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/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86
  87 #ifdef CONFIG_IPV6_PRIVACY
  88 #include <linux/random.h>
  89 #endif
  90
  91 #include <linux/uaccess.h>
  92 #include <asm/unaligned.h>
  93
  94 #include <linux/proc_fs.h>
  95 #include <linux/seq_file.h>
  96 #include <linux/export.h>
  97
  98 /* Set to 3 to get tracing... */
  99 #define ACONF_DEBUG 2
 100
 101 #if ACONF_DEBUG >= 3
 102 #define ADBG(x) printk x
 103 #else
 104 #define ADBG(x)
 105 #endif
 106
 107 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 108
 109 static inline u32 cstamp_delta(unsigned long cstamp)
 110 {
 111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112 }
 113
 114 #ifdef CONFIG_SYSCTL
 115 static void addrconf_sysctl_register(struct inet6_dev *idev);
 116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117 #else
 118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 119 {
 120 }
 121
 122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 123 {
 124 }
 125 #endif
 126
 127 #ifdef CONFIG_IPV6_PRIVACY
 128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130 static void ipv6_regen_rndid(unsigned long data);
 131 #endif
 132
 133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134 static int ipv6_count_addresses(struct inet6_dev *idev);
 135
 136 /*
 137  *      Configured unicast address hash table
 138  */
 139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 140 static DEFINE_SPINLOCK(addrconf_hash_lock);
 141
 142 static void addrconf_verify(unsigned long);
 143
 144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 145 static DEFINE_SPINLOCK(addrconf_verify_lock);
 146
 147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 149
 150 static void addrconf_type_change(struct net_device *dev,
 151                                  unsigned long event);
 152 static int addrconf_ifdown(struct net_device *dev, int how);
 153
 154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 155                                                   int plen,
 156                                                   const struct net_device *dev,
 157                                                   u32 flags, u32 noflags);
 158
 159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 160 static void addrconf_dad_timer(unsigned long data);
 161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 162 static void addrconf_dad_run(struct inet6_dev *idev);
 163 static void addrconf_rs_timer(unsigned long data);
 164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 166
 167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 168                                 struct prefix_info *pinfo);
 169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 170                                struct net_device *dev);
 171
 172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 173         .forwarding             = 0,
 174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 175         .mtu6                   = IPV6_MIN_MTU,
 176         .accept_ra              = 1,
 177         .accept_redirects       = 1,
 178         .autoconf               = 1,
 179         .force_mld_version      = 0,
 180         .dad_transmits          = 1,
 181         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 182         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 183         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 184 #ifdef CONFIG_IPV6_PRIVACY
 185         .use_tempaddr           = 0,
 186         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 187         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 188         .regen_max_retry        = REGEN_MAX_RETRY,
 189         .max_desync_factor      = MAX_DESYNC_FACTOR,
 190 #endif
 191         .max_addresses          = IPV6_MAX_ADDRESSES,
 192         .accept_ra_defrtr       = 1,
 193         .accept_ra_pinfo        = 1,
 194 #ifdef CONFIG_IPV6_ROUTER_PREF
 195         .accept_ra_rtr_pref     = 1,
 196         .rtr_probe_interval     = 60 * HZ,
 197 #ifdef CONFIG_IPV6_ROUTE_INFO
 198         .accept_ra_rt_info_max_plen = 0,
 199 #endif
 200 #endif
 201         .proxy_ndp              = 0,
 202         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 203         .disable_ipv6           = 0,
 204         .accept_dad             = 1,
 205 };
 206
 207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 208         .forwarding             = 0,
 209         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 210         .mtu6                   = IPV6_MIN_MTU,
 211         .accept_ra              = 1,
 212         .accept_redirects       = 1,
 213         .autoconf               = 1,
 214         .dad_transmits          = 1,
 215         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 216         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 217         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 218 #ifdef CONFIG_IPV6_PRIVACY
 219         .use_tempaddr           = 0,
 220         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 221         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 222         .regen_max_retry        = REGEN_MAX_RETRY,
 223         .max_desync_factor      = MAX_DESYNC_FACTOR,
 224 #endif
 225         .max_addresses          = IPV6_MAX_ADDRESSES,
 226         .accept_ra_defrtr       = 1,
 227         .accept_ra_pinfo        = 1,
 228 #ifdef CONFIG_IPV6_ROUTER_PREF
 229         .accept_ra_rtr_pref     = 1,
 230         .rtr_probe_interval     = 60 * HZ,
 231 #ifdef CONFIG_IPV6_ROUTE_INFO
 232         .accept_ra_rt_info_max_plen = 0,
 233 #endif
 234 #endif
 235         .proxy_ndp              = 0,
 236         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 237         .disable_ipv6           = 0,
 238         .accept_dad             = 1,
 239 };
 240
 241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 246 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 247 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 248 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 249
 250 /* Check if a valid qdisc is available */
 251 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 252 {
 253         return !qdisc_tx_is_noop(dev);
 254 }
 255
 256 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 257 {
 258         if (del_timer(&ifp->timer))
 259                 __in6_ifa_put(ifp);
 260 }
 261
 262 enum addrconf_timer_t {
 263         AC_NONE,
 264         AC_DAD,
 265         AC_RS,
 266 };
 267
 268 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 269                                enum addrconf_timer_t what,
 270                                unsigned long when)
 271 {
 272         if (!del_timer(&ifp->timer))
 273                 in6_ifa_hold(ifp);
 274
 275         switch (what) {
 276         case AC_DAD:
 277                 ifp->timer.function = addrconf_dad_timer;
 278                 break;
 279         case AC_RS:
 280                 ifp->timer.function = addrconf_rs_timer;
 281                 break;
 282         default:
 283                 break;
 284         }
 285         ifp->timer.expires = jiffies + when;
 286         add_timer(&ifp->timer);
 287 }
 288
 289 static int snmp6_alloc_dev(struct inet6_dev *idev)
 290 {
 291         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 292                           sizeof(struct ipstats_mib),
 293                           __alignof__(struct ipstats_mib)) < 0)
 294                 goto err_ip;
 295         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 296                                         GFP_KERNEL);
 297         if (!idev->stats.icmpv6dev)
 298                 goto err_icmp;
 299         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 300                                            GFP_KERNEL);
 301         if (!idev->stats.icmpv6msgdev)
 302                 goto err_icmpmsg;
 303
 304         return 0;
 305
 306 err_icmpmsg:
 307         kfree(idev->stats.icmpv6dev);
 308 err_icmp:
 309         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 310 err_ip:
 311         return -ENOMEM;
 312 }
 313
 314 static void snmp6_free_dev(struct inet6_dev *idev)
 315 {
 316         kfree(idev->stats.icmpv6msgdev);
 317         kfree(idev->stats.icmpv6dev);
 318         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 319 }
 320
 321 /* Nobody refers to this device, we may destroy it. */
 322
 323 void in6_dev_finish_destroy(struct inet6_dev *idev)
 324 {
 325         struct net_device *dev = idev->dev;
 326
 327         WARN_ON(!list_empty(&idev->addr_list));
 328         WARN_ON(idev->mc_list != NULL);
 329
 330 #ifdef NET_REFCNT_DEBUG
 331         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 332 #endif
 333         dev_put(dev);
 334         if (!idev->dead) {
 335                 pr_warn("Freeing alive inet6 device %p\n", idev);
 336                 return;
 337         }
 338         snmp6_free_dev(idev);
 339         kfree_rcu(idev, rcu);
 340 }
 341 EXPORT_SYMBOL(in6_dev_finish_destroy);
 342
 343 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 344 {
 345         struct inet6_dev *ndev;
 346
 347         ASSERT_RTNL();
 348
 349         if (dev->mtu < IPV6_MIN_MTU)
 350                 return NULL;
 351
 352         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 353
 354         if (ndev == NULL)
 355                 return NULL;
 356
 357         rwlock_init(&ndev->lock);
 358         ndev->dev = dev;
 359         INIT_LIST_HEAD(&ndev->addr_list);
 360
 361         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 362         ndev->cnf.mtu6 = dev->mtu;
 363         ndev->cnf.sysctl = NULL;
 364         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 365         if (ndev->nd_parms == NULL) {
 366                 kfree(ndev);
 367                 return NULL;
 368         }
 369         if (ndev->cnf.forwarding)
 370                 dev_disable_lro(dev);
 371         /* We refer to the device */
 372         dev_hold(dev);
 373
 374         if (snmp6_alloc_dev(ndev) < 0) {
 375                 ADBG((KERN_WARNING
 376                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 377                         __func__, dev->name));
 378                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 379                 dev_put(dev);
 380                 kfree(ndev);
 381                 return NULL;
 382         }
 383
 384         if (snmp6_register_dev(ndev) < 0) {
 385                 ADBG((KERN_WARNING
 386                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 387                         __func__, dev->name));
 388                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 389                 ndev->dead = 1;
 390                 in6_dev_finish_destroy(ndev);
 391                 return NULL;
 392         }
 393
 394         /* One reference from device.  We must do this before
 395          * we invoke __ipv6_regen_rndid().
