if (onlink)
1337                                 break;
1338                 }
1339                 read_unlock_bh(&idev->lock);
1340         }
1341         rcu_read_unlock();
1342         return onlink;
1343 }
1344
1345 EXPORT_SYMBOL(ipv6_chk_prefix);
1346
1347 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1348                                      struct net_device *dev, int strict)
1349 {
1350         struct inet6_ifaddr *ifp, *result = NULL;
1351         unsigned int hash = ipv6_addr_hash(addr);
1352         struct hlist_node *node;
1353
1354         rcu_read_lock_bh();
1355         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1356                 if (!net_eq(dev_net(ifp->idev->dev), net))
1357                         continue;
1358                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1359                         if (dev == NULL || ifp->idev->dev == dev ||
1360                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1361                                 result = ifp;
1362                                 in6_ifa_hold(ifp);
1363                                 break;
1364                         }
1365                 }
1366         }
1367         rcu_read_unlock_bh();
1368
1369         return result;
1370 }
1371
1372 /* Gets referenced address, destroys ifaddr */
1373
1374 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1375 {
1376         if (ifp->flags&IFA_F_PERMANENT) {
1377                 spin_lock_bh(&ifp->lock);
1378                 addrconf_del_timer(ifp);
1379                 ifp->flags |= IFA_F_TENTATIVE;
1380                 if (dad_failed)
1381                         ifp->flags |= IFA_F_DADFAILED;
1382                 spin_unlock_bh(&ifp->lock);
1383                 if (dad_failed)
1384                         ipv6_ifa_notify(0, ifp);
1385                 in6_ifa_put(ifp);
1386 #ifdef CONFIG_IPV6_PRIVACY
1387         } else if (ifp->flags&IFA_F_TEMPORARY) {
1388                 struct inet6_ifaddr *ifpub;
1389                 spin_lock_bh(&ifp->lock);
1390                 ifpub = ifp->ifpub;
1391                 if (ifpub) {
1392                         in6_ifa_hold(ifpub);
1393                         spin_unlock_bh(&ifp->lock);
1394                         ipv6_create_tempaddr(ifpub, ifp);
1395                         in6_ifa_put(ifpub);
1396                 } else {
1397                         spin_unlock_bh(&ifp->lock);
1398                 }
1399                 ipv6_del_addr(ifp);
1400 #endif
1401         } else
1402                 ipv6_del_addr(ifp);
1403 }
1404
1405 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1406 {
1407         int err = -ENOENT;
1408
1409         spin_lock(&ifp->state_lock);
1410         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1411                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1412                 err = 0;
1413         }
1414         spin_unlock(&ifp->state_lock);
1415
1416         return err;
1417 }
1418
1419 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1420 {
1421         struct inet6_dev *idev = ifp->idev;
1422
1423         if (addrconf_dad_end(ifp)) {
1424                 in6_ifa_put(ifp);
1425                 return;
1426         }
1427
1428         if (net_ratelimit())
1429                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1430                         ifp->idev->dev->name, &ifp->addr);
1431
1432         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1433                 struct in6_addr addr;
1434
1435                 addr.s6_addr32[0] = htonl(0xfe800000);
1436                 addr.s6_addr32[1] = 0;
1437
1438                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1439                     ipv6_addr_equal(&ifp->addr, &addr)) {
1440                         /* DAD failed for link-local based on MAC address */
1441                         idev->cnf.disable_ipv6 = 1;
1442
1443                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1444                                 ifp->idev->dev->name);
1445                 }
1446         }
1447
1448         addrconf_dad_stop(ifp, 1);
1449 }
1450
1451 /* Join to solicited addr multicast group. */
1452
1453 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1454 {
1455         struct in6_addr maddr;
1456
1457         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1458                 return;
1459
1460         addrconf_addr_solict_mult(addr, &maddr);
1461         ipv6_dev_mc_inc(dev, &maddr);
1462 }
1463
1464 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1465 {
1466         struct in6_addr maddr;
1467
1468         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1469                 return;
1470
1471         addrconf_addr_solict_mult(addr, &maddr);
1472         __ipv6_dev_mc_dec(idev, &maddr);
1473 }
1474
1475 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1476 {
1477         struct in6_addr addr;
1478         if (ifp->prefix_len == 127) /* RFC 6164 */
1479                 return;
1480         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1481         if (ipv6_addr_any(&addr))
1482                 return;
1483         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1484 }
1485
1486 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1487 {
1488         struct in6_addr addr;
1489         if (ifp->prefix_len == 127) /* RFC 6164 */
1490                 return;
1491         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1492         if (ipv6_addr_any(&addr))
1493                 return;
1494         __ipv6_dev_ac_dec(ifp->idev, &addr);
1495 }
1496
1497 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1498 {
1499         if (dev->addr_len != ETH_ALEN)
1500                 return -1;
1501         memcpy(eui, dev->dev_addr, 3);
1502         memcpy(eui + 5, dev->dev_addr + 3, 3);
1503
1504         /*
1505          * The zSeries OSA network cards can be shared among various
1506          * OS instances, but the OSA cards have only one MAC address.
1507          * This leads to duplicate address conflicts in conjunction
1508          * with IPv6 if more than one instance uses the same card.
1509          *
1510          * The driver for these cards can deliver a unique 16-bit
1511          * identifier for each instance sharing the same card.  It is
1512          * placed instead of 0xFFFE in the interface identifier.  The
1513          * "u" bit of the interface identifier is not inverted in this
1514          * case.  Hence the resulting interface identifier has local
1515          * scope according to RFC2373.
1516          */
1517         if (dev->dev_id) {
1518                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1519                 eui[4] = dev->dev_id & 0xFF;
1520         } else {
1521                 eui[3] = 0xFF;
1522                 eui[4] = 0xFE;
1523                 eui[0] ^= 2;
1524         }
1525         return 0;
1526 }
1527
1528 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1529 {
1530         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1531         if (dev->addr_len != ARCNET_ALEN)
1532                 return -1;
1533         memset(eui, 0, 7);
1534         eui[7] = *(u8*)dev->dev_addr;
1535         return 0;
1536 }
1537
1538 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1539 {
1540         if (dev->addr_len != INFINIBAND_ALEN)
1541                 return -1;
1542         memcpy(eui, dev->dev_addr + 12, 8);
1543         eui[0] |= 2;
1544         return 0;
1545 }
1546
1547 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1548 {
1549         if (addr == 0)
1550                 return -1;
1551         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1552                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1553                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1554                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1555                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1556                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1557         eui[1] = 0;
1558         eui[2] = 0x5E;
1559         eui[3] = 0xFE;
1560         memcpy(eui + 4, &addr, 4);
1561         return 0;
1562 }
1563
1564 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1565 {
1566         if (dev->priv_flags & IFF_ISATAP)
1567                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1568         return -1;
1569 }
1570
1571 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1572 {
1573         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1574 }
1575
1576 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1577 {
1578         switch (dev->type) {
1579         case ARPHRD_ETHER:
1580         case ARPHRD_FDDI:
1581         case ARPHRD_IEEE802_TR:
1582                 return addrconf_ifid_eui48(eui, dev);
1583         case ARPHRD_ARCNET:
1584                 return addrconf_ifid_arcnet(eui, dev);
1585         case ARPHRD_INFINIBAND:
1586                 return addrconf_ifid_infiniband(eui, dev);
1587         case ARPHRD_SIT:
1588                 return addrconf_ifid_sit(eui, dev);
1589         case ARPHRD_IPGRE:
1590                 return addrconf_ifid_gre(eui, dev);
1591         }
1592         return -1;
1593 }
1594
1595 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1596 {
1597         int err = -1;
1598         struct inet6_ifaddr *ifp;
1599
1600         read_lock_bh(&idev->lock);
1601         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1602                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1603                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1604                         err = 0;
1605                         break;
1606                 }
1607         }
1608         read_unlock_bh(&idev->lock);
1609         return err;
1610 }
1611
1612 #ifdef CONFIG_IPV6_PRIVACY
1613 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1614 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1615 {
1616 regen:
1617         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1618         idev->rndid[0] &= ~0x02;
1619
1620         /*
1621          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1622          * check if generated address is not inappropriate
1623          *
1624          *  - Reserved subnet anycast (RFC 2526)
1625          *      11111101 11....11 1xxxxxxx
1626          *  - ISATAP (RFC4214) 6.1
1627          *      00-00-5E-FE-xx-xx-xx-xx
1628          *  - value 0
1629          *  - XXX: already assigned to an address on the device
1630          */
1631         if (idev->rndid[0] == 0xfd &&
1632             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1633             (idev->rndid[7]&0x80))
1634                 goto regen;
1635         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1636                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1637                         goto regen;
1638                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1639                         goto regen;
1640         }
1641
1642         return 0;
1643 }
1644
1645 static void ipv6_regen_rndid(unsigned long data)
1646 {
1647         struct inet6_dev *idev = (struct inet6_dev *) data;
1648         unsigned long expires;
1649
1650         rcu_read_lock_bh();
1651         write_lock_bh(&idev->lock);
1652
1653         if (idev->dead)
1654                 goto out;
1655
1656         if (__ipv6_regen_rndid(idev) < 0)
1657                 goto out;
1658
1659         expires = jiffies +
1660                 idev->cnf.temp_prefered_lft * HZ -
1661                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1662                 idev->cnf.max_desync_factor * HZ;
1663         if (time_before(expires, jiffies)) {
1664                 printk(KERN_WARNING
1665                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1666                         idev->dev->name);
1667                 goto out;
1668         }
1669
1670         if (!mod_timer(&idev->regen_timer, expires))
1671                 in6_dev_hold(idev);
1672
1673 out:
1674         write_unlock_bh(&idev->lock);
1675         rcu_read_unlock_bh();
1676         in6_dev_put(idev);
1677 }
1678
1679 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1680         int ret = 0;
1681
1682         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1683                 ret = __ipv6_regen_rndid(idev);
1684         return ret;
1685 }
1686 #endif
1687
1688 /*
1689  *      Add prefix route.
1690  */
1691
1692 static void
1693 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1694                       unsigned long expires, u32 flags)
1695 {
1696         struct fib6_config cfg = {
1697                 .fc_table = RT6_TABLE_PREFIX,
1698                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1699                 .fc_ifindex = dev->ifindex,
1700                 .fc_expires = expires,
1701                 .fc_dst_len = plen,
1702                 .fc_flags = RTF_UP | flags,
1703                 .fc_nlinfo.nl_net = dev_net(dev),
1704                 .fc_protocol = RTPROT_KERNEL,
1705         };
1706
1707         ipv6_addr_copy(&cfg.fc_dst, pfx);
1708
1709         /* Prevent useless cloning on PtP SIT.
1710            This thing is done here expecting that the whole
1711            class of non-broadcast devices need not cloning.
