net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/ipv6.h>
  74 #include <net/protocol.h>
  75 #include <net/ndisc.h>
  76 #include <net/ip6_route.h>
  77 #include <net/addrconf.h>
  78 #include <net/tcp.h>
  79 #include <net/ip.h>
  80 #include <net/netlink.h>
  81 #include <net/pkt_sched.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85
  86 #ifdef CONFIG_IPV6_PRIVACY
  87 #include <linux/random.h>
  88 #endif
  89
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 /* Set to 3 to get tracing... */
  98 #define ACONF_DEBUG 2
  99
 100 #if ACONF_DEBUG >= 3
 101 #define ADBG(x) printk x
 102 #else
 103 #define ADBG(x)
 104 #endif
 105
 106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 107
 108 static inline u32 cstamp_delta(unsigned long cstamp)
 109 {
 110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 111 }
 112
 113 #ifdef CONFIG_SYSCTL
 114 static void addrconf_sysctl_register(struct inet6_dev *idev);
 115 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 116 #else
 117 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 118 {
 119 }
 120
 121 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 122 {
 123 }
 124 #endif
 125
 126 #ifdef CONFIG_IPV6_PRIVACY
 127 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 128 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 129 static void ipv6_regen_rndid(unsigned long data);
 130 #endif
 131
 132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 133 static int ipv6_count_addresses(struct inet6_dev *idev);
 134
 135 /*
 136  *      Configured unicast address hash table
 137  */
 138 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 139 static DEFINE_SPINLOCK(addrconf_hash_lock);
 140
 141 static void addrconf_verify(unsigned long);
 142
 143 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 144 static DEFINE_SPINLOCK(addrconf_verify_lock);
 145
 146 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 147 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 148
 149 static void addrconf_type_change(struct net_device *dev,
 150                                  unsigned long event);
 151 static int addrconf_ifdown(struct net_device *dev, int how);
 152
 153 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 154                                                   int plen,
 155                                                   const struct net_device *dev,
 156                                                   u32 flags, u32 noflags);
 157
 158 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 159 static void addrconf_dad_timer(unsigned long data);
 160 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 161 static void addrconf_dad_run(struct inet6_dev *idev);
 162 static void addrconf_rs_timer(unsigned long data);
 163 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 164 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165
 166 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 167                                 struct prefix_info *pinfo);
 168 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 169                                struct net_device *dev);
 170
 171 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 172
 173 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 174         .forwarding             = 0,
 175         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 176         .mtu6                   = IPV6_MIN_MTU,
 177         .accept_ra              = 1,
 178         .accept_redirects       = 1,
 179         .autoconf               = 1,
 180         .force_mld_version      = 0,
 181         .dad_transmits          = 1,
 182         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 183         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 184         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 185 #ifdef CONFIG_IPV6_PRIVACY
 186         .use_tempaddr           = 0,
 187         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 188         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 189         .regen_max_retry        = REGEN_MAX_RETRY,
 190         .max_desync_factor      = MAX_DESYNC_FACTOR,
 191 #endif
 192         .max_addresses          = IPV6_MAX_ADDRESSES,
 193         .accept_ra_defrtr       = 1,
 194         .accept_ra_pinfo        = 1,
 195 #ifdef CONFIG_IPV6_ROUTER_PREF
 196         .accept_ra_rtr_pref     = 1,
 197         .rtr_probe_interval     = 60 * HZ,
 198 #ifdef CONFIG_IPV6_ROUTE_INFO
 199         .accept_ra_rt_info_max_plen = 0,
 200 #endif
 201 #endif
 202         .proxy_ndp              = 0,
 203         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 204         .disable_ipv6           = 0,
 205         .accept_dad             = 1,
 206 };
 207
 208 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 209         .forwarding             = 0,
 210         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 211         .mtu6                   = IPV6_MIN_MTU,
 212         .accept_ra              = 1,
 213         .accept_redirects       = 1,
 214         .autoconf               = 1,
 215         .dad_transmits          = 1,
 216         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 217         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 218         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 219 #ifdef CONFIG_IPV6_PRIVACY
 220         .use_tempaddr           = 0,
 221         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 222         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 223         .regen_max_retry        = REGEN_MAX_RETRY,
 224         .max_desync_factor      = MAX_DESYNC_FACTOR,
 225 #endif
 226         .max_addresses          = IPV6_MAX_ADDRESSES,
 227         .accept_ra_defrtr       = 1,
 228         .accept_ra_pinfo        = 1,
 229 #ifdef CONFIG_IPV6_ROUTER_PREF
 230         .accept_ra_rtr_pref     = 1,
 231         .rtr_probe_interval     = 60 * HZ,
 232 #ifdef CONFIG_IPV6_ROUTE_INFO
 233         .accept_ra_rt_info_max_plen = 0,
 234 #endif
 235 #endif
 236         .proxy_ndp              = 0,
 237         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 238         .disable_ipv6           = 0,
 239         .accept_dad             = 1,
 240 };
 241
 242 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 243 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 244 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 245 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 246 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 247 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 248 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 249 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 250
 251 /* Check if a valid qdisc is available */
 252 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 253 {
 254         return !qdisc_tx_is_noop(dev);
 255 }
 256
 257 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 258 {
 259         if (del_timer(&ifp->timer))
 260                 __in6_ifa_put(ifp);
 261 }
 262
 263 enum addrconf_timer_t {
 264         AC_NONE,
 265         AC_DAD,
 266         AC_RS,
 267 };
 268
 269 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 270                                enum addrconf_timer_t what,
 271                                unsigned long when)
 272 {
 273         if (!del_timer(&ifp->timer))
 274                 in6_ifa_hold(ifp);
 275
 276         switch (what) {
 277         case AC_DAD:
 278                 ifp->timer.function = addrconf_dad_timer;
 279                 break;
 280         case AC_RS:
 281                 ifp->timer.function = addrconf_rs_timer;
 282                 break;
 283         default:
 284                 break;
 285         }
 286         ifp->timer.expires = jiffies + when;
 287         add_timer(&ifp->timer);
 288 }
 289
 290 static int snmp6_alloc_dev(struct inet6_dev *idev)
 291 {
 292         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 293                           sizeof(struct ipstats_mib),
 294                           __alignof__(struct ipstats_mib)) < 0)
 295                 goto err_ip;
 296         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 297                                         GFP_KERNEL);
 298         if (!idev->stats.icmpv6dev)
 299                 goto err_icmp;
 300         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 301                                            GFP_KERNEL);
 302         if (!idev->stats.icmpv6msgdev)
 303                 goto err_icmpmsg;
 304
 305         return 0;
 306
 307 err_icmpmsg:
 308         kfree(idev->stats.icmpv6dev);
 309 err_icmp:
 310         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 311 err_ip:
 312         return -ENOMEM;
 313 }
 314
 315 static void snmp6_free_dev(struct inet6_dev *idev)
 316 {
 317         kfree(idev->stats.icmpv6msgdev);
 318         kfree(idev->stats.icmpv6dev);
 319         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 320 }
 321
 322 /* Nobody refers to this device, we may destroy it. */
 323
 324 void in6_dev_finish_destroy(struct inet6_dev *idev)
 325 {
 326         struct net_device *dev = idev->dev;
 327
 328         WARN_ON(!list_empty(&idev->addr_list));
 329         WARN_ON(idev->mc_list != NULL);
 330
 331 #ifdef NET_REFCNT_DEBUG
 332         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 333 #endif
 334         dev_put(dev);
 335         if (!idev->dead) {
 336                 pr_warn("Freeing alive inet6 device %p\n", idev);
 337                 return;
 338         }
 339         snmp6_free_dev(idev);
 340         kfree_rcu(idev, rcu);
 341 }
 342 EXPORT_SYMBOL(in6_dev_finish_destroy);
 343
 344 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 345 {
 346         struct inet6_dev *ndev;
 347
 348         ASSERT_RTNL();
 349
 350         if (dev->mtu < IPV6_MIN_MTU)
 351                 return NULL;
 352
 353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 354
 355         if (ndev == NULL)
 356                 return NULL;
 357
 358         rwlock_init(&ndev->lock);
 359         ndev->dev = dev;
 360         INIT_LIST_HEAD(&ndev->addr_list);
 361
 362         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 363         ndev->cnf.mtu6 = dev->mtu;
 364         ndev->cnf.sysctl = NULL;
 365         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 366         if (ndev->nd_parms == NULL) {
 367                 kfree(ndev);
 368                 return NULL;
 369         }
 370         if (ndev->cnf.forwarding)
 371                 dev_disable_lro(dev);
 372         /* We refer to the device */
 373         dev_hold(dev);
 374
 375         if (snmp6_alloc_dev(ndev) < 0) {
 376                 ADBG((KERN_WARNING
 377                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 378                         __func__, dev->name));
 379                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 380                 dev_put(dev);
 381                 kfree(ndev);
 382                 return NULL;
 383         }
 384
 385         if (snmp6_register_dev(ndev) < 0) {
 386                 ADBG((KERN_WARNING
 387                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 388                         __func__, dev->name));
 389                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 390                 ndev->dead = 1;
 391                 in6_dev_finish_destroy(ndev);
 392                 return NULL;
 393         }
 394
 395         /* One reference from device.  We must do this before
 396          * we invoke __ipv6_regen_rndid().
 397          */
 398         in6_dev_hold(ndev);
 399
 400         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 401                 ndev->cnf.accept_dad = -1;
 402
 403 #if IS_ENABLED(CONFIG_IPV6_SIT)
 404         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 405                 pr_info("%s: Disabled Multicast RS\n", 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 interface-local all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 436
 437         /* Join all-node multicast group */
 438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 439
 440         /* Join all-router multicast group if forwarding is set */
 441         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 442                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 443
 444         return ndev;
 445 }
 446
 447 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 448 {
 449         struct inet6_dev *idev;
 450
 451         ASSERT_RTNL();
 452
 453         idev = __in6_dev_get(dev);
 454         if (!idev) {
 455                 idev = ipv6_add_dev(dev);
 456                 if (!idev)
 457                         return NULL;
 458         }
 459
 460         if (dev->flags&IFF_UP)
 461                 ipv6_mc_up(idev);
 462         return idev;
 463 }
 464
 465 static int inet6_netconf_msgsize_devconf(int type)
 466 {
 467         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 468                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 469
 470         /* type -1 is used for ALL */
 471         if (type == -1 || type == NETCONFA_FORWARDING)
 472                 size += nla_total_size(4);
 473 #ifdef CONFIG_IPV6_MROUTE
 474         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 475                 size += nla_total_size(4);
 476 #endif
 477
 478         return size;
 479 }
 480
 481 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 482                                       struct ipv6_devconf *devconf, u32 portid,
 483                                       u32 seq, int event, unsigned int flags,
 484                                       int type)
 485 {
 486         struct nlmsghdr  *nlh;
 487         struct netconfmsg *ncm;
 488
 489         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 490                         flags);
 491         if (nlh == NULL)
 492                 return -EMSGSIZE;
 493
 494         ncm = nlmsg_data(nlh);
 495         ncm->ncm_family = AF_INET6;
 496
 497         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 498                 goto nla_put_failure;
 499
 500         /* type -1 is used for ALL */
 501         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 502             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 503                 goto nla_put_failure;
 504 #ifdef CONFIG_IPV6_MROUTE
 505         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 506             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 507                         devconf->mc_forwarding) < 0)
 508                 goto nla_put_failure;
 509 #endif
 510         return nlmsg_end(skb, nlh);
 511
 512 nla_put_failure:
 513         nlmsg_cancel(skb, nlh);
 514         return -EMSGSIZE;
 515 }
 516
 517 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 518                                   struct ipv6_devconf *devconf)
 519 {
 520         struct sk_buff *skb;
 521         int err = -ENOBUFS;
 522
 523         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 524         if (skb == NULL)
 525                 goto errout;
 526
 527         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 528                                          RTM_NEWNETCONF, 0, type);
 529         if (err < 0) {
 530                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 531                 WARN_ON(err == -EMSGSIZE);
 532                 kfree_skb(skb);
 533                 goto errout;
 534         }
 535         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 536         return;
 537 errout:
 538         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 539 }
 540
 541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 542         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 543         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 544 };
 545
 546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 547                                      struct nlmsghdr *nlh)
 548 {
 549         struct net *net = sock_net(in_skb->sk);
 550         struct nlattr *tb[NETCONFA_MAX+1];
 551         struct netconfmsg *ncm;
 552         struct sk_buff *skb;
 553         struct ipv6_devconf *devconf;
 554         struct inet6_dev *in6_dev;
 555         struct net_device *dev;
 556         int ifindex;
 557         int err;
 558
 559         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 560                           devconf_ipv6_policy);
 561         if (err < 0)
 562                 goto errout;
 563
 564         err = EINVAL;
 565         if (!tb[NETCONFA_IFINDEX])
 566                 goto errout;
 567
 568         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 569         switch (ifindex) {
 570         case NETCONFA_IFINDEX_ALL:
 571                 devconf = net->ipv6.devconf_all;
 572                 break;
 573         case NETCONFA_IFINDEX_DEFAULT:
 574                 devconf = net->ipv6.devconf_dflt;
 575                 break;
 576         default:
 577                 dev = __dev_get_by_index(net, ifindex);
 578                 if (dev == NULL)
 579                         goto errout;
 580                 in6_dev = __in6_dev_get(dev);
 581                 if (in6_dev == NULL)
 582                         goto errout;
 583                 devconf = &in6_dev->cnf;
 584                 break;
 585         }
 586
 587         err = -ENOBUFS;
 588         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 589         if (skb == NULL)
 590                 goto errout;
 591
 592         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 593                                          NETLINK_CB(in_skb).portid,
 594                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 595                                          -1);
 596         if (err < 0) {
 597                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 598                 WARN_ON(err == -EMSGSIZE);
 599                 kfree_skb(skb);
 600                 goto errout;
 601         }
 602         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 603 errout:
 604         return err;
 605 }
 606
 607 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 608                                       struct netlink_callback *cb)
 609 {
 610         struct net *net = sock_net(skb->sk);
 611         int h, s_h;
 612         int idx, s_idx;
 613         struct net_device *dev;
 614         struct inet6_dev *idev;
 615         struct hlist_head *head;
 616
 617         s_h = cb->args[0];
 618         s_idx = idx = cb->args[1];
 619
 620         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 621                 idx = 0;
 622                 head = &net->dev_index_head[h];
 623                 rcu_read_lock();
 624                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 625                           net->dev_base_seq;
 626                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 627                         if (idx < s_idx)
 628                                 goto cont;
 629                         idev = __in6_dev_get(dev);
 630                         if (!idev)
 631                                 goto cont;
 632
 633                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 634                                                        &idev->cnf,
 635                                                        NETLINK_CB(cb->skb).portid,
 636                                                        cb->nlh->nlmsg_seq,
 637                                                        RTM_NEWNETCONF,
 638                                                        NLM_F_MULTI,
 639                                                        -1) <= 0) {
 640                                 rcu_read_unlock();
 641                                 goto done;
 642                         }
 643                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 644 cont:
 645                         idx++;
 646                 }
 647                 rcu_read_unlock();
 648         }
 649         if (h == NETDEV_HASHENTRIES) {
 650                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 651                                                net->ipv6.devconf_all,
 652                                                NETLINK_CB(cb->skb).portid,
 653                                                cb->nlh->nlmsg_seq,
 654                                                RTM_NEWNETCONF, NLM_F_MULTI,
 655                                                -1) <= 0)
 656                         goto done;
 657                 else
 658                         h++;
 659         }
 660         if (h == NETDEV_HASHENTRIES + 1) {
 661                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 662                                                net->ipv6.devconf_dflt,
 663                                                NETLINK_CB(cb->skb).portid,
 664                                                cb->nlh->nlmsg_seq,
 665                                                RTM_NEWNETCONF, NLM_F_MULTI,
 666                                                -1) <= 0)
 667                         goto done;
 668                 else
 669                         h++;
 670         }
 671 done:
 672         cb->args[0] = h;
 673         cb->args[1] = idx;
 674
 675         return skb->len;
 676 }
 677
 678 #ifdef CONFIG_SYSCTL
 679 static void dev_forward_change(struct inet6_dev *idev)
 680 {
 681         struct net_device *dev;
 682         struct inet6_ifaddr *ifa;
 683
 684         if (!idev)
 685                 return;
 686         dev = idev->dev;
 687         if (idev->cnf.forwarding)
 688                 dev_disable_lro(dev);
 689         if (dev->flags & IFF_MULTICAST) {
 690                 if (idev->cnf.forwarding) {
 691                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 692                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 693                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 694                 } else {
 695                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 696                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 697                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 698                 }
 699         }
 700
 701         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 702                 if (ifa->flags&IFA_F_TENTATIVE)
 703                         continue;
 704                 if (idev->cnf.forwarding)
 705                         addrconf_join_anycast(ifa);
 706                 else
 707                         addrconf_leave_anycast(ifa);
 708         }
 709         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 710                                      dev->ifindex, &idev->cnf);
 711 }
 712
 713
 714 static void addrconf_forward_change(struct net *net, __s32 newf)
 715 {
 716         struct net_device *dev;
 717         struct inet6_dev *idev;
 718
 719         for_each_netdev(net, dev) {
 720                 idev = __in6_dev_get(dev);
 721                 if (idev) {
 722                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 723                         idev->cnf.forwarding = newf;
 724                         if (changed)
 725                                 dev_forward_change(idev);
 726                 }
 727         }
 728 }
 729
 730 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 731 {
 732         struct net *net;
 733         int old;
 734
 735         if (!rtnl_trylock())
 736                 return restart_syscall();
 737
 738         net = (struct net *)table->extra2;
 739         old = *p;
 740         *p = newf;
 741
 742         if (p == &net->ipv6.devconf_dflt->forwarding) {
 743                 if ((!newf) ^ (!old))
 744                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 745                                                      NETCONFA_IFINDEX_DEFAULT,
 746                                                      net->ipv6.devconf_dflt);
 747                 rtnl_unlock();
 748                 return 0;
 749         }
 750
 751         if (p == &net->ipv6.devconf_all->forwarding) {
 752                 net->ipv6.devconf_dflt->forwarding = newf;
 753                 addrconf_forward_change(net, newf);
 754                 if ((!newf) ^ (!old))
 755                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 756                                                      NETCONFA_IFINDEX_ALL,
 757                                                      net->ipv6.devconf_all);
 758         } else if ((!newf) ^ (!old))
 759                 dev_forward_change((struct inet6_dev *)table->extra1);
 760         rtnl_unlock();
 761
 762         if (newf)
 763                 rt6_purge_dflt_routers(net);
 764         return 1;
 765 }
 766 #endif
 767
 768 /* Nobody refers to this ifaddr, destroy it */
 769 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 770 {
 771         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 772
 773 #ifdef NET_REFCNT_DEBUG
 774         pr_debug("%s\n", __func__);
 775 #endif
 776
 777         in6_dev_put(ifp->idev);
 778
 779         if (del_timer(&ifp->timer))
 780                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 781
 782         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 783                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 784                 return;
 785         }
 786         ip6_rt_put(ifp->rt);
 787
 788         kfree_rcu(ifp, rcu);
 789 }
 790
 791 static void
 792 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 793 {
 794         struct list_head *p;
 795         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 796
 797         /*
 798          * Each device address list is sorted in order of scope -
 799          * global before linklocal.