 396          */
 397         in6_dev_hold(ndev);
 398
 399         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 400                 ndev->cnf.accept_dad = -1;
 401
 402 #if IS_ENABLED(CONFIG_IPV6_SIT)
 403         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 404                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 405                 ndev->cnf.rtr_solicits = 0;
 406         }
 407 #endif
 408
 409 #ifdef CONFIG_IPV6_PRIVACY
 410         INIT_LIST_HEAD(&ndev->tempaddr_list);
 411         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 412         if ((dev->flags&IFF_LOOPBACK) ||
 413             dev->type == ARPHRD_TUNNEL ||
 414             dev->type == ARPHRD_TUNNEL6 ||
 415             dev->type == ARPHRD_SIT ||
 416             dev->type == ARPHRD_NONE) {
 417                 ndev->cnf.use_tempaddr = -1;
 418         } else {
 419                 in6_dev_hold(ndev);
 420                 ipv6_regen_rndid((unsigned long) ndev);
 421         }
 422 #endif
 423         ndev->token = in6addr_any;
 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         ifa->tokenized = false;
 880
 881         ifa->rt = rt;
 882
 883         ifa->idev = idev;
 884         in6_dev_hold(idev);
 885         /* For caller */
 886         in6_ifa_hold(ifa);
 887
 888         /* Add to big hash table */
 889         hash = inet6_addr_hash(addr);
 890
 891         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 892         spin_unlock(&addrconf_hash_lock);
 893
 894         write_lock(&idev->lock);
 895         /* Add to inet6_dev unicast addr list. */
 896         ipv6_link_dev_addr(idev, ifa);
 897
 898 #ifdef CONFIG_IPV6_PRIVACY
 899         if (ifa->flags&IFA_F_TEMPORARY) {
 900                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 901                 in6_ifa_hold(ifa);
 902         }
 903 #endif
 904
 905         in6_ifa_hold(ifa);
 906         write_unlock(&idev->lock);
 907 out2:
 908         rcu_read_unlock_bh();
 909
 910         if (likely(err == 0))
 911                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 912         else {
 913                 kfree(ifa);
 914                 ifa = ERR_PTR(err);
 915         }
 916
 917         return ifa;
 918 out:
 919         spin_unlock(&addrconf_hash_lock);
 920         goto out2;
 921 }
 922
 923 /* This function wants to get referenced ifp and releases it before return */
 924
 925 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 926 {
 927         struct inet6_ifaddr *ifa, *ifn;
 928         struct inet6_dev *idev = ifp->idev;
 929         int state;
 930         int deleted = 0, onlink = 0;
 931         unsigned long expires = jiffies;
 932
 933         spin_lock_bh(&ifp->state_lock);
 934         state = ifp->state;
 935         ifp->state = INET6_IFADDR_STATE_DEAD;
 936         spin_unlock_bh(&ifp->state_lock);
 937
 938         if (state == INET6_IFADDR_STATE_DEAD)
 939                 goto out;
 940
 941         spin_lock_bh(&addrconf_hash_lock);
 942         hlist_del_init_rcu(&ifp->addr_lst);
 943         spin_unlock_bh(&addrconf_hash_lock);
 944
 945         write_lock_bh(&idev->lock);
 946 #ifdef CONFIG_IPV6_PRIVACY
 947         if (ifp->flags&IFA_F_TEMPORARY) {
 948                 list_del(&ifp->tmp_list);
 949                 if (ifp->ifpub) {
 950                         in6_ifa_put(ifp->ifpub);
 951                         ifp->ifpub = NULL;
 952                 }
 953                 __in6_ifa_put(ifp);
 954         }
 955 #endif
 956
 957         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 958                 if (ifa == ifp) {
 959                         list_del_init(&ifp->if_list);
 960                         __in6_ifa_put(ifp);
 961
 962                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 963                                 break;
 964                         deleted = 1;
 965                         continue;
 966                 } else if (ifp->flags & IFA_F_PERMANENT) {
 967                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 968                                               ifp->prefix_len)) {
 969                                 if (ifa->flags & IFA_F_PERMANENT) {
 970                                         onlink = 1;
 971                                         if (deleted)
 972                                                 break;
 973                                 } else {
 974                                         unsigned long lifetime;
 975
 976                                         if (!onlink)
 977                                                 onlink = -1;
 978
 979                                         spin_lock(&ifa->lock);
 980
 981                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 982                                         /*
 983                                          * Note: Because this address is
 984                                          * not permanent, lifetime <
 985                                          * LONG_MAX / HZ here.
 986                                          */
 987                                         if (time_before(expires,
 988                                                         ifa->tstamp + lifetime * HZ))
 989                                                 expires = ifa->tstamp + lifetime * HZ;
 990                                         spin_unlock(&ifa->lock);
 991                                 }
 992                         }
 993                 }
 994         }
 995         write_unlock_bh(&idev->lock);
 996
 997         addrconf_del_timer(ifp);
 998
 999         ipv6_ifa_notify(RTM_DELADDR, ifp);
1000
1001         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1002
1003         /*
1004          * Purge or update corresponding prefix
1005          *
1006          * 1) we don't purge prefix here if address was not permanent.
1007          *    prefix is managed by its own lifetime.
1008          * 2) if there're no addresses, delete prefix.
1009          * 3) if there're still other permanent address(es),
1010          *    corresponding prefix is still permanent.
1011          * 4) otherwise, update prefix lifetime to the
1012          *    longest valid lifetime among the corresponding
1013          *    addresses on the device.
1014          *    Note: subsequent RA will update lifetime.
1015          *
1016          * --yoshfuji
1017          */
1018         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1019                 struct in6_addr prefix;
1020                 struct rt6_info *rt;
1021
1022                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1023
1024                 rt = addrconf_get_prefix_route(&prefix,
1025                                                ifp->prefix_len,
1026                                                ifp->idev->dev,
1027                                                0, RTF_GATEWAY | RTF_DEFAULT);
1028
1029                 if (rt) {
1030                         if (onlink == 0) {
1031                                 ip6_del_rt(rt);
1032                                 rt = NULL;
1033                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1034                                 rt6_set_expires(rt, expires);
1035                         }
1036                 }
1037                 ip6_rt_put(rt);
1038         }
1039
1040         /* clean up prefsrc entries */
1041         rt6_remove_prefsrc(ifp);
1042 out:
1043         in6_ifa_put(ifp);
1044 }
1045
1046 #ifdef CONFIG_IPV6_PRIVACY
1047 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1048 {
1049         struct inet6_dev *idev = ifp->idev;
1050         struct in6_addr addr, *tmpaddr;
1051         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1052         unsigned long regen_advance;
1053         int tmp_plen;
1054         int ret = 0;
1055         int max_addresses;
1056         u32 addr_flags;
1057         unsigned long now = jiffies;
1058
1059         write_lock(&idev->lock);
1060         if (ift) {
1061                 spin_lock_bh(&ift->lock);
1062                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1063                 spin_unlock_bh(&ift->lock);
1064                 tmpaddr = &addr;
1065         } else {
1066                 tmpaddr = NULL;
1067         }
1068 retry:
1069         in6_dev_hold(idev);
1070         if (idev->cnf.use_tempaddr <= 0) {
1071                 write_unlock(&idev->lock);
1072                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1073                 in6_dev_put(idev);
1074                 ret = -1;
1075                 goto out;
1076         }
1077         spin_lock_bh(&ifp->lock);
1078         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1079                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1080                 spin_unlock_bh(&ifp->lock);
1081                 write_unlock(&idev->lock);
1082                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1083                         __func__);
1084                 in6_dev_put(idev);
1085                 ret = -1;
1086                 goto out;
1087         }
1088         in6_ifa_hold(ifp);
1089         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1090         __ipv6_try_regen_rndid(idev, tmpaddr);
1091         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1092         age = (now - ifp->tstamp) / HZ;
1093         tmp_valid_lft = min_t(__u32,
1094                               ifp->valid_lft,
1095                               idev->cnf.temp_valid_lft + age);
1096         tmp_prefered_lft = min_t(__u32,
1097                                  ifp->prefered_lft,
1098                                  idev->cnf.temp_prefered_lft + age -
1099                                  idev->cnf.max_desync_factor);
1100         tmp_plen = ifp->prefix_len;
1101         max_addresses = idev->cnf.max_addresses;
1102         tmp_tstamp = ifp->tstamp;
1103         spin_unlock_bh(&ifp->lock);
1104
1105         regen_advance = idev->cnf.regen_max_retry *
1106                         idev->cnf.dad_transmits *
1107                         idev->nd_parms->retrans_time / HZ;
1108         write_unlock(&idev->lock);
1109
1110         /* A temporary address is created only if this calculated Preferred
1111          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1112          * an implementation must not create a temporary address with a zero
1113          * Preferred Lifetime.