1712          */
1713 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1714         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1715                 cfg.fc_flags |= RTF_NONEXTHOP;
1716 #endif
1717
1718         ip6_route_add(&cfg);
1719 }
1720
1721
1722 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1723                                                   int plen,
1724                                                   const struct net_device *dev,
1725                                                   u32 flags, u32 noflags)
1726 {
1727         struct fib6_node *fn;
1728         struct rt6_info *rt = NULL;
1729         struct fib6_table *table;
1730
1731         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1732         if (table == NULL)
1733                 return NULL;
1734
1735         write_lock_bh(&table->tb6_lock);
1736         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1737         if (!fn)
1738                 goto out;
1739         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1740                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1741                         continue;
1742                 if ((rt->rt6i_flags & flags) != flags)
1743                         continue;
1744                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1745                         continue;
1746                 dst_hold(&rt->dst);
1747                 break;
1748         }
1749 out:
1750         write_unlock_bh(&table->tb6_lock);
1751         return rt;
1752 }
1753
1754
1755 /* Create "default" multicast route to the interface */
1756
1757 static void addrconf_add_mroute(struct net_device *dev)
1758 {
1759         struct fib6_config cfg = {
1760                 .fc_table = RT6_TABLE_LOCAL,
1761                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1762                 .fc_ifindex = dev->ifindex,
1763                 .fc_dst_len = 8,
1764                 .fc_flags = RTF_UP,
1765                 .fc_nlinfo.nl_net = dev_net(dev),
1766         };
1767
1768         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1769
1770         ip6_route_add(&cfg);
1771 }
1772
1773 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1774 static void sit_route_add(struct net_device *dev)
1775 {
1776         struct fib6_config cfg = {
1777                 .fc_table = RT6_TABLE_MAIN,
1778                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1779                 .fc_ifindex = dev->ifindex,
1780                 .fc_dst_len = 96,
1781                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1782                 .fc_nlinfo.nl_net = dev_net(dev),
1783         };
1784
1785         /* prefix length - 96 bits "::d.d.d.d" */
1786         ip6_route_add(&cfg);
1787 }
1788 #endif
1789
1790 static void addrconf_add_lroute(struct net_device *dev)
1791 {
1792         struct in6_addr addr;
1793
1794         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1795         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1796 }
1797
1798 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1799 {
1800         struct inet6_dev *idev;
1801
1802         ASSERT_RTNL();
1803
1804         idev = ipv6_find_idev(dev);
1805         if (!idev)
1806                 return ERR_PTR(-ENOBUFS);
1807
1808         if (idev->cnf.disable_ipv6)
1809                 return ERR_PTR(-EACCES);
1810
1811         /* Add default multicast route */
1812         if (!(dev->flags & IFF_LOOPBACK))
1813                 addrconf_add_mroute(dev);
1814
1815         /* Add link local route */
1816         addrconf_add_lroute(dev);
1817         return idev;
1818 }
1819
1820 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1821 {
1822         struct prefix_info *pinfo;
1823         __u32 valid_lft;
1824         __u32 prefered_lft;
1825         int addr_type;
1826         struct inet6_dev *in6_dev;
1827         struct net *net = dev_net(dev);
1828
1829         pinfo = (struct prefix_info *) opt;
1830
1831         if (len < sizeof(struct prefix_info)) {
1832                 ADBG(("addrconf: prefix option too short\n"));
1833                 return;
1834         }
1835
1836         /*
1837          *      Validation checks ([ADDRCONF], page 19)
1838          */
1839
1840         addr_type = ipv6_addr_type(&pinfo->prefix);
1841
1842         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1843                 return;
1844
1845         valid_lft = ntohl(pinfo->valid);
1846         prefered_lft = ntohl(pinfo->prefered);
1847
1848         if (prefered_lft > valid_lft) {
1849                 if (net_ratelimit())
1850                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1851                 return;
1852         }
1853
1854         in6_dev = in6_dev_get(dev);
1855
1856         if (in6_dev == NULL) {
1857                 if (net_ratelimit())
1858                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1859                 return;
1860         }
1861
1862         /*
1863          *      Two things going on here:
1864          *      1) Add routes for on-link prefixes
1865          *      2) Configure prefixes with the auto flag set
1866          */
1867
1868         if (pinfo->onlink) {
1869                 struct rt6_info *rt;
1870                 unsigned long rt_expires;
1871
1872                 /* Avoid arithmetic overflow. Really, we could
1873                  * save rt_expires in seconds, likely valid_lft,
1874                  * but it would require division in fib gc, that it
1875                  * not good.
1876                  */
1877                 if (HZ > USER_HZ)
1878                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1879                 else
1880                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1881
1882                 if (addrconf_finite_timeout(rt_expires))
1883                         rt_expires *= HZ;
1884
1885                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1886                                                pinfo->prefix_len,
1887                                                dev,
1888                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1889                                                RTF_GATEWAY | RTF_DEFAULT);
1890
1891                 if (rt) {
1892                         /* Autoconf prefix route */
1893                         if (valid_lft == 0) {
1894                                 ip6_del_rt(rt);
1895                                 rt = NULL;
1896                         } else if (addrconf_finite_timeout(rt_expires)) {
1897                                 /* not infinity */
1898                                 rt->rt6i_expires = jiffies + rt_expires;
1899                                 rt->rt6i_flags |= RTF_EXPIRES;
1900                         } else {
1901                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1902                                 rt->rt6i_expires = 0;
1903                         }
1904                 } else if (valid_lft) {
1905                         clock_t expires = 0;
1906                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1907                         if (addrconf_finite_timeout(rt_expires)) {
1908                                 /* not infinity */
1909                                 flags |= RTF_EXPIRES;
1910                                 expires = jiffies_to_clock_t(rt_expires);
1911                         }
1912                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1913                                               dev, expires, flags);
1914                 }
1915                 if (rt)
1916                         dst_release(&rt->dst);
1917         }
1918
1919         /* Try to figure out our local address for this prefix */
1920
1921         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1922                 struct inet6_ifaddr * ifp;
1923                 struct in6_addr addr;
1924                 int create = 0, update_lft = 0;
1925
1926                 if (pinfo->prefix_len == 64) {
1927                         memcpy(&addr, &pinfo->prefix, 8);
1928                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1929                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1930                                 in6_dev_put(in6_dev);
1931                                 return;
1932                         }
1933                         goto ok;
1934                 }
1935                 if (net_ratelimit())
1936                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1937                                pinfo->prefix_len);
1938                 in6_dev_put(in6_dev);
1939                 return;
1940
1941 ok:
1942
1943                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1944
1945                 if (ifp == NULL && valid_lft) {
1946                         int max_addresses = in6_dev->cnf.max_addresses;
1947                         u32 addr_flags = 0;
1948
1949 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1950                         if (in6_dev->cnf.optimistic_dad &&
1951                             !net->ipv6.devconf_all->forwarding)
1952                                 addr_flags = IFA_F_OPTIMISTIC;
1953 #endif
1954
1955                         /* Do not allow to create too much of autoconfigured
1956                          * addresses; this would be too easy way to crash kernel.
1957                          */
1958                         if (!max_addresses ||
1959                             ipv6_count_addresses(in6_dev) < max_addresses)
1960                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1961                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1962                                                     addr_flags);
1963
1964                         if (!ifp || IS_ERR(ifp)) {
1965                                 in6_dev_put(in6_dev);
1966                                 return;
1967                         }
1968
1969                         update_lft = create = 1;
1970                         ifp->cstamp = jiffies;
1971                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1972                 }
1973
1974                 if (ifp) {
1975                         int flags;
1976                         unsigned long now;
1977 #ifdef CONFIG_IPV6_PRIVACY
1978                         struct inet6_ifaddr *ift;
1979 #endif
1980                         u32 stored_lft;
1981
1982                         /* update lifetime (RFC2462 5.5.3 e) */
1983                         spin_lock(&ifp->lock);
1984                         now = jiffies;
1985                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1986                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1987                         else
1988                                 stored_lft = 0;
1989                         if (!update_lft && stored_lft) {
1990                                 if (valid_lft > MIN_VALID_LIFETIME ||
1991                                     valid_lft > stored_lft)
1992                                         update_lft = 1;
1993                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1994                                         /* valid_lft <= stored_lft is always true */
1995                                         /*
1996                                          * RFC 4862 Section 5.5.3e:
1997                                          * "Note that the preferred lifetime of
1998                                          *  the corresponding address is always
1999                                          *  reset to the Preferred Lifetime in
2000                                          *  the received Prefix Information
2001                                          *  option, regardless of whether the
2002                                          *  valid lifetime is also reset or
2003                                          *  ignored."
2004                                          *
2005                                          *  So if the preferred lifetime in
2006                                          *  this advertisement is different
2007                                          *  than what we have stored, but the
2008                                          *  valid lifetime is invalid, just
2009                                          *  reset prefered_lft.
2010                                          *
2011                                          *  We must set the valid lifetime
2012                                          *  to the stored lifetime since we'll
2013                                          *  be updating the timestamp below,
2014                                          *  else we'll set it back to the
2015                                          *  minimum.
2016                                          */
2017                                         if (prefered_lft != ifp->prefered_lft) {
2018                                                 valid_lft = stored_lft;
2019                                                 update_lft = 1;
2020                                         }
2021                                 } else {
2022                                         valid_lft = MIN_VALID_LIFETIME;
2023                                         if (valid_lft < prefered_lft)
2024                                                 prefered_lft = valid_lft;
2025                                         update_lft = 1;
2026                                 }
2027                         }
2028
2029                         if (update_lft) {
2030                                 ifp->valid_lft = valid_lft;
2031                                 ifp->prefered_lft = prefered_lft;
2032                                 ifp->tstamp = now;
2033                                 flags = ifp->flags;
2034                                 ifp->flags &= ~IFA_F_DEPRECATED;
2035                                 spin_unlock(&ifp->lock);
2036
2037                                 if (!(flags&IFA_F_TENTATIVE))
2038                                         ipv6_ifa_notify(0, ifp);
2039                         } else
2040                                 spin_unlock(&ifp->lock);
2041
2042 #ifdef CONFIG_IPV6_PRIVACY
2043                         read_lock_bh(&in6_dev->lock);
2044                         /* update all temporary addresses in the list */
2045                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2046                                             tmp_list) {
2047                                 int age, max_valid, max_prefered;
2048
2049                                 if (ifp != ift->ifpub)
2050                                         continue;
2051
2052                                 /*
2053                                  * RFC 4941 section 3.3:
2054                                  * If a received option will extend the lifetime
2055                                  * of a public address, the lifetimes of
2056                                  * temporary addresses should be extended,
2057                                  * subject to the overall constraint that no
2058                                  * temporary addresses should ever remain
2059                                  * "valid" or "preferred" for a time longer than
2060                                  * (TEMP_VALID_LIFETIME) or
2061                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2062                                  * respectively.
2063                                  */
2064                                 age = (now - ift->cstamp) / HZ;
2065                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2066                                 if (max_valid < 0)
2067                                         max_valid = 0;
2068
2069                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2070                                                in6_dev->cnf.max_desync_factor -
2071                                                age;
2072                                 if (max_prefered < 0)
2073                                         max_prefered = 0;
2074
2075                                 if (valid_lft > max_valid)
2076                                         valid_lft = max_valid;
2077
2078                                 if (prefered_lft > max_prefered)
2079                                         prefered_lft = max_prefered;
2080
2081                                 spin_lock(&ift->lock);
2082                                 flags = ift->flags;
2083                                 ift->valid_lft = valid_lft;
2084                                 ift->prefered_lft = prefered_lft;
2085                                 ift->tstamp = now;
2086                                 if (prefered_lft > 0)
2087                                         ift->flags &= ~IFA_F_DEPRECATED;
2088
2089                                 spin_unlock(&ift->lock);
2090                                 if (!(flags&IFA_F_TENTATIVE))
2091                                         ipv6_ifa_notify(0, ift);
2092                         }
2093
2094                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2095                                 /*
2096                                  * When a new public address is created as
2097                                  * described in [ADDRCONF], also create a new
2098                                  * temporary address. Also create a temporary
2099                                  * address if it's enabled but no temporary
2100                                  * address currently exists.
2101                                  */
2102                                 read_unlock_bh(&in6_dev->lock);
2103                                 ipv6_create_tempaddr(ifp, NULL);
2104                         } else {
2105                                 read_unlock_bh(&in6_dev->lock);
2106                         }
2107 #endif
2108                         in6_ifa_put(ifp);
2109                         addrconf_verify(0);
2110                 }
2111         }
2112         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2113         in6_dev_put(in6_dev);
2114 }
2115
2116 /*
2117  *      Set destination address.
2118  *      Special case for SIT interfaces where we create a new "virtual"
2119  *      device.