 800          */
 801         list_for_each(p, &idev->addr_list) {
 802                 struct inet6_ifaddr *ifa
 803                         = list_entry(p, struct inet6_ifaddr, if_list);
 804                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 805                         break;
 806         }
 807
 808         list_add_tail(&ifp->if_list, p);
 809 }
 810
 811 static u32 inet6_addr_hash(const struct in6_addr *addr)
 812 {
 813         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 814 }
 815
 816 /* On success it returns ifp with increased reference count */
 817
 818 static struct inet6_ifaddr *
 819 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 820               int scope, u32 flags)
 821 {
 822         struct inet6_ifaddr *ifa = NULL;
 823         struct rt6_info *rt;
 824         unsigned int hash;
 825         int err = 0;
 826         int addr_type = ipv6_addr_type(addr);
 827
 828         if (addr_type == IPV6_ADDR_ANY ||
 829             addr_type & IPV6_ADDR_MULTICAST ||
 830             (!(idev->dev->flags & IFF_LOOPBACK) &&
 831              addr_type & IPV6_ADDR_LOOPBACK))
 832                 return ERR_PTR(-EADDRNOTAVAIL);
 833
 834         rcu_read_lock_bh();
 835         if (idev->dead) {
 836                 err = -ENODEV;                  /*XXX*/
 837                 goto out2;
 838         }
 839
 840         if (idev->cnf.disable_ipv6) {
 841                 err = -EACCES;
 842                 goto out2;
 843         }
 844
 845         spin_lock(&addrconf_hash_lock);
 846
 847         /* Ignore adding duplicate addresses on an interface */
 848         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 849                 ADBG(("ipv6_add_addr: already assigned\n"));
 850                 err = -EEXIST;
 851                 goto out;
 852         }
 853
 854         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 855
 856         if (ifa == NULL) {
 857                 ADBG(("ipv6_add_addr: malloc failed\n"));
 858                 err = -ENOBUFS;
 859                 goto out;
 860         }
 861
 862         rt = addrconf_dst_alloc(idev, addr, false);
 863         if (IS_ERR(rt)) {
 864                 err = PTR_ERR(rt);
 865                 goto out;
 866         }
 867
 868         ifa->addr = *addr;
 869
 870         spin_lock_init(&ifa->lock);
 871         spin_lock_init(&ifa->state_lock);
 872         init_timer(&ifa->timer);
 873         INIT_HLIST_NODE(&ifa->addr_lst);
 874         ifa->timer.data = (unsigned long) ifa;
 875         ifa->scope = scope;
 876         ifa->prefix_len = pfxlen;
 877         ifa->flags = flags | IFA_F_TENTATIVE;
 878         ifa->cstamp = ifa->tstamp = jiffies;
 879
 880         ifa->rt = rt;
 881
 882         ifa->idev = idev;
 883         in6_dev_hold(idev);
 884         /* For caller */
 885         in6_ifa_hold(ifa);
 886
 887         /* Add to big hash table */
 888         hash = inet6_addr_hash(addr);
 889
 890         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 891         spin_unlock(&addrconf_hash_lock);
 892
 893         write_lock(&idev->lock);
 894         /* Add to inet6_dev unicast addr list. */
 895         ipv6_link_dev_addr(idev, ifa);
 896
 897 #ifdef CONFIG_IPV6_PRIVACY
 898         if (ifa->flags&IFA_F_TEMPORARY) {
 899                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 900                 in6_ifa_hold(ifa);
 901         }
 902 #endif
 903
 904         in6_ifa_hold(ifa);
 905         write_unlock(&idev->lock);
 906 out2:
 907         rcu_read_unlock_bh();
 908
 909         if (likely(err == 0))
 910                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 911         else {
 912                 kfree(ifa);
 913                 ifa = ERR_PTR(err);
 914         }
 915
 916         return ifa;
 917 out:
 918         spin_unlock(&addrconf_hash_lock);
 919         goto out2;
 920 }
 921
 922 /* This function wants to get referenced ifp and releases it before return */
 923
 924 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 925 {
 926         struct inet6_ifaddr *ifa, *ifn;
 927         struct inet6_dev *idev = ifp->idev;
 928         int state;
 929         int deleted = 0, onlink = 0;
 930         unsigned long expires = jiffies;
 931
 932         spin_lock_bh(&ifp->state_lock);
 933         state = ifp->state;
 934         ifp->state = INET6_IFADDR_STATE_DEAD;
 935         spin_unlock_bh(&ifp->state_lock);
 936
 937         if (state == INET6_IFADDR_STATE_DEAD)
 938                 goto out;
 939
 940         spin_lock_bh(&addrconf_hash_lock);
 941         hlist_del_init_rcu(&ifp->addr_lst);
 942         spin_unlock_bh(&addrconf_hash_lock);
 943
 944         write_lock_bh(&idev->lock);
 945 #ifdef CONFIG_IPV6_PRIVACY
 946         if (ifp->flags&IFA_F_TEMPORARY) {
 947                 list_del(&ifp->tmp_list);
 948                 if (ifp->ifpub) {
 949                         in6_ifa_put(ifp->ifpub);
 950                         ifp->ifpub = NULL;
 951                 }
 952                 __in6_ifa_put(ifp);
 953         }
 954 #endif
 955
 956         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 957                 if (ifa == ifp) {
 958                         list_del_init(&ifp->if_list);
 959                         __in6_ifa_put(ifp);
 960
 961                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 962                                 break;
 963                         deleted = 1;
 964                         continue;
 965                 } else if (ifp->flags & IFA_F_PERMANENT) {
 966                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 967                                               ifp->prefix_len)) {
 968                                 if (ifa->flags & IFA_F_PERMANENT) {
 969                                         onlink = 1;
 970                                         if (deleted)
 971                                                 break;
 972                                 } else {
 973                                         unsigned long lifetime;
 974
 975                                         if (!onlink)
 976                                                 onlink = -1;
 977
 978                                         spin_lock(&ifa->lock);
 979
 980                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 981                                         /*
 982                                          * Note: Because this address is
 983                                          * not permanent, lifetime <
 984                                          * LONG_MAX / HZ here.
 985                                          */
 986                                         if (time_before(expires,
 987                                                         ifa->tstamp + lifetime * HZ))
 988                                                 expires = ifa->tstamp + lifetime * HZ;
 989                                         spin_unlock(&ifa->lock);
 990                                 }
 991                         }
 992                 }
 993         }
 994         write_unlock_bh(&idev->lock);
 995
 996         addrconf_del_timer(ifp);
 997
 998         ipv6_ifa_notify(RTM_DELADDR, ifp);
 999
1000         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
1001
1002         /*
1003          * Purge or update corresponding prefix
1004          *
1005          * 1) we don't purge prefix here if address was not permanent.
1006          *    prefix is managed by its own lifetime.
1007          * 2) if there're no addresses, delete prefix.
1008          * 3) if there're still other permanent address(es),
1009          *    corresponding prefix is still permanent.
1010          * 4) otherwise, update prefix lifetime to the
1011          *    longest valid lifetime among the corresponding
1012          *    addresses on the device.
1013          *    Note: subsequent RA will update lifetime.
1014          *
1015          * --yoshfuji
1016          */
1017         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1018                 struct in6_addr prefix;
1019                 struct rt6_info *rt;
1020
1021                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1022
1023                 rt = addrconf_get_prefix_route(&prefix,
1024                                                ifp->prefix_len,
1025                                                ifp->idev->dev,
1026                                                0, RTF_GATEWAY | RTF_DEFAULT);
1027
1028                 if (rt) {
1029                         if (onlink == 0) {
1030                                 ip6_del_rt(rt);
1031                                 rt = NULL;
1032                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1033                                 rt6_set_expires(rt, expires);
1034                         }
1035                 }
1036                 ip6_rt_put(rt);
1037         }
1038
1039         /* clean up prefsrc entries */
1040         rt6_remove_prefsrc(ifp);
1041 out:
1042         in6_ifa_put(ifp);
1043 }
1044
1045 #ifdef CONFIG_IPV6_PRIVACY
1046 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1047 {
1048         struct inet6_dev *idev = ifp->idev;
1049         struct in6_addr addr, *tmpaddr;
1050         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1051         unsigned long regen_advance;
1052         int tmp_plen;
1053         int ret = 0;
1054         int max_addresses;
1055         u32 addr_flags;
1056         unsigned long now = jiffies;
1057
1058         write_lock(&idev->lock);
1059         if (ift) {
1060                 spin_lock_bh(&ift->lock);
1061                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1062                 spin_unlock_bh(&ift->lock);
1063                 tmpaddr = &addr;
1064         } else {
1065                 tmpaddr = NULL;
1066         }
1067 retry:
1068         in6_dev_hold(idev);
1069         if (idev->cnf.use_tempaddr <= 0) {
1070                 write_unlock(&idev->lock);
1071                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1072                 in6_dev_put(idev);
1073                 ret = -1;
1074                 goto out;
1075         }
1076         spin_lock_bh(&ifp->lock);
1077         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1078                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1079                 spin_unlock_bh(&ifp->lock);
1080                 write_unlock(&idev->lock);
1081                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1082                         __func__);
1083                 in6_dev_put(idev);
1084                 ret = -1;
1085                 goto out;
1086         }
1087         in6_ifa_hold(ifp);
1088         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1089         __ipv6_try_regen_rndid(idev, tmpaddr);
1090         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1091         age = (now - ifp->tstamp) / HZ;
1092         tmp_valid_lft = min_t(__u32,
1093                               ifp->valid_lft,
1094                               idev->cnf.temp_valid_lft + age);
1095         tmp_prefered_lft = min_t(__u32,
1096                                  ifp->prefered_lft,
1097                                  idev->cnf.temp_prefered_lft + age -
1098                                  idev->cnf.max_desync_factor);
1099         tmp_plen = ifp->prefix_len;
1100         max_addresses = idev->cnf.max_addresses;
1101         tmp_tstamp = ifp->tstamp;
1102         spin_unlock_bh(&ifp->lock);
1103
1104         regen_advance = idev->cnf.regen_max_retry *
1105                         idev->cnf.dad_transmits *
1106                         idev->nd_parms->retrans_time / HZ;
1107         write_unlock(&idev->lock);
1108
1109         /* A temporary address is created only if this calculated Preferred
1110          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1111          * an implementation must not create a temporary address with a zero
1112          * Preferred Lifetime.
1113          */
1114         if (tmp_prefered_lft <= regen_advance) {
1115                 in6_ifa_put(ifp);
1116                 in6_dev_put(idev);
1117                 ret = -1;
1118                 goto out;
1119         }
1120
1121         addr_flags = IFA_F_TEMPORARY;
1122         /* set in addrconf_prefix_rcv() */
1123         if (ifp->flags & IFA_F_OPTIMISTIC)
1124                 addr_flags |= IFA_F_OPTIMISTIC;
1125
1126         ift = !max_addresses ||
1127               ipv6_count_addresses(idev) < max_addresses ?
1128                 ipv6_add_addr(idev, &addr, tmp_plen,
1129                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1130                               addr_flags) : NULL;
1131         if (IS_ERR_OR_NULL(ift)) {
1132                 in6_ifa_put(ifp);
1133                 in6_dev_put(idev);
1134                 pr_info("%s: retry temporary address regeneration\n", __func__);
1135                 tmpaddr = &addr;
1136                 write_lock(&idev->lock);
1137                 goto retry;
1138         }
1139
1140         spin_lock_bh(&ift->lock);
1141         ift->ifpub = ifp;
1142         ift->valid_lft = tmp_valid_lft;
1143         ift->prefered_lft = tmp_prefered_lft;
1144         ift->cstamp = now;
1145         ift->tstamp = tmp_tstamp;
1146         spin_unlock_bh(&ift->lock);
1147
1148         addrconf_dad_start(ift);
1149         in6_ifa_put(ift);
1150         in6_dev_put(idev);
1151 out:
1152         return ret;
1153 }
1154 #endif
1155
1156 /*
1157  *      Choose an appropriate source address (RFC3484)
1158  */
1159 enum {
1160         IPV6_SADDR_RULE_INIT = 0,
1161         IPV6_SADDR_RULE_LOCAL,
1162         IPV6_SADDR_RULE_SCOPE,
1163         IPV6_SADDR_RULE_PREFERRED,
1164 #ifdef CONFIG_IPV6_MIP6
1165         IPV6_SADDR_RULE_HOA,
1166 #endif
1167         IPV6_SADDR_RULE_OIF,
1168         IPV6_SADDR_RULE_LABEL,
1169 #ifdef CONFIG_IPV6_PRIVACY
1170         IPV6_SADDR_RULE_PRIVACY,
1171 #endif
1172         IPV6_SADDR_RULE_ORCHID,
1173         IPV6_SADDR_RULE_PREFIX,
1174         IPV6_SADDR_RULE_MAX
1175 };
1176
1177 struct ipv6_saddr_score {
1178         int                     rule;
1179         int                     addr_type;
1180         struct inet6_ifaddr     *ifa;
1181         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1182         int                     scopedist;
1183         int                     matchlen;
1184 };
1185
1186 struct ipv6_saddr_dst {
1187         const struct in6_addr *addr;
1188         int ifindex;
1189         int scope;
1190         int label;
1191         unsigned int prefs;
1192 };
1193
1194 static inline int ipv6_saddr_preferred(int type)
1195 {
1196         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1197                 return 1;
1198         return 0;
1199 }
1200
1201 static int ipv6_get_saddr_eval(struct net *net,
1202                                struct ipv6_saddr_score *score,
1203                                struct ipv6_saddr_dst *dst,
1204                                int i)
1205 {
1206         int ret;
1207
1208         if (i <= score->rule) {
1209                 switch (i) {
1210                 case IPV6_SADDR_RULE_SCOPE:
1211                         ret = score->scopedist;
1212                         break;
1213                 case IPV6_SADDR_RULE_PREFIX:
1214                         ret = score->matchlen;
1215                         break;
1216                 default:
1217                         ret = !!test_bit(i, score->scorebits);
1218                 }
1219                 goto out;
1220         }
1221
1222         switch (i) {
1223         case IPV6_SADDR_RULE_INIT:
1224                 /* Rule 0: remember if hiscore is not ready yet */
1225                 ret = !!score->ifa;
1226                 break;
1227         case IPV6_SADDR_RULE_LOCAL:
1228                 /* Rule 1: Prefer same address */
1229                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1230                 break;
1231         case IPV6_SADDR_RULE_SCOPE:
1232                 /* Rule 2: Prefer appropriate scope
1233                  *
1234                  *      ret
1235                  *       ^
1236                  *    -1 |  d 15
1237                  *    ---+--+-+---> scope
1238                  *       |
1239                  *       |             d is scope of the destination.
1240                  *  B-d  |  \
1241                  *       |   \      <- smaller scope is better if
1242                  *  B-15 |    \        if scope is enough for destinaion.
1243                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1244                  * d-C-1 | /
1245                  *       |/         <- greater is better
1246                  *   -C  /             if scope is not enough for destination.
1247                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1248                  *
1249                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1250                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1251                  * Assume B = 0 and we get C > 29.
1252                  */
1253                 ret = __ipv6_addr_src_scope(score->addr_type);
1254                 if (ret >= dst->scope)
1255                         ret = -ret;
1256                 else
1257                         ret -= 128;     /* 30 is enough */
1258                 score->scopedist = ret;
1259                 break;
1260         case IPV6_SADDR_RULE_PREFERRED:
1261                 /* Rule 3: Avoid deprecated and optimistic addresses */
1262                 ret = ipv6_saddr_preferred(score->addr_type) ||
1263                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1264                 break;
1265 #ifdef CONFIG_IPV6_MIP6
1266         case IPV6_SADDR_RULE_HOA:
1267             {
1268                 /* Rule 4: Prefer home address */
1269                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1270                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1271                 break;
1272             }
1273 #endif
1274         case IPV6_SADDR_RULE_OIF:
1275                 /* Rule 5: Prefer outgoing interface */
1276                 ret = (!dst->ifindex ||
1277                        dst->ifindex == score->ifa->idev->dev->ifindex);
1278                 break;
1279         case IPV6_SADDR_RULE_LABEL:
1280                 /* Rule 6: Prefer matching label */
1281                 ret = ipv6_addr_label(net,
1282                                       &score->ifa->addr, score->addr_type,
1283                                       score->ifa->idev->dev->ifindex) == dst->label;
1284                 break;
1285 #ifdef CONFIG_IPV6_PRIVACY
1286         case IPV6_SADDR_RULE_PRIVACY:
1287             {
1288                 /* Rule 7: Prefer public address
1289                  * Note: prefer temporary address if use_tempaddr >= 2
1290                  */
1291                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1292                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1293                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1294                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1295                 break;
1296             }
1297 #endif
1298         case IPV6_SADDR_RULE_ORCHID:
1299                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1300                  *          non-ORCHID vs non-ORCHID
1301                  */
1302                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1303                         ipv6_addr_orchid(dst->addr));
1304                 break;
1305         case IPV6_SADDR_RULE_PREFIX:
1306                 /* Rule 8: Use longest matching prefix */
1307                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1308                 if (ret > score->ifa->prefix_len)
1309                         ret = score->ifa->prefix_len;
1310                 score->matchlen = ret;
1311                 break;
1312         default:
1313                 ret = 0;
1314         }
1315
1316         if (ret)
1317                 __set_bit(i, score->scorebits);
1318         score->rule = i;
1319 out:
1320         return ret;
1321 }
1322
1323 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1324                        const struct in6_addr *daddr, unsigned int prefs,
1325                        struct in6_addr *saddr)
1326 {
1327         struct ipv6_saddr_score scores[2],
1328                                 *score = &scores[0], *hiscore = &scores[1];
1329         struct ipv6_saddr_dst dst;
1330         struct net_device *dev;
1331         int dst_type;
1332
1333         dst_type = __ipv6_addr_type(daddr);
1334         dst.addr = daddr;
1335         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1336         dst.scope = __ipv6_addr_src_scope(dst_type);
1337         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1338         dst.prefs = prefs;
1339
1340         hiscore->rule = -1;
1341         hiscore->ifa = NULL;
1342
1343         rcu_read_lock();
1344
1345         for_each_netdev_rcu(net, dev) {
1346                 struct inet6_dev *idev;
1347
1348                 /* Candidate Source Address (section 4)
1349                  *  - multicast and link-local destination address,
1350                  *    the set of candidate source address MUST only
1351                  *    include addresses assigned to interfaces
1352                  *    belonging to the same link as the outgoing
1353                  *    interface.