1114          * Use age calculation as in addrconf_verify to avoid unnecessary
1115          * temporary addresses being generated.
1116          */
1117         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1118         if (tmp_prefered_lft <= regen_advance + age) {
1119                 in6_ifa_put(ifp);
1120                 in6_dev_put(idev);
1121                 ret = -1;
1122                 goto out;
1123         }
1124
1125         addr_flags = IFA_F_TEMPORARY;
1126         /* set in addrconf_prefix_rcv() */
1127         if (ifp->flags & IFA_F_OPTIMISTIC)
1128                 addr_flags |= IFA_F_OPTIMISTIC;
1129
1130         ift = ipv6_add_addr(idev, &addr, tmp_plen,
1131                             ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1132                             addr_flags);
1133         if (IS_ERR(ift)) {
1134                 in6_ifa_put(ifp);
1135                 in6_dev_put(idev);
1136                 pr_info("%s: retry temporary address regeneration\n", __func__);
1137                 tmpaddr = &addr;
1138                 write_lock(&idev->lock);
1139                 goto retry;
1140         }
1141
1142         spin_lock_bh(&ift->lock);
1143         ift->ifpub = ifp;
1144         ift->valid_lft = tmp_valid_lft;
1145         ift->prefered_lft = tmp_prefered_lft;
1146         ift->cstamp = now;
1147         ift->tstamp = tmp_tstamp;
1148         spin_unlock_bh(&ift->lock);
1149
1150         addrconf_dad_start(ift);
1151         in6_ifa_put(ift);
1152         in6_dev_put(idev);
1153 out:
1154         return ret;
1155 }
1156 #endif
1157
1158 /*
1159  *      Choose an appropriate source address (RFC3484)
1160  */
1161 enum {
1162         IPV6_SADDR_RULE_INIT = 0,
1163         IPV6_SADDR_RULE_LOCAL,
1164         IPV6_SADDR_RULE_SCOPE,
1165         IPV6_SADDR_RULE_PREFERRED,
1166 #ifdef CONFIG_IPV6_MIP6
1167         IPV6_SADDR_RULE_HOA,
1168 #endif
1169         IPV6_SADDR_RULE_OIF,
1170         IPV6_SADDR_RULE_LABEL,
1171 #ifdef CONFIG_IPV6_PRIVACY
1172         IPV6_SADDR_RULE_PRIVACY,
1173 #endif
1174         IPV6_SADDR_RULE_ORCHID,
1175         IPV6_SADDR_RULE_PREFIX,
1176         IPV6_SADDR_RULE_MAX
1177 };
1178
1179 struct ipv6_saddr_score {
1180         int                     rule;
1181         int                     addr_type;
1182         struct inet6_ifaddr     *ifa;
1183         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1184         int                     scopedist;
1185         int                     matchlen;
1186 };
1187
1188 struct ipv6_saddr_dst {
1189         const struct in6_addr *addr;
1190         int ifindex;
1191         int scope;
1192         int label;
1193         unsigned int prefs;
1194 };
1195
1196 static inline int ipv6_saddr_preferred(int type)
1197 {
1198         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1199                 return 1;
1200         return 0;
1201 }
1202
1203 static int ipv6_get_saddr_eval(struct net *net,
1204                                struct ipv6_saddr_score *score,
1205                                struct ipv6_saddr_dst *dst,
1206                                int i)
1207 {
1208         int ret;
1209
1210         if (i <= score->rule) {
1211                 switch (i) {
1212                 case IPV6_SADDR_RULE_SCOPE:
1213                         ret = score->scopedist;
1214                         break;
1215                 case IPV6_SADDR_RULE_PREFIX:
1216                         ret = score->matchlen;
1217                         break;
1218                 default:
1219                         ret = !!test_bit(i, score->scorebits);
1220                 }
1221                 goto out;
1222         }
1223
1224         switch (i) {
1225         case IPV6_SADDR_RULE_INIT:
1226                 /* Rule 0: remember if hiscore is not ready yet */
1227                 ret = !!score->ifa;
1228                 break;
1229         case IPV6_SADDR_RULE_LOCAL:
1230                 /* Rule 1: Prefer same address */
1231                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1232                 break;
1233         case IPV6_SADDR_RULE_SCOPE:
1234                 /* Rule 2: Prefer appropriate scope
1235                  *
1236                  *      ret
1237                  *       ^
1238                  *    -1 |  d 15
1239                  *    ---+--+-+---> scope
1240                  *       |
1241                  *       |             d is scope of the destination.
1242                  *  B-d  |  \
1243                  *       |   \      <- smaller scope is better if
1244                  *  B-15 |    \        if scope is enough for destinaion.
1245                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1246                  * d-C-1 | /
1247                  *       |/         <- greater is better
1248                  *   -C  /             if scope is not enough for destination.
1249                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1250                  *
1251                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1252                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1253                  * Assume B = 0 and we get C > 29.
1254                  */
1255                 ret = __ipv6_addr_src_scope(score->addr_type);
1256                 if (ret >= dst->scope)
1257                         ret = -ret;
1258                 else
1259                         ret -= 128;     /* 30 is enough */
1260                 score->scopedist = ret;
1261                 break;
1262         case IPV6_SADDR_RULE_PREFERRED:
1263                 /* Rule 3: Avoid deprecated and optimistic addresses */
1264                 ret = ipv6_saddr_preferred(score->addr_type) ||
1265                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1266                 break;
1267 #ifdef CONFIG_IPV6_MIP6
1268         case IPV6_SADDR_RULE_HOA:
1269             {
1270                 /* Rule 4: Prefer home address */
1271                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1272                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1273                 break;
1274             }
1275 #endif
1276         case IPV6_SADDR_RULE_OIF:
1277                 /* Rule 5: Prefer outgoing interface */
1278                 ret = (!dst->ifindex ||
1279                        dst->ifindex == score->ifa->idev->dev->ifindex);
1280                 break;
1281         case IPV6_SADDR_RULE_LABEL:
1282                 /* Rule 6: Prefer matching label */
1283                 ret = ipv6_addr_label(net,
1284                                       &score->ifa->addr, score->addr_type,
1285                                       score->ifa->idev->dev->ifindex) == dst->label;
1286                 break;
1287 #ifdef CONFIG_IPV6_PRIVACY
1288         case IPV6_SADDR_RULE_PRIVACY:
1289             {
1290                 /* Rule 7: Prefer public address
1291                  * Note: prefer temporary address if use_tempaddr >= 2
1292                  */
1293                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1294                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1295                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1296                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1297                 break;
1298             }
1299 #endif
1300         case IPV6_SADDR_RULE_ORCHID:
1301                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1302                  *          non-ORCHID vs non-ORCHID
1303                  */
1304                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1305                         ipv6_addr_orchid(dst->addr));
1306                 break;
1307         case IPV6_SADDR_RULE_PREFIX:
1308                 /* Rule 8: Use longest matching prefix */
1309                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1310                 if (ret > score->ifa->prefix_len)
1311                         ret = score->ifa->prefix_len;
1312                 score->matchlen = ret;
1313                 break;
1314         default:
1315                 ret = 0;
1316         }
1317
1318         if (ret)
1319                 __set_bit(i, score->scorebits);
1320         score->rule = i;
1321 out:
1322         return ret;
1323 }
1324
1325 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1326                        const struct in6_addr *daddr, unsigned int prefs,
1327                        struct in6_addr *saddr)
1328 {
1329         struct ipv6_saddr_score scores[2],
1330                                 *score = &scores[0], *hiscore = &scores[1];
1331         struct ipv6_saddr_dst dst;
1332         struct net_device *dev;
1333         int dst_type;
1334
1335         dst_type = __ipv6_addr_type(daddr);
1336         dst.addr = daddr;
1337         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1338         dst.scope = __ipv6_addr_src_scope(dst_type);
1339         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1340         dst.prefs = prefs;
1341
1342         hiscore->rule = -1;
1343         hiscore->ifa = NULL;
1344
1345         rcu_read_lock();
1346
1347         for_each_netdev_rcu(net, dev) {
1348                 struct inet6_dev *idev;
1349
1350                 /* Candidate Source Address (section 4)
1351                  *  - multicast and link-local destination address,
1352                  *    the set of candidate source address MUST only
1353                  *    include addresses assigned to interfaces
1354                  *    belonging to the same link as the outgoing
1355                  *    interface.
1356                  * (- For site-local destination addresses, the
1357                  *    set of candidate source addresses MUST only
1358                  *    include addresses assigned to interfaces
1359                  *    belonging to the same site as the outgoing
1360                  *    interface.)