2120  */
2121 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2122 {
2123         struct in6_ifreq ireq;
2124         struct net_device *dev;
2125         int err = -EINVAL;
2126
2127         rtnl_lock();
2128
2129         err = -EFAULT;
2130         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2131                 goto err_exit;
2132
2133         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2134
2135         err = -ENODEV;
2136         if (dev == NULL)
2137                 goto err_exit;
2138
2139 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2140         if (dev->type == ARPHRD_SIT) {
2141                 const struct net_device_ops *ops = dev->netdev_ops;
2142                 struct ifreq ifr;
2143                 struct ip_tunnel_parm p;
2144
2145                 err = -EADDRNOTAVAIL;
2146                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2147                         goto err_exit;
2148
2149                 memset(&p, 0, sizeof(p));
2150                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2151                 p.iph.saddr = 0;
2152                 p.iph.version = 4;
2153                 p.iph.ihl = 5;
2154                 p.iph.protocol = IPPROTO_IPV6;
2155                 p.iph.ttl = 64;
2156                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2157
2158                 if (ops->ndo_do_ioctl) {
2159                         mm_segment_t oldfs = get_fs();
2160
2161                         set_fs(KERNEL_DS);
2162                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2163                         set_fs(oldfs);
2164                 } else
2165                         err = -EOPNOTSUPP;
2166
2167                 if (err == 0) {
2168                         err = -ENOBUFS;
2169                         dev = __dev_get_by_name(net, p.name);
2170                         if (!dev)
2171                                 goto err_exit;
2172                         err = dev_open(dev);
2173                 }
2174         }
2175 #endif
2176
2177 err_exit:
2178         rtnl_unlock();
2179         return err;
2180 }
2181
2182 /*
2183  *      Manual configuration of address on an interface
2184  */
2185 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2186                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2187                           __u32 valid_lft)
2188 {
2189         struct inet6_ifaddr *ifp;
2190         struct inet6_dev *idev;
2191         struct net_device *dev;
2192         int scope;
2193         u32 flags;
2194         clock_t expires;
2195         unsigned long timeout;
2196
2197         ASSERT_RTNL();
2198
2199         if (plen > 128)
2200                 return -EINVAL;
2201
2202         /* check the lifetime */
2203         if (!valid_lft || prefered_lft > valid_lft)
2204                 return -EINVAL;
2205
2206         dev = __dev_get_by_index(net, ifindex);
2207         if (!dev)
2208                 return -ENODEV;
2209
2210         idev = addrconf_add_dev(dev);
2211         if (IS_ERR(idev))
2212                 return PTR_ERR(idev);
2213
2214         scope = ipv6_addr_scope(pfx);
2215
2216         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2217         if (addrconf_finite_timeout(timeout)) {
2218                 expires = jiffies_to_clock_t(timeout * HZ);
2219                 valid_lft = timeout;
2220                 flags = RTF_EXPIRES;
2221         } else {
2222                 expires = 0;
2223                 flags = 0;
2224                 ifa_flags |= IFA_F_PERMANENT;
2225         }
2226
2227         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2228         if (addrconf_finite_timeout(timeout)) {
2229                 if (timeout == 0)
2230                         ifa_flags |= IFA_F_DEPRECATED;
2231                 prefered_lft = timeout;
2232         }
2233
2234         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2235
2236         if (!IS_ERR(ifp)) {
2237                 spin_lock_bh(&ifp->lock);
2238                 ifp->valid_lft = valid_lft;
2239                 ifp->prefered_lft = prefered_lft;
2240                 ifp->tstamp = jiffies;
2241                 spin_unlock_bh(&ifp->lock);
2242
2243                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2244                                       expires, flags);
2245                 /*
2246                  * Note that section 3.1 of RFC 4429 indicates
2247                  * that the Optimistic flag should not be set for
2248                  * manually configured addresses
2249                  */
2250                 addrconf_dad_start(ifp, 0);
2251                 in6_ifa_put(ifp);
2252                 addrconf_verify(0);
2253                 return 0;
2254         }
2255
2256         return PTR_ERR(ifp);
2257 }
2258
2259 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2260                           unsigned int plen)
2261 {
2262         struct inet6_ifaddr *ifp;
2263         struct inet6_dev *idev;
2264         struct net_device *dev;
2265
2266         if (plen > 128)
2267                 return -EINVAL;
2268
2269         dev = __dev_get_by_index(net, ifindex);
2270         if (!dev)
2271                 return -ENODEV;
2272
2273         if ((idev = __in6_dev_get(dev)) == NULL)
2274                 return -ENXIO;
2275
2276         read_lock_bh(&idev->lock);
2277         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2278                 if (ifp->prefix_len == plen &&
2279                     ipv6_addr_equal(pfx, &ifp->addr)) {
2280                         in6_ifa_hold(ifp);
2281                         read_unlock_bh(&idev->lock);
2282
2283                         ipv6_del_addr(ifp);
2284
2285                         /* If the last address is deleted administratively,
2286                            disable IPv6 on this interface.
2287                          */
2288                         if (list_empty(&idev->addr_list))
2289                                 addrconf_ifdown(idev->dev, 1);
2290                         return 0;
2291                 }
2292         }
2293         read_unlock_bh(&idev->lock);
2294         return -EADDRNOTAVAIL;
2295 }
2296
2297
2298 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2299 {
2300         struct in6_ifreq ireq;
2301         int err;
2302
2303         if (!capable(CAP_NET_ADMIN))
2304                 return -EPERM;
2305
2306         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2307                 return -EFAULT;
2308
2309         rtnl_lock();
2310         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2311                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2312                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2313         rtnl_unlock();
2314         return err;
2315 }
2316
2317 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2318 {
2319         struct in6_ifreq ireq;
2320         int err;
2321
2322         if (!capable(CAP_NET_ADMIN))
2323                 return -EPERM;
2324
2325         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2326                 return -EFAULT;
2327
2328         rtnl_lock();
2329         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2330                              ireq.ifr6_prefixlen);
2331         rtnl_unlock();
2332         return err;
2333 }
2334
2335 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2336                      int plen, int scope)
2337 {
2338         struct inet6_ifaddr *ifp;
2339
2340         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2341         if (!IS_ERR(ifp)) {
2342                 spin_lock_bh(&ifp->lock);
2343                 ifp->flags &= ~IFA_F_TENTATIVE;
2344                 spin_unlock_bh(&ifp->lock);
2345                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2346                 in6_ifa_put(ifp);
2347         }
2348 }
2349
2350 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2351 static void sit_add_v4_addrs(struct inet6_dev *idev)
2352 {
2353         struct in6_addr addr;
2354         struct net_device *dev;
2355         struct net *net = dev_net(idev->dev);
2356         int scope;
2357
2358         ASSERT_RTNL();
2359
2360         memset(&addr, 0, sizeof(struct in6_addr));
2361         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2362
2363         if (idev->dev->flags&IFF_POINTOPOINT) {
2364                 addr.s6_addr32[0] = htonl(0xfe800000);
2365                 scope = IFA_LINK;
2366         } else {
2367                 scope = IPV6_ADDR_COMPATv4;
2368         }
2369
2370         if (addr.s6_addr32[3]) {
2371                 add_addr(idev, &addr, 128, scope);
2372                 return;
2373         }
2374
2375         for_each_netdev(net, dev) {
2376                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2377                 if (in_dev && (dev->flags & IFF_UP)) {
2378                         struct in_ifaddr * ifa;
2379
2380                         int flag = scope;
2381
2382                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2383                                 int plen;
2384
2385                                 addr.s6_addr32[3] = ifa->ifa_local;
2386
2387                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2388                                         continue;
2389                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2390                                         if (idev->dev->flags&IFF_POINTOPOINT)
2391                                                 continue;
2392                                         flag |= IFA_HOST;
2393                                 }
2394                                 if (idev->dev->flags&IFF_POINTOPOINT)
2395                                         plen = 64;
2396                                 else
2397                                         plen = 96;
2398
2399                                 add_addr(idev, &addr, plen, flag);
2400                         }
2401                 }
2402         }
2403 }
2404 #endif
2405
2406 static void init_loopback(struct net_device *dev)
2407 {
2408         struct inet6_dev  *idev;
2409
2410         /* ::1 */
2411
2412         ASSERT_RTNL();
2413
2414         if ((idev = ipv6_find_idev(dev)) == NULL) {
2415                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2416                 return;
2417         }
2418
2419         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2420 }
2421
2422 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2423 {
2424         struct inet6_ifaddr * ifp;
2425         u32 addr_flags = IFA_F_PERMANENT;
2426
2427 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2428         if (idev->cnf.optimistic_dad &&
2429             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2430                 addr_flags |= IFA_F_OPTIMISTIC;
2431 #endif
2432
2433
2434         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2435         if (!IS_ERR(ifp)) {
2436                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2437                 addrconf_dad_start(ifp, 0);
2438                 in6_ifa_put(ifp);
2439         }
2440 }
2441
2442 static void addrconf_dev_config(struct net_device *dev)
2443 {
2444         struct in6_addr addr;
2445         struct inet6_dev    * idev;
2446
2447         ASSERT_RTNL();
2448
2449         if ((dev->type != ARPHRD_ETHER) &&
2450             (dev->type != ARPHRD_FDDI) &&
2451             (dev->type != ARPHRD_IEEE802_TR) &&
2452             (dev->type != ARPHRD_ARCNET) &&
2453             (dev->type != ARPHRD_INFINIBAND)) {
2454                 /* Alas, we support only Ethernet autoconfiguration. */
2455                 return;
2456         }
2457
2458         idev = addrconf_add_dev(dev);
2459         if (IS_ERR(idev))
2460                 return;
2461
2462         memset(&addr, 0, sizeof(struct in6_addr));
2463         addr.s6_addr32[0] = htonl(0xFE800000);
2464
2465         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2466                 addrconf_add_linklocal(idev, &addr);
2467 }
2468
2469 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2470 static void addrconf_sit_config(struct net_device *dev)
2471 {
2472         struct inet6_dev *idev;
2473
2474         ASSERT_RTNL();
2475
2476         /*
2477          * Configure the tunnel with one of our IPv4
2478          * addresses... we should configure all of
2479          * our v4 addrs in the tunnel
2480          */
2481
2482         if ((idev = ipv6_find_idev(dev)) == NULL) {
2483                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2484                 return;
2485         }
2486
2487         if (dev->priv_flags & IFF_ISATAP) {
2488                 struct in6_addr addr;
2489
2490                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2491                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2492                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2493                         addrconf_add_linklocal(idev, &addr);
2494                 return;
2495         }
2496
2497         sit_add_v4_addrs(idev);
2498
2499         if (dev->flags&IFF_POINTOPOINT) {
2500                 addrconf_add_mroute(dev);
2501                 addrconf_add_lroute(dev);
2502         } else
2503                 sit_route_add(dev);
2504 }
2505 #endif
2506
2507 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2508 static void addrconf_gre_config(struct net_device *dev)
2509 {
2510         struct inet6_dev *idev;
2511         struct in6_addr addr;
2512
2513         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2514
2515         ASSERT_RTNL();
2516
2517         if ((idev = ipv6_find_idev(dev)) == NULL) {
2518                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2519                 return;
2520         }
2521
2522         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2523         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2524
2525         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2526                 addrconf_add_linklocal(idev, &addr);
2527 }
2528 #endif
2529
2530 static inline int
2531 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2532 {
2533         struct in6_addr lladdr;
2534
2535         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2536                 addrconf_add_linklocal(idev, &lladdr);
2537                 return 0;
2538         }
2539         return -1;
2540 }
2541
2542 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2543 {
2544         struct net_device *link_dev;
2545         struct net *net = dev_net(idev->dev);
2546
2547         /* first try to inherit the link-local address from the link device */
2548         if (idev->dev->iflink &&
2549             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2550                 if (!ipv6_inherit_linklocal(idev, link_dev))
2551                         return;
2552         }
2553         /* then try to inherit it from any device */
2554         for_each_netdev(net, link_dev) {
2555                 if (!ipv6_inherit_linklocal(idev, link_dev))
2556                         return;
2557         }
2558         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2559 }
2560
2561 /*
2562  * Autoconfigure tunnel with a link-local address so routing protocols,
2563  * DHCPv6, MLD etc. can be run over the virtual link
2564  */
2565
2566 static void addrconf_ip6_tnl_config(struct net_device *dev)
2567 {
2568         struct inet6_dev *idev;
2569
2570         ASSERT_RTNL();
2571
2572         idev = addrconf_add_dev(dev);
2573         if (IS_ERR(idev)) {
2574                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2575                 return;
2576         }
2577         ip6_tnl_add_linklocal(idev);
2578 }
2579
2580 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2581                            void * data)
2582 {
2583         struct net_device *dev = (struct net_device *) data;
2584         struct inet6_dev *idev = __in6_dev_get(dev);
2585         int run_pending = 0;
2586         int err;
2587
2588         switch (event) {
2589         case NETDEV_REGISTER:
2590                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2591                         idev = ipv6_add_dev(dev);
2592                         if (!idev)
2593                                 return notifier_from_errno(-ENOMEM);
2594                 }
2595                 break;
2596
2597         case NETDEV_UP:
2598         case NETDEV_CHANGE:
2599                 if (dev->flags & IFF_SLAVE)
2600                         break;
2601
2602                 if (event == NETDEV_UP) {
2603                         if (!addrconf_qdisc_ok(dev)) {
2604                                 /* device is not ready yet. */
2605                                 printk(KERN_INFO
2606                                         "ADDRCONF(NETDEV_UP): %s: "
2607                                         "link is not ready\n",
2608                                         dev->name);
2609                                 break;
2610                         }
2611
2612                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2613                                 idev = ipv6_add_dev(dev);
2614
2615                         if (idev) {
2616                                 idev->if_flags |= IF_READY;
2617                                 run_pending = 1;
2618                         }
2619                 } else {
2620                         if (!addrconf_qdisc_ok(dev)) {
2621                                 /* device is still not ready. */
2622                                 break;
2623                         }
2624
2625                         if (idev) {
2626                                 if (idev->if_flags & IF_READY)
2627                                         /* device is already configured. */
2628                                         break;
2629                                 idev->if_flags |= IF_READY;
2630                         }
2631
2632                         printk(KERN_INFO
2633                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2634                                         "link becomes ready\n",
2635                                         dev->name);
2636
2637                         run_pending = 1;
2638                 }
2639
2640                 switch (dev->type) {
2641 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2642                 case ARPHRD_SIT:
2643                         addrconf_sit_config(dev);
2644                         break;
2645 #endif
2646 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2647                 case ARPHRD_IPGRE:
2648                         addrconf_gre_config(dev);
2649                         break;
2650 #endif
2651                 case ARPHRD_TUNNEL6:
2652                         addrconf_ip6_tnl_config(dev);
2653                         break;
2654                 case ARPHRD_LOOPBACK:
2655                         init_loopback(dev);
2656                         break;
2657
2658                 default:
2659                         addrconf_dev_config(dev);
2660                         break;
2661                 }
2662
2663                 if (idev) {
2664                         if (run_pending)
2665                                 addrconf_dad_run(idev);
2666
2667                         /*
2668                          * If the MTU changed during the interface down,
2669                          * when the interface up, the changed MTU must be
2670                          * reflected in the idev as well as routers.