1354                  * (- For site-local destination addresses, the
1355                  *    set of candidate source addresses MUST only
1356                  *    include addresses assigned to interfaces
1357                  *    belonging to the same site as the outgoing
1358                  *    interface.)
1359                  */
1360                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1361                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1362                     dst.ifindex && dev->ifindex != dst.ifindex)
1363                         continue;
1364
1365                 idev = __in6_dev_get(dev);
1366                 if (!idev)
1367                         continue;
1368
1369                 read_lock_bh(&idev->lock);
1370                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1371                         int i;
1372
1373                         /*
1374                          * - Tentative Address (RFC2462 section 5.4)
1375                          *  - A tentative address is not considered
1376                          *    "assigned to an interface" in the traditional
1377                          *    sense, unless it is also flagged as optimistic.
1378                          * - Candidate Source Address (section 4)
1379                          *  - In any case, anycast addresses, multicast
1380                          *    addresses, and the unspecified address MUST
1381                          *    NOT be included in a candidate set.
1382                          */
1383                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1384                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1385                                 continue;
1386
1387                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1388
1389                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1390                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1391                                 LIMIT_NETDEBUG(KERN_DEBUG
1392                                                "ADDRCONF: unspecified / multicast address "
1393                                                "assigned as unicast address on %s",
1394                                                dev->name);
1395                                 continue;
1396                         }
1397
1398                         score->rule = -1;
1399                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1400
1401                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1402                                 int minihiscore, miniscore;
1403
1404                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1405                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1406
1407                                 if (minihiscore > miniscore) {
1408                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1409                                             score->scopedist > 0) {
1410                                                 /*
1411                                                  * special case:
1412                                                  * each remaining entry
1413                                                  * has too small (not enough)
1414                                                  * scope, because ifa entries
1415                                                  * are sorted by their scope
1416                                                  * values.
1417                                                  */
1418                                                 goto try_nextdev;
1419                                         }
1420                                         break;
1421                                 } else if (minihiscore < miniscore) {
1422                                         if (hiscore->ifa)
1423                                                 in6_ifa_put(hiscore->ifa);
1424
1425                                         in6_ifa_hold(score->ifa);
1426
1427                                         swap(hiscore, score);
1428
1429                                         /* restore our iterator */
1430                                         score->ifa = hiscore->ifa;
1431
1432                                         break;
1433                                 }
1434                         }
1435                 }
1436 try_nextdev:
1437                 read_unlock_bh(&idev->lock);
1438         }
1439         rcu_read_unlock();
1440
1441         if (!hiscore->ifa)
1442                 return -EADDRNOTAVAIL;
1443
1444         *saddr = hiscore->ifa->addr;
1445         in6_ifa_put(hiscore->ifa);
1446         return 0;
1447 }
1448 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1449
1450 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1451                     unsigned char banned_flags)
1452 {
1453         struct inet6_dev *idev;
1454         int err = -EADDRNOTAVAIL;
1455
1456         rcu_read_lock();
1457         idev = __in6_dev_get(dev);
1458         if (idev) {
1459                 struct inet6_ifaddr *ifp;
1460
1461                 read_lock_bh(&idev->lock);
1462                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1463                         if (ifp->scope == IFA_LINK &&
1464                             !(ifp->flags & banned_flags)) {
1465                                 *addr = ifp->addr;
1466                                 err = 0;
1467                                 break;
1468                         }
1469                 }
1470                 read_unlock_bh(&idev->lock);
1471         }
1472         rcu_read_unlock();
1473         return err;
1474 }
1475
1476 static int ipv6_count_addresses(struct inet6_dev *idev)
1477 {
1478         int cnt = 0;
1479         struct inet6_ifaddr *ifp;
1480
1481         read_lock_bh(&idev->lock);
1482         list_for_each_entry(ifp, &idev->addr_list, if_list)
1483                 cnt++;
1484         read_unlock_bh(&idev->lock);
1485         return cnt;
1486 }
1487
1488 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1489                   struct net_device *dev, int strict)
1490 {
1491         struct inet6_ifaddr *ifp;
1492         unsigned int hash = inet6_addr_hash(addr);
1493
1494         rcu_read_lock_bh();
1495         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1496                 if (!net_eq(dev_net(ifp->idev->dev), net))
1497                         continue;
1498                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1499                     !(ifp->flags&IFA_F_TENTATIVE) &&
1500                     (dev == NULL || ifp->idev->dev == dev ||
1501                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1502                         rcu_read_unlock_bh();
1503                         return 1;
1504                 }
1505         }
1506
1507         rcu_read_unlock_bh();
1508         return 0;
1509 }
1510 EXPORT_SYMBOL(ipv6_chk_addr);
1511
1512 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1513                                struct net_device *dev)
1514 {
1515         unsigned int hash = inet6_addr_hash(addr);
1516         struct inet6_ifaddr *ifp;
1517
1518         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1519                 if (!net_eq(dev_net(ifp->idev->dev), net))
1520                         continue;
1521                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1522                         if (dev == NULL || ifp->idev->dev == dev)
1523                                 return true;
1524                 }
1525         }
1526         return false;
1527 }
1528
1529 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1530 {
1531         struct inet6_dev *idev;
1532         struct inet6_ifaddr *ifa;
1533         int     onlink;
1534
1535         onlink = 0;
1536         rcu_read_lock();
1537         idev = __in6_dev_get(dev);
1538         if (idev) {
1539                 read_lock_bh(&idev->lock);
1540                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1541                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1542                                                    ifa->prefix_len);
1543                         if (onlink)
1544                                 break;
1545                 }
1546                 read_unlock_bh(&idev->lock);
1547         }
1548         rcu_read_unlock();
1549         return onlink;
1550 }
1551 EXPORT_SYMBOL(ipv6_chk_prefix);
1552
1553 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1554                                      struct net_device *dev, int strict)
1555 {
1556         struct inet6_ifaddr *ifp, *result = NULL;
1557         unsigned int hash = inet6_addr_hash(addr);
1558
1559         rcu_read_lock_bh();
1560         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1561                 if (!net_eq(dev_net(ifp->idev->dev), net))
1562                         continue;
1563                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1564                         if (dev == NULL || ifp->idev->dev == dev ||
1565                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1566                                 result = ifp;
1567                                 in6_ifa_hold(ifp);
1568                                 break;
1569                         }
1570                 }
1571         }
1572         rcu_read_unlock_bh();
1573
1574         return result;
1575 }
1576
1577 /* Gets referenced address, destroys ifaddr */
1578
1579 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1580 {
1581         if (ifp->flags&IFA_F_PERMANENT) {
1582                 spin_lock_bh(&ifp->lock);
1583                 addrconf_del_timer(ifp);
1584                 ifp->flags |= IFA_F_TENTATIVE;
1585                 if (dad_failed)
1586                         ifp->flags |= IFA_F_DADFAILED;
1587                 spin_unlock_bh(&ifp->lock);
1588                 if (dad_failed)
1589                         ipv6_ifa_notify(0, ifp);
1590                 in6_ifa_put(ifp);
1591 #ifdef CONFIG_IPV6_PRIVACY
1592         } else if (ifp->flags&IFA_F_TEMPORARY) {
1593                 struct inet6_ifaddr *ifpub;
1594                 spin_lock_bh(&ifp->lock);
1595                 ifpub = ifp->ifpub;
1596                 if (ifpub) {
1597                         in6_ifa_hold(ifpub);
1598                         spin_unlock_bh(&ifp->lock);
1599                         ipv6_create_tempaddr(ifpub, ifp);
1600                         in6_ifa_put(ifpub);
1601                 } else {
1602                         spin_unlock_bh(&ifp->lock);
1603                 }
1604                 ipv6_del_addr(ifp);
1605 #endif
1606         } else
1607                 ipv6_del_addr(ifp);
1608 }
1609
1610 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1611 {
1612         int err = -ENOENT;
1613
1614         spin_lock(&ifp->state_lock);
1615         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1616                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1617                 err = 0;
1618         }
1619         spin_unlock(&ifp->state_lock);
1620
1621         return err;
1622 }
1623
1624 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1625 {
1626         struct inet6_dev *idev = ifp->idev;
1627
1628         if (addrconf_dad_end(ifp)) {
1629                 in6_ifa_put(ifp);
1630                 return;
1631         }
1632
1633         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1634                              ifp->idev->dev->name, &ifp->addr);
1635
1636         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1637                 struct in6_addr addr;
1638
1639                 addr.s6_addr32[0] = htonl(0xfe800000);
1640                 addr.s6_addr32[1] = 0;
1641
1642                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1643                     ipv6_addr_equal(&ifp->addr, &addr)) {
1644                         /* DAD failed for link-local based on MAC address */
1645                         idev->cnf.disable_ipv6 = 1;
1646
1647                         pr_info("%s: IPv6 being disabled!\n",
1648                                 ifp->idev->dev->name);
1649                 }
1650         }
1651
1652         addrconf_dad_stop(ifp, 1);
1653 }
1654
1655 /* Join to solicited addr multicast group. */
1656
1657 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1658 {
1659         struct in6_addr maddr;
1660
1661         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1662                 return;
1663
1664         addrconf_addr_solict_mult(addr, &maddr);
1665         ipv6_dev_mc_inc(dev, &maddr);
1666 }
1667
1668 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1669 {
1670         struct in6_addr maddr;
1671
1672         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1673                 return;
1674
1675         addrconf_addr_solict_mult(addr, &maddr);
1676         __ipv6_dev_mc_dec(idev, &maddr);
1677 }
1678
1679 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1680 {
1681         struct in6_addr addr;
1682         if (ifp->prefix_len == 127) /* RFC 6164 */
1683                 return;
1684         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1685         if (ipv6_addr_any(&addr))
1686                 return;
1687         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1688 }
1689
1690 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1691 {
1692         struct in6_addr addr;
1693         if (ifp->prefix_len == 127) /* RFC 6164 */
1694                 return;
1695         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1696         if (ipv6_addr_any(&addr))
1697                 return;
1698         __ipv6_dev_ac_dec(ifp->idev, &addr);
1699 }
1700
1701 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1702 {
1703         if (dev->addr_len != ETH_ALEN)
1704                 return -1;
1705         memcpy(eui, dev->dev_addr, 3);
1706         memcpy(eui + 5, dev->dev_addr + 3, 3);
1707
1708         /*
1709          * The zSeries OSA network cards can be shared among various
1710          * OS instances, but the OSA cards have only one MAC address.
1711          * This leads to duplicate address conflicts in conjunction
1712          * with IPv6 if more than one instance uses the same card.
1713          *
1714          * The driver for these cards can deliver a unique 16-bit
1715          * identifier for each instance sharing the same card.  It is
1716          * placed instead of 0xFFFE in the interface identifier.  The
1717          * "u" bit of the interface identifier is not inverted in this
1718          * case.  Hence the resulting interface identifier has local
1719          * scope according to RFC2373.
1720          */
1721         if (dev->dev_id) {
1722                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1723                 eui[4] = dev->dev_id & 0xFF;
1724         } else {
1725                 eui[3] = 0xFF;
1726                 eui[4] = 0xFE;
1727                 eui[0] ^= 2;
1728         }
1729         return 0;
1730 }
1731
1732 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1733 {
1734         if (dev->addr_len != IEEE802154_ADDR_LEN)
1735                 return -1;
1736         memcpy(eui, dev->dev_addr, 8);
1737         eui[0] ^= 2;
1738         return 0;
1739 }
1740
1741 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1742 {
1743         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1744         if (dev->addr_len != ARCNET_ALEN)
1745                 return -1;
1746         memset(eui, 0, 7);
1747         eui[7] = *(u8 *)dev->dev_addr;
1748         return 0;
1749 }
1750
1751 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1752 {
1753         if (dev->addr_len != INFINIBAND_ALEN)
1754                 return -1;
1755         memcpy(eui, dev->dev_addr + 12, 8);
1756         eui[0] |= 2;
1757         return 0;
1758 }
1759
1760 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1761 {
1762         if (addr == 0)
1763                 return -1;
1764         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1765                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1766                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1767                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1768                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1769                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1770         eui[1] = 0;
1771         eui[2] = 0x5E;
1772         eui[3] = 0xFE;
1773         memcpy(eui + 4, &addr, 4);
1774         return 0;
1775 }
1776
1777 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1778 {
1779         if (dev->priv_flags & IFF_ISATAP)
1780                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1781         return -1;
1782 }
1783
1784 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1785 {
1786         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1787 }
1788
1789 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1790 {
1791         switch (dev->type) {
1792         case ARPHRD_ETHER:
1793         case ARPHRD_FDDI:
1794                 return addrconf_ifid_eui48(eui, dev);
1795         case ARPHRD_ARCNET:
1796                 return addrconf_ifid_arcnet(eui, dev);
1797         case ARPHRD_INFINIBAND:
1798                 return addrconf_ifid_infiniband(eui, dev);
1799         case ARPHRD_SIT:
1800                 return addrconf_ifid_sit(eui, dev);
1801         case ARPHRD_IPGRE:
1802                 return addrconf_ifid_gre(eui, dev);
1803         case ARPHRD_IEEE802154:
1804                 return addrconf_ifid_eui64(eui, dev);
1805         }
1806         return -1;
1807 }
1808
1809 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1810 {
1811         int err = -1;
1812         struct inet6_ifaddr *ifp;
1813
1814         read_lock_bh(&idev->lock);
1815         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1816                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1817                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1818                         err = 0;
1819                         break;
1820                 }
1821         }
1822         read_unlock_bh(&idev->lock);
1823         return err;
1824 }
1825
1826 #ifdef CONFIG_IPV6_PRIVACY
1827 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1828 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1829 {
1830 regen:
1831         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1832         idev->rndid[0] &= ~0x02;
1833
1834         /*
1835          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1836          * check if generated address is not inappropriate
1837          *
1838          *  - Reserved subnet anycast (RFC 2526)
1839          *      11111101 11....11 1xxxxxxx
1840          *  - ISATAP (RFC4214) 6.1
1841          *      00-00-5E-FE-xx-xx-xx-xx
1842          *  - value 0
1843          *  - XXX: already assigned to an address on the device
1844          */
1845         if (idev->rndid[0] == 0xfd &&
1846             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1847             (idev->rndid[7]&0x80))
1848                 goto regen;
1849         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1850                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1851                         goto regen;
1852                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1853                         goto regen;
1854         }
1855 }
1856
1857 static void ipv6_regen_rndid(unsigned long data)
1858 {
1859         struct inet6_dev *idev = (struct inet6_dev *) data;
1860         unsigned long expires;
1861
1862         rcu_read_lock_bh();
1863         write_lock_bh(&idev->lock);
1864
1865         if (idev->dead)
1866                 goto out;
1867
1868         __ipv6_regen_rndid(idev);
1869
1870         expires = jiffies +
1871                 idev->cnf.temp_prefered_lft * HZ -
1872                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1873                 idev->cnf.max_desync_factor * HZ;
1874         if (time_before(expires, jiffies)) {
1875                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1876                         __func__, idev->dev->name);
1877                 goto out;
1878         }
1879
1880         if (!mod_timer(&idev->regen_timer, expires))
1881                 in6_dev_hold(idev);
1882
1883 out:
1884         write_unlock_bh(&idev->lock);
1885         rcu_read_unlock_bh();
1886         in6_dev_put(idev);
1887 }
1888
1889 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1890 {
1891         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1892                 __ipv6_regen_rndid(idev);
1893 }
1894 #endif
1895
1896 /*
1897  *      Add prefix route.
1898  */
1899
1900 static void
1901 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1902                       unsigned long expires, u32 flags)
1903 {
1904         struct fib6_config cfg = {
1905                 .fc_table = RT6_TABLE_PREFIX,
1906                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1907                 .fc_ifindex = dev->ifindex,
1908                 .fc_expires = expires,
1909                 .fc_dst_len = plen,
1910                 .fc_flags = RTF_UP | flags,
1911                 .fc_nlinfo.nl_net = dev_net(dev),
1912                 .fc_protocol = RTPROT_KERNEL,
1913         };
1914
1915         cfg.fc_dst = *pfx;
1916
1917         /* Prevent useless cloning on PtP SIT.
1918            This thing is done here expecting that the whole
1919            class of non-broadcast devices need not cloning.