1361                  */
1362                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1363                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1364                     dst.ifindex && dev->ifindex != dst.ifindex)
1365                         continue;
1366
1367                 idev = __in6_dev_get(dev);
1368                 if (!idev)
1369                         continue;
1370
1371                 read_lock_bh(&idev->lock);
1372                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1373                         int i;
1374
1375                         /*
1376                          * - Tentative Address (RFC2462 section 5.4)
1377                          *  - A tentative address is not considered
1378                          *    "assigned to an interface" in the traditional
1379                          *    sense, unless it is also flagged as optimistic.
1380                          * - Candidate Source Address (section 4)
1381                          *  - In any case, anycast addresses, multicast
1382                          *    addresses, and the unspecified address MUST
1383                          *    NOT be included in a candidate set.
1384                          */
1385                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1386                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1387                                 continue;
1388
1389                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1390
1391                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1392                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1393                                 LIMIT_NETDEBUG(KERN_DEBUG
1394                                                "ADDRCONF: unspecified / multicast address "
1395                                                "assigned as unicast address on %s",
1396                                                dev->name);
1397                                 continue;
1398                         }
1399
1400                         score->rule = -1;
1401                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1402
1403                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1404                                 int minihiscore, miniscore;
1405
1406                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1407                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1408
1409                                 if (minihiscore > miniscore) {
1410                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1411                                             score->scopedist > 0) {
1412                                                 /*
1413                                                  * special case:
1414                                                  * each remaining entry
1415                                                  * has too small (not enough)
1416                                                  * scope, because ifa entries
1417                                                  * are sorted by their scope
1418                                                  * values.
1419                                                  */
1420                                                 goto try_nextdev;
1421                                         }
1422                                         break;
1423                                 } else if (minihiscore < miniscore) {
1424                                         if (hiscore->ifa)
1425                                                 in6_ifa_put(hiscore->ifa);
1426
1427                                         in6_ifa_hold(score->ifa);
1428
1429                                         swap(hiscore, score);
1430
1431                                         /* restore our iterator */
1432                                         score->ifa = hiscore->ifa;
1433
1434                                         break;
1435                                 }
1436                         }
1437                 }
1438 try_nextdev:
1439                 read_unlock_bh(&idev->lock);
1440         }
1441         rcu_read_unlock();
1442
1443         if (!hiscore->ifa)
1444                 return -EADDRNOTAVAIL;
1445
1446         *saddr = hiscore->ifa->addr;
1447         in6_ifa_put(hiscore->ifa);
1448         return 0;
1449 }
1450 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1451
1452 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1453                       unsigned char banned_flags)
1454 {
1455         struct inet6_ifaddr *ifp;
1456         int err = -EADDRNOTAVAIL;
1457
1458         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1459                 if (ifp->scope == IFA_LINK &&
1460                     !(ifp->flags & banned_flags)) {
1461                         *addr = ifp->addr;
1462                         err = 0;
1463                         break;
1464                 }
1465         }
1466         return err;
1467 }
1468
1469 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1470                     unsigned char banned_flags)
1471 {
1472         struct inet6_dev *idev;
1473         int err = -EADDRNOTAVAIL;
1474
1475         rcu_read_lock();
1476         idev = __in6_dev_get(dev);
1477         if (idev) {
1478                 read_lock_bh(&idev->lock);
1479                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1480                 read_unlock_bh(&idev->lock);
1481         }
1482         rcu_read_unlock();
1483         return err;
1484 }
1485
1486 static int ipv6_count_addresses(struct inet6_dev *idev)
1487 {
1488         int cnt = 0;
1489         struct inet6_ifaddr *ifp;
1490
1491         read_lock_bh(&idev->lock);
1492         list_for_each_entry(ifp, &idev->addr_list, if_list)
1493                 cnt++;
1494         read_unlock_bh(&idev->lock);
1495         return cnt;
1496 }
1497
1498 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1499                   const struct net_device *dev, int strict)
1500 {
1501         struct inet6_ifaddr *ifp;
1502         unsigned int hash = inet6_addr_hash(addr);
1503
1504         rcu_read_lock_bh();
1505         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1506                 if (!net_eq(dev_net(ifp->idev->dev), net))
1507                         continue;
1508                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1509                     !(ifp->flags&IFA_F_TENTATIVE) &&
1510                     (dev == NULL || ifp->idev->dev == dev ||
1511                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1512                         rcu_read_unlock_bh();
1513                         return 1;
1514                 }
1515         }
1516
1517         rcu_read_unlock_bh();
1518         return 0;
1519 }
1520 EXPORT_SYMBOL(ipv6_chk_addr);
1521
1522 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1523                                struct net_device *dev)
1524 {
1525         unsigned int hash = inet6_addr_hash(addr);
1526         struct inet6_ifaddr *ifp;
1527
1528         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1529                 if (!net_eq(dev_net(ifp->idev->dev), net))
1530                         continue;
1531                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1532                         if (dev == NULL || ifp->idev->dev == dev)
1533                                 return true;
1534                 }
1535         }
1536         return false;
1537 }
1538
1539 /* Compares an address/prefix_len with addresses on device @dev.
1540  * If one is found it returns true.
1541  */
1542 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1543         const unsigned int prefix_len, struct net_device *dev)
1544 {
1545         struct inet6_dev *idev;
1546         struct inet6_ifaddr *ifa;
1547         bool ret = false;
1548
1549         rcu_read_lock();
1550         idev = __in6_dev_get(dev);
1551         if (idev) {
1552                 read_lock_bh(&idev->lock);
1553                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1554                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1555                         if (ret)
1556                                 break;
1557                 }
1558                 read_unlock_bh(&idev->lock);
1559         }
1560         rcu_read_unlock();
1561
1562         return ret;
1563 }
1564 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1565
1566 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1567 {
1568         struct inet6_dev *idev;
1569         struct inet6_ifaddr *ifa;
1570         int     onlink;
1571
1572         onlink = 0;
1573         rcu_read_lock();
1574         idev = __in6_dev_get(dev);
1575         if (idev) {
1576                 read_lock_bh(&idev->lock);
1577                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1578                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1579                                                    ifa->prefix_len);
1580                         if (onlink)
1581                                 break;
1582                 }
1583                 read_unlock_bh(&idev->lock);
1584         }
1585         rcu_read_unlock();
1586         return onlink;
1587 }
1588 EXPORT_SYMBOL(ipv6_chk_prefix);
1589
1590 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1591                                      struct net_device *dev, int strict)
1592 {
1593         struct inet6_ifaddr *ifp, *result = NULL;
1594         unsigned int hash = inet6_addr_hash(addr);
1595
1596         rcu_read_lock_bh();
1597         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1598                 if (!net_eq(dev_net(ifp->idev->dev), net))
1599                         continue;
1600                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1601                         if (dev == NULL || ifp->idev->dev == dev ||
1602                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1603                                 result = ifp;
1604                                 in6_ifa_hold(ifp);
1605                                 break;
1606                         }
1607                 }
1608         }
1609         rcu_read_unlock_bh();
1610
1611         return result;
1612 }
1613
1614 /* Gets referenced address, destroys ifaddr */
1615
1616 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1617 {
1618         if (ifp->flags&IFA_F_PERMANENT) {
1619                 spin_lock_bh(&ifp->lock);
1620                 addrconf_del_timer(ifp);
1621                 ifp->flags |= IFA_F_TENTATIVE;
1622                 if (dad_failed)
1623                         ifp->flags |= IFA_F_DADFAILED;
1624                 spin_unlock_bh(&ifp->lock);
1625                 if (dad_failed)
1626                         ipv6_ifa_notify(0, ifp);
1627                 in6_ifa_put(ifp);
1628 #ifdef CONFIG_IPV6_PRIVACY
1629         } else if (ifp->flags&IFA_F_TEMPORARY) {
1630                 struct inet6_ifaddr *ifpub;
1631                 spin_lock_bh(&ifp->lock);
1632                 ifpub = ifp->ifpub;
1633                 if (ifpub) {
1634                         in6_ifa_hold(ifpub);
1635                         spin_unlock_bh(&ifp->lock);
1636                         ipv6_create_tempaddr(ifpub, ifp);
1637                         in6_ifa_put(ifpub);
1638                 } else {
1639                         spin_unlock_bh(&ifp->lock);
1640                 }
1641                 ipv6_del_addr(ifp);
1642 #endif
1643         } else
1644                 ipv6_del_addr(ifp);
1645 }
1646
1647 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1648 {
1649         int err = -ENOENT;
1650
1651         spin_lock(&ifp->state_lock);
1652         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1653                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1654                 err = 0;
1655         }
1656         spin_unlock(&ifp->state_lock);
1657
1658         return err;
1659 }
1660
1661 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1662 {
1663         struct inet6_dev *idev = ifp->idev;
1664
1665         if (addrconf_dad_end(ifp)) {
1666                 in6_ifa_put(ifp);
1667                 return;
1668         }
1669
1670         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1671                              ifp->idev->dev->name, &ifp->addr);
1672
1673         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1674                 struct in6_addr addr;
1675
1676                 addr.s6_addr32[0] = htonl(0xfe800000);
1677                 addr.s6_addr32[1] = 0;
1678
1679                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1680                     ipv6_addr_equal(&ifp->addr, &addr)) {
1681                         /* DAD failed for link-local based on MAC address */
1682                         idev->cnf.disable_ipv6 = 1;
1683
1684                         pr_info("%s: IPv6 being disabled!\n",
1685                                 ifp->idev->dev->name);
1686                 }
1687         }
1688
1689         addrconf_dad_stop(ifp, 1);
1690 }
1691
1692 /* Join to solicited addr multicast group. */
1693
1694 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1695 {
1696         struct in6_addr maddr;
1697
1698         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1699                 return;
1700
1701         addrconf_addr_solict_mult(addr, &maddr);
1702         ipv6_dev_mc_inc(dev, &maddr);
1703 }
1704
1705 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1706 {
1707         struct in6_addr maddr;
1708
1709         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1710                 return;
1711
1712         addrconf_addr_solict_mult(addr, &maddr);
1713         __ipv6_dev_mc_dec(idev, &maddr);
1714 }
1715
1716 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1717 {
1718         struct in6_addr addr;
1719         if (ifp->prefix_len == 127) /* RFC 6164 */
1720                 return;
1721         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1722         if (ipv6_addr_any(&addr))
1723                 return;
1724         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1725 }
1726
1727 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1728 {
1729         struct in6_addr addr;
1730         if (ifp->prefix_len == 127) /* RFC 6164 */
1731                 return;
1732         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1733         if (ipv6_addr_any(&addr))
1734                 return;
1735         __ipv6_dev_ac_dec(ifp->idev, &addr);
1736 }
1737
1738 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1739 {
1740         if (dev->addr_len != ETH_ALEN)
1741                 return -1;
1742         memcpy(eui, dev->dev_addr, 3);
1743         memcpy(eui + 5, dev->dev_addr + 3, 3);
1744
1745         /*
1746          * The zSeries OSA network cards can be shared among various
1747          * OS instances, but the OSA cards have only one MAC address.