2671                          */
2672                         if (idev->cnf.mtu6 != dev->mtu &&
2673                             dev->mtu >= IPV6_MIN_MTU) {
2674                                 rt6_mtu_change(dev, dev->mtu);
2675                                 idev->cnf.mtu6 = dev->mtu;
2676                         }
2677                         idev->tstamp = jiffies;
2678                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2679
2680                         /*
2681                          * If the changed mtu during down is lower than
2682                          * IPV6_MIN_MTU stop IPv6 on this interface.
2683                          */
2684                         if (dev->mtu < IPV6_MIN_MTU)
2685                                 addrconf_ifdown(dev, 1);
2686                 }
2687                 break;
2688
2689         case NETDEV_CHANGEMTU:
2690                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2691                         rt6_mtu_change(dev, dev->mtu);
2692                         idev->cnf.mtu6 = dev->mtu;
2693                         break;
2694                 }
2695
2696                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2697                         idev = ipv6_add_dev(dev);
2698                         if (idev)
2699                                 break;
2700                 }
2701
2702                 /*
2703                  * MTU falled under IPV6_MIN_MTU.
2704                  * Stop IPv6 on this interface.
2705                  */
2706
2707         case NETDEV_DOWN:
2708         case NETDEV_UNREGISTER:
2709                 /*
2710                  *      Remove all addresses from this interface.
2711                  */
2712                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2713                 break;
2714
2715         case NETDEV_CHANGENAME:
2716                 if (idev) {
2717                         snmp6_unregister_dev(idev);
2718                         addrconf_sysctl_unregister(idev);
2719                         addrconf_sysctl_register(idev);
2720                         err = snmp6_register_dev(idev);
2721                         if (err)
2722                                 return notifier_from_errno(err);
2723                 }
2724                 break;
2725
2726         case NETDEV_PRE_TYPE_CHANGE:
2727         case NETDEV_POST_TYPE_CHANGE:
2728                 addrconf_type_change(dev, event);
2729                 break;
2730         }
2731
2732         return NOTIFY_OK;
2733 }
2734
2735 /*
2736  *      addrconf module should be notified of a device going up
2737  */
2738 static struct notifier_block ipv6_dev_notf = {
2739         .notifier_call = addrconf_notify,
2740 };
2741
2742 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2743 {
2744         struct inet6_dev *idev;
2745         ASSERT_RTNL();
2746
2747         idev = __in6_dev_get(dev);
2748
2749         if (event == NETDEV_POST_TYPE_CHANGE)
2750                 ipv6_mc_remap(idev);
2751         else if (event == NETDEV_PRE_TYPE_CHANGE)
2752                 ipv6_mc_unmap(idev);
2753 }
2754
2755 static int addrconf_ifdown(struct net_device *dev, int how)
2756 {
2757         struct net *net = dev_net(dev);
2758         struct inet6_dev *idev;
2759         struct inet6_ifaddr *ifa;
2760         int state, i;
2761
2762         ASSERT_RTNL();
2763
2764         rt6_ifdown(net, dev);
2765         neigh_ifdown(&nd_tbl, dev);
2766
2767         idev = __in6_dev_get(dev);
2768         if (idev == NULL)
2769                 return -ENODEV;
2770
2771         /*
2772          * Step 1: remove reference to ipv6 device from parent device.
2773          *         Do not dev_put!
2774          */
2775         if (how) {
2776                 idev->dead = 1;
2777
2778                 /* protected by rtnl_lock */
2779                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2780
2781                 /* Step 1.5: remove snmp6 entry */
2782                 snmp6_unregister_dev(idev);
2783
2784         }
2785
2786         /* Step 2: clear hash table */
2787         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2788                 struct hlist_head *h = &inet6_addr_lst[i];
2789                 struct hlist_node *n;
2790
2791                 spin_lock_bh(&addrconf_hash_lock);
2792         restart:
2793                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2794                         if (ifa->idev == idev) {
2795                                 hlist_del_init_rcu(&ifa->addr_lst);
2796                                 addrconf_del_timer(ifa);
2797                                 goto restart;
2798                         }
2799                 }
2800                 spin_unlock_bh(&addrconf_hash_lock);
2801         }
2802
2803         write_lock_bh(&idev->lock);
2804
2805         /* Step 2: clear flags for stateless addrconf */
2806         if (!how)
2807                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2808
2809 #ifdef CONFIG_IPV6_PRIVACY
2810         if (how && del_timer(&idev->regen_timer))
2811                 in6_dev_put(idev);
2812
2813         /* Step 3: clear tempaddr list */
2814         while (!list_empty(&idev->tempaddr_list)) {
2815                 ifa = list_first_entry(&idev->tempaddr_list,
2816                                        struct inet6_ifaddr, tmp_list);
2817                 list_del(&ifa->tmp_list);
2818                 write_unlock_bh(&idev->lock);
2819                 spin_lock_bh(&ifa->lock);
2820
2821                 if (ifa->ifpub) {
2822                         in6_ifa_put(ifa->ifpub);
2823                         ifa->ifpub = NULL;
2824                 }
2825                 spin_unlock_bh(&ifa->lock);
2826                 in6_ifa_put(ifa);
2827                 write_lock_bh(&idev->lock);
2828         }
2829 #endif
2830
2831         while (!list_empty(&idev->addr_list)) {
2832                 ifa = list_first_entry(&idev->addr_list,
2833                                        struct inet6_ifaddr, if_list);
2834                 addrconf_del_timer(ifa);
2835
2836                 list_del(&ifa->if_list);
2837
2838                 write_unlock_bh(&idev->lock);
2839
2840                 spin_lock_bh(&ifa->state_lock);
2841                 state = ifa->state;
2842                 ifa->state = INET6_IFADDR_STATE_DEAD;
2843                 spin_unlock_bh(&ifa->state_lock);
2844
2845                 if (state != INET6_IFADDR_STATE_DEAD) {
2846                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2847                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2848                 }
2849                 in6_ifa_put(ifa);
2850
2851                 write_lock_bh(&idev->lock);
2852         }
2853
2854         write_unlock_bh(&idev->lock);
2855
2856         /* Step 5: Discard multicast list */
2857         if (how)
2858                 ipv6_mc_destroy_dev(idev);
2859         else
2860                 ipv6_mc_down(idev);
2861
2862         idev->tstamp = jiffies;
2863
2864         /* Last: Shot the device (if unregistered) */
2865         if (how) {
2866                 addrconf_sysctl_unregister(idev);
2867                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2868                 neigh_ifdown(&nd_tbl, dev);
2869                 in6_dev_put(idev);
2870         }
2871         return 0;
2872 }
2873
2874 static void addrconf_rs_timer(unsigned long data)
2875 {
2876         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2877         struct inet6_dev *idev = ifp->idev;
2878
2879         read_lock(&idev->lock);
2880         if (idev->dead || !(idev->if_flags & IF_READY))
2881                 goto out;
2882
2883         if (idev->cnf.forwarding)
2884                 goto out;
2885
2886         /* Announcement received after solicitation was sent */
2887         if (idev->if_flags & IF_RA_RCVD)
2888                 goto out;
2889
2890         spin_lock(&ifp->lock);
2891         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2892                 /* The wait after the last probe can be shorter */
2893                 addrconf_mod_timer(ifp, AC_RS,
2894                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2895                                    idev->cnf.rtr_solicit_delay :
2896                                    idev->cnf.rtr_solicit_interval);
2897                 spin_unlock(&ifp->lock);
2898
2899                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2900         } else {
2901                 spin_unlock(&ifp->lock);
2902                 /*
2903                  * Note: we do not support deprecated "all on-link"
2904                  * assumption any longer.
2905                  */
2906                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2907                        idev->dev->name);
2908         }
2909
2910 out:
2911         read_unlock(&idev->lock);
2912         in6_ifa_put(ifp);
2913 }
2914
2915 /*
2916  *      Duplicate Address Detection
2917  */
2918 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2919 {
2920         unsigned long rand_num;
2921         struct inet6_dev *idev = ifp->idev;
2922
2923         if (ifp->flags & IFA_F_OPTIMISTIC)
2924                 rand_num = 0;
2925         else
2926                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2927
2928         ifp->probes = idev->cnf.dad_transmits;
2929         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2930 }
2931
2932 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2933 {
2934         struct inet6_dev *idev = ifp->idev;
2935         struct net_device *dev = idev->dev;
2936
2937         addrconf_join_solict(dev, &ifp->addr);
2938
2939         net_srandom(ifp->addr.s6_addr32[3]);
2940
2941         read_lock_bh(&idev->lock);
2942         spin_lock(&ifp->lock);
2943         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2944                 goto out;
2945
2946         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2947             idev->cnf.accept_dad < 1 ||
2948             !(ifp->flags&IFA_F_TENTATIVE) ||
2949             ifp->flags & IFA_F_NODAD) {
2950                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2951                 spin_unlock(&ifp->lock);
2952                 read_unlock_bh(&idev->lock);
2953
2954                 addrconf_dad_completed(ifp);
2955                 return;
2956         }
2957
2958         if (!(idev->if_flags & IF_READY)) {
2959                 spin_unlock(&ifp->lock);
2960                 read_unlock_bh(&idev->lock);
2961                 /*
2962                  * If the device is not ready:
2963                  * - keep it tentative if it is a permanent address.
2964                  * - otherwise, kill it.
2965                  */
2966                 in6_ifa_hold(ifp);
2967                 addrconf_dad_stop(ifp, 0);
2968                 return;
2969         }
2970
2971         /*
2972          * Optimistic nodes can start receiving
2973          * Frames right away
2974          */
2975         if (ifp->flags & IFA_F_OPTIMISTIC)
2976                 ip6_ins_rt(ifp->rt);
2977
2978         addrconf_dad_kick(ifp);
2979 out:
2980         spin_unlock(&ifp->lock);
2981         read_unlock_bh(&idev->lock);
2982 }
2983
2984 static void addrconf_dad_timer(unsigned long data)
2985 {
2986         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2987         struct inet6_dev *idev = ifp->idev;
2988         struct in6_addr mcaddr;
2989
2990         if (!ifp->probes && addrconf_dad_end(ifp))
2991                 goto out;
2992
2993         read_lock(&idev->lock);
2994         if (idev->dead || !(idev->if_flags & IF_READY)) {
2995                 read_unlock(&idev->lock);
2996                 goto out;
2997         }
2998
2999         spin_lock(&ifp->lock);
3000         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3001                 spin_unlock(&ifp->lock);
3002                 read_unlock(&idev->lock);
3003                 goto out;
3004         }
3005
3006         if (ifp->probes == 0) {
3007                 /*
3008                  * DAD was successful
3009                  */
3010
3011                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3012                 spin_unlock(&ifp->lock);
3013                 read_unlock(&idev->lock);
3014
3015                 addrconf_dad_completed(ifp);
3016
3017                 goto out;
3018         }
3019
3020         ifp->probes--;
3021         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3022         spin_unlock(&ifp->lock);
3023         read_unlock(&idev->lock);
3024
3025         /* send a neighbour solicitation for our addr */
3026         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3027         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3028 out:
3029         in6_ifa_put(ifp);
3030 }
3031
3032 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3033 {
3034         struct net_device *dev = ifp->idev->dev;
3035
3036         /*
3037          *      Configure the address for reception. Now it is valid.
3038          */
3039
3040         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3041
3042         /* If added prefix is link local and we are prepared to process
3043            router advertisements, start sending router solicitations.