1920          */
1921 #if IS_ENABLED(CONFIG_IPV6_SIT)
1922         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1923                 cfg.fc_flags |= RTF_NONEXTHOP;
1924 #endif
1925
1926         ip6_route_add(&cfg);
1927 }
1928
1929
1930 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1931                                                   int plen,
1932                                                   const struct net_device *dev,
1933                                                   u32 flags, u32 noflags)
1934 {
1935         struct fib6_node *fn;
1936         struct rt6_info *rt = NULL;
1937         struct fib6_table *table;
1938
1939         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1940         if (table == NULL)
1941                 return NULL;
1942
1943         read_lock_bh(&table->tb6_lock);
1944         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1945         if (!fn)
1946                 goto out;
1947         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1948                 if (rt->dst.dev->ifindex != dev->ifindex)
1949                         continue;
1950                 if ((rt->rt6i_flags & flags) != flags)
1951                         continue;
1952                 if ((rt->rt6i_flags & noflags) != 0)
1953                         continue;
1954                 dst_hold(&rt->dst);
1955                 break;
1956         }
1957 out:
1958         read_unlock_bh(&table->tb6_lock);
1959         return rt;
1960 }
1961
1962
1963 /* Create "default" multicast route to the interface */
1964
1965 static void addrconf_add_mroute(struct net_device *dev)
1966 {
1967         struct fib6_config cfg = {
1968                 .fc_table = RT6_TABLE_LOCAL,
1969                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1970                 .fc_ifindex = dev->ifindex,
1971                 .fc_dst_len = 8,
1972                 .fc_flags = RTF_UP,
1973                 .fc_nlinfo.nl_net = dev_net(dev),
1974         };
1975
1976         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1977
1978         ip6_route_add(&cfg);
1979 }
1980
1981 #if IS_ENABLED(CONFIG_IPV6_SIT)
1982 static void sit_route_add(struct net_device *dev)
1983 {
1984         struct fib6_config cfg = {
1985                 .fc_table = RT6_TABLE_MAIN,
1986                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1987                 .fc_ifindex = dev->ifindex,
1988                 .fc_dst_len = 96,
1989                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1990                 .fc_nlinfo.nl_net = dev_net(dev),
1991         };
1992
1993         /* prefix length - 96 bits "::d.d.d.d" */
1994         ip6_route_add(&cfg);
1995 }
1996 #endif
1997
1998 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1999 {
2000         struct inet6_dev *idev;
2001
2002         ASSERT_RTNL();
2003
2004         idev = ipv6_find_idev(dev);
2005         if (!idev)
2006                 return ERR_PTR(-ENOBUFS);
2007
2008         if (idev->cnf.disable_ipv6)
2009                 return ERR_PTR(-EACCES);
2010
2011         /* Add default multicast route */
2012         if (!(dev->flags & IFF_LOOPBACK))
2013                 addrconf_add_mroute(dev);
2014
2015         return idev;
2016 }
2017
2018 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2019 {
2020         struct prefix_info *pinfo;
2021         __u32 valid_lft;
2022         __u32 prefered_lft;
2023         int addr_type;
2024         struct inet6_dev *in6_dev;
2025         struct net *net = dev_net(dev);
2026
2027         pinfo = (struct prefix_info *) opt;
2028
2029         if (len < sizeof(struct prefix_info)) {
2030                 ADBG(("addrconf: prefix option too short\n"));
2031                 return;
2032         }
2033
2034         /*
2035          *      Validation checks ([ADDRCONF], page 19)
2036          */
2037
2038         addr_type = ipv6_addr_type(&pinfo->prefix);
2039
2040         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2041                 return;
2042
2043         valid_lft = ntohl(pinfo->valid);
2044         prefered_lft = ntohl(pinfo->prefered);
2045
2046         if (prefered_lft > valid_lft) {
2047                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2048                 return;
2049         }
2050
2051         in6_dev = in6_dev_get(dev);
2052
2053         if (in6_dev == NULL) {
2054                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2055                                     dev->name);
2056                 return;
2057         }
2058
2059         /*
2060          *      Two things going on here:
2061          *      1) Add routes for on-link prefixes
2062          *      2) Configure prefixes with the auto flag set
2063          */
2064
2065         if (pinfo->onlink) {
2066                 struct rt6_info *rt;
2067                 unsigned long rt_expires;
2068
2069                 /* Avoid arithmetic overflow. Really, we could
2070                  * save rt_expires in seconds, likely valid_lft,
2071                  * but it would require division in fib gc, that it
2072                  * not good.
2073                  */
2074                 if (HZ > USER_HZ)
2075                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2076                 else
2077                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2078
2079                 if (addrconf_finite_timeout(rt_expires))
2080                         rt_expires *= HZ;
2081
2082                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2083                                                pinfo->prefix_len,
2084                                                dev,
2085                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2086                                                RTF_GATEWAY | RTF_DEFAULT);
2087
2088                 if (rt) {
2089                         /* Autoconf prefix route */
2090                         if (valid_lft == 0) {
2091                                 ip6_del_rt(rt);
2092                                 rt = NULL;
2093                         } else if (addrconf_finite_timeout(rt_expires)) {
2094                                 /* not infinity */
2095                                 rt6_set_expires(rt, jiffies + rt_expires);
2096                         } else {
2097                                 rt6_clean_expires(rt);
2098                         }
2099                 } else if (valid_lft) {
2100                         clock_t expires = 0;
2101                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2102                         if (addrconf_finite_timeout(rt_expires)) {
2103                                 /* not infinity */
2104                                 flags |= RTF_EXPIRES;
2105                                 expires = jiffies_to_clock_t(rt_expires);
2106                         }
2107                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2108                                               dev, expires, flags);
2109                 }
2110                 ip6_rt_put(rt);
2111         }
2112
2113         /* Try to figure out our local address for this prefix */
2114
2115         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2116                 struct inet6_ifaddr *ifp;
2117                 struct in6_addr addr;
2118                 int create = 0, update_lft = 0;
2119
2120                 if (pinfo->prefix_len == 64) {
2121                         memcpy(&addr, &pinfo->prefix, 8);
2122                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2123                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2124                                 in6_dev_put(in6_dev);
2125                                 return;
2126                         }
2127                         goto ok;
2128                 }
2129                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2130                                     pinfo->prefix_len);
2131                 in6_dev_put(in6_dev);
2132                 return;
2133
2134 ok:
2135
2136                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2137
2138                 if (ifp == NULL && valid_lft) {
2139                         int max_addresses = in6_dev->cnf.max_addresses;
2140                         u32 addr_flags = 0;
2141
2142 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2143                         if (in6_dev->cnf.optimistic_dad &&
2144                             !net->ipv6.devconf_all->forwarding && sllao)
2145                                 addr_flags = IFA_F_OPTIMISTIC;
2146 #endif
2147
2148                         /* Do not allow to create too much of autoconfigured
2149                          * addresses; this would be too easy way to crash kernel.
2150                          */
2151                         if (!max_addresses ||
2152                             ipv6_count_addresses(in6_dev) < max_addresses)
2153                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2154                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2155                                                     addr_flags);
2156
2157                         if (IS_ERR_OR_NULL(ifp)) {
2158                                 in6_dev_put(in6_dev);
2159                                 return;
2160                         }
2161
2162                         update_lft = create = 1;
2163                         ifp->cstamp = jiffies;
2164                         addrconf_dad_start(ifp);
2165                 }
2166
2167                 if (ifp) {
2168                         int flags;
2169                         unsigned long now;
2170 #ifdef CONFIG_IPV6_PRIVACY
2171                         struct inet6_ifaddr *ift;
2172 #endif
2173                         u32 stored_lft;
2174
2175                         /* update lifetime (RFC2462 5.5.3 e) */
2176                         spin_lock(&ifp->lock);
2177                         now = jiffies;
2178                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2179                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2180                         else
2181                                 stored_lft = 0;
2182                         if (!update_lft && stored_lft) {
2183                                 if (valid_lft > MIN_VALID_LIFETIME ||
2184                                     valid_lft > stored_lft)
2185                                         update_lft = 1;
2186                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2187                                         /* valid_lft <= stored_lft is always true */
2188                                         /*
2189                                          * RFC 4862 Section 5.5.3e:
2190                                          * "Note that the preferred lifetime of
2191                                          *  the corresponding address is always
2192                                          *  reset to the Preferred Lifetime in
2193                                          *  the received Prefix Information
2194                                          *  option, regardless of whether the
2195                                          *  valid lifetime is also reset or
2196                                          *  ignored."
2197                                          *
2198                                          *  So if the preferred lifetime in
2199                                          *  this advertisement is different
2200                                          *  than what we have stored, but the
2201                                          *  valid lifetime is invalid, just
2202                                          *  reset prefered_lft.
2203                                          *
2204                                          *  We must set the valid lifetime
2205                                          *  to the stored lifetime since we'll
2206                                          *  be updating the timestamp below,
2207                                          *  else we'll set it back to the
2208                                          *  minimum.
2209                                          */
2210                                         if (prefered_lft != ifp->prefered_lft) {
2211                                                 valid_lft = stored_lft;
2212                                                 update_lft = 1;
2213                                         }
2214                                 } else {
2215                                         valid_lft = MIN_VALID_LIFETIME;
2216                                         if (valid_lft < prefered_lft)
2217                                                 prefered_lft = valid_lft;
2218                                         update_lft = 1;
2219                                 }
2220                         }
2221
2222                         if (update_lft) {
2223                                 ifp->valid_lft = valid_lft;
2224                                 ifp->prefered_lft = prefered_lft;
2225                                 ifp->tstamp = now;
2226                                 flags = ifp->flags;
2227                                 ifp->flags &= ~IFA_F_DEPRECATED;
2228                                 spin_unlock(&ifp->lock);
2229
2230                                 if (!(flags&IFA_F_TENTATIVE))
2231                                         ipv6_ifa_notify(0, ifp);
2232                         } else
2233                                 spin_unlock(&ifp->lock);
2234
2235 #ifdef CONFIG_IPV6_PRIVACY
2236                         read_lock_bh(&in6_dev->lock);
2237                         /* update all temporary addresses in the list */
2238                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2239                                             tmp_list) {
2240                                 int age, max_valid, max_prefered;
2241
2242                                 if (ifp != ift->ifpub)
2243                                         continue;
2244
2245                                 /*
2246                                  * RFC 4941 section 3.3:
2247                                  * If a received option will extend the lifetime
2248                                  * of a public address, the lifetimes of
2249                                  * temporary addresses should be extended,
2250                                  * subject to the overall constraint that no
2251                                  * temporary addresses should ever remain
2252                                  * "valid" or "preferred" for a time longer than
2253                                  * (TEMP_VALID_LIFETIME) or
2254                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2255                                  * respectively.
2256                                  */
2257                                 age = (now - ift->cstamp) / HZ;
2258                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2259                                 if (max_valid < 0)
2260                                         max_valid = 0;
2261
2262                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2263                                                in6_dev->cnf.max_desync_factor -
2264                                                age;
2265                                 if (max_prefered < 0)
2266                                         max_prefered = 0;
2267
2268                                 if (valid_lft > max_valid)
2269                                         valid_lft = max_valid;
2270
2271                                 if (prefered_lft > max_prefered)
2272                                         prefered_lft = max_prefered;
2273
2274                                 spin_lock(&ift->lock);
2275                                 flags = ift->flags;
2276                                 ift->valid_lft = valid_lft;
2277                                 ift->prefered_lft = prefered_lft;
2278                                 ift->tstamp = now;
2279                                 if (prefered_lft > 0)
2280                                         ift->flags &= ~IFA_F_DEPRECATED;
2281
2282                                 spin_unlock(&ift->lock);
2283                                 if (!(flags&IFA_F_TENTATIVE))
2284                                         ipv6_ifa_notify(0, ift);
2285                         }
2286
2287                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2288                                 /*
2289                                  * When a new public address is created as
2290                                  * described in [ADDRCONF], also create a new
2291                                  * temporary address. Also create a temporary
2292                                  * address if it's enabled but no temporary
2293                                  * address currently exists.
2294                                  */
2295                                 read_unlock_bh(&in6_dev->lock);
2296                                 ipv6_create_tempaddr(ifp, NULL);
2297                         } else {
2298                                 read_unlock_bh(&in6_dev->lock);
2299                         }
2300 #endif
2301                         in6_ifa_put(ifp);
2302                         addrconf_verify(0);
2303                 }
2304         }
2305         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2306         in6_dev_put(in6_dev);
2307 }
2308
2309 /*
2310  *      Set destination address.
2311  *      Special case for SIT interfaces where we create a new "virtual"
2312  *      device.
2313  */
2314 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2315 {
2316         struct in6_ifreq ireq;
2317         struct net_device *dev;
2318         int err = -EINVAL;
2319
2320         rtnl_lock();
2321
2322         err = -EFAULT;
2323         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2324                 goto err_exit;
2325
2326         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2327
2328         err = -ENODEV;
2329         if (dev == NULL)
2330                 goto err_exit;
2331
2332 #if IS_ENABLED(CONFIG_IPV6_SIT)
2333         if (dev->type == ARPHRD_SIT) {
2334                 const struct net_device_ops *ops = dev->netdev_ops;
2335                 struct ifreq ifr;
2336                 struct ip_tunnel_parm p;
2337
2338                 err = -EADDRNOTAVAIL;
2339                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2340                         goto err_exit;
2341
2342                 memset(&p, 0, sizeof(p));
2343                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2344                 p.iph.saddr = 0;
2345                 p.iph.version = 4;
2346                 p.iph.ihl = 5;
2347                 p.iph.protocol = IPPROTO_IPV6;
2348                 p.iph.ttl = 64;
2349                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2350
2351                 if (ops->ndo_do_ioctl) {
2352                         mm_segment_t oldfs = get_fs();
2353
2354                         set_fs(KERNEL_DS);
2355                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2356                         set_fs(oldfs);
2357                 } else
2358                         err = -EOPNOTSUPP;
2359
2360                 if (err == 0) {
2361                         err = -ENOBUFS;
2362                         dev = __dev_get_by_name(net, p.name);
2363                         if (!dev)
2364                                 goto err_exit;
2365                         err = dev_open(dev);
2366                 }
2367         }
2368 #endif
2369
2370 err_exit:
2371         rtnl_unlock();
2372         return err;
2373 }
2374
2375 /*
2376  *      Manual configuration of address on an interface
2377  */
2378 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2379                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2380                           __u32 valid_lft)
2381 {
2382         struct inet6_ifaddr *ifp;
2383         struct inet6_dev *idev;
2384         struct net_device *dev;
2385         int scope;
2386         u32 flags;
2387         clock_t expires;
2388         unsigned long timeout;
2389
2390         ASSERT_RTNL();
2391
2392         if (plen > 128)
2393                 return -EINVAL;
2394
2395         /* check the lifetime */
2396         if (!valid_lft || prefered_lft > valid_lft)
2397                 return -EINVAL;
2398
2399         dev = __dev_get_by_index(net, ifindex);
2400         if (!dev)
2401                 return -ENODEV;
2402
2403         idev = addrconf_add_dev(dev);
2404         if (IS_ERR(idev))
2405                 return PTR_ERR(idev);
2406
2407         scope = ipv6_addr_scope(pfx);
2408
2409         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2410         if (addrconf_finite_timeout(timeout)) {
2411                 expires = jiffies_to_clock_t(timeout * HZ);
2412                 valid_lft = timeout;
2413                 flags = RTF_EXPIRES;
2414         } else {
2415                 expires = 0;
2416                 flags = 0;
2417                 ifa_flags |= IFA_F_PERMANENT;
2418         }
2419
2420         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2421         if (addrconf_finite_timeout(timeout)) {
2422                 if (timeout == 0)
2423                         ifa_flags |= IFA_F_DEPRECATED;
2424                 prefered_lft = timeout;
2425         }
2426
2427         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2428
2429         if (!IS_ERR(ifp)) {
2430                 spin_lock_bh(&ifp->lock);
2431                 ifp->valid_lft = valid_lft;
2432                 ifp->prefered_lft = prefered_lft;
2433                 ifp->tstamp = jiffies;
2434                 spin_unlock_bh(&ifp->lock);
2435
2436                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2437                                       expires, flags);
2438                 /*
2439                  * Note that section 3.1 of RFC 4429 indicates
2440                  * that the Optimistic flag should not be set for
2441                  * manually configured addresses
2442                  */
2443                 addrconf_dad_start(ifp);
2444                 in6_ifa_put(ifp);
2445                 addrconf_verify(0);
2446                 return 0;
2447         }
2448
2449         return PTR_ERR(ifp);
2450 }
2451
2452 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2453                           unsigned int plen)
2454 {
2455         struct inet6_ifaddr *ifp;
2456         struct inet6_dev *idev;
2457         struct net_device *dev;
2458
2459         if (plen > 128)
2460                 return -EINVAL;
2461
2462         dev = __dev_get_by_index(net, ifindex);
2463         if (!dev)
2464                 return -ENODEV;
2465
2466         if ((idev = __in6_dev_get(dev)) == NULL)
2467                 return -ENXIO;
2468
2469         read_lock_bh(&idev->lock);
2470         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2471                 if (ifp->prefix_len == plen &&
2472                     ipv6_addr_equal(pfx, &ifp->addr)) {
2473                         in6_ifa_hold(ifp);
2474                         read_unlock_bh(&idev->lock);
2475
2476                         ipv6_del_addr(ifp);
2477
2478                         /* If the last address is deleted administratively,
2479                            disable IPv6 on this interface.