1748          * This leads to duplicate address conflicts in conjunction
1749          * with IPv6 if more than one instance uses the same card.
1750          *
1751          * The driver for these cards can deliver a unique 16-bit
1752          * identifier for each instance sharing the same card.  It is
1753          * placed instead of 0xFFFE in the interface identifier.  The
1754          * "u" bit of the interface identifier is not inverted in this
1755          * case.  Hence the resulting interface identifier has local
1756          * scope according to RFC2373.
1757          */
1758         if (dev->dev_id) {
1759                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1760                 eui[4] = dev->dev_id & 0xFF;
1761         } else {
1762                 eui[3] = 0xFF;
1763                 eui[4] = 0xFE;
1764                 eui[0] ^= 2;
1765         }
1766         return 0;
1767 }
1768
1769 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1770 {
1771         if (dev->addr_len != IEEE802154_ADDR_LEN)
1772                 return -1;
1773         memcpy(eui, dev->dev_addr, 8);
1774         eui[0] ^= 2;
1775         return 0;
1776 }
1777
1778 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1779 {
1780         union fwnet_hwaddr *ha;
1781
1782         if (dev->addr_len != FWNET_ALEN)
1783                 return -1;
1784
1785         ha = (union fwnet_hwaddr *)dev->dev_addr;
1786
1787         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1788         eui[0] ^= 2;
1789         return 0;
1790 }
1791
1792 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1793 {
1794         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1795         if (dev->addr_len != ARCNET_ALEN)
1796                 return -1;
1797         memset(eui, 0, 7);
1798         eui[7] = *(u8 *)dev->dev_addr;
1799         return 0;
1800 }
1801
1802 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1803 {
1804         if (dev->addr_len != INFINIBAND_ALEN)
1805                 return -1;
1806         memcpy(eui, dev->dev_addr + 12, 8);
1807         eui[0] |= 2;
1808         return 0;
1809 }
1810
1811 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1812 {
1813         if (addr == 0)
1814                 return -1;
1815         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1816                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1817                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1818                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1819                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1820                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1821         eui[1] = 0;
1822         eui[2] = 0x5E;
1823         eui[3] = 0xFE;
1824         memcpy(eui + 4, &addr, 4);
1825         return 0;
1826 }
1827
1828 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1829 {
1830         if (dev->priv_flags & IFF_ISATAP)
1831                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1832         return -1;
1833 }
1834
1835 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1836 {
1837         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1838 }
1839
1840 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1841 {
1842         switch (dev->type) {
1843         case ARPHRD_ETHER:
1844         case ARPHRD_FDDI:
1845                 return addrconf_ifid_eui48(eui, dev);
1846         case ARPHRD_ARCNET:
1847                 return addrconf_ifid_arcnet(eui, dev);
1848         case ARPHRD_INFINIBAND:
1849                 return addrconf_ifid_infiniband(eui, dev);
1850         case ARPHRD_SIT:
1851                 return addrconf_ifid_sit(eui, dev);
1852         case ARPHRD_IPGRE:
1853                 return addrconf_ifid_gre(eui, dev);
1854         case ARPHRD_IEEE802154:
1855                 return addrconf_ifid_eui64(eui, dev);
1856         case ARPHRD_IEEE1394:
1857                 return addrconf_ifid_ieee1394(eui, dev);
1858         }
1859         return -1;
1860 }
1861
1862 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1863 {
1864         int err = -1;
1865         struct inet6_ifaddr *ifp;
1866
1867         read_lock_bh(&idev->lock);
1868         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1869                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1870                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1871                         err = 0;
1872                         break;
1873                 }
1874         }
1875         read_unlock_bh(&idev->lock);
1876         return err;
1877 }
1878
1879 #ifdef CONFIG_IPV6_PRIVACY
1880 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1881 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1882 {
1883 regen:
1884         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1885         idev->rndid[0] &= ~0x02;
1886
1887         /*
1888          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1889          * check if generated address is not inappropriate
1890          *
1891          *  - Reserved subnet anycast (RFC 2526)
1892          *      11111101 11....11 1xxxxxxx
1893          *  - ISATAP (RFC4214) 6.1
1894          *      00-00-5E-FE-xx-xx-xx-xx
1895          *  - value 0
1896          *  - XXX: already assigned to an address on the device
1897          */
1898         if (idev->rndid[0] == 0xfd &&
1899             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1900             (idev->rndid[7]&0x80))
1901                 goto regen;
1902         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1903                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1904                         goto regen;
1905                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1906                         goto regen;
1907         }
1908 }
1909
1910 static void ipv6_regen_rndid(unsigned long data)
1911 {
1912         struct inet6_dev *idev = (struct inet6_dev *) data;
1913         unsigned long expires;
1914
1915         rcu_read_lock_bh();
1916         write_lock_bh(&idev->lock);
1917
1918         if (idev->dead)
1919                 goto out;
1920
1921         __ipv6_regen_rndid(idev);
1922
1923         expires = jiffies +
1924                 idev->cnf.temp_prefered_lft * HZ -
1925                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1926                 idev->cnf.max_desync_factor * HZ;
1927         if (time_before(expires, jiffies)) {
1928                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1929                         __func__, idev->dev->name);
1930                 goto out;
1931         }
1932
1933         if (!mod_timer(&idev->regen_timer, expires))
1934                 in6_dev_hold(idev);
1935
1936 out:
1937         write_unlock_bh(&idev->lock);
1938         rcu_read_unlock_bh();
1939         in6_dev_put(idev);
1940 }
1941
1942 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1943 {
1944         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1945                 __ipv6_regen_rndid(idev);
1946 }
1947 #endif
1948
1949 /*
1950  *      Add prefix route.
1951  */
1952
1953 static void
1954 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1955                       unsigned long expires, u32 flags)
1956 {
1957         struct fib6_config cfg = {
1958                 .fc_table = RT6_TABLE_PREFIX,
1959                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1960                 .fc_ifindex = dev->ifindex,
1961                 .fc_expires = expires,
1962                 .fc_dst_len = plen,
1963                 .fc_flags = RTF_UP | flags,
1964                 .fc_nlinfo.nl_net = dev_net(dev),
1965                 .fc_protocol = RTPROT_KERNEL,
1966         };
1967
1968         cfg.fc_dst = *pfx;
1969
1970         /* Prevent useless cloning on PtP SIT.
1971            This thing is done here expecting that the whole
1972            class of non-broadcast devices need not cloning.