3044          */
3045
3046         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3047              ifp->idev->cnf.accept_ra == 2) &&
3048             ifp->idev->cnf.rtr_solicits > 0 &&
3049             (dev->flags&IFF_LOOPBACK) == 0 &&
3050             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3051                 /*
3052                  *      If a host as already performed a random delay
3053                  *      [...] as part of DAD [...] there is no need
3054                  *      to delay again before sending the first RS
3055                  */
3056                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3057
3058                 spin_lock_bh(&ifp->lock);
3059                 ifp->probes = 1;
3060                 ifp->idev->if_flags |= IF_RS_SENT;
3061                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3062                 spin_unlock_bh(&ifp->lock);
3063         }
3064 }
3065
3066 static void addrconf_dad_run(struct inet6_dev *idev)
3067 {
3068         struct inet6_ifaddr *ifp;
3069
3070         read_lock_bh(&idev->lock);
3071         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3072                 spin_lock(&ifp->lock);
3073                 if (ifp->flags & IFA_F_TENTATIVE &&
3074                     ifp->state == INET6_IFADDR_STATE_DAD)
3075                         addrconf_dad_kick(ifp);
3076                 spin_unlock(&ifp->lock);
3077         }
3078         read_unlock_bh(&idev->lock);
3079 }
3080
3081 #ifdef CONFIG_PROC_FS
3082 struct if6_iter_state {
3083         struct seq_net_private p;
3084         int bucket;
3085 };
3086
3087 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3088 {
3089         struct inet6_ifaddr *ifa = NULL;
3090         struct if6_iter_state *state = seq->private;
3091         struct net *net = seq_file_net(seq);
3092
3093         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3094                 struct hlist_node *n;
3095                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3096                                          addr_lst)
3097                         if (net_eq(dev_net(ifa->idev->dev), net))
3098                                 return ifa;
3099         }
3100         return NULL;
3101 }
3102
3103 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3104                                          struct inet6_ifaddr *ifa)
3105 {
3106         struct if6_iter_state *state = seq->private;
3107         struct net *net = seq_file_net(seq);
3108         struct hlist_node *n = &ifa->addr_lst;
3109
3110         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3111                 if (net_eq(dev_net(ifa->idev->dev), net))
3112                         return ifa;
3113
3114         while (++state->bucket < IN6_ADDR_HSIZE) {
3115                 hlist_for_each_entry_rcu_bh(ifa, n,
3116                                      &inet6_addr_lst[state->bucket], addr_lst) {
3117                         if (net_eq(dev_net(ifa->idev->dev), net))
3118                                 return ifa;
3119                 }
3120         }
3121
3122         return NULL;
3123 }
3124
3125 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3126 {
3127         struct inet6_ifaddr *ifa = if6_get_first(seq);
3128
3129         if (ifa)
3130                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3131                         --pos;
3132         return pos ? NULL : ifa;
3133 }
3134
3135 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3136         __acquires(rcu_bh)
3137 {
3138         rcu_read_lock_bh();
3139         return if6_get_idx(seq, *pos);
3140 }
3141
3142 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3143 {
3144         struct inet6_ifaddr *ifa;
3145
3146         ifa = if6_get_next(seq, v);
3147         ++*pos;
3148         return ifa;
3149 }
3150
3151 static void if6_seq_stop(struct seq_file *seq, void *v)
3152         __releases(rcu_bh)
3153 {
3154         rcu_read_unlock_bh();
3155 }
3156
3157 static int if6_seq_show(struct seq_file *seq, void *v)
3158 {
3159         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3160         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3161                    &ifp->addr,
3162                    ifp->idev->dev->ifindex,
3163                    ifp->prefix_len,
3164                    ifp->scope,
3165                    ifp->flags,
3166                    ifp->idev->dev->name);
3167         return 0;
3168 }
3169
3170 static const struct seq_operations if6_seq_ops = {
3171         .start  = if6_seq_start,
3172         .next   = if6_seq_next,
3173         .show   = if6_seq_show,
3174         .stop   = if6_seq_stop,
3175 };
3176
3177 static int if6_seq_open(struct inode *inode, struct file *file)
3178 {
3179         return seq_open_net(inode, file, &if6_seq_ops,
3180                             sizeof(struct if6_iter_state));
3181 }
3182
3183 static const struct file_operations if6_fops = {
3184         .owner          = THIS_MODULE,
3185         .open           = if6_seq_open,
3186         .read           = seq_read,
3187         .llseek         = seq_lseek,
3188         .release        = seq_release_net,
3189 };
3190
3191 static int __net_init if6_proc_net_init(struct net *net)
3192 {
3193         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3194                 return -ENOMEM;
3195         return 0;
3196 }
3197
3198 static void __net_exit if6_proc_net_exit(struct net *net)
3199 {
3200        proc_net_remove(net, "if_inet6");
3201 }
3202
3203 static struct pernet_operations if6_proc_net_ops = {
3204        .init = if6_proc_net_init,
3205        .exit = if6_proc_net_exit,
3206 };
3207
3208 int __init if6_proc_init(void)
3209 {
3210         return register_pernet_subsys(&if6_proc_net_ops);
3211 }
3212
3213 void if6_proc_exit(void)
3214 {
3215         unregister_pernet_subsys(&if6_proc_net_ops);
3216 }
3217 #endif  /* CONFIG_PROC_FS */
3218
3219 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3220 /* Check if address is a home address configured on any interface. */
3221 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3222 {
3223         int ret = 0;
3224         struct inet6_ifaddr *ifp = NULL;
3225         struct hlist_node *n;
3226         unsigned int hash = ipv6_addr_hash(addr);
3227
3228         rcu_read_lock_bh();
3229         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3230                 if (!net_eq(dev_net(ifp->idev->dev), net))
3231                         continue;
3232                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3233                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3234                         ret = 1;
3235                         break;
3236                 }
3237         }
3238         rcu_read_unlock_bh();
3239         return ret;
3240 }
3241 #endif
3242
3243 /*
3244  *      Periodic address status verification
3245  */
3246
3247 static void addrconf_verify(unsigned long foo)
3248 {
3249         unsigned long now, next, next_sec, next_sched;
3250         struct inet6_ifaddr *ifp;
3251         struct hlist_node *node;
3252         int i;
3253
3254         rcu_read_lock_bh();
3255         spin_lock(&addrconf_verify_lock);
3256         now = jiffies;
3257         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3258
3259         del_timer(&addr_chk_timer);
3260
3261         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3262 restart:
3263                 hlist_for_each_entry_rcu_bh(ifp, node,
3264                                          &inet6_addr_lst[i], addr_lst) {
3265                         unsigned long age;
3266
3267                         if (ifp->flags & IFA_F_PERMANENT)
3268                                 continue;
3269
3270                         spin_lock(&ifp->lock);
3271                         /* We try to batch several events at once. */
3272                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3273
3274                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3275                             age >= ifp->valid_lft) {
3276                                 spin_unlock(&ifp->lock);
3277                                 in6_ifa_hold(ifp);
3278                                 ipv6_del_addr(ifp);
3279                                 goto restart;
3280                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3281                                 spin_unlock(&ifp->lock);
3282                                 continue;
3283                         } else if (age >= ifp->prefered_lft) {
3284                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3285                                 int deprecate = 0;
3286
3287                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3288                                         deprecate = 1;
3289                                         ifp->flags |= IFA_F_DEPRECATED;
3290                                 }
3291
3292                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3293                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3294
3295                                 spin_unlock(&ifp->lock);
3296
3297                                 if (deprecate) {
3298                                         in6_ifa_hold(ifp);
3299
3300                                         ipv6_ifa_notify(0, ifp);
3301                                         in6_ifa_put(ifp);
3302                                         goto restart;
3303                                 }
3304 #ifdef CONFIG_IPV6_PRIVACY
3305                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3306                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3307                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3308                                         ifp->idev->cnf.dad_transmits *
3309                                         ifp->idev->nd_parms->retrans_time / HZ;
3310
3311                                 if (age >= ifp->prefered_lft - regen_advance) {
3312                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3313                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3314                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3315                                         if (!ifp->regen_count && ifpub) {
3316                                                 ifp->regen_count++;
3317                                                 in6_ifa_hold(ifp);
3318                                                 in6_ifa_hold(ifpub);
3319                                                 spin_unlock(&ifp->lock);
3320
3321                                                 spin_lock(&ifpub->lock);
3322                                                 ifpub->regen_count = 0;
3323                                                 spin_unlock(&ifpub->lock);
3324                                                 ipv6_create_tempaddr(ifpub, ifp);
3325                                                 in6_ifa_put(ifpub);
3326                                                 in6_ifa_put(ifp);
3327                                                 goto restart;
3328                                         }
3329                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3330                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3331                                 spin_unlock(&ifp->lock);
3332 #endif
3333                         } else {
3334                                 /* ifp->prefered_lft <= ifp->valid_lft */
3335                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3336                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3337                                 spin_unlock(&ifp->lock);
3338                         }
3339                 }
3340         }
3341
3342         next_sec = round_jiffies_up(next);
3343         next_sched = next;
3344
3345         /* If rounded timeout is accurate enough, accept it. */
3346         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3347                 next_sched = next_sec;
3348
3349         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3350         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3351                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3352
3353         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3354               now, next, next_sec, next_sched));
3355
3356         addr_chk_timer.expires = next_sched;
3357         add_timer(&addr_chk_timer);
3358         spin_unlock(&addrconf_verify_lock);
3359         rcu_read_unlock_bh();
3360 }
3361
3362 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3363 {
3364         struct in6_addr *pfx = NULL;
3365
3366         if (addr)
3367                 pfx = nla_data(addr);
3368
3369         if (local) {
3370                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3371                         pfx = NULL;
3372                 else
3373                         pfx = nla_data(local);
3374         }
3375
3376         return pfx;
3377 }
3378
3379 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3380         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3381         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3382         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3383 };
3384
3385 static int
3386 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3387 {
3388         struct net *net = sock_net(skb->sk);
3389         struct ifaddrmsg *ifm;
3390         struct nlattr *tb[IFA_MAX+1];
3391         struct in6_addr *pfx;
3392         int err;
3393
3394         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3395         if (err < 0)
3396                 return err;
3397
3398         ifm = nlmsg_data(nlh);
3399         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3400         if (pfx == NULL)
3401                 return -EINVAL;
3402
3403         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3404 }
3405
3406 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3407                              u32 prefered_lft, u32 valid_lft)
3408 {
3409         u32 flags;
3410         clock_t expires;
3411         unsigned long timeout;
3412
3413         if (!valid_lft || (prefered_lft > valid_lft))
3414                 return -EINVAL;
3415
3416         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3417         if (addrconf_finite_timeout(timeout)) {
3418                 expires = jiffies_to_clock_t(timeout * HZ);
3419                 valid_lft = timeout;
3420                 flags = RTF_EXPIRES;
3421         } else {
3422                 expires = 0;
3423                 flags = 0;
3424                 ifa_flags |= IFA_F_PERMANENT;
3425         }
3426
3427         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3428         if (addrconf_finite_timeout(timeout)) {
3429                 if (timeout == 0)
3430                         ifa_flags |= IFA_F_DEPRECATED;
3431                 prefered_lft = timeout;
3432         }
3433
3434         spin_lock_bh(&ifp->lock);
3435         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3436         ifp->tstamp = jiffies;
3437         ifp->valid_lft = valid_lft;
3438         ifp->prefered_lft = prefered_lft;
3439
3440         spin_unlock_bh(&ifp->lock);
3441         if (!(ifp->flags&IFA_F_TENTATIVE))
3442                 ipv6_ifa_notify(0, ifp);
3443
3444         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3445                               expires, flags);
3446         addrconf_verify(0);
3447
3448         return 0;
3449 }
3450
3451 static int
3452 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3453 {
3454         struct net *net = sock_net(skb->sk);
3455         struct ifaddrmsg *ifm;
3456         struct nlattr *tb[IFA_MAX+1];
3457         struct in6_addr *pfx;
3458         struct inet6_ifaddr *ifa;
3459         struct net_device *dev;
3460         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3461         u8 ifa_flags;
3462         int err;
3463
3464         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3465         if (err < 0)
3466                 return err;
3467
3468         ifm = nlmsg_data(nlh);
3469         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3470         if (pfx == NULL)
3471                 return -EINVAL;
3472
3473         if (tb[IFA_CACHEINFO]) {
3474                 struct ifa_cacheinfo *ci;
3475
3476                 ci = nla_data(tb[IFA_CACHEINFO]);
3477                 valid_lft = ci->ifa_valid;
3478                 preferred_lft = ci->ifa_prefered;
3479         } else {
3480                 preferred_lft = INFINITY_LIFE_TIME;
3481                 valid_lft = INFINITY_LIFE_TIME;
3482         }
3483
3484         dev =  __dev_get_by_index(net, ifm->ifa_index);
3485         if (dev == NULL)
3486                 return -ENODEV;
3487
3488         /* We ignore other flags so far. */
3489         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3490
3491         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3492         if (ifa == NULL) {
3493                 /*
3494                  * It would be best to check for !NLM_F_CREATE here but
3495                  * userspace alreay relies on not having to provide this.