2480                          */
2481                         if (list_empty(&idev->addr_list))
2482                                 addrconf_ifdown(idev->dev, 1);
2483                         return 0;
2484                 }
2485         }
2486         read_unlock_bh(&idev->lock);
2487         return -EADDRNOTAVAIL;
2488 }
2489
2490
2491 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2492 {
2493         struct in6_ifreq ireq;
2494         int err;
2495
2496         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2497                 return -EPERM;
2498
2499         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2500                 return -EFAULT;
2501
2502         rtnl_lock();
2503         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2504                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2505                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2506         rtnl_unlock();
2507         return err;
2508 }
2509
2510 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2511 {
2512         struct in6_ifreq ireq;
2513         int err;
2514
2515         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2516                 return -EPERM;
2517
2518         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2519                 return -EFAULT;
2520
2521         rtnl_lock();
2522         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2523                              ireq.ifr6_prefixlen);
2524         rtnl_unlock();
2525         return err;
2526 }
2527
2528 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2529                      int plen, int scope)
2530 {
2531         struct inet6_ifaddr *ifp;
2532
2533         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2534         if (!IS_ERR(ifp)) {
2535                 spin_lock_bh(&ifp->lock);
2536                 ifp->flags &= ~IFA_F_TENTATIVE;
2537                 spin_unlock_bh(&ifp->lock);
2538                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2539                 in6_ifa_put(ifp);
2540         }
2541 }
2542
2543 #if IS_ENABLED(CONFIG_IPV6_SIT)
2544 static void sit_add_v4_addrs(struct inet6_dev *idev)
2545 {
2546         struct in6_addr addr;
2547         struct net_device *dev;
2548         struct net *net = dev_net(idev->dev);
2549         int scope;
2550
2551         ASSERT_RTNL();
2552
2553         memset(&addr, 0, sizeof(struct in6_addr));
2554         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2555
2556         if (idev->dev->flags&IFF_POINTOPOINT) {
2557                 addr.s6_addr32[0] = htonl(0xfe800000);
2558                 scope = IFA_LINK;
2559         } else {
2560                 scope = IPV6_ADDR_COMPATv4;
2561         }
2562
2563         if (addr.s6_addr32[3]) {
2564                 add_addr(idev, &addr, 128, scope);
2565                 return;
2566         }
2567
2568         for_each_netdev(net, dev) {
2569                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2570                 if (in_dev && (dev->flags & IFF_UP)) {
2571                         struct in_ifaddr *ifa;
2572
2573                         int flag = scope;
2574
2575                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2576                                 int plen;
2577
2578                                 addr.s6_addr32[3] = ifa->ifa_local;
2579
2580                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2581                                         continue;
2582                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2583                                         if (idev->dev->flags&IFF_POINTOPOINT)
2584                                                 continue;
2585                                         flag |= IFA_HOST;
2586                                 }
2587                                 if (idev->dev->flags&IFF_POINTOPOINT)
2588                                         plen = 64;
2589                                 else
2590                                         plen = 96;
2591
2592                                 add_addr(idev, &addr, plen, flag);
2593                         }
2594                 }
2595         }
2596 }
2597 #endif
2598
2599 static void init_loopback(struct net_device *dev)
2600 {
2601         struct inet6_dev  *idev;
2602
2603         /* ::1 */
2604
2605         ASSERT_RTNL();
2606
2607         if ((idev = ipv6_find_idev(dev)) == NULL) {
2608                 pr_debug("%s: add_dev failed\n", __func__);
2609                 return;
2610         }
2611
2612         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2613 }
2614
2615 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2616 {
2617         struct inet6_ifaddr *ifp;
2618         u32 addr_flags = IFA_F_PERMANENT;
2619
2620 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2621         if (idev->cnf.optimistic_dad &&
2622             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2623                 addr_flags |= IFA_F_OPTIMISTIC;
2624 #endif
2625
2626
2627         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2628         if (!IS_ERR(ifp)) {
2629                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2630                 addrconf_dad_start(ifp);
2631                 in6_ifa_put(ifp);
2632         }
2633 }
2634
2635 static void addrconf_dev_config(struct net_device *dev)
2636 {
2637         struct in6_addr addr;
2638         struct inet6_dev *idev;
2639
2640         ASSERT_RTNL();
2641
2642         if ((dev->type != ARPHRD_ETHER) &&
2643             (dev->type != ARPHRD_FDDI) &&
2644             (dev->type != ARPHRD_ARCNET) &&
2645             (dev->type != ARPHRD_INFINIBAND) &&
2646             (dev->type != ARPHRD_IEEE802154)) {
2647                 /* Alas, we support only Ethernet autoconfiguration. */
2648                 return;
2649         }
2650
2651         idev = addrconf_add_dev(dev);
2652         if (IS_ERR(idev))
2653                 return;
2654
2655         memset(&addr, 0, sizeof(struct in6_addr));
2656         addr.s6_addr32[0] = htonl(0xFE800000);
2657
2658         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2659                 addrconf_add_linklocal(idev, &addr);
2660 }
2661
2662 #if IS_ENABLED(CONFIG_IPV6_SIT)
2663 static void addrconf_sit_config(struct net_device *dev)
2664 {
2665         struct inet6_dev *idev;
2666
2667         ASSERT_RTNL();
2668
2669         /*
2670          * Configure the tunnel with one of our IPv4
2671          * addresses... we should configure all of
2672          * our v4 addrs in the tunnel
2673          */
2674
2675         if ((idev = ipv6_find_idev(dev)) == NULL) {
2676                 pr_debug("%s: add_dev failed\n", __func__);
2677                 return;
2678         }
2679
2680         if (dev->priv_flags & IFF_ISATAP) {
2681                 struct in6_addr addr;
2682
2683                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2684                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2685                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2686                         addrconf_add_linklocal(idev, &addr);
2687                 return;
2688         }
2689
2690         sit_add_v4_addrs(idev);
2691
2692         if (dev->flags&IFF_POINTOPOINT)
2693                 addrconf_add_mroute(dev);
2694         else
2695                 sit_route_add(dev);
2696 }
2697 #endif
2698
2699 #if IS_ENABLED(CONFIG_NET_IPGRE)
2700 static void addrconf_gre_config(struct net_device *dev)
2701 {
2702         struct inet6_dev *idev;
2703         struct in6_addr addr;
2704
2705         pr_info("%s(%s)\n", __func__, dev->name);
2706
2707         ASSERT_RTNL();
2708
2709         if ((idev = ipv6_find_idev(dev)) == NULL) {
2710                 pr_debug("%s: add_dev failed\n", __func__);
2711                 return;
2712         }
2713
2714         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2715         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2716
2717         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2718                 addrconf_add_linklocal(idev, &addr);
2719 }
2720 #endif
2721
2722 static inline int
2723 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2724 {
2725         struct in6_addr lladdr;
2726
2727         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2728                 addrconf_add_linklocal(idev, &lladdr);
2729                 return 0;
2730         }
2731         return -1;
2732 }
2733
2734 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2735 {
2736         struct net_device *link_dev;
2737         struct net *net = dev_net(idev->dev);
2738
2739         /* first try to inherit the link-local address from the link device */
2740         if (idev->dev->iflink &&
2741             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2742                 if (!ipv6_inherit_linklocal(idev, link_dev))
2743                         return;
2744         }
2745         /* then try to inherit it from any device */
2746         for_each_netdev(net, link_dev) {
2747                 if (!ipv6_inherit_linklocal(idev, link_dev))
2748                         return;
2749         }
2750         pr_debug("init ip6-ip6: add_linklocal failed\n");
2751 }
2752
2753 /*
2754  * Autoconfigure tunnel with a link-local address so routing protocols,
2755  * DHCPv6, MLD etc. can be run over the virtual link
2756  */
2757
2758 static void addrconf_ip6_tnl_config(struct net_device *dev)
2759 {
2760         struct inet6_dev *idev;
2761
2762         ASSERT_RTNL();
2763
2764         idev = addrconf_add_dev(dev);
2765         if (IS_ERR(idev)) {
2766                 pr_debug("init ip6-ip6: add_dev failed\n");
2767                 return;
2768         }
2769         ip6_tnl_add_linklocal(idev);
2770 }
2771
2772 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2773                            void *data)
2774 {
2775         struct net_device *dev = (struct net_device *) data;
2776         struct inet6_dev *idev = __in6_dev_get(dev);
2777         int run_pending = 0;
2778         int err;
2779
2780         switch (event) {
2781         case NETDEV_REGISTER:
2782                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2783                         idev = ipv6_add_dev(dev);
2784                         if (!idev)
2785                                 return notifier_from_errno(-ENOMEM);
2786                 }
2787                 break;
2788
2789         case NETDEV_UP:
2790         case NETDEV_CHANGE:
2791                 if (dev->flags & IFF_SLAVE)
2792                         break;
2793
2794                 if (event == NETDEV_UP) {
2795                         if (!addrconf_qdisc_ok(dev)) {
2796                                 /* device is not ready yet. */
2797                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2798                                         dev->name);
2799                                 break;
2800                         }
2801
2802                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2803                                 idev = ipv6_add_dev(dev);
2804
2805                         if (idev) {
2806                                 idev->if_flags |= IF_READY;
2807                                 run_pending = 1;
2808                         }
2809                 } else {
2810                         if (!addrconf_qdisc_ok(dev)) {
2811                                 /* device is still not ready. */
2812                                 break;
2813                         }
2814
2815                         if (idev) {
2816                                 if (idev->if_flags & IF_READY)
2817                                         /* device is already configured. */
2818                                         break;
2819                                 idev->if_flags |= IF_READY;
2820                         }
2821
2822                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2823                                 dev->name);
2824
2825                         run_pending = 1;
2826                 }
2827
2828                 switch (dev->type) {
2829 #if IS_ENABLED(CONFIG_IPV6_SIT)
2830                 case ARPHRD_SIT:
2831                         addrconf_sit_config(dev);
2832                         break;
2833 #endif
2834 #if IS_ENABLED(CONFIG_NET_IPGRE)
2835                 case ARPHRD_IPGRE:
2836                         addrconf_gre_config(dev);
2837                         break;
2838 #endif
2839                 case ARPHRD_TUNNEL6:
2840                         addrconf_ip6_tnl_config(dev);
2841                         break;
2842                 case ARPHRD_LOOPBACK:
2843                         init_loopback(dev);
2844                         break;
2845
2846                 default:
2847                         addrconf_dev_config(dev);
2848                         break;
2849                 }
2850
2851                 if (idev) {
2852                         if (run_pending)
2853                                 addrconf_dad_run(idev);
2854
2855                         /*
2856                          * If the MTU changed during the interface down,
2857                          * when the interface up, the changed MTU must be
2858                          * reflected in the idev as well as routers.
2859                          */
2860                         if (idev->cnf.mtu6 != dev->mtu &&
2861                             dev->mtu >= IPV6_MIN_MTU) {
2862                                 rt6_mtu_change(dev, dev->mtu);
2863                                 idev->cnf.mtu6 = dev->mtu;
2864                         }
2865                         idev->tstamp = jiffies;
2866                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2867
2868                         /*
2869                          * If the changed mtu during down is lower than
2870                          * IPV6_MIN_MTU stop IPv6 on this interface.
2871                          */
2872                         if (dev->mtu < IPV6_MIN_MTU)
2873                                 addrconf_ifdown(dev, 1);
2874                 }
2875                 break;
2876
2877         case NETDEV_CHANGEMTU:
2878                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2879                         rt6_mtu_change(dev, dev->mtu);
2880                         idev->cnf.mtu6 = dev->mtu;
2881                         break;
2882                 }
2883
2884                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2885                         idev = ipv6_add_dev(dev);
2886                         if (idev)
2887                                 break;
2888                 }
2889
2890                 /*
2891                  * MTU falled under IPV6_MIN_MTU.
2892                  * Stop IPv6 on this interface.
2893                  */
2894
2895         case NETDEV_DOWN:
2896         case NETDEV_UNREGISTER:
2897                 /*
2898                  *      Remove all addresses from this interface.
2899                  */
2900                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2901                 break;
2902
2903         case NETDEV_CHANGENAME:
2904                 if (idev) {
2905                         snmp6_unregister_dev(idev);
2906                         addrconf_sysctl_unregister(idev);
2907                         addrconf_sysctl_register(idev);
2908                         err = snmp6_register_dev(idev);
2909                         if (err)
2910                                 return notifier_from_errno(err);
2911                 }
2912                 break;
2913
2914         case NETDEV_PRE_TYPE_CHANGE:
2915         case NETDEV_POST_TYPE_CHANGE:
2916                 addrconf_type_change(dev, event);
2917                 break;
2918         }
2919
2920         return NOTIFY_OK;
2921 }
2922
2923 /*
2924  *      addrconf module should be notified of a device going up
2925  */
2926 static struct notifier_block ipv6_dev_notf = {
2927         .notifier_call = addrconf_notify,
2928 };
2929
2930 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2931 {
2932         struct inet6_dev *idev;
2933         ASSERT_RTNL();
2934
2935         idev = __in6_dev_get(dev);
2936
2937         if (event == NETDEV_POST_TYPE_CHANGE)
2938                 ipv6_mc_remap(idev);
2939         else if (event == NETDEV_PRE_TYPE_CHANGE)
2940                 ipv6_mc_unmap(idev);
2941 }
2942
2943 static int addrconf_ifdown(struct net_device *dev, int how)
2944 {
2945         struct net *net = dev_net(dev);
2946         struct inet6_dev *idev;
2947         struct inet6_ifaddr *ifa;
2948         int state, i;
2949
2950         ASSERT_RTNL();
2951
2952         rt6_ifdown(net, dev);
2953         neigh_ifdown(&nd_tbl, dev);
2954
2955         idev = __in6_dev_get(dev);
2956         if (idev == NULL)
2957                 return -ENODEV;
2958
2959         /*
2960          * Step 1: remove reference to ipv6 device from parent device.
2961          *         Do not dev_put!
2962          */
2963         if (how) {
2964                 idev->dead = 1;
2965
2966                 /* protected by rtnl_lock */
2967                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2968
2969                 /* Step 1.5: remove snmp6 entry */
2970                 snmp6_unregister_dev(idev);
2971
2972         }
2973
2974         /* Step 2: clear hash table */
2975         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2976                 struct hlist_head *h = &inet6_addr_lst[i];
2977
2978                 spin_lock_bh(&addrconf_hash_lock);
2979         restart:
2980                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2981                         if (ifa->idev == idev) {
2982                                 hlist_del_init_rcu(&ifa->addr_lst);
2983                                 addrconf_del_timer(ifa);
2984                                 goto restart;
2985                         }
2986                 }
2987                 spin_unlock_bh(&addrconf_hash_lock);
2988         }
2989
2990         write_lock_bh(&idev->lock);
2991
2992         /* Step 2: clear flags for stateless addrconf */
2993         if (!how)
2994                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2995
2996 #ifdef CONFIG_IPV6_PRIVACY
2997         if (how && del_timer(&idev->regen_timer))
2998                 in6_dev_put(idev);
2999
3000         /* Step 3: clear tempaddr list */
3001         while (!list_empty(&idev->tempaddr_list)) {
3002                 ifa = list_first_entry(&idev->tempaddr_list,
3003                                        struct inet6_ifaddr, tmp_list);
3004                 list_del(&ifa->tmp_list);
3005                 write_unlock_bh(&idev->lock);
3006                 spin_lock_bh(&ifa->lock);
3007
3008                 if (ifa->ifpub) {
3009                         in6_ifa_put(ifa->ifpub);
3010                         ifa->ifpub = NULL;
3011                 }
3012                 spin_unlock_bh(&ifa->lock);
3013                 in6_ifa_put(ifa);
3014                 write_lock_bh(&idev->lock);
3015         }
3016 #endif
3017
3018         while (!list_empty(&idev->addr_list)) {
3019                 ifa = list_first_entry(&idev->addr_list,
3020                                        struct inet6_ifaddr, if_list);
3021                 addrconf_del_timer(ifa);
3022
3023                 list_del(&ifa->if_list);
3024
3025                 write_unlock_bh(&idev->lock);
3026
3027                 spin_lock_bh(&ifa->state_lock);
3028                 state = ifa->state;
3029                 ifa->state = INET6_IFADDR_STATE_DEAD;
3030                 spin_unlock_bh(&ifa->state_lock);
3031
3032                 if (state != INET6_IFADDR_STATE_DEAD) {
3033                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3034                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
3035                 }
3036                 in6_ifa_put(ifa);
3037
3038                 write_lock_bh(&idev->lock);
3039         }
3040
3041         write_unlock_bh(&idev->lock);
3042
3043         /* Step 5: Discard multicast list */
3044         if (how)
3045                 ipv6_mc_destroy_dev(idev);
3046         else
3047                 ipv6_mc_down(idev);
3048
3049         idev->tstamp = jiffies;
3050
3051         /* Last: Shot the device (if unregistered) */
3052         if (how) {
3053                 addrconf_sysctl_unregister(idev);
3054                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3055                 neigh_ifdown(&nd_tbl, dev);
3056                 in6_dev_put(idev);
3057         }
3058         return 0;
3059 }
3060
3061 static void addrconf_rs_timer(unsigned long data)
3062 {
3063         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3064         struct inet6_dev *idev = ifp->idev;
3065
3066         read_lock(&idev->lock);
3067         if (idev->dead || !(idev->if_flags & IF_READY))
3068                 goto out;
3069
3070         if (!ipv6_accept_ra(idev))
3071                 goto out;
3072
3073         /* Announcement received after solicitation was sent */
3074         if (idev->if_flags & IF_RA_RCVD)
3075                 goto out;
3076
3077         spin_lock(&ifp->lock);
3078         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3079                 /* The wait after the last probe can be shorter */
3080                 addrconf_mod_timer(ifp, AC_RS,
3081                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3082                                    idev->cnf.rtr_solicit_delay :
3083                                    idev->cnf.rtr_solicit_interval);
3084                 spin_unlock(&ifp->lock);
3085
3086                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3087         } else {
3088                 spin_unlock(&ifp->lock);
3089                 /*
3090                  * Note: we do not support deprecated "all on-link"
3091                  * assumption any longer.
3092                  */
3093                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3094         }
3095
3096 out:
3097         read_unlock(&idev->lock);
3098         in6_ifa_put(ifp);
3099 }
3100
3101 /*
3102  *      Duplicate Address Detection
3103  */
3104 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3105 {
3106         unsigned long rand_num;
3107         struct inet6_dev *idev = ifp->idev;
3108
3109         if (ifp->flags & IFA_F_OPTIMISTIC)
3110                 rand_num = 0;
3111         else
3112                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3113
3114         ifp->probes = idev->cnf.dad_transmits;
3115         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3116 }
3117
3118 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3119 {
3120         struct inet6_dev *idev = ifp->idev;
3121         struct net_device *dev = idev->dev;
3122
3123         addrconf_join_solict(dev, &ifp->addr);
3124
3125         net_srandom(ifp->addr.s6_addr32[3]);
3126
3127         read_lock_bh(&idev->lock);
3128         spin_lock(&ifp->lock);
3129         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3130                 goto out;
3131
3132         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3133             idev->cnf.accept_dad < 1 ||
3134             !(ifp->flags&IFA_F_TENTATIVE) ||
3135             ifp->flags & IFA_F_NODAD) {
3136                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3137                 spin_unlock(&ifp->lock);
3138                 read_unlock_bh(&idev->lock);
3139
3140                 addrconf_dad_completed(ifp);
3141                 return;
3142         }
3143
3144         if (!(idev->if_flags & IF_READY)) {
3145                 spin_unlock(&ifp->lock);
3146                 read_unlock_bh(&idev->lock);
3147                 /*
3148                  * If the device is not ready:
3149                  * - keep it tentative if it is a permanent address.
3150                  * - otherwise, kill it.