1973          */
1974 #if IS_ENABLED(CONFIG_IPV6_SIT)
1975         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1976                 cfg.fc_flags |= RTF_NONEXTHOP;
1977 #endif
1978
1979         ip6_route_add(&cfg);
1980 }
1981
1982
1983 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1984                                                   int plen,
1985                                                   const struct net_device *dev,
1986                                                   u32 flags, u32 noflags)
1987 {
1988         struct fib6_node *fn;
1989         struct rt6_info *rt = NULL;
1990         struct fib6_table *table;
1991
1992         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1993         if (table == NULL)
1994                 return NULL;
1995
1996         read_lock_bh(&table->tb6_lock);
1997         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1998         if (!fn)
1999                 goto out;
2000         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2001                 if (rt->dst.dev->ifindex != dev->ifindex)
2002                         continue;
2003                 if ((rt->rt6i_flags & flags) != flags)
2004                         continue;
2005                 if ((rt->rt6i_flags & noflags) != 0)
2006                         continue;
2007                 dst_hold(&rt->dst);
2008                 break;
2009         }
2010 out:
2011         read_unlock_bh(&table->tb6_lock);
2012         return rt;
2013 }
2014
2015
2016 /* Create "default" multicast route to the interface */
2017
2018 static void addrconf_add_mroute(struct net_device *dev)
2019 {
2020         struct fib6_config cfg = {
2021                 .fc_table = RT6_TABLE_LOCAL,
2022                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2023                 .fc_ifindex = dev->ifindex,
2024                 .fc_dst_len = 8,
2025                 .fc_flags = RTF_UP,
2026                 .fc_nlinfo.nl_net = dev_net(dev),
2027         };
2028
2029         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2030
2031         ip6_route_add(&cfg);
2032 }
2033
2034 #if IS_ENABLED(CONFIG_IPV6_SIT)
2035 static void sit_route_add(struct net_device *dev)
2036 {
2037         struct fib6_config cfg = {
2038                 .fc_table = RT6_TABLE_MAIN,
2039                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2040                 .fc_ifindex = dev->ifindex,
2041                 .fc_dst_len = 96,
2042                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2043                 .fc_nlinfo.nl_net = dev_net(dev),
2044         };
2045
2046         /* prefix length - 96 bits "::d.d.d.d" */
2047         ip6_route_add(&cfg);
2048 }
2049 #endif
2050
2051 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2052 {
2053         struct inet6_dev *idev;
2054
2055         ASSERT_RTNL();
2056
2057         idev = ipv6_find_idev(dev);
2058         if (!idev)
2059                 return ERR_PTR(-ENOBUFS);
2060
2061         if (idev->cnf.disable_ipv6)
2062                 return ERR_PTR(-EACCES);
2063
2064         /* Add default multicast route */
2065         if (!(dev->flags & IFF_LOOPBACK))
2066                 addrconf_add_mroute(dev);
2067
2068         return idev;
2069 }
2070
2071 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2072 {
2073         struct prefix_info *pinfo;
2074         __u32 valid_lft;
2075         __u32 prefered_lft;
2076         int addr_type;
2077         struct inet6_dev *in6_dev;
2078         struct net *net = dev_net(dev);
2079
2080         pinfo = (struct prefix_info *) opt;
2081
2082         if (len < sizeof(struct prefix_info)) {
2083                 ADBG(("addrconf: prefix option too short\n"));
2084                 return;
2085         }
2086
2087         /*
2088          *      Validation checks ([ADDRCONF], page 19)
2089          */
2090
2091         addr_type = ipv6_addr_type(&pinfo->prefix);
2092
2093         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2094                 return;
2095
2096         valid_lft = ntohl(pinfo->valid);
2097         prefered_lft = ntohl(pinfo->prefered);
2098
2099         if (prefered_lft > valid_lft) {
2100                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2101                 return;
2102         }
2103
2104         in6_dev = in6_dev_get(dev);
2105
2106         if (in6_dev == NULL) {
2107                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2108                                     dev->name);
2109                 return;
2110         }
2111
2112         /*
2113          *      Two things going on here:
2114          *      1) Add routes for on-link prefixes
2115          *      2) Configure prefixes with the auto flag set
2116          */
2117
2118         if (pinfo->onlink) {
2119                 struct rt6_info *rt;
2120                 unsigned long rt_expires;
2121
2122                 /* Avoid arithmetic overflow. Really, we could
2123                  * save rt_expires in seconds, likely valid_lft,
2124                  * but it would require division in fib gc, that it
2125                  * not good.
2126                  */
2127                 if (HZ > USER_HZ)
2128                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2129                 else
2130                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2131
2132                 if (addrconf_finite_timeout(rt_expires))
2133                         rt_expires *= HZ;
2134
2135                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2136                                                pinfo->prefix_len,
2137                                                dev,
2138                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2139                                                RTF_GATEWAY | RTF_DEFAULT);
2140
2141                 if (rt) {
2142                         /* Autoconf prefix route */
2143                         if (valid_lft == 0) {
2144                                 ip6_del_rt(rt);
2145                                 rt = NULL;
2146                         } else if (addrconf_finite_timeout(rt_expires)) {
2147                                 /* not infinity */
2148                                 rt6_set_expires(rt, jiffies + rt_expires);
2149                         } else {
2150                                 rt6_clean_expires(rt);
2151                         }
2152                 } else if (valid_lft) {
2153                         clock_t expires = 0;
2154                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2155                         if (addrconf_finite_timeout(rt_expires)) {
2156                                 /* not infinity */
2157                                 flags |= RTF_EXPIRES;
2158                                 expires = jiffies_to_clock_t(rt_expires);
2159                         }
2160                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2161                                               dev, expires, flags);
2162                 }
2163                 ip6_rt_put(rt);
2164         }
2165
2166         /* Try to figure out our local address for this prefix */
2167
2168         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2169                 struct inet6_ifaddr *ifp;
2170                 struct in6_addr addr;
2171                 int create = 0, update_lft = 0;
2172                 bool tokenized = false;
2173
2174                 if (pinfo->prefix_len == 64) {
2175                         memcpy(&addr, &pinfo->prefix, 8);
2176
2177                         if (!ipv6_addr_any(&in6_dev->token)) {
2178                                 read_lock_bh(&in6_dev->lock);
2179                                 memcpy(addr.s6_addr + 8,
2180                                        in6_dev->token.s6_addr + 8, 8);
2181                                 read_unlock_bh(&in6_dev->lock);
2182                                 tokenized = true;
2183                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2184                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2185                                 in6_dev_put(in6_dev);
2186                                 return;
2187                         }
2188                         goto ok;
2189                 }
2190                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2191                                     pinfo->prefix_len);
2192                 in6_dev_put(in6_dev);
2193                 return;
2194
2195 ok:
2196
2197                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2198
2199                 if (ifp == NULL && valid_lft) {
2200                         int max_addresses = in6_dev->cnf.max_addresses;
2201                         u32 addr_flags = 0;
2202
2203 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2204                         if (in6_dev->cnf.optimistic_dad &&
2205                             !net->ipv6.devconf_all->forwarding && sllao)
2206                                 addr_flags = IFA_F_OPTIMISTIC;
2207 #endif
2208
2209                         /* Do not allow to create too much of autoconfigured
2210                          * addresses; this would be too easy way to crash kernel.
2211                          */
2212                         if (!max_addresses ||
2213                             ipv6_count_addresses(in6_dev) < max_addresses)
2214                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2215                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2216                                                     addr_flags);
2217
2218                         if (IS_ERR_OR_NULL(ifp)) {
2219                                 in6_dev_put(in6_dev);
2220                                 return;
2221                         }
2222
2223                         update_lft = create = 1;
2224                         ifp->cstamp = jiffies;
2225                         ifp->tokenized = tokenized;
2226                         addrconf_dad_start(ifp);
2227                 }
2228
2229                 if (ifp) {
2230                         int flags;
2231                         unsigned long now;
2232 #ifdef CONFIG_IPV6_PRIVACY
2233                         struct inet6_ifaddr *ift;
2234 #endif
2235                         u32 stored_lft;
2236
2237                         /* update lifetime (RFC2462 5.5.3 e) */
2238                         spin_lock(&ifp->lock);
2239                         now = jiffies;
2240                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2241                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2242                         else
2243                                 stored_lft = 0;
2244                         if (!update_lft && stored_lft) {
2245                                 if (valid_lft > MIN_VALID_LIFETIME ||
2246                                     valid_lft > stored_lft)
2247                                         update_lft = 1;
2248                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2249                                         /* valid_lft <= stored_lft is always true */
2250                                         /*
2251                                          * RFC 4862 Section 5.5.3e:
2252                                          * "Note that the preferred lifetime of
2253                                          *  the corresponding address is always
2254                                          *  reset to the Preferred Lifetime in
2255                                          *  the received Prefix Information
2256                                          *  option, regardless of whether the
2257                                          *  valid lifetime is also reset or
2258                                          *  ignored."
2259                                          *
2260                                          *  So if the preferred lifetime in
2261                                          *  this advertisement is different
2262                                          *  than what we have stored, but the
2263                                          *  valid lifetime is invalid, just
2264                                          *  reset prefered_lft.