3496                  */
3497                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3498                                       ifm->ifa_prefixlen, ifa_flags,
3499                                       preferred_lft, valid_lft);
3500         }
3501
3502         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3503             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3504                 err = -EEXIST;
3505         else
3506                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3507
3508         in6_ifa_put(ifa);
3509
3510         return err;
3511 }
3512
3513 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3514                           u8 scope, int ifindex)
3515 {
3516         struct ifaddrmsg *ifm;
3517
3518         ifm = nlmsg_data(nlh);
3519         ifm->ifa_family = AF_INET6;
3520         ifm->ifa_prefixlen = prefixlen;
3521         ifm->ifa_flags = flags;
3522         ifm->ifa_scope = scope;
3523         ifm->ifa_index = ifindex;
3524 }
3525
3526 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3527                          unsigned long tstamp, u32 preferred, u32 valid)
3528 {
3529         struct ifa_cacheinfo ci;
3530
3531         ci.cstamp = cstamp_delta(cstamp);
3532         ci.tstamp = cstamp_delta(tstamp);
3533         ci.ifa_prefered = preferred;
3534         ci.ifa_valid = valid;
3535
3536         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3537 }
3538
3539 static inline int rt_scope(int ifa_scope)
3540 {
3541         if (ifa_scope & IFA_HOST)
3542                 return RT_SCOPE_HOST;
3543         else if (ifa_scope & IFA_LINK)
3544                 return RT_SCOPE_LINK;
3545         else if (ifa_scope & IFA_SITE)
3546                 return RT_SCOPE_SITE;
3547         else
3548                 return RT_SCOPE_UNIVERSE;
3549 }
3550
3551 static inline int inet6_ifaddr_msgsize(void)
3552 {
3553         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3554                + nla_total_size(16) /* IFA_ADDRESS */
3555                + nla_total_size(sizeof(struct ifa_cacheinfo));
3556 }
3557
3558 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3559                              u32 pid, u32 seq, int event, unsigned int flags)
3560 {
3561         struct nlmsghdr  *nlh;
3562         u32 preferred, valid;
3563
3564         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3565         if (nlh == NULL)
3566                 return -EMSGSIZE;
3567
3568         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3569                       ifa->idev->dev->ifindex);
3570
3571         if (!(ifa->flags&IFA_F_PERMANENT)) {
3572                 preferred = ifa->prefered_lft;
3573                 valid = ifa->valid_lft;
3574                 if (preferred != INFINITY_LIFE_TIME) {
3575                         long tval = (jiffies - ifa->tstamp)/HZ;
3576                         if (preferred > tval)
3577                                 preferred -= tval;
3578                         else
3579                                 preferred = 0;
3580                         if (valid != INFINITY_LIFE_TIME) {
3581                                 if (valid > tval)
3582                                         valid -= tval;
3583                                 else
3584                                         valid = 0;
3585                         }
3586                 }
3587         } else {
3588                 preferred = INFINITY_LIFE_TIME;
3589                 valid = INFINITY_LIFE_TIME;
3590         }
3591
3592         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3593             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3594                 nlmsg_cancel(skb, nlh);
3595                 return -EMSGSIZE;
3596         }
3597
3598         return nlmsg_end(skb, nlh);
3599 }
3600
3601 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3602                                 u32 pid, u32 seq, int event, u16 flags)
3603 {
3604         struct nlmsghdr  *nlh;
3605         u8 scope = RT_SCOPE_UNIVERSE;
3606         int ifindex = ifmca->idev->dev->ifindex;
3607
3608         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3609                 scope = RT_SCOPE_SITE;
3610
3611         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3612         if (nlh == NULL)
3613                 return -EMSGSIZE;
3614
3615         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3616         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3617             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3618                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3619                 nlmsg_cancel(skb, nlh);
3620                 return -EMSGSIZE;
3621         }
3622
3623         return nlmsg_end(skb, nlh);
3624 }
3625
3626 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3627                                 u32 pid, u32 seq, int event, unsigned int flags)
3628 {
3629         struct nlmsghdr  *nlh;
3630         u8 scope = RT_SCOPE_UNIVERSE;
3631         int ifindex = ifaca->aca_idev->dev->ifindex;
3632
3633         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3634                 scope = RT_SCOPE_SITE;
3635
3636         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3637         if (nlh == NULL)
3638                 return -EMSGSIZE;
3639
3640         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3641         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3642             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3643                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3644                 nlmsg_cancel(skb, nlh);
3645                 return -EMSGSIZE;
3646         }
3647
3648         return nlmsg_end(skb, nlh);
3649 }
3650
3651 enum addr_type_t {
3652         UNICAST_ADDR,
3653         MULTICAST_ADDR,
3654         ANYCAST_ADDR,
3655 };
3656
3657 /* called with rcu_read_lock() */
3658 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3659                           struct netlink_callback *cb, enum addr_type_t type,
3660                           int s_ip_idx, int *p_ip_idx)
3661 {
3662         struct ifmcaddr6 *ifmca;
3663         struct ifacaddr6 *ifaca;
3664         int err = 1;
3665         int ip_idx = *p_ip_idx;
3666
3667         read_lock_bh(&idev->lock);
3668         switch (type) {
3669         case UNICAST_ADDR: {
3670                 struct inet6_ifaddr *ifa;
3671
3672                 /* unicast address incl. temp addr */
3673                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3674                         if (++ip_idx < s_ip_idx)
3675                                 continue;
3676                         err = inet6_fill_ifaddr(skb, ifa,
3677                                                 NETLINK_CB(cb->skb).pid,
3678                                                 cb->nlh->nlmsg_seq,
3679                                                 RTM_NEWADDR,
3680                                                 NLM_F_MULTI);
3681                         if (err <= 0)
3682                                 break;
3683                 }
3684                 break;
3685         }
3686         case MULTICAST_ADDR:
3687                 /* multicast address */
3688                 for (ifmca = idev->mc_list; ifmca;
3689                      ifmca = ifmca->next, ip_idx++) {
3690                         if (ip_idx < s_ip_idx)
3691                                 continue;
3692                         err = inet6_fill_ifmcaddr(skb, ifmca,
3693                                                   NETLINK_CB(cb->skb).pid,
3694                                                   cb->nlh->nlmsg_seq,
3695                                                   RTM_GETMULTICAST,
3696                                                   NLM_F_MULTI);
3697                         if (err <= 0)
3698                                 break;
3699                 }
3700                 break;
3701         case ANYCAST_ADDR:
3702                 /* anycast address */
3703                 for (ifaca = idev->ac_list; ifaca;
3704                      ifaca = ifaca->aca_next, ip_idx++) {
3705                         if (ip_idx < s_ip_idx)
3706                                 continue;
3707                         err = inet6_fill_ifacaddr(skb, ifaca,
3708                                                   NETLINK_CB(cb->skb).pid,
3709                                                   cb->nlh->nlmsg_seq,
3710                                                   RTM_GETANYCAST,
3711                                                   NLM_F_MULTI);
3712                         if (err <= 0)
3713                                 break;
3714                 }
3715                 break;
3716         default:
3717                 break;
3718         }
3719         read_unlock_bh(&idev->lock);
3720         *p_ip_idx = ip_idx;
3721         return err;
3722 }
3723
3724 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3725                            enum addr_type_t type)
3726 {
3727         struct net *net = sock_net(skb->sk);
3728         int h, s_h;
3729         int idx, ip_idx;
3730         int s_idx, s_ip_idx;
3731         struct net_device *dev;
3732         struct inet6_dev *idev;
3733         struct hlist_head *head;
3734         struct hlist_node *node;
3735
3736         s_h = cb->args[0];
3737         s_idx = idx = cb->args[1];
3738         s_ip_idx = ip_idx = cb->args[2];
3739
3740         rcu_read_lock();
3741         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3742                 idx = 0;
3743                 head = &net->dev_index_head[h];
3744                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3745                         if (idx < s_idx)
3746                                 goto cont;
3747                         if (h > s_h || idx > s_idx)
3748                                 s_ip_idx = 0;
3749                         ip_idx = 0;
3750                         idev = __in6_dev_get(dev);
3751                         if (!idev)
3752                                 goto cont;
3753
3754                         if (in6_dump_addrs(idev, skb, cb, type,
3755                                            s_ip_idx, &ip_idx) <= 0)
3756                                 goto done;
3757 cont:
3758                         idx++;
3759                 }
3760         }
3761 done:
3762         rcu_read_unlock();
3763         cb->args[0] = h;
3764         cb->args[1] = idx;
3765         cb->args[2] = ip_idx;
3766
3767         return skb->len;
3768 }
3769
3770 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3771 {
3772         enum addr_type_t type = UNICAST_ADDR;
3773
3774         return inet6_dump_addr(skb, cb, type);
3775 }
3776
3777 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3778 {
3779         enum addr_type_t type = MULTICAST_ADDR;
3780
3781         return inet6_dump_addr(skb, cb, type);
3782 }
3783
3784
3785 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3786 {
3787         enum addr_type_t type = ANYCAST_ADDR;
3788
3789         return inet6_dump_addr(skb, cb, type);
3790 }
3791
3792 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3793                              void *arg)
3794 {
3795         struct net *net = sock_net(in_skb->sk);
3796         struct ifaddrmsg *ifm;
3797         struct nlattr *tb[IFA_MAX+1];
3798         struct in6_addr *addr = NULL;
3799         struct net_device *dev = NULL;
3800         struct inet6_ifaddr *ifa;
3801         struct sk_buff *skb;
3802         int err;
3803
3804         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3805         if (err < 0)
3806                 goto errout;
3807
3808         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3809         if (addr == NULL) {
3810                 err = -EINVAL;
3811                 goto errout;
3812         }
3813
3814         ifm = nlmsg_data(nlh);
3815         if (ifm->ifa_index)
3816                 dev = __dev_get_by_index(net, ifm->ifa_index);
3817
3818         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3819         if (!ifa) {
3820                 err = -EADDRNOTAVAIL;
3821                 goto errout;
3822         }
3823
3824         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3825         if (!skb) {
3826                 err = -ENOBUFS;
3827                 goto errout_ifa;
3828         }
3829
3830         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3831                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3832         if (err < 0) {
3833                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3834                 WARN_ON(err == -EMSGSIZE);
3835                 kfree_skb(skb);
3836                 goto errout_ifa;
3837         }
3838         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3839 errout_ifa:
3840         in6_ifa_put(ifa);
3841 errout:
3842         return err;
3843 }
3844
3845 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3846 {
3847         struct sk_buff *skb;
3848         struct net *net = dev_net(ifa->idev->dev);
3849         int err = -ENOBUFS;
3850
3851         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3852         if (skb == NULL)
3853                 goto errout;
3854
3855         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3856         if (err < 0) {
3857                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3858                 WARN_ON(err == -EMSGSIZE);
3859                 kfree_skb(skb);
3860                 goto errout;
3861         }
3862         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3863         return;
3864 errout:
3865         if (err < 0)
3866                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3867 }
3868
3869 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3870                                 __s32 *array, int bytes)
3871 {
3872         BUG_ON(bytes < (DEVCONF_MAX * 4));
3873
3874         memset(array, 0, bytes);
3875         array[DEVCONF_FORWARDING] = cnf->forwarding;
3876         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3877         array[DEVCONF_MTU6] = cnf->mtu6;
3878         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3879         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3880         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3881         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3882         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3883         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3884                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3885         array[DEVCONF_RTR_SOLICIT_DELAY] =
3886                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3887         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3888 #ifdef CONFIG_IPV6_PRIVACY
3889         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3890         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3891         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3892         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3893         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3894 #endif
3895         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3896         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3897         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3898 #ifdef CONFIG_IPV6_ROUTER_PREF
3899         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3900         array[DEVCONF_RTR_PROBE_INTERVAL] =
3901                 jiffies_to_msecs(cnf->rtr_probe_interval);
3902 #ifdef CONFIG_IPV6_ROUTE_INFO
3903         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3904 #endif
3905 #endif
3906         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3907         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3908 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3909         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3910 #endif
3911 #ifdef CONFIG_IPV6_MROUTE
3912         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3913 #endif
3914         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3915         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3916         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3917 }
3918
3919 static inline size_t inet6_ifla6_size(void)
3920 {
3921         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3922              + nla_total_size(sizeof(struct ifla_cacheinfo))
3923              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3924              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3925              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3926 }
3927
3928 static inline size_t inet6_if_nlmsg_size(void)
3929 {
3930         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3931                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3932                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3933                + nla_total_size(4) /* IFLA_MTU */
3934                + nla_total_size(4) /* IFLA_LINK */
3935                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3936 }
3937
3938 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3939                                       int items, int bytes)
3940 {
3941         int i;
3942         int pad = bytes - sizeof(u64) * items;