3151                  */
3152                 in6_ifa_hold(ifp);
3153                 addrconf_dad_stop(ifp, 0);
3154                 return;
3155         }
3156
3157         /*
3158          * Optimistic nodes can start receiving
3159          * Frames right away
3160          */
3161         if (ifp->flags & IFA_F_OPTIMISTIC)
3162                 ip6_ins_rt(ifp->rt);
3163
3164         addrconf_dad_kick(ifp);
3165 out:
3166         spin_unlock(&ifp->lock);
3167         read_unlock_bh(&idev->lock);
3168 }
3169
3170 static void addrconf_dad_timer(unsigned long data)
3171 {
3172         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3173         struct inet6_dev *idev = ifp->idev;
3174         struct in6_addr mcaddr;
3175
3176         if (!ifp->probes && addrconf_dad_end(ifp))
3177                 goto out;
3178
3179         read_lock(&idev->lock);
3180         if (idev->dead || !(idev->if_flags & IF_READY)) {
3181                 read_unlock(&idev->lock);
3182                 goto out;
3183         }
3184
3185         spin_lock(&ifp->lock);
3186         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3187                 spin_unlock(&ifp->lock);
3188                 read_unlock(&idev->lock);
3189                 goto out;
3190         }
3191
3192         if (ifp->probes == 0) {
3193                 /*
3194                  * DAD was successful
3195                  */
3196
3197                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3198                 spin_unlock(&ifp->lock);
3199                 read_unlock(&idev->lock);
3200
3201                 addrconf_dad_completed(ifp);
3202
3203                 goto out;
3204         }
3205
3206         ifp->probes--;
3207         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3208         spin_unlock(&ifp->lock);
3209         read_unlock(&idev->lock);
3210
3211         /* send a neighbour solicitation for our addr */
3212         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3213         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3214 out:
3215         in6_ifa_put(ifp);
3216 }
3217
3218 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3219 {
3220         struct net_device *dev = ifp->idev->dev;
3221
3222         /*
3223          *      Configure the address for reception. Now it is valid.
3224          */
3225
3226         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3227
3228         /* If added prefix is link local and we are prepared to process
3229            router advertisements, start sending router solicitations.
3230          */
3231
3232         if (ipv6_accept_ra(ifp->idev) &&
3233             ifp->idev->cnf.rtr_solicits > 0 &&
3234             (dev->flags&IFF_LOOPBACK) == 0 &&
3235             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3236                 /*
3237                  *      If a host as already performed a random delay
3238                  *      [...] as part of DAD [...] there is no need
3239                  *      to delay again before sending the first RS
3240                  */
3241                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3242
3243                 spin_lock_bh(&ifp->lock);
3244                 ifp->probes = 1;
3245                 ifp->idev->if_flags |= IF_RS_SENT;
3246                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3247                 spin_unlock_bh(&ifp->lock);
3248         }
3249 }
3250
3251 static void addrconf_dad_run(struct inet6_dev *idev)
3252 {
3253         struct inet6_ifaddr *ifp;
3254
3255         read_lock_bh(&idev->lock);
3256         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3257                 spin_lock(&ifp->lock);
3258                 if (ifp->flags & IFA_F_TENTATIVE &&
3259                     ifp->state == INET6_IFADDR_STATE_DAD)
3260                         addrconf_dad_kick(ifp);
3261                 spin_unlock(&ifp->lock);
3262         }
3263         read_unlock_bh(&idev->lock);
3264 }
3265
3266 #ifdef CONFIG_PROC_FS
3267 struct if6_iter_state {
3268         struct seq_net_private p;
3269         int bucket;
3270         int offset;
3271 };
3272
3273 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3274 {
3275         struct inet6_ifaddr *ifa = NULL;
3276         struct if6_iter_state *state = seq->private;
3277         struct net *net = seq_file_net(seq);
3278         int p = 0;
3279
3280         /* initial bucket if pos is 0 */
3281         if (pos == 0) {
3282                 state->bucket = 0;
3283                 state->offset = 0;
3284         }
3285
3286         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3287                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3288                                          addr_lst) {
3289                         if (!net_eq(dev_net(ifa->idev->dev), net))
3290                                 continue;
3291                         /* sync with offset */
3292                         if (p < state->offset) {
3293                                 p++;
3294                                 continue;
3295                         }
3296                         state->offset++;
3297                         return ifa;
3298                 }
3299
3300                 /* prepare for next bucket */
3301                 state->offset = 0;
3302                 p = 0;
3303         }
3304         return NULL;
3305 }
3306
3307 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3308                                          struct inet6_ifaddr *ifa)
3309 {
3310         struct if6_iter_state *state = seq->private;
3311         struct net *net = seq_file_net(seq);
3312
3313         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3314                 if (!net_eq(dev_net(ifa->idev->dev), net))
3315                         continue;
3316                 state->offset++;
3317                 return ifa;
3318         }
3319
3320         while (++state->bucket < IN6_ADDR_HSIZE) {
3321                 state->offset = 0;
3322                 hlist_for_each_entry_rcu_bh(ifa,
3323                                      &inet6_addr_lst[state->bucket], addr_lst) {
3324                         if (!net_eq(dev_net(ifa->idev->dev), net))
3325                                 continue;
3326                         state->offset++;
3327                         return ifa;
3328                 }
3329         }
3330
3331         return NULL;
3332 }
3333
3334 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3335         __acquires(rcu_bh)
3336 {
3337         rcu_read_lock_bh();
3338         return if6_get_first(seq, *pos);
3339 }
3340
3341 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3342 {
3343         struct inet6_ifaddr *ifa;
3344
3345         ifa = if6_get_next(seq, v);
3346         ++*pos;
3347         return ifa;
3348 }
3349
3350 static void if6_seq_stop(struct seq_file *seq, void *v)
3351         __releases(rcu_bh)
3352 {
3353         rcu_read_unlock_bh();
3354 }
3355
3356 static int if6_seq_show(struct seq_file *seq, void *v)
3357 {
3358         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3359         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3360                    &ifp->addr,
3361                    ifp->idev->dev->ifindex,
3362                    ifp->prefix_len,
3363                    ifp->scope,
3364                    ifp->flags,
3365                    ifp->idev->dev->name);
3366         return 0;
3367 }
3368
3369 static const struct seq_operations if6_seq_ops = {
3370         .start  = if6_seq_start,
3371         .next   = if6_seq_next,
3372         .show   = if6_seq_show,
3373         .stop   = if6_seq_stop,
3374 };
3375
3376 static int if6_seq_open(struct inode *inode, struct file *file)
3377 {
3378         return seq_open_net(inode, file, &if6_seq_ops,
3379                             sizeof(struct if6_iter_state));
3380 }
3381
3382 static const struct file_operations if6_fops = {
3383         .owner          = THIS_MODULE,
3384         .open           = if6_seq_open,
3385         .read           = seq_read,
3386         .llseek         = seq_lseek,
3387         .release        = seq_release_net,
3388 };
3389
3390 static int __net_init if6_proc_net_init(struct net *net)
3391 {
3392         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3393                 return -ENOMEM;
3394         return 0;
3395 }
3396
3397 static void __net_exit if6_proc_net_exit(struct net *net)
3398 {
3399         remove_proc_entry("if_inet6", net->proc_net);
3400 }
3401
3402 static struct pernet_operations if6_proc_net_ops = {
3403        .init = if6_proc_net_init,
3404        .exit = if6_proc_net_exit,
3405 };
3406
3407 int __init if6_proc_init(void)
3408 {
3409         return register_pernet_subsys(&if6_proc_net_ops);
3410 }
3411
3412 void if6_proc_exit(void)
3413 {
3414         unregister_pernet_subsys(&if6_proc_net_ops);
3415 }
3416 #endif  /* CONFIG_PROC_FS */
3417
3418 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3419 /* Check if address is a home address configured on any interface. */
3420 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3421 {
3422         int ret = 0;
3423         struct inet6_ifaddr *ifp = NULL;
3424         unsigned int hash = inet6_addr_hash(addr);
3425
3426         rcu_read_lock_bh();
3427         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3428                 if (!net_eq(dev_net(ifp->idev->dev), net))
3429                         continue;
3430                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3431                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3432                         ret = 1;
3433                         break;
3434                 }
3435         }
3436         rcu_read_unlock_bh();
3437         return ret;
3438 }
3439 #endif
3440
3441 /*
3442  *      Periodic address status verification
3443  */
3444
3445 static void addrconf_verify(unsigned long foo)
3446 {
3447         unsigned long now, next, next_sec, next_sched;
3448         struct inet6_ifaddr *ifp;
3449         int i;
3450
3451         rcu_read_lock_bh();
3452         spin_lock(&addrconf_verify_lock);
3453         now = jiffies;
3454         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3455
3456         del_timer(&addr_chk_timer);
3457
3458         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3459 restart:
3460                 hlist_for_each_entry_rcu_bh(ifp,
3461                                          &inet6_addr_lst[i], addr_lst) {
3462                         unsigned long age;
3463
3464                         if (ifp->flags & IFA_F_PERMANENT)
3465                                 continue;
3466
3467                         spin_lock(&ifp->lock);
3468                         /* We try to batch several events at once. */
3469                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3470
3471                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3472                             age >= ifp->valid_lft) {
3473                                 spin_unlock(&ifp->lock);
3474                                 in6_ifa_hold(ifp);
3475                                 ipv6_del_addr(ifp);
3476                                 goto restart;
3477                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3478                                 spin_unlock(&ifp->lock);
3479                                 continue;
3480                         } else if (age >= ifp->prefered_lft) {
3481                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3482                                 int deprecate = 0;
3483
3484                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3485                                         deprecate = 1;
3486                                         ifp->flags |= IFA_F_DEPRECATED;
3487                                 }
3488
3489                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3490                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3491
3492                                 spin_unlock(&ifp->lock);
3493
3494                                 if (deprecate) {
3495                                         in6_ifa_hold(ifp);
3496
3497                                         ipv6_ifa_notify(0, ifp);
3498                                         in6_ifa_put(ifp);
3499                                         goto restart;
3500                                 }
3501 #ifdef CONFIG_IPV6_PRIVACY
3502                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3503                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3504                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3505                                         ifp->idev->cnf.dad_transmits *
3506                                         ifp->idev->nd_parms->retrans_time / HZ;
3507
3508                                 if (age >= ifp->prefered_lft - regen_advance) {
3509                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3510                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3511                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3512                                         if (!ifp->regen_count && ifpub) {
3513                                                 ifp->regen_count++;
3514                                                 in6_ifa_hold(ifp);
3515                                                 in6_ifa_hold(ifpub);
3516                                                 spin_unlock(&ifp->lock);
3517
3518                                                 spin_lock(&ifpub->lock);
3519                                                 ifpub->regen_count = 0;
3520                                                 spin_unlock(&ifpub->lock);
3521                                                 ipv6_create_tempaddr(ifpub, ifp);
3522                                                 in6_ifa_put(ifpub);
3523                                                 in6_ifa_put(ifp);
3524                                                 goto restart;
3525                                         }
3526                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3527                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3528                                 spin_unlock(&ifp->lock);
3529 #endif
3530                         } else {
3531                                 /* ifp->prefered_lft <= ifp->valid_lft */
3532                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3533                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3534                                 spin_unlock(&ifp->lock);
3535                         }
3536                 }
3537         }
3538
3539         next_sec = round_jiffies_up(next);
3540         next_sched = next;
3541
3542         /* If rounded timeout is accurate enough, accept it. */
3543         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3544                 next_sched = next_sec;
3545
3546         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3547         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3548                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3549
3550         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3551               now, next, next_sec, next_sched));
3552
3553         addr_chk_timer.expires = next_sched;
3554         add_timer(&addr_chk_timer);
3555         spin_unlock(&addrconf_verify_lock);
3556         rcu_read_unlock_bh();
3557 }
3558
3559 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3560 {
3561         struct in6_addr *pfx = NULL;
3562
3563         if (addr)
3564                 pfx = nla_data(addr);
3565
3566         if (local) {
3567                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3568                         pfx = NULL;
3569                 else
3570                         pfx = nla_data(local);
3571         }
3572
3573         return pfx;
3574 }
3575
3576 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3577         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3578         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3579         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3580 };
3581
3582 static int
3583 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3584 {
3585         struct net *net = sock_net(skb->sk);
3586         struct ifaddrmsg *ifm;
3587         struct nlattr *tb[IFA_MAX+1];
3588         struct in6_addr *pfx;
3589         int err;
3590
3591         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3592         if (err < 0)
3593                 return err;
3594
3595         ifm = nlmsg_data(nlh);
3596         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3597         if (pfx == NULL)
3598                 return -EINVAL;
3599
3600         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3601 }
3602
3603 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3604                              u32 prefered_lft, u32 valid_lft)
3605 {
3606         u32 flags;
3607         clock_t expires;
3608         unsigned long timeout;
3609
3610         if (!valid_lft || (prefered_lft > valid_lft))
3611                 return -EINVAL;
3612
3613         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3614         if (addrconf_finite_timeout(timeout)) {
3615                 expires = jiffies_to_clock_t(timeout * HZ);
3616                 valid_lft = timeout;
3617                 flags = RTF_EXPIRES;
3618         } else {
3619                 expires = 0;
3620                 flags = 0;
3621                 ifa_flags |= IFA_F_PERMANENT;
3622         }
3623
3624         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3625         if (addrconf_finite_timeout(timeout)) {
3626                 if (timeout == 0)
3627                         ifa_flags |= IFA_F_DEPRECATED;
3628                 prefered_lft = timeout;
3629         }
3630
3631         spin_lock_bh(&ifp->lock);
3632         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3633         ifp->tstamp = jiffies;
3634         ifp->valid_lft = valid_lft;
3635         ifp->prefered_lft = prefered_lft;
3636
3637         spin_unlock_bh(&ifp->lock);
3638         if (!(ifp->flags&IFA_F_TENTATIVE))
3639                 ipv6_ifa_notify(0, ifp);
3640
3641         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3642                               expires, flags);
3643         addrconf_verify(0);
3644
3645         return 0;
3646 }
3647
3648 static int
3649 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3650 {
3651         struct net *net = sock_net(skb->sk);
3652         struct ifaddrmsg *ifm;
3653         struct nlattr *tb[IFA_MAX+1];
3654         struct in6_addr *pfx;
3655         struct inet6_ifaddr *ifa;
3656         struct net_device *dev;
3657         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3658         u8 ifa_flags;
3659         int err;
3660
3661         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3662         if (err < 0)
3663                 return err;
3664
3665         ifm = nlmsg_data(nlh);
3666         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3667         if (pfx == NULL)
3668                 return -EINVAL;
3669
3670         if (tb[IFA_CACHEINFO]) {
3671                 struct ifa_cacheinfo *ci;
3672
3673                 ci = nla_data(tb[IFA_CACHEINFO]);
3674                 valid_lft = ci->ifa_valid;
3675                 preferred_lft = ci->ifa_prefered;
3676         } else {
3677                 preferred_lft = INFINITY_LIFE_TIME;
3678                 valid_lft = INFINITY_LIFE_TIME;
3679         }
3680
3681         dev =  __dev_get_by_index(net, ifm->ifa_index);
3682         if (dev == NULL)
3683                 return -ENODEV;
3684
3685         /* We ignore other flags so far. */
3686         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3687
3688         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3689         if (ifa == NULL) {
3690                 /*
3691                  * It would be best to check for !NLM_F_CREATE here but
3692                  * userspace alreay relies on not having to provide this.