2265                                          *
2266                                          *  We must set the valid lifetime
2267                                          *  to the stored lifetime since we'll
2268                                          *  be updating the timestamp below,
2269                                          *  else we'll set it back to the
2270                                          *  minimum.
2271                                          */
2272                                         if (prefered_lft != ifp->prefered_lft) {
2273                                                 valid_lft = stored_lft;
2274                                                 update_lft = 1;
2275                                         }
2276                                 } else {
2277                                         valid_lft = MIN_VALID_LIFETIME;
2278                                         if (valid_lft < prefered_lft)
2279                                                 prefered_lft = valid_lft;
2280                                         update_lft = 1;
2281                                 }
2282                         }
2283
2284                         if (update_lft) {
2285                                 ifp->valid_lft = valid_lft;
2286                                 ifp->prefered_lft = prefered_lft;
2287                                 ifp->tstamp = now;
2288                                 flags = ifp->flags;
2289                                 ifp->flags &= ~IFA_F_DEPRECATED;
2290                                 spin_unlock(&ifp->lock);
2291
2292                                 if (!(flags&IFA_F_TENTATIVE))
2293                                         ipv6_ifa_notify(0, ifp);
2294                         } else
2295                                 spin_unlock(&ifp->lock);
2296
2297 #ifdef CONFIG_IPV6_PRIVACY
2298                         read_lock_bh(&in6_dev->lock);
2299                         /* update all temporary addresses in the list */
2300                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2301                                             tmp_list) {
2302                                 int age, max_valid, max_prefered;
2303
2304                                 if (ifp != ift->ifpub)
2305                                         continue;
2306
2307                                 /*
2308                                  * RFC 4941 section 3.3:
2309                                  * If a received option will extend the lifetime
2310                                  * of a public address, the lifetimes of
2311                                  * temporary addresses should be extended,
2312                                  * subject to the overall constraint that no
2313                                  * temporary addresses should ever remain
2314                                  * "valid" or "preferred" for a time longer than
2315                                  * (TEMP_VALID_LIFETIME) or
2316                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2317                                  * respectively.
2318                                  */
2319                                 age = (now - ift->cstamp) / HZ;
2320                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2321                                 if (max_valid < 0)
2322                                         max_valid = 0;
2323
2324                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2325                                                in6_dev->cnf.max_desync_factor -
2326                                                age;
2327                                 if (max_prefered < 0)
2328                                         max_prefered = 0;
2329
2330                                 if (valid_lft > max_valid)
2331                                         valid_lft = max_valid;
2332
2333                                 if (prefered_lft > max_prefered)
2334                                         prefered_lft = max_prefered;
2335
2336                                 spin_lock(&ift->lock);
2337                                 flags = ift->flags;
2338                                 ift->valid_lft = valid_lft;
2339                                 ift->prefered_lft = prefered_lft;
2340                                 ift->tstamp = now;
2341                                 if (prefered_lft > 0)
2342                                         ift->flags &= ~IFA_F_DEPRECATED;
2343
2344                                 spin_unlock(&ift->lock);
2345                                 if (!(flags&IFA_F_TENTATIVE))
2346                                         ipv6_ifa_notify(0, ift);
2347                         }
2348
2349                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2350                                 /*
2351                                  * When a new public address is created as
2352                                  * described in [ADDRCONF], also create a new
2353                                  * temporary address. Also create a temporary
2354                                  * address if it's enabled but no temporary
2355                                  * address currently exists.
2356                                  */
2357                                 read_unlock_bh(&in6_dev->lock);
2358                                 ipv6_create_tempaddr(ifp, NULL);
2359                         } else {
2360                                 read_unlock_bh(&in6_dev->lock);
2361                         }
2362 #endif
2363                         in6_ifa_put(ifp);
2364                         addrconf_verify(0);
2365                 }
2366         }
2367         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2368         in6_dev_put(in6_dev);
2369 }
2370
2371 /*
2372  *      Set destination address.
2373  *      Special case for SIT interfaces where we create a new "virtual"
2374  *      device.
2375  */
2376 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2377 {
2378         struct in6_ifreq ireq;
2379         struct net_device *dev;
2380         int err = -EINVAL;
2381
2382         rtnl_lock();
2383
2384         err = -EFAULT;
2385         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2386                 goto err_exit;
2387
2388         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2389
2390         err = -ENODEV;
2391         if (dev == NULL)
2392                 goto err_exit;
2393
2394 #if IS_ENABLED(CONFIG_IPV6_SIT)
2395         if (dev->type == ARPHRD_SIT) {
2396                 const struct net_device_ops *ops = dev->netdev_ops;
2397                 struct ifreq ifr;
2398                 struct ip_tunnel_parm p;
2399
2400                 err = -EADDRNOTAVAIL;
2401                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2402                         goto err_exit;
2403
2404                 memset(&p, 0, sizeof(p));
2405                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2406                 p.iph.saddr = 0;
2407                 p.iph.version = 4;
2408                 p.iph.ihl = 5;
2409                 p.iph.protocol = IPPROTO_IPV6;
2410                 p.iph.ttl = 64;
2411                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2412
2413                 if (ops->ndo_do_ioctl) {
2414                         mm_segment_t oldfs = get_fs();
2415
2416                         set_fs(KERNEL_DS);
2417                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2418                         set_fs(oldfs);
2419                 } else
2420                         err = -EOPNOTSUPP;
2421
2422                 if (err == 0) {
2423                         err = -ENOBUFS;
2424                         dev = __dev_get_by_name(net, p.name);
2425                         if (!dev)
2426                                 goto err_exit;
2427                         err = dev_open(dev);
2428                 }
2429         }
2430 #endif
2431
2432 err_exit:
2433         rtnl_unlock();
2434         return err;
2435 }
2436
2437 /*
2438  *      Manual configuration of address on an interface
2439  */
2440 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2441                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2442                           __u32 valid_lft)
2443 {
2444         struct inet6_ifaddr *ifp;
2445         struct inet6_dev *idev;
2446         struct net_device *dev;
2447         int scope;
2448         u32 flags;
2449         clock_t expires;
2450         unsigned long timeout;
2451
2452         ASSERT_RTNL();
2453
2454         if (plen > 128)
2455                 return -EINVAL;
2456
2457         /* check the lifetime */
2458         if (!valid_lft || prefered_lft > valid_lft)
2459                 return -EINVAL;
2460
2461         dev = __dev_get_by_index(net, ifindex);
2462         if (!dev)
2463                 return -ENODEV;
2464
2465         idev = addrconf_add_dev(dev);
2466         if (IS_ERR(idev))
2467                 return PTR_ERR(idev);
2468
2469         scope = ipv6_addr_scope(pfx);
2470
2471         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2472         if (addrconf_finite_timeout(timeout)) {
2473                 expires = jiffies_to_clock_t(timeout * HZ);
2474                 valid_lft = timeout;
2475                 flags = RTF_EXPIRES;
2476         } else {
2477                 expires = 0;
2478                 flags = 0;
2479                 ifa_flags |= IFA_F_PERMANENT;
2480         }
2481
2482         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2483         if (addrconf_finite_timeout(timeout)) {
2484                 if (timeout == 0)
2485                         ifa_flags |= IFA_F_DEPRECATED;
2486                 prefered_lft = timeout;
2487         }
2488
2489         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2490
2491         if (!IS_ERR(ifp)) {
2492                 spin_lock_bh(&ifp->lock);
2493                 ifp->valid_lft = valid_lft;
2494                 ifp->prefered_lft = prefered_lft;
2495                 ifp->tstamp = jiffies;
2496                 spin_unlock_bh(&ifp->lock);
2497
2498                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2499                                       expires, flags);
2500                 /*
2501                  * Note that section 3.1 of RFC 4429 indicates
2502                  * that the Optimistic flag should not be set for
2503                  * manually configured addresses
2504                  */
2505                 addrconf_dad_start(ifp);
2506                 in6_ifa_put(ifp);
2507                 addrconf_verify(0);
2508                 return 0;
2509         }
2510
2511         return PTR_ERR(ifp);
2512 }
2513
2514 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2515                           unsigned int plen)
2516 {
2517         struct inet6_ifaddr *ifp;
2518         struct inet6_dev *idev;
2519         struct net_device *dev;
2520
2521         if (plen > 128)
2522                 return -EINVAL;
2523
2524         dev = __dev_get_by_index(net, ifindex);
2525         if (!dev)
2526                 return -ENODEV;
2527
2528         if ((idev = __in6_dev_get(dev)) == NULL)
2529                 return -ENXIO;
2530
2531         read_lock_bh(&idev->lock);
2532         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2533                 if (ifp->prefix_len == plen &&
2534                     ipv6_addr_equal(pfx, &ifp->addr)) {
2535                         in6_ifa_hold(ifp);
2536                         read_unlock_bh(&idev->lock);
2537
2538                         ipv6_del_addr(ifp);
2539
2540                         /* If the last address is deleted administratively,
2541                            disable IPv6 on this interface.