3943         BUG_ON(pad < 0);
3944
3945         /* Use put_unaligned() because stats may not be aligned for u64. */
3946         put_unaligned(items, &stats[0]);
3947         for (i = 1; i < items; i++)
3948                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3949
3950         memset(&stats[items], 0, pad);
3951 }
3952
3953 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3954                                       int items, int bytes, size_t syncpoff)
3955 {
3956         int i;
3957         int pad = bytes - sizeof(u64) * items;
3958         BUG_ON(pad < 0);
3959
3960         /* Use put_unaligned() because stats may not be aligned for u64. */
3961         put_unaligned(items, &stats[0]);
3962         for (i = 1; i < items; i++)
3963                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3964
3965         memset(&stats[items], 0, pad);
3966 }
3967
3968 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3969                              int bytes)
3970 {
3971         switch (attrtype) {
3972         case IFLA_INET6_STATS:
3973                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3974                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3975                 break;
3976         case IFLA_INET6_ICMP6STATS:
3977                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3978                 break;
3979         }
3980 }
3981
3982 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3983 {
3984         struct nlattr *nla;
3985         struct ifla_cacheinfo ci;
3986
3987         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3988
3989         ci.max_reasm_len = IPV6_MAXPLEN;
3990         ci.tstamp = cstamp_delta(idev->tstamp);
3991         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3992         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3993         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3994
3995         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3996         if (nla == NULL)
3997                 goto nla_put_failure;
3998         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3999
4000         /* XXX - MC not implemented */
4001
4002         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4003         if (nla == NULL)
4004                 goto nla_put_failure;
4005         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4006
4007         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4008         if (nla == NULL)
4009                 goto nla_put_failure;
4010         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4011
4012         return 0;
4013
4014 nla_put_failure:
4015         return -EMSGSIZE;
4016 }
4017
4018 static size_t inet6_get_link_af_size(const struct net_device *dev)
4019 {
4020         if (!__in6_dev_get(dev))
4021                 return 0;
4022
4023         return inet6_ifla6_size();
4024 }
4025
4026 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4027 {
4028         struct inet6_dev *idev = __in6_dev_get(dev);
4029
4030         if (!idev)
4031                 return -ENODATA;
4032
4033         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4034                 return -EMSGSIZE;
4035
4036         return 0;
4037 }
4038
4039 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4040                              u32 pid, u32 seq, int event, unsigned int flags)
4041 {
4042         struct net_device *dev = idev->dev;
4043         struct ifinfomsg *hdr;
4044         struct nlmsghdr *nlh;
4045         void *protoinfo;
4046
4047         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4048         if (nlh == NULL)
4049                 return -EMSGSIZE;
4050
4051         hdr = nlmsg_data(nlh);
4052         hdr->ifi_family = AF_INET6;
4053         hdr->__ifi_pad = 0;
4054         hdr->ifi_type = dev->type;
4055         hdr->ifi_index = dev->ifindex;
4056         hdr->ifi_flags = dev_get_flags(dev);
4057         hdr->ifi_change = 0;
4058
4059         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4060
4061         if (dev->addr_len)
4062                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4063
4064         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4065         if (dev->ifindex != dev->iflink)
4066                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4067
4068         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4069         if (protoinfo == NULL)
4070                 goto nla_put_failure;
4071
4072         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4073                 goto nla_put_failure;
4074
4075         nla_nest_end(skb, protoinfo);
4076         return nlmsg_end(skb, nlh);
4077
4078 nla_put_failure:
4079         nlmsg_cancel(skb, nlh);
4080         return -EMSGSIZE;
4081 }
4082
4083 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4084 {
4085         struct net *net = sock_net(skb->sk);
4086         int h, s_h;
4087         int idx = 0, s_idx;
4088         struct net_device *dev;
4089         struct inet6_dev *idev;
4090         struct hlist_head *head;
4091         struct hlist_node *node;
4092
4093         s_h = cb->args[0];
4094         s_idx = cb->args[1];
4095
4096         rcu_read_lock();
4097         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4098                 idx = 0;
4099                 head = &net->dev_index_head[h];
4100                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4101                         if (idx < s_idx)
4102                                 goto cont;
4103                         idev = __in6_dev_get(dev);
4104                         if (!idev)
4105                                 goto cont;
4106                         if (inet6_fill_ifinfo(skb, idev,
4107                                               NETLINK_CB(cb->skb).pid,
4108                                               cb->nlh->nlmsg_seq,
4109                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4110                                 goto out;
4111 cont:
4112                         idx++;
4113                 }
4114         }
4115 out:
4116         rcu_read_unlock();
4117         cb->args[1] = idx;
4118         cb->args[0] = h;
4119
4120         return skb->len;
4121 }
4122
4123 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4124 {
4125         struct sk_buff *skb;
4126         struct net *net = dev_net(idev->dev);
4127         int err = -ENOBUFS;
4128
4129         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4130         if (skb == NULL)
4131                 goto errout;
4132
4133         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4134         if (err < 0) {
4135                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4136                 WARN_ON(err == -EMSGSIZE);
4137                 kfree_skb(skb);
4138                 goto errout;
4139         }
4140         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4141         return;
4142 errout:
4143         if (err < 0)
4144                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4145 }
4146
4147 static inline size_t inet6_prefix_nlmsg_size(void)
4148 {
4149         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4150                + nla_total_size(sizeof(struct in6_addr))
4151                + nla_total_size(sizeof(struct prefix_cacheinfo));
4152 }
4153
4154 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4155                              struct prefix_info *pinfo, u32 pid, u32 seq,
4156                              int event, unsigned int flags)
4157 {
4158         struct prefixmsg *pmsg;
4159         struct nlmsghdr *nlh;
4160         struct prefix_cacheinfo ci;
4161
4162         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4163         if (nlh == NULL)
4164                 return -EMSGSIZE;
4165
4166         pmsg = nlmsg_data(nlh);
4167         pmsg->prefix_family = AF_INET6;
4168         pmsg->prefix_pad1 = 0;
4169         pmsg->prefix_pad2 = 0;
4170         pmsg->prefix_ifindex = idev->dev->ifindex;
4171         pmsg->prefix_len = pinfo->prefix_len;
4172         pmsg->prefix_type = pinfo->type;
4173         pmsg->prefix_pad3 = 0;
4174         pmsg->prefix_flags = 0;
4175         if (pinfo->onlink)
4176                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4177         if (pinfo->autoconf)
4178                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4179
4180         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4181
4182         ci.preferred_time = ntohl(pinfo->prefered);
4183         ci.valid_time = ntohl(pinfo->valid);
4184         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4185
4186         return nlmsg_end(skb, nlh);
4187
4188 nla_put_failure:
4189         nlmsg_cancel(skb, nlh);
4190         return -EMSGSIZE;
4191 }
4192
4193 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4194                          struct prefix_info *pinfo)
4195 {
4196         struct sk_buff *skb;
4197         struct net *net = dev_net(idev->dev);
4198         int err = -ENOBUFS;
4199
4200         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4201         if (skb == NULL)
4202                 goto errout;
4203
4204         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4205         if (err < 0) {
4206                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4207                 WARN_ON(err == -EMSGSIZE);
4208                 kfree_skb(skb);
4209                 goto errout;
4210         }
4211         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4212         return;
4213 errout:
4214         if (err < 0)
4215                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4216 }
4217
4218 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4219 {
4220         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4221
4222         switch (event) {
4223         case RTM_NEWADDR:
4224                 /*
4225                  * If the address was optimistic
4226                  * we inserted the route at the start of
4227                  * our DAD process, so we don't need
4228                  * to do it again
4229                  */
4230                 if (!(ifp->rt->rt6i_node))
4231                         ip6_ins_rt(ifp->rt);
4232                 if (ifp->idev->cnf.forwarding)
4233                         addrconf_join_anycast(ifp);
4234                 break;
4235         case RTM_DELADDR:
4236                 if (ifp->idev->cnf.forwarding)
4237                         addrconf_leave_anycast(ifp);
4238                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4239                 dst_hold(&ifp->rt->dst);
4240
4241                 if (ip6_del_rt(ifp->rt))
4242                         dst_free(&ifp->rt->dst);
4243                 break;
4244         }
4245 }
4246
4247 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4248 {
4249         rcu_read_lock_bh();
4250         if (likely(ifp->idev->dead == 0))
4251                 __ipv6_ifa_notify(event, ifp);
4252         rcu_read_unlock_bh();
4253 }
4254
4255 #ifdef CONFIG_SYSCTL
4256
4257 static
4258 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4259                            void __user *buffer, size_t *lenp, loff_t *ppos)
4260 {
4261         int *valp = ctl->data;
4262         int val = *valp;
4263         loff_t pos = *ppos;
4264         int ret;
4265
4266         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4267
4268         if (write)
4269                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4270         if (ret)
4271                 *ppos = pos;
4272         return ret;
4273 }
4274
4275 static void dev_disable_change(struct inet6_dev *idev)
4276 {
4277         if (!idev || !idev->dev)
4278                 return;
4279
4280         if (idev->cnf.disable_ipv6)
4281                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4282         else
4283                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4284 }
4285
4286 static void addrconf_disable_change(struct net *net, __s32 newf)
4287 {
4288         struct net_device *dev;
4289         struct inet6_dev *idev;
4290
4291         rcu_read_lock();
4292         for_each_netdev_rcu(net, dev) {
4293                 idev = __in6_dev_get(dev);
4294                 if (idev) {
4295                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4296                         idev->cnf.disable_ipv6 = newf;
4297                         if (changed)
4298                                 dev_disable_change(idev);
4299                 }
4300         }
4301         rcu_read_unlock();
4302 }
4303
4304 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4305 {
4306         struct net *net;
4307
4308         net = (struct net *)table->extra2;
4309
4310         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4311                 return 0;
4312
4313         if (!rtnl_trylock()) {
4314                 /* Restore the original values before restarting */
4315                 *p = old;
4316                 return restart_syscall();
4317         }
4318
4319         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4320                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4321                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4322                 addrconf_disable_change(net, newf);
4323         } else if ((!*p) ^ (!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         int ret;
4338
4339         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4340
4341         if (write)
4342                 ret = addrconf_disable_ipv6(ctl, valp, val);
4343         if (ret)
4344                 *ppos = pos;
4345         return ret;
4346 }
4347
4348 static struct addrconf_sysctl_table
4349 {
4350         struct ctl_table_header *sysctl_header;
4351         ctl_table addrconf_vars[DEVCONF_MAX+1];
4352         char *dev_name;
4353 } addrconf_sysctl __read_mostly = {
4354         .sysctl_header = NULL,
4355         .addrconf_vars = {
4356                 {
4357                         .procname       = "forwarding",
4358                         .data           = &ipv6_devconf.forwarding,
4359                         .maxlen         = sizeof(int),
4360                         .mode           = 0644,
4361                         .proc_handler   = addrconf_sysctl_forward,
4362                 },
4363                 {
4364                         .procname       = "hop_limit",
4365                         .data           = &ipv6_devconf.hop_limit,
4366                         .maxlen         = sizeof(int),
4367                         .mode           = 0644,
4368                         .proc_handler   = proc_dointvec,
4369                 },
4370                 {
4371                         .procname       = "mtu",
4372                         .data           = &ipv6_devconf.mtu6,
4373                         .maxlen         = sizeof(int),
4374                         .mode           = 0644,
4375                         .proc_handler   = proc_dointvec,
4376                 },
4377                 {
4378                         .procname       = "accept_ra",
4379                         .data           = &ipv6_devconf.accept_ra,
4380                         .maxlen         = sizeof(int),
4381                         .mode           = 0644,
4382                         .proc_handler   = proc_dointvec,
4383                 },
4384                 {
4385                         .procname       = "accept_redirects",
4386                         .data           = &ipv6_devconf.accept_redirects,
4387                         .maxlen         = sizeof(int),
4388                         .mode           = 0644,
4389                         .proc_handler   = proc_dointvec,
4390                 },
4391                 {
4392                         .procname       = "autoconf",
4393                         .data           = &ipv6_devconf.autoconf,
4394                         .maxlen         = sizeof(int),
4395                         .mode           = 0644,
4396                         .proc_handler   = proc_dointvec,
4397                 },
4398                 {
4399                         .procname       = "dad_transmits",
4400                         .data           = &ipv6_devconf.dad_transmits,
4401                         .maxlen         = sizeof(int),
4402                         .mode           = 0644,
4403                         .proc_handler   = proc_dointvec,
4404                 },
4405                 {
4406                         .procname       = "router_solicitations",
4407                         .data           = &ipv6_devconf.rtr_solicits,
4408                         .maxlen         = sizeof(int),
4409                         .mode           = 0644,
4410                         .proc_handler   = proc_dointvec,
4411                 },
4412                 {
4413                         .procname       = "router_solicitation_interval",
4414                         .data           = &ipv6_devconf.rtr_solicit_interval,
4415                         .maxlen         = sizeof(int),
4416                         .mode           = 0644,
4417                         .proc_handler   = proc_dointvec_jiffies,
4418                 },
4419                 {
4420                         .procname       = "router_solicitation_delay",