3693                  */
3694                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3695                                       ifm->ifa_prefixlen, ifa_flags,
3696                                       preferred_lft, valid_lft);
3697         }
3698
3699         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3700             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3701                 err = -EEXIST;
3702         else
3703                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3704
3705         in6_ifa_put(ifa);
3706
3707         return err;
3708 }
3709
3710 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3711                           u8 scope, int ifindex)
3712 {
3713         struct ifaddrmsg *ifm;
3714
3715         ifm = nlmsg_data(nlh);
3716         ifm->ifa_family = AF_INET6;
3717         ifm->ifa_prefixlen = prefixlen;
3718         ifm->ifa_flags = flags;
3719         ifm->ifa_scope = scope;
3720         ifm->ifa_index = ifindex;
3721 }
3722
3723 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3724                          unsigned long tstamp, u32 preferred, u32 valid)
3725 {
3726         struct ifa_cacheinfo ci;
3727
3728         ci.cstamp = cstamp_delta(cstamp);
3729         ci.tstamp = cstamp_delta(tstamp);
3730         ci.ifa_prefered = preferred;
3731         ci.ifa_valid = valid;
3732
3733         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3734 }
3735
3736 static inline int rt_scope(int ifa_scope)
3737 {
3738         if (ifa_scope & IFA_HOST)
3739                 return RT_SCOPE_HOST;
3740         else if (ifa_scope & IFA_LINK)
3741                 return RT_SCOPE_LINK;
3742         else if (ifa_scope & IFA_SITE)
3743                 return RT_SCOPE_SITE;
3744         else
3745                 return RT_SCOPE_UNIVERSE;
3746 }
3747
3748 static inline int inet6_ifaddr_msgsize(void)
3749 {
3750         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3751                + nla_total_size(16) /* IFA_ADDRESS */
3752                + nla_total_size(sizeof(struct ifa_cacheinfo));
3753 }
3754
3755 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3756                              u32 portid, u32 seq, int event, unsigned int flags)
3757 {
3758         struct nlmsghdr  *nlh;
3759         u32 preferred, valid;
3760
3761         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3762         if (nlh == NULL)
3763                 return -EMSGSIZE;
3764
3765         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3766                       ifa->idev->dev->ifindex);
3767
3768         if (!(ifa->flags&IFA_F_PERMANENT)) {
3769                 preferred = ifa->prefered_lft;
3770                 valid = ifa->valid_lft;
3771                 if (preferred != INFINITY_LIFE_TIME) {
3772                         long tval = (jiffies - ifa->tstamp)/HZ;
3773                         if (preferred > tval)
3774                                 preferred -= tval;
3775                         else
3776                                 preferred = 0;
3777                         if (valid != INFINITY_LIFE_TIME) {
3778                                 if (valid > tval)
3779                                         valid -= tval;
3780                                 else
3781                                         valid = 0;
3782                         }
3783                 }
3784         } else {
3785                 preferred = INFINITY_LIFE_TIME;
3786                 valid = INFINITY_LIFE_TIME;
3787         }
3788
3789         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3790             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3791                 nlmsg_cancel(skb, nlh);
3792                 return -EMSGSIZE;
3793         }
3794
3795         return nlmsg_end(skb, nlh);
3796 }
3797
3798 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3799                                 u32 portid, u32 seq, int event, u16 flags)
3800 {
3801         struct nlmsghdr  *nlh;
3802         u8 scope = RT_SCOPE_UNIVERSE;
3803         int ifindex = ifmca->idev->dev->ifindex;
3804
3805         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3806                 scope = RT_SCOPE_SITE;
3807
3808         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3809         if (nlh == NULL)
3810                 return -EMSGSIZE;
3811
3812         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3813         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3814             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3815                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3816                 nlmsg_cancel(skb, nlh);
3817                 return -EMSGSIZE;
3818         }
3819
3820         return nlmsg_end(skb, nlh);
3821 }
3822
3823 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3824                                 u32 portid, u32 seq, int event, unsigned int flags)
3825 {
3826         struct nlmsghdr  *nlh;
3827         u8 scope = RT_SCOPE_UNIVERSE;
3828         int ifindex = ifaca->aca_idev->dev->ifindex;
3829
3830         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3831                 scope = RT_SCOPE_SITE;
3832
3833         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3834         if (nlh == NULL)
3835                 return -EMSGSIZE;
3836
3837         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3838         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3839             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3840                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3841                 nlmsg_cancel(skb, nlh);
3842                 return -EMSGSIZE;
3843         }
3844
3845         return nlmsg_end(skb, nlh);
3846 }
3847
3848 enum addr_type_t {
3849         UNICAST_ADDR,
3850         MULTICAST_ADDR,
3851         ANYCAST_ADDR,
3852 };
3853
3854 /* called with rcu_read_lock() */
3855 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3856                           struct netlink_callback *cb, enum addr_type_t type,
3857                           int s_ip_idx, int *p_ip_idx)
3858 {
3859         struct ifmcaddr6 *ifmca;
3860         struct ifacaddr6 *ifaca;
3861         int err = 1;
3862         int ip_idx = *p_ip_idx;
3863
3864         read_lock_bh(&idev->lock);
3865         switch (type) {
3866         case UNICAST_ADDR: {
3867                 struct inet6_ifaddr *ifa;
3868
3869                 /* unicast address incl. temp addr */
3870                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3871                         if (++ip_idx < s_ip_idx)
3872                                 continue;
3873                         err = inet6_fill_ifaddr(skb, ifa,
3874                                                 NETLINK_CB(cb->skb).portid,
3875                                                 cb->nlh->nlmsg_seq,
3876                                                 RTM_NEWADDR,
3877                                                 NLM_F_MULTI);
3878                         if (err <= 0)
3879                                 break;
3880                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3881                 }
3882                 break;
3883         }
3884         case MULTICAST_ADDR:
3885                 /* multicast address */
3886                 for (ifmca = idev->mc_list; ifmca;
3887                      ifmca = ifmca->next, ip_idx++) {
3888                         if (ip_idx < s_ip_idx)
3889                                 continue;
3890                         err = inet6_fill_ifmcaddr(skb, ifmca,
3891                                                   NETLINK_CB(cb->skb).portid,
3892                                                   cb->nlh->nlmsg_seq,
3893                                                   RTM_GETMULTICAST,
3894                                                   NLM_F_MULTI);
3895                         if (err <= 0)
3896                                 break;
3897                 }
3898                 break;
3899         case ANYCAST_ADDR:
3900                 /* anycast address */
3901                 for (ifaca = idev->ac_list; ifaca;
3902                      ifaca = ifaca->aca_next, ip_idx++) {
3903                         if (ip_idx < s_ip_idx)
3904                                 continue;
3905                         err = inet6_fill_ifacaddr(skb, ifaca,
3906                                                   NETLINK_CB(cb->skb).portid,
3907                                                   cb->nlh->nlmsg_seq,
3908                                                   RTM_GETANYCAST,
3909                                                   NLM_F_MULTI);
3910                         if (err <= 0)
3911                                 break;
3912                 }
3913                 break;
3914         default:
3915                 break;
3916         }
3917         read_unlock_bh(&idev->lock);
3918         *p_ip_idx = ip_idx;
3919         return err;
3920 }
3921
3922 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3923                            enum addr_type_t type)
3924 {
3925         struct net *net = sock_net(skb->sk);
3926         int h, s_h;
3927         int idx, ip_idx;
3928         int s_idx, s_ip_idx;
3929         struct net_device *dev;
3930         struct inet6_dev *idev;
3931         struct hlist_head *head;
3932
3933         s_h = cb->args[0];
3934         s_idx = idx = cb->args[1];
3935         s_ip_idx = ip_idx = cb->args[2];
3936
3937         rcu_read_lock();
3938         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3939         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3940                 idx = 0;
3941                 head = &net->dev_index_head[h];
3942                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3943                         if (idx < s_idx)
3944                                 goto cont;
3945                         if (h > s_h || idx > s_idx)
3946                                 s_ip_idx = 0;
3947                         ip_idx = 0;
3948                         idev = __in6_dev_get(dev);
3949                         if (!idev)
3950                                 goto cont;
3951
3952                         if (in6_dump_addrs(idev, skb, cb, type,
3953                                            s_ip_idx, &ip_idx) <= 0)
3954                                 goto done;
3955 cont:
3956                         idx++;
3957                 }
3958         }
3959 done:
3960         rcu_read_unlock();
3961         cb->args[0] = h;
3962         cb->args[1] = idx;
3963         cb->args[2] = ip_idx;
3964
3965         return skb->len;
3966 }
3967
3968 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3969 {
3970         enum addr_type_t type = UNICAST_ADDR;
3971
3972         return inet6_dump_addr(skb, cb, type);
3973 }
3974
3975 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3976 {
3977         enum addr_type_t type = MULTICAST_ADDR;
3978
3979         return inet6_dump_addr(skb, cb, type);
3980 }
3981
3982
3983 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3984 {
3985         enum addr_type_t type = ANYCAST_ADDR;
3986
3987         return inet6_dump_addr(skb, cb, type);
3988 }
3989
3990 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3991 {
3992         struct net *net = sock_net(in_skb->sk);
3993         struct ifaddrmsg *ifm;
3994         struct nlattr *tb[IFA_MAX+1];
3995         struct in6_addr *addr = NULL;
3996         struct net_device *dev = NULL;
3997         struct inet6_ifaddr *ifa;
3998         struct sk_buff *skb;
3999         int err;
4000
4001         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4002         if (err < 0)
4003                 goto errout;
4004
4005         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4006         if (addr == NULL) {
4007                 err = -EINVAL;
4008                 goto errout;
4009         }
4010
4011         ifm = nlmsg_data(nlh);
4012         if (ifm->ifa_index)
4013                 dev = __dev_get_by_index(net, ifm->ifa_index);
4014
4015         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4016         if (!ifa) {
4017                 err = -EADDRNOTAVAIL;
4018                 goto errout;
4019         }
4020
4021         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4022         if (!skb) {
4023                 err = -ENOBUFS;
4024                 goto errout_ifa;
4025         }
4026
4027         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4028                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4029         if (err < 0) {
4030                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4031                 WARN_ON(err == -EMSGSIZE);
4032                 kfree_skb(skb);
4033                 goto errout_ifa;
4034         }
4035         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4036 errout_ifa:
4037         in6_ifa_put(ifa);
4038 errout:
4039         return err;
4040 }
4041
4042 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4043 {
4044         struct sk_buff *skb;
4045         struct net *net = dev_net(ifa->idev->dev);
4046         int err = -ENOBUFS;
4047
4048         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4049         if (skb == NULL)
4050                 goto errout;
4051
4052         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4053         if (err < 0) {
4054                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4055                 WARN_ON(err == -EMSGSIZE);
4056                 kfree_skb(skb);
4057                 goto errout;
4058         }
4059         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4060         return;
4061 errout:
4062         if (err < 0)
4063                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4064 }
4065
4066 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4067                                 __s32 *array, int bytes)
4068 {
4069         BUG_ON(bytes < (DEVCONF_MAX * 4));
4070
4071         memset(array, 0, bytes);
4072         array[DEVCONF_FORWARDING] = cnf->forwarding;
4073         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4074         array[DEVCONF_MTU6] = cnf->mtu6;
4075         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4076         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4077         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4078         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4079         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4080         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4081                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4082         array[DEVCONF_RTR_SOLICIT_DELAY] =
4083                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4084         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4085 #ifdef CONFIG_IPV6_PRIVACY
4086         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4087         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4088         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4089         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4090         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4091 #endif
4092         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4093         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4094         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4095 #ifdef CONFIG_IPV6_ROUTER_PREF
4096         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4097         array[DEVCONF_RTR_PROBE_INTERVAL] =
4098                 jiffies_to_msecs(cnf->rtr_probe_interval);
4099 #ifdef CONFIG_IPV6_ROUTE_INFO
4100         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4101 #endif
4102 #endif
4103         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4104         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4105 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4106         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4107 #endif
4108 #ifdef CONFIG_IPV6_MROUTE
4109         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4110 #endif
4111         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4112         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4113         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4114         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4115 }
4116
4117 static inline size_t inet6_ifla6_size(void)
4118 {
4119         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4120              + nla_total_size(sizeof(struct ifla_cacheinfo))
4121              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4122              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4123              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4124 }
4125
4126 static inline size_t inet6_if_nlmsg_size(void)
4127 {
4128         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4129                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4130                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4131                + nla_total_size(4) /* IFLA_MTU */
4132                + nla_total_size(4) /* IFLA_LINK */
4133                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4134 }
4135
4136 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4137                                       int items, int bytes)
4138 {
4139         int i;
4140         int pad = bytes - sizeof(u64) * items;
4141         BUG_ON(pad < 0);
4142
4143         /* Use put_unaligned() because stats may not be aligned for u64. */
4144         put_unaligned(items, &stats[0]);
4145         for (i = 1; i < items; i++)
4146                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4147
4148         memset(&stats[items], 0, pad);
4149 }
4150
4151 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4152                                       int items, int bytes, size_t syncpoff)
4153 {
4154         int i;
4155         int pad = bytes - sizeof(u64) * items;
4156         BUG_ON(pad < 0);
4157
4158         /* Use put_unaligned() because stats may not be aligned for u64. */
4159         put_unaligned(items, &stats[0]);
4160         for (i = 1; i < items; i++)
4161                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4162
4163         memset(&stats[items], 0, pad);
4164 }
4165
4166 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4167                              int bytes)
4168 {
4169         switch (attrtype) {
4170         case IFLA_INET6_STATS:
4171                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4172                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4173                 break;
4174         case IFLA_INET6_ICMP6STATS:
4175                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4176                 break;
4177         }
4178 }
4179
4180 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4181 {
4182         struct nlattr *nla;
4183         struct ifla_cacheinfo ci;
4184
4185         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4186                 goto nla_put_failure;
4187         ci.max_reasm_len = IPV6_MAXPLEN;
4188         ci.tstamp = cstamp_delta(idev->tstamp);
4189         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4190         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4191         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4192                 goto nla_put_failure;
4193         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4194         if (nla == NULL)
4195                 goto nla_put_failure;
4196         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4197
4198         /* XXX - MC not implemented */
4199
4200         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4201         if (nla == NULL)
4202                 goto nla_put_failure;
4203         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4204
4205         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4206         if (nla == NULL)
4207                 goto nla_put_failure;
4208         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4209
4210         return 0;
4211
4212 nla_put_failure:
4213         return -EMSGSIZE;
4214 }
4215
4216 static size_t inet6_get_link_af_size(const struct net_device *dev)
4217 {
4218         if (!__in6_dev_get(dev))
4219                 return 0;
4220
4221         return inet6_ifla6_size();
4222 }
4223
4224 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4225 {
4226         struct inet6_dev *idev = __in6_dev_get(dev);
4227
4228         if (!idev)
4229                 return -ENODATA;
4230
4231         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4232                 return -EMSGSIZE;
4233
4234         return 0;
4235 }
4236
4237 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4238                              u32 portid, u32 seq, int event, unsigned int flags)
4239 {
4240         struct net_device *dev = idev->dev;
4241         struct ifinfomsg *hdr;
4242         struct nlmsghdr *nlh;
4243         void *protoinfo;
4244
4245         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4246         if (nlh == NULL)
4247                 return -EMSGSIZE;
4248
4249         hdr = nlmsg_data(nlh);
4250         hdr->ifi_family = AF_INET6;
4251         hdr->__ifi_pad = 0;
4252         hdr->ifi_type = dev->type;
4253         hdr->ifi_index = dev->ifindex;
4254         hdr->ifi_flags = dev_get_flags(dev);
4255         hdr->ifi_change = 0;
4256
4257         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4258             (dev->addr_len &&
4259              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4260             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4261             (dev->ifindex != dev->iflink &&
4262              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4263                 goto nla_put_failure;
4264         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4265         if (protoinfo == NULL)
4266                 goto nla_put_failure;
4267
4268         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4269                 goto nla_put_failure;
4270
4271         nla_nest_end(skb, protoinfo);
4272         return nlmsg_end(skb, nlh);
4273
4274 nla_put_failure:
4275         nlmsg_cancel(skb, nlh);
4276         return -EMSGSIZE;
4277 }
4278
4279 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4280 {
4281         struct net *net = sock_net(skb->sk);
4282         int h, s_h;
4283         int idx = 0, s_idx;
4284         struct net_device *dev;
4285         struct inet6_dev *idev;
4286         struct hlist_head *head;
4287
4288         s_h = cb->args[0];
4289         s_idx = cb->args[1];
4290
4291         rcu_read_lock();
4292         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4293                 idx = 0;
4294                 head = &net->dev_index_head[h];
4295                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4296                         if (idx < s_idx)
4297                                 goto cont;
4298                         idev = __in6_dev_get(dev);
4299                         if (!idev)
4300                                 goto cont;
4301                         if (inet6_fill_ifinfo(skb, idev,
4302                                               NETLINK_CB(cb->skb).portid,
4303                                               cb->nlh->nlmsg_seq,
4304                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4305                                 goto out;
4306 cont:
4307                         idx++;
4308                 }
4309         }
4310 out:
4311         rcu_read_unlock();
4312         cb->args[1] = idx;
4313         cb->args[0] = h;
4314
4315         return skb->len;
4316 }
4317
4318 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4319 {
4320         struct sk_buff *skb;
4321         struct net *net = dev_net(idev->dev);
4322         int err = -ENOBUFS;
4323
4324         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4325         if (skb == NULL)
4326                 goto errout;
4327
4328         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4329         if (err < 0) {
4330                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4331                 WARN_ON(err == -EMSGSIZE);
4332                 kfree_skb(skb);
4333                 goto errout;
4334         }
4335         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4336         return;
4337 errout:
4338         if (err < 0)
4339                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4340 }
4341
4342 static inline size_t inet6_prefix_nlmsg_size(void)
4343 {
4344         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4345                + nla_total_size(sizeof(struct in6_addr))
4346                + nla_total_size(sizeof(struct prefix_cacheinfo));
4347 }
4348
4349 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4350                              struct prefix_info *pinfo, u32 portid, u32 seq,
4351                              int event, unsigned int flags)
4352 {
4353         struct prefixmsg *pmsg;
4354         struct nlmsghdr *nlh;
4355         struct prefix_cacheinfo ci;
4356
4357         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4358         if (nlh == NULL)
4359                 return -EMSGSIZE;
4360
4361         pmsg = nlmsg_data(nlh);
4362         pmsg->prefix_family = AF_INET6;
4363         pmsg->prefix_pad1 = 0;
4364         pmsg->prefix_pad2 = 0;
4365         pmsg->prefix_ifindex = idev->dev->ifindex;
4366         pmsg->prefix_len = pinfo->prefix_len;
4367         pmsg->prefix_type = pinfo->type;
4368         pmsg->prefix_pad3 = 0;
4369         pmsg->prefix_flags = 0;
4370         if (pinfo->onlink)
4371                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4372         if (pinfo->autoconf)
4373                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4374
4375         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4376                 goto nla_put_failure;
4377         ci.preferred_time = ntohl(pinfo->prefered);
4378         ci.valid_time = ntohl(pinfo->valid);
4379         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4380                 goto nla_put_failure;
4381         return nlmsg_end(skb, nlh);
4382
4383 nla_put_failure:
4384         nlmsg_cancel(skb, nlh);
4385         return -EMSGSIZE;
4386 }
4387
4388 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4389                          struct prefix_info *pinfo)
4390 {
4391         struct sk_buff *skb;
4392         struct net *net = dev_net(idev->dev);
4393         int err = -ENOBUFS;
4394
4395         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4396         if (skb == NULL)
4397                 goto errout;
4398
4399         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4400         if (err < 0) {
4401                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4402                 WARN_ON(err == -EMSGSIZE);
4403                 kfree_skb(skb);
4404                 goto errout;
4405         }
4406         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4407         return;
4408 errout:
4409         if (err < 0)
4410                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4411 }
4412
4413 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4414 {
4415         struct net *net = dev_net(ifp->idev->dev);
4416
4417         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4418
4419         switch (event) {
4420         case RTM_NEWADDR:
4421                 /*
4422                  * If the address was optimistic
4423                  * we inserted the route at the start of
4424                  * our DAD process, so we don't need
4425                  * to do it again
4426                  */
4427                 if (!(ifp->rt->rt6i_node))
4428                         ip6_ins_rt(ifp->rt);
4429                 if (ifp->idev->cnf.forwarding)
4430                         addrconf_join_anycast(ifp);
4431                 break;
4432         case RTM_DELADDR:
4433                 if (ifp->idev->cnf.forwarding)
4434                         addrconf_leave_anycast(ifp);
4435                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4436                 dst_hold(&ifp->rt->dst);
4437
4438                 if (ip6_del_rt(ifp->rt))
4439                         dst_free(&ifp->rt->dst);
4440                 break;
4441         }
4442         atomic_inc(&net->ipv6.dev_addr_genid);
4443 }
4444
4445 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4446 {
4447         rcu_read_lock_bh();
4448         if (likely(ifp->idev->dead == 0))
4449                 __ipv6_ifa_notify(event, ifp);
4450         rcu_read_unlock_bh();
4451 }
4452
4453 #ifdef CONFIG_SYSCTL
4454
4455 static
4456 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4457                            void __user *buffer, size_t *lenp, loff_t *ppos)
4458 {
4459         int *valp = ctl->data;
4460         int val = *valp;
4461         loff_t pos = *ppos;
4462         ctl_table lctl;
4463         int ret;
4464
4465         /*
4466          * ctl->data points to idev->cnf.forwarding, we should
4467          * not modify it until we get the rtnl lock.