4421                         .data           = &ipv6_devconf.rtr_solicit_delay,
4422                         .maxlen         = sizeof(int),
4423                         .mode           = 0644,
4424                         .proc_handler   = proc_dointvec_jiffies,
4425                 },
4426                 {
4427                         .procname       = "force_mld_version",
4428                         .data           = &ipv6_devconf.force_mld_version,
4429                         .maxlen         = sizeof(int),
4430                         .mode           = 0644,
4431                         .proc_handler   = proc_dointvec,
4432                 },
4433 #ifdef CONFIG_IPV6_PRIVACY
4434                 {
4435                         .procname       = "use_tempaddr",
4436                         .data           = &ipv6_devconf.use_tempaddr,
4437                         .maxlen         = sizeof(int),
4438                         .mode           = 0644,
4439                         .proc_handler   = proc_dointvec,
4440                 },
4441                 {
4442                         .procname       = "temp_valid_lft",
4443                         .data           = &ipv6_devconf.temp_valid_lft,
4444                         .maxlen         = sizeof(int),
4445                         .mode           = 0644,
4446                         .proc_handler   = proc_dointvec,
4447                 },
4448                 {
4449                         .procname       = "temp_prefered_lft",
4450                         .data           = &ipv6_devconf.temp_prefered_lft,
4451                         .maxlen         = sizeof(int),
4452                         .mode           = 0644,
4453                         .proc_handler   = proc_dointvec,
4454                 },
4455                 {
4456                         .procname       = "regen_max_retry",
4457                         .data           = &ipv6_devconf.regen_max_retry,
4458                         .maxlen         = sizeof(int),
4459                         .mode           = 0644,
4460                         .proc_handler   = proc_dointvec,
4461                 },
4462                 {
4463                         .procname       = "max_desync_factor",
4464                         .data           = &ipv6_devconf.max_desync_factor,
4465                         .maxlen         = sizeof(int),
4466                         .mode           = 0644,
4467                         .proc_handler   = proc_dointvec,
4468                 },
4469 #endif
4470                 {
4471                         .procname       = "max_addresses",
4472                         .data           = &ipv6_devconf.max_addresses,
4473                         .maxlen         = sizeof(int),
4474                         .mode           = 0644,
4475                         .proc_handler   = proc_dointvec,
4476                 },
4477                 {
4478                         .procname       = "accept_ra_defrtr",
4479                         .data           = &ipv6_devconf.accept_ra_defrtr,
4480                         .maxlen         = sizeof(int),
4481                         .mode           = 0644,
4482                         .proc_handler   = proc_dointvec,
4483                 },
4484                 {
4485                         .procname       = "accept_ra_pinfo",
4486                         .data           = &ipv6_devconf.accept_ra_pinfo,
4487                         .maxlen         = sizeof(int),
4488                         .mode           = 0644,
4489                         .proc_handler   = proc_dointvec,
4490                 },
4491 #ifdef CONFIG_IPV6_ROUTER_PREF
4492                 {
4493                         .procname       = "accept_ra_rtr_pref",
4494                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4495                         .maxlen         = sizeof(int),
4496                         .mode           = 0644,
4497                         .proc_handler   = proc_dointvec,
4498                 },
4499                 {
4500                         .procname       = "router_probe_interval",
4501                         .data           = &ipv6_devconf.rtr_probe_interval,
4502                         .maxlen         = sizeof(int),
4503                         .mode           = 0644,
4504                         .proc_handler   = proc_dointvec_jiffies,
4505                 },
4506 #ifdef CONFIG_IPV6_ROUTE_INFO
4507                 {
4508                         .procname       = "accept_ra_rt_info_max_plen",
4509                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4510                         .maxlen         = sizeof(int),
4511                         .mode           = 0644,
4512                         .proc_handler   = proc_dointvec,
4513                 },
4514 #endif
4515 #endif
4516                 {
4517                         .procname       = "proxy_ndp",
4518                         .data           = &ipv6_devconf.proxy_ndp,
4519                         .maxlen         = sizeof(int),
4520                         .mode           = 0644,
4521                         .proc_handler   = proc_dointvec,
4522                 },
4523                 {
4524                         .procname       = "accept_source_route",
4525                         .data           = &ipv6_devconf.accept_source_route,
4526                         .maxlen         = sizeof(int),
4527                         .mode           = 0644,
4528                         .proc_handler   = proc_dointvec,
4529                 },
4530 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4531                 {
4532                         .procname       = "optimistic_dad",
4533                         .data           = &ipv6_devconf.optimistic_dad,
4534                         .maxlen         = sizeof(int),
4535                         .mode           = 0644,
4536                         .proc_handler   = proc_dointvec,
4537
4538                 },
4539 #endif
4540 #ifdef CONFIG_IPV6_MROUTE
4541                 {
4542                         .procname       = "mc_forwarding",
4543                         .data           = &ipv6_devconf.mc_forwarding,
4544                         .maxlen         = sizeof(int),
4545                         .mode           = 0444,
4546                         .proc_handler   = proc_dointvec,
4547                 },
4548 #endif
4549                 {
4550                         .procname       = "disable_ipv6",
4551                         .data           = &ipv6_devconf.disable_ipv6,
4552                         .maxlen         = sizeof(int),
4553                         .mode           = 0644,
4554                         .proc_handler   = addrconf_sysctl_disable,
4555                 },
4556                 {
4557                         .procname       = "accept_dad",
4558                         .data           = &ipv6_devconf.accept_dad,
4559                         .maxlen         = sizeof(int),
4560                         .mode           = 0644,
4561                         .proc_handler   = proc_dointvec,
4562                 },
4563                 {
4564                         .procname       = "force_tllao",
4565                         .data           = &ipv6_devconf.force_tllao,
4566                         .maxlen         = sizeof(int),
4567                         .mode           = 0644,
4568                         .proc_handler   = proc_dointvec
4569                 },
4570                 {
4571                         /* sentinel */
4572                 }
4573         },
4574 };
4575
4576 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4577                 struct inet6_dev *idev, struct ipv6_devconf *p)
4578 {
4579         int i;
4580         struct addrconf_sysctl_table *t;
4581
4582 #define ADDRCONF_CTL_PATH_DEV   3
4583
4584         struct ctl_path addrconf_ctl_path[] = {
4585                 { .procname = "net", },
4586                 { .procname = "ipv6", },
4587                 { .procname = "conf", },
4588                 { /* to be set */ },
4589                 { },
4590         };
4591
4592
4593         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4594         if (t == NULL)
4595                 goto out;
4596
4597         for (i = 0; t->addrconf_vars[i].data; i++) {
4598                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4599                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4600                 t->addrconf_vars[i].extra2 = net;
4601         }
4602
4603         /*
4604          * Make a copy of dev_name, because '.procname' is regarded as const
4605          * by sysctl and we wouldn't want anyone to change it under our feet
4606          * (see SIOCSIFNAME).
4607          */
4608         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4609         if (!t->dev_name)
4610                 goto free;
4611
4612         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4613
4614         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4615                         t->addrconf_vars);
4616         if (t->sysctl_header == NULL)
4617                 goto free_procname;
4618
4619         p->sysctl = t;
4620         return 0;
4621
4622 free_procname:
4623         kfree(t->dev_name);
4624 free:
4625         kfree(t);
4626 out:
4627         return -ENOBUFS;
4628 }
4629
4630 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4631 {
4632         struct addrconf_sysctl_table *t;
4633
4634         if (p->sysctl == NULL)
4635                 return;
4636
4637         t = p->sysctl;
4638         p->sysctl = NULL;
4639         unregister_net_sysctl_table(t->sysctl_header);
4640         kfree(t->dev_name);
4641         kfree(t);
4642 }
4643
4644 static void addrconf_sysctl_register(struct inet6_dev *idev)
4645 {
4646         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4647                               &ndisc_ifinfo_sysctl_change);
4648         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4649                                         idev, &idev->cnf);
4650 }
4651
4652 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4653 {
4654         __addrconf_sysctl_unregister(&idev->cnf);
4655         neigh_sysctl_unregister(idev->nd_parms);
4656 }
4657
4658
4659 #endif
4660
4661 static int __net_init addrconf_init_net(struct net *net)
4662 {
4663         int err;
4664         struct ipv6_devconf *all, *dflt;
4665
4666         err = -ENOMEM;
4667         all = &ipv6_devconf;
4668         dflt = &ipv6_devconf_dflt;
4669
4670         if (!net_eq(net, &init_net)) {
4671                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4672                 if (all == NULL)
4673                         goto err_alloc_all;
4674
4675                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4676                 if (dflt == NULL)
4677                         goto err_alloc_dflt;
4678         } else {
4679                 /* these will be inherited by all namespaces */
4680                 dflt->autoconf = ipv6_defaults.autoconf;
4681                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4682         }
4683
4684         net->ipv6.devconf_all = all;
4685         net->ipv6.devconf_dflt = dflt;
4686
4687 #ifdef CONFIG_SYSCTL
4688         err = __addrconf_sysctl_register(net, "all", NULL, all);
4689         if (err < 0)
4690                 goto err_reg_all;
4691
4692         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4693         if (err < 0)
4694                 goto err_reg_dflt;
4695 #endif
4696         return 0;
4697
4698 #ifdef CONFIG_SYSCTL
4699 err_reg_dflt:
4700         __addrconf_sysctl_unregister(all);
4701 err_reg_all:
4702         kfree(dflt);
4703 #endif
4704 err_alloc_dflt:
4705         kfree(all);
4706 err_alloc_all:
4707         return err;
4708 }
4709
4710 static void __net_exit addrconf_exit_net(struct net *net)
4711 {
4712 #ifdef CONFIG_SYSCTL
4713         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4714         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4715 #endif
4716         if (!net_eq(net, &init_net)) {
4717                 kfree(net->ipv6.devconf_dflt);
4718                 kfree(net->ipv6.devconf_all);
4719         }
4720 }
4721
4722 static struct pernet_operations addrconf_ops = {
4723         .init = addrconf_init_net,
4724         .exit = addrconf_exit_net,
4725 };
4726
4727 /*
4728  *      Device notifier
4729  */
4730
4731 int register_inet6addr_notifier(struct notifier_block *nb)
4732 {
4733         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4734 }
4735 EXPORT_SYMBOL(register_inet6addr_notifier);
4736
4737 int unregister_inet6addr_notifier(struct notifier_block *nb)
4738 {
4739         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4740 }
4741 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4742
4743 static struct rtnl_af_ops inet6_ops = {
4744         .family           = AF_INET6,
4745         .fill_link_af     = inet6_fill_link_af,
4746         .get_link_af_size = inet6_get_link_af_size,
4747 };
4748
4749 /*
4750  *      Init / cleanup code
4751  */
4752
4753 int __init addrconf_init(void)
4754 {
4755         int i, err;
4756
4757         err = ipv6_addr_label_init();
4758         if (err < 0) {
4759                 printk(KERN_CRIT "IPv6 Addrconf:"
4760                        " cannot initialize default policy table: %d.\n", err);
4761                 goto out;
4762         }
4763
4764         err = register_pernet_subsys(&addrconf_ops);
4765         if (err < 0)
4766                 goto out_addrlabel;
4767
4768         /* The addrconf netdev notifier requires that loopback_dev
4769          * has it's ipv6 private information allocated and setup
4770          * before it can bring up and give link-local addresses
4771          * to other devices which are up.
4772          *
4773          * Unfortunately, loopback_dev is not necessarily the first
4774          * entry in the global dev_base list of net devices.  In fact,
4775          * it is likely to be the very last entry on that list.
4776          * So this causes the notifier registry below to try and
4777          * give link-local addresses to all devices besides loopback_dev
4778          * first, then loopback_dev, which cases all the non-loopback_dev
4779          * devices to fail to get a link-local address.
4780          *
4781          * So, as a temporary fix, allocate the ipv6 structure for
4782          * loopback_dev first by hand.
4783          * Longer term, all of the dependencies ipv6 has upon the loopback
4784          * device and it being up should be removed.
4785          */
4786         rtnl_lock();
4787         if (!ipv6_add_dev(init_net.loopback_dev))
4788                 err = -ENOMEM;
4789         rtnl_unlock();
4790         if (err)
4791                 goto errlo;
4792
4793         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4794                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4795
4796         register_netdevice_notifier(&ipv6_dev_notf);
4797
4798         addrconf_verify(0);
4799
4800         err = rtnl_af_register(&inet6_ops);
4801         if (err < 0)
4802                 goto errout_af;
4803
4804         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4805                               NULL);
4806         if (err < 0)
4807                 goto errout;
4808
4809         /* Only the first call to __rtnl_register can fail */
4810         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4811         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4812         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4813                         inet6_dump_ifaddr, NULL);
4814         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4815                         inet6_dump_ifmcaddr, NULL);
4816         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4817                         inet6_dump_ifacaddr, NULL);
4818
4819         ipv6_addr_label_rtnl_register();
4820
4821         return 0;
4822 errout:
4823         rtnl_af_unregister(&inet6_ops);
4824 errout_af:
4825         unregister_netdevice_notifier(&ipv6_dev_notf);
4826 errlo:
4827         unregister_pernet_subsys(&addrconf_ops);
4828 out_addrlabel:
4829         ipv6_addr_label_cleanup();
4830 out:
4831         return err;
4832 }
4833
4834 void addrconf_cleanup(void)
4835 {
4836         struct net_device *dev;
4837         int i;
4838
4839         unregister_netdevice_notifier(&ipv6_dev_notf);
4840         unregister_pernet_subsys(&addrconf_ops);
4841         ipv6_addr_label_cleanup();
4842
4843         rtnl_lock();
4844
4845         __rtnl_af_unregister(&inet6_ops);
4846
4847         /* clean dev list */
4848         for_each_netdev(&init_net, dev) {
4849                 if (__in6_dev_get(dev) == NULL)
4850                         continue;
4851                 addrconf_ifdown(dev, 1);
4852         }
4853         addrconf_ifdown(init_net.loopback_dev, 2);
4854
4855         /*
4856          *      Check hash table.
4857          */
4858         spin_lock_bh(&addrconf_hash_lock);
4859         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4860                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4861         spin_unlock_bh(&addrconf_hash_lock);
4862
4863         del_timer(&addr_chk_timer);
4864         rtnl_unlock();
4865 }