4468          */
4469         lctl = *ctl;
4470         lctl.data = &val;
4471
4472         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4473
4474         if (write)
4475                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4476         if (ret)
4477                 *ppos = pos;
4478         return ret;
4479 }
4480
4481 static void dev_disable_change(struct inet6_dev *idev)
4482 {
4483         if (!idev || !idev->dev)
4484                 return;
4485
4486         if (idev->cnf.disable_ipv6)
4487                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4488         else
4489                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4490 }
4491
4492 static void addrconf_disable_change(struct net *net, __s32 newf)
4493 {
4494         struct net_device *dev;
4495         struct inet6_dev *idev;
4496
4497         rcu_read_lock();
4498         for_each_netdev_rcu(net, dev) {
4499                 idev = __in6_dev_get(dev);
4500                 if (idev) {
4501                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4502                         idev->cnf.disable_ipv6 = newf;
4503                         if (changed)
4504                                 dev_disable_change(idev);
4505                 }
4506         }
4507         rcu_read_unlock();
4508 }
4509
4510 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4511 {
4512         struct net *net;
4513         int old;
4514
4515         if (!rtnl_trylock())
4516                 return restart_syscall();
4517
4518         net = (struct net *)table->extra2;
4519         old = *p;
4520         *p = newf;
4521
4522         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4523                 rtnl_unlock();
4524                 return 0;
4525         }
4526
4527         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4528                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4529                 addrconf_disable_change(net, newf);
4530         } else if ((!newf) ^ (!old))
4531                 dev_disable_change((struct inet6_dev *)table->extra1);
4532
4533         rtnl_unlock();
4534         return 0;
4535 }
4536
4537 static
4538 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4539                             void __user *buffer, size_t *lenp, loff_t *ppos)
4540 {
4541         int *valp = ctl->data;
4542         int val = *valp;
4543         loff_t pos = *ppos;
4544         ctl_table lctl;
4545         int ret;
4546
4547         /*
4548          * ctl->data points to idev->cnf.disable_ipv6, we should
4549          * not modify it until we get the rtnl lock.
4550          */
4551         lctl = *ctl;
4552         lctl.data = &val;
4553
4554         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4555
4556         if (write)
4557                 ret = addrconf_disable_ipv6(ctl, valp, val);
4558         if (ret)
4559                 *ppos = pos;
4560         return ret;
4561 }
4562
4563 static struct addrconf_sysctl_table
4564 {
4565         struct ctl_table_header *sysctl_header;
4566         ctl_table addrconf_vars[DEVCONF_MAX+1];
4567 } addrconf_sysctl __read_mostly = {
4568         .sysctl_header = NULL,
4569         .addrconf_vars = {
4570                 {
4571                         .procname       = "forwarding",
4572                         .data           = &ipv6_devconf.forwarding,
4573                         .maxlen         = sizeof(int),
4574                         .mode           = 0644,
4575                         .proc_handler   = addrconf_sysctl_forward,
4576                 },
4577                 {
4578                         .procname       = "hop_limit",
4579                         .data           = &ipv6_devconf.hop_limit,
4580                         .maxlen         = sizeof(int),
4581                         .mode           = 0644,
4582                         .proc_handler   = proc_dointvec,
4583                 },
4584                 {
4585                         .procname       = "mtu",
4586                         .data           = &ipv6_devconf.mtu6,
4587                         .maxlen         = sizeof(int),
4588                         .mode           = 0644,
4589                         .proc_handler   = proc_dointvec,
4590                 },
4591                 {
4592                         .procname       = "accept_ra",
4593                         .data           = &ipv6_devconf.accept_ra,
4594                         .maxlen         = sizeof(int),
4595                         .mode           = 0644,
4596                         .proc_handler   = proc_dointvec,
4597                 },
4598                 {
4599                         .procname       = "accept_redirects",
4600                         .data           = &ipv6_devconf.accept_redirects,
4601                         .maxlen         = sizeof(int),
4602                         .mode           = 0644,
4603                         .proc_handler   = proc_dointvec,
4604                 },
4605                 {
4606                         .procname       = "autoconf",
4607                         .data           = &ipv6_devconf.autoconf,
4608                         .maxlen         = sizeof(int),
4609                         .mode           = 0644,
4610                         .proc_handler   = proc_dointvec,
4611                 },
4612                 {
4613                         .procname       = "dad_transmits",
4614                         .data           = &ipv6_devconf.dad_transmits,
4615                         .maxlen         = sizeof(int),
4616                         .mode           = 0644,
4617                         .proc_handler   = proc_dointvec,
4618                 },
4619                 {
4620                         .procname       = "router_solicitations",
4621                         .data           = &ipv6_devconf.rtr_solicits,
4622                         .maxlen         = sizeof(int),
4623                         .mode           = 0644,
4624                         .proc_handler   = proc_dointvec,
4625                 },
4626                 {
4627                         .procname       = "router_solicitation_interval",
4628                         .data           = &ipv6_devconf.rtr_solicit_interval,
4629                         .maxlen         = sizeof(int),
4630                         .mode           = 0644,
4631                         .proc_handler   = proc_dointvec_jiffies,
4632                 },
4633                 {
4634                         .procname       = "router_solicitation_delay",
4635                         .data           = &ipv6_devconf.rtr_solicit_delay,
4636                         .maxlen         = sizeof(int),
4637                         .mode           = 0644,
4638                         .proc_handler   = proc_dointvec_jiffies,
4639                 },
4640                 {
4641                         .procname       = "force_mld_version",
4642                         .data           = &ipv6_devconf.force_mld_version,
4643                         .maxlen         = sizeof(int),
4644                         .mode           = 0644,
4645                         .proc_handler   = proc_dointvec,
4646                 },
4647 #ifdef CONFIG_IPV6_PRIVACY
4648                 {
4649                         .procname       = "use_tempaddr",
4650                         .data           = &ipv6_devconf.use_tempaddr,
4651                         .maxlen         = sizeof(int),
4652                         .mode           = 0644,
4653                         .proc_handler   = proc_dointvec,
4654                 },
4655                 {
4656                         .procname       = "temp_valid_lft",
4657                         .data           = &ipv6_devconf.temp_valid_lft,
4658                         .maxlen         = sizeof(int),
4659                         .mode           = 0644,
4660                         .proc_handler   = proc_dointvec,
4661                 },
4662                 {
4663                         .procname       = "temp_prefered_lft",
4664                         .data           = &ipv6_devconf.temp_prefered_lft,
4665                         .maxlen         = sizeof(int),
4666                         .mode           = 0644,
4667                         .proc_handler   = proc_dointvec,
4668                 },
4669                 {
4670                         .procname       = "regen_max_retry",
4671                         .data           = &ipv6_devconf.regen_max_retry,
4672                         .maxlen         = sizeof(int),
4673                         .mode           = 0644,
4674                         .proc_handler   = proc_dointvec,
4675                 },
4676                 {
4677                         .procname       = "max_desync_factor",
4678                         .data           = &ipv6_devconf.max_desync_factor,
4679                         .maxlen         = sizeof(int),
4680                         .mode           = 0644,
4681                         .proc_handler   = proc_dointvec,
4682                 },
4683 #endif
4684                 {
4685                         .procname       = "max_addresses",
4686                         .data           = &ipv6_devconf.max_addresses,
4687                         .maxlen         = sizeof(int),
4688                         .mode           = 0644,
4689                         .proc_handler   = proc_dointvec,
4690                 },
4691                 {
4692                         .procname       = "accept_ra_defrtr",
4693                         .data           = &ipv6_devconf.accept_ra_defrtr,
4694                         .maxlen         = sizeof(int),
4695                         .mode           = 0644,
4696                         .proc_handler   = proc_dointvec,
4697                 },
4698                 {
4699                         .procname       = "accept_ra_pinfo",
4700                         .data           = &ipv6_devconf.accept_ra_pinfo,
4701                         .maxlen         = sizeof(int),
4702                         .mode           = 0644,
4703                         .proc_handler   = proc_dointvec,
4704                 },
4705 #ifdef CONFIG_IPV6_ROUTER_PREF
4706                 {
4707                         .procname       = "accept_ra_rtr_pref",
4708                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4709                         .maxlen         = sizeof(int),
4710                         .mode           = 0644,
4711                         .proc_handler   = proc_dointvec,
4712                 },
4713                 {
4714                         .procname       = "router_probe_interval",
4715                         .data           = &ipv6_devconf.rtr_probe_interval,
4716                         .maxlen         = sizeof(int),
4717                         .mode           = 0644,
4718                         .proc_handler   = proc_dointvec_jiffies,
4719                 },
4720 #ifdef CONFIG_IPV6_ROUTE_INFO
4721                 {
4722                         .procname       = "accept_ra_rt_info_max_plen",
4723                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4724                         .maxlen         = sizeof(int),
4725                         .mode           = 0644,
4726                         .proc_handler   = proc_dointvec,
4727                 },
4728 #endif
4729 #endif
4730                 {
4731                         .procname       = "proxy_ndp",
4732                         .data           = &ipv6_devconf.proxy_ndp,
4733                         .maxlen         = sizeof(int),
4734                         .mode           = 0644,
4735                         .proc_handler   = proc_dointvec,
4736                 },
4737                 {
4738                         .procname       = "accept_source_route",
4739                         .data           = &ipv6_devconf.accept_source_route,
4740                         .maxlen         = sizeof(int),
4741                         .mode           = 0644,
4742                         .proc_handler   = proc_dointvec,
4743                 },
4744 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4745                 {
4746                         .procname       = "optimistic_dad",
4747                         .data           = &ipv6_devconf.optimistic_dad,
4748                         .maxlen         = sizeof(int),
4749                         .mode           = 0644,
4750                         .proc_handler   = proc_dointvec,
4751
4752                 },
4753 #endif
4754 #ifdef CONFIG_IPV6_MROUTE
4755                 {
4756                         .procname       = "mc_forwarding",
4757                         .data           = &ipv6_devconf.mc_forwarding,
4758                         .maxlen         = sizeof(int),
4759                         .mode           = 0444,
4760                         .proc_handler   = proc_dointvec,
4761                 },
4762 #endif
4763                 {
4764                         .procname       = "disable_ipv6",
4765                         .data           = &ipv6_devconf.disable_ipv6,
4766                         .maxlen         = sizeof(int),
4767                         .mode           = 0644,
4768                         .proc_handler   = addrconf_sysctl_disable,
4769                 },
4770                 {
4771                         .procname       = "accept_dad",
4772                         .data           = &ipv6_devconf.accept_dad,
4773                         .maxlen         = sizeof(int),
4774                         .mode           = 0644,
4775                         .proc_handler   = proc_dointvec,
4776                 },
4777                 {
4778                         .procname       = "force_tllao",
4779                         .data           = &ipv6_devconf.force_tllao,
4780                         .maxlen         = sizeof(int),
4781                         .mode           = 0644,
4782                         .proc_handler   = proc_dointvec
4783                 },
4784                 {
4785                         .procname       = "ndisc_notify",
4786                         .data           = &ipv6_devconf.ndisc_notify,
4787                         .maxlen         = sizeof(int),
4788                         .mode           = 0644,
4789                         .proc_handler   = proc_dointvec
4790                 },
4791                 {
4792                         /* sentinel */
4793                 }
4794         },
4795 };
4796
4797 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4798                 struct inet6_dev *idev, struct ipv6_devconf *p)
4799 {
4800         int i;
4801         struct addrconf_sysctl_table *t;
4802         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4803
4804         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4805         if (t == NULL)
4806                 goto out;
4807
4808         for (i = 0; t->addrconf_vars[i].data; i++) {
4809                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4810                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4811                 t->addrconf_vars[i].extra2 = net;
4812         }
4813
4814         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4815
4816         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4817         if (t->sysctl_header == NULL)
4818                 goto free;
4819
4820         p->sysctl = t;
4821         return 0;
4822
4823 free:
4824         kfree(t);
4825 out:
4826         return -ENOBUFS;
4827 }
4828
4829 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4830 {
4831         struct addrconf_sysctl_table *t;
4832
4833         if (p->sysctl == NULL)
4834                 return;
4835
4836         t = p->sysctl;
4837         p->sysctl = NULL;
4838         unregister_net_sysctl_table(t->sysctl_header);
4839         kfree(t);
4840 }
4841
4842 static void addrconf_sysctl_register(struct inet6_dev *idev)
4843 {
4844         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4845                               &ndisc_ifinfo_sysctl_change);
4846         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4847                                         idev, &idev->cnf);
4848 }
4849
4850 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4851 {
4852         __addrconf_sysctl_unregister(&idev->cnf);
4853         neigh_sysctl_unregister(idev->nd_parms);
4854 }
4855
4856
4857 #endif
4858
4859 static int __net_init addrconf_init_net(struct net *net)
4860 {
4861         int err;
4862         struct ipv6_devconf *all, *dflt;
4863
4864         err = -ENOMEM;
4865         all = &ipv6_devconf;
4866         dflt = &ipv6_devconf_dflt;
4867
4868         if (!net_eq(net, &init_net)) {
4869                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4870                 if (all == NULL)
4871                         goto err_alloc_all;
4872
4873                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4874                 if (dflt == NULL)
4875                         goto err_alloc_dflt;
4876         } else {
4877                 /* these will be inherited by all namespaces */
4878                 dflt->autoconf = ipv6_defaults.autoconf;
4879                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4880         }
4881
4882         net->ipv6.devconf_all = all;
4883         net->ipv6.devconf_dflt = dflt;
4884
4885 #ifdef CONFIG_SYSCTL
4886         err = __addrconf_sysctl_register(net, "all", NULL, all);
4887         if (err < 0)
4888                 goto err_reg_all;
4889
4890         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4891         if (err < 0)
4892                 goto err_reg_dflt;
4893 #endif
4894         return 0;
4895
4896 #ifdef CONFIG_SYSCTL
4897 err_reg_dflt:
4898         __addrconf_sysctl_unregister(all);
4899 err_reg_all:
4900         kfree(dflt);
4901 #endif
4902 err_alloc_dflt:
4903         kfree(all);
4904 err_alloc_all:
4905         return err;
4906 }
4907
4908 static void __net_exit addrconf_exit_net(struct net *net)
4909 {
4910 #ifdef CONFIG_SYSCTL
4911         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4912         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4913 #endif
4914         if (!net_eq(net, &init_net)) {
4915                 kfree(net->ipv6.devconf_dflt);
4916                 kfree(net->ipv6.devconf_all);
4917         }
4918 }
4919
4920 static struct pernet_operations addrconf_ops = {
4921         .init = addrconf_init_net,
4922         .exit = addrconf_exit_net,
4923 };
4924
4925 /*
4926  *      Device notifier
4927  */
4928
4929 int register_inet6addr_notifier(struct notifier_block *nb)
4930 {
4931         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4932 }
4933 EXPORT_SYMBOL(register_inet6addr_notifier);
4934
4935 int unregister_inet6addr_notifier(struct notifier_block *nb)
4936 {
4937         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4938 }
4939 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4940
4941 static struct rtnl_af_ops inet6_ops = {
4942         .family           = AF_INET6,
4943         .fill_link_af     = inet6_fill_link_af,
4944         .get_link_af_size = inet6_get_link_af_size,
4945 };
4946
4947 /*
4948  *      Init / cleanup code
4949  */
4950
4951 int __init addrconf_init(void)
4952 {
4953         int i, err;
4954
4955         err = ipv6_addr_label_init();
4956         if (err < 0) {
4957                 pr_crit("%s: cannot initialize default policy table: %d\n",
4958                         __func__, err);
4959                 goto out;
4960         }
4961
4962         err = register_pernet_subsys(&addrconf_ops);
4963         if (err < 0)
4964                 goto out_addrlabel;
4965
4966         /* The addrconf netdev notifier requires that loopback_dev
4967          * has it's ipv6 private information allocated and setup
4968          * before it can bring up and give link-local addresses
4969          * to other devices which are up.
4970          *
4971          * Unfortunately, loopback_dev is not necessarily the first
4972          * entry in the global dev_base list of net devices.  In fact,
4973          * it is likely to be the very last entry on that list.
4974          * So this causes the notifier registry below to try and
4975          * give link-local addresses to all devices besides loopback_dev
4976          * first, then loopback_dev, which cases all the non-loopback_dev
4977          * devices to fail to get a link-local address.
4978          *
4979          * So, as a temporary fix, allocate the ipv6 structure for
4980          * loopback_dev first by hand.
4981          * Longer term, all of the dependencies ipv6 has upon the loopback
4982          * device and it being up should be removed.
4983          */
4984         rtnl_lock();
4985         if (!ipv6_add_dev(init_net.loopback_dev))
4986                 err = -ENOMEM;
4987         rtnl_unlock();
4988         if (err)
4989                 goto errlo;
4990
4991         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4992                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4993
4994         register_netdevice_notifier(&ipv6_dev_notf);
4995
4996         addrconf_verify(0);
4997
4998         err = rtnl_af_register(&inet6_ops);
4999         if (err < 0)
5000                 goto errout_af;
5001
5002         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5003                               NULL);
5004         if (err < 0)
5005                 goto errout;
5006
5007         /* Only the first call to __rtnl_register can fail */
5008         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5009         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5010         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5011                         inet6_dump_ifaddr, NULL);
5012         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5013                         inet6_dump_ifmcaddr, NULL);
5014         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5015                         inet6_dump_ifacaddr, NULL);
5016         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5017                         inet6_netconf_dump_devconf, NULL);
5018
5019         ipv6_addr_label_rtnl_register();
5020
5021         return 0;
5022 errout:
5023         rtnl_af_unregister(&inet6_ops);
5024 errout_af:
5025         unregister_netdevice_notifier(&ipv6_dev_notf);
5026 errlo:
5027         unregister_pernet_subsys(&addrconf_ops);
5028 out_addrlabel:
5029         ipv6_addr_label_cleanup();
5030 out:
5031         return err;
5032 }
5033
5034 void addrconf_cleanup(void)
5035 {
5036         struct net_device *dev;
5037         int i;
5038
5039         unregister_netdevice_notifier(&ipv6_dev_notf);
5040         unregister_pernet_subsys(&addrconf_ops);
5041         ipv6_addr_label_cleanup();
5042
5043         rtnl_lock();
5044
5045         __rtnl_af_unregister(&inet6_ops);
5046
5047         /* clean dev list */
5048         for_each_netdev(&init_net, dev) {
5049                 if (__in6_dev_get(dev) == NULL)
5050                         continue;
5051                 addrconf_ifdown(dev, 1);
5052         }
5053         addrconf_ifdown(init_net.loopback_dev, 2);
5054
5055         /*
5056          *      Check hash table.
5057          */
5058         spin_lock_bh(&addrconf_hash_lock);
5059         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5060                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5061         spin_unlock_bh(&addrconf_hash_lock);
5062
5063         del_timer(&addr_chk_timer);
5064         rtnl_unlock();
5065 }