2542                          */
2543                         if (list_empty(&idev->addr_list))
2544                                 addrconf_ifdown(idev->dev, 1);
2545                         return 0;
2546                 }
2547         }
2548         read_unlock_bh(&idev->lock);
2549         return -EADDRNOTAVAIL;
2550 }
2551
2552
2553 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2554 {
2555         struct in6_ifreq ireq;
2556         int err;
2557
2558         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2559                 return -EPERM;
2560
2561         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2562                 return -EFAULT;
2563
2564         rtnl_lock();
2565         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2566                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2567                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2568         rtnl_unlock();
2569         return err;
2570 }
2571
2572 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2573 {
2574         struct in6_ifreq ireq;
2575         int err;
2576
2577         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2578                 return -EPERM;
2579
2580         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2581                 return -EFAULT;
2582
2583         rtnl_lock();
2584         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2585                              ireq.ifr6_prefixlen);
2586         rtnl_unlock();
2587         return err;
2588 }
2589
2590 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2591                      int plen, int scope)
2592 {
2593         struct inet6_ifaddr *ifp;
2594
2595         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2596         if (!IS_ERR(ifp)) {
2597                 spin_lock_bh(&ifp->lock);
2598                 ifp->flags &= ~IFA_F_TENTATIVE;
2599                 spin_unlock_bh(&ifp->lock);
2600                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2601                 in6_ifa_put(ifp);
2602         }
2603 }
2604
2605 #if IS_ENABLED(CONFIG_IPV6_SIT)
2606 static void sit_add_v4_addrs(struct inet6_dev *idev)
2607 {
2608         struct in6_addr addr;
2609         struct net_device *dev;
2610         struct net *net = dev_net(idev->dev);
2611         int scope;
2612
2613         ASSERT_RTNL();
2614
2615         memset(&addr, 0, sizeof(struct in6_addr));
2616         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2617
2618         if (idev->dev->flags&IFF_POINTOPOINT) {
2619                 addr.s6_addr32[0] = htonl(0xfe800000);
2620                 scope = IFA_LINK;
2621         } else {
2622                 scope = IPV6_ADDR_COMPATv4;
2623         }
2624
2625         if (addr.s6_addr32[3]) {
2626                 add_addr(idev, &addr, 128, scope);
2627                 return;
2628         }
2629
2630         for_each_netdev(net, dev) {
2631                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2632                 if (in_dev && (dev->flags & IFF_UP)) {
2633                         struct in_ifaddr *ifa;
2634
2635                         int flag = scope;
2636
2637                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2638                                 int plen;
2639
2640                                 addr.s6_addr32[3] = ifa->ifa_local;
2641
2642                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2643                                         continue;
2644                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2645                                         if (idev->dev->flags&IFF_POINTOPOINT)
2646                                                 continue;
2647                                         flag |= IFA_HOST;
2648                                 }
2649                                 if (idev->dev->flags&IFF_POINTOPOINT)
2650                                         plen = 64;
2651                                 else
2652                                         plen = 96;
2653
2654                                 add_addr(idev, &addr, plen, flag);
2655                         }
2656                 }
2657         }
2658 }
2659 #endif
2660
2661 static void init_loopback(struct net_device *dev)
2662 {
2663         struct inet6_dev  *idev;
2664         struct net_device *sp_dev;
2665         struct inet6_ifaddr *sp_ifa;
2666         struct rt6_info *sp_rt;
2667
2668         /* ::1 */
2669
2670         ASSERT_RTNL();
2671
2672         if ((idev = ipv6_find_idev(dev)) == NULL) {
2673                 pr_debug("%s: add_dev failed\n", __func__);
2674                 return;
2675         }
2676
2677         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2678
2679         /* Add routes to other interface's IPv6 addresses */
2680         for_each_netdev(dev_net(dev), sp_dev) {
2681                 if (!strcmp(sp_dev->name, dev->name))
2682                         continue;
2683
2684                 idev = __in6_dev_get(sp_dev);
2685                 if (!idev)
2686                         continue;
2687
2688                 read_lock_bh(&idev->lock);
2689                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2690
2691                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2692                                 continue;
2693
2694                         if (sp_ifa->rt)
2695                                 continue;
2696
2697                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2698
2699                         /* Failure cases are ignored */
2700                         if (!IS_ERR(sp_rt)) {
2701                                 sp_ifa->rt = sp_rt;
2702                                 ip6_ins_rt(sp_rt);
2703                         }
2704                 }
2705                 read_unlock_bh(&idev->lock);
2706         }
2707 }
2708
2709 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2710 {
2711         struct inet6_ifaddr *ifp;
2712         u32 addr_flags = IFA_F_PERMANENT;
2713
2714 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2715         if (idev->cnf.optimistic_dad &&
2716             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2717                 addr_flags |= IFA_F_OPTIMISTIC;
2718 #endif
2719
2720
2721         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2722         if (!IS_ERR(ifp)) {
2723                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2724                 addrconf_dad_start(ifp);
2725                 in6_ifa_put(ifp);
2726         }
2727 }
2728
2729 static void addrconf_dev_config(struct net_device *dev)
2730 {
2731         struct in6_addr addr;
2732         struct inet6_dev *idev;
2733
2734         ASSERT_RTNL();
2735
2736         if ((dev->type != ARPHRD_ETHER) &&
2737             (dev->type != ARPHRD_FDDI) &&
2738             (dev->type != ARPHRD_ARCNET) &&
2739             (dev->type != ARPHRD_INFINIBAND) &&
2740             (dev->type != ARPHRD_IEEE802154) &&
2741             (dev->type != ARPHRD_IEEE1394)) {
2742                 /* Alas, we support only Ethernet autoconfiguration. */
2743                 return;
2744         }
2745
2746         idev = addrconf_add_dev(dev);
2747         if (IS_ERR(idev))
2748                 return;
2749
2750         memset(&addr, 0, sizeof(struct in6_addr));
2751         addr.s6_addr32[0] = htonl(0xFE800000);
2752
2753         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2754                 addrconf_add_linklocal(idev, &addr);
2755 }
2756
2757 #if IS_ENABLED(CONFIG_IPV6_SIT)
2758 static void addrconf_sit_config(struct net_device *dev)
2759 {
2760         struct inet6_dev *idev;
2761
2762         ASSERT_RTNL();
2763
2764         /*
2765          * Configure the tunnel with one of our IPv4
2766          * addresses... we should configure all of
2767          * our v4 addrs in the tunnel
2768          */
2769
2770         if ((idev = ipv6_find_idev(dev)) == NULL) {
2771                 pr_debug("%s: add_dev failed\n", __func__);
2772                 return;
2773         }
2774
2775         if (dev->priv_flags & IFF_ISATAP) {
2776                 struct in6_addr addr;
2777
2778                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2779                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2780                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2781                         addrconf_add_linklocal(idev, &addr);
2782                 return;
2783         }
2784
2785         sit_add_v4_addrs(idev);
2786
2787         if (dev->flags&IFF_POINTOPOINT)
2788                 addrconf_add_mroute(dev);
2789         else
2790                 sit_route_add(dev);
2791 }
2792 #endif
2793
2794 #if IS_ENABLED(CONFIG_NET_IPGRE)
2795 static void addrconf_gre_config(struct net_device *dev)
2796 {
2797         struct inet6_dev *idev;
2798         struct in6_addr addr;
2799
2800         pr_info("%s(%s)\n", __func__, dev->name);
2801
2802         ASSERT_RTNL();
2803
2804         if ((idev = ipv6_find_idev(dev)) == NULL) {
2805                 pr_debug("%s: add_dev failed\n", __func__);
2806                 return;
2807         }
2808
2809         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2810         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2811
2812         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2813                 addrconf_add_linklocal(idev, &addr);
2814 }
2815 #endif
2816
2817 static inline int
2818 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2819 {
2820         struct in6_addr lladdr;
2821
2822         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2823                 addrconf_add_linklocal(idev, &lladdr);
2824                 return 0;
2825         }
2826         return -1;
2827 }
2828
2829 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2830 {
2831         struct net_device *link_dev;
2832         struct net *net = dev_net(idev->dev);
2833
2834         /* first try to inherit the link-local address from the link device */
2835         if (idev->dev->iflink &&
2836             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2837                 if (!ipv6_inherit_linklocal(idev, link_dev))
2838                         return;
2839         }
2840         /* then try to inherit it from any device */
2841         for_each_netdev(net, link_dev) {
2842                 if (!ipv6_inherit_linklocal(idev, link_dev))
2843                         return;
2844         }
2845         pr_debug("init ip6-ip6: add_linklocal failed\n");
2846 }
2847
2848 /*
2849  * Autoconfigure tunnel with a link-local address so routing protocols,
2850  * DHCPv6, MLD etc. can be run over the virtual link
2851  */
2852
2853 static void addrconf_ip6_tnl_config(struct net_device *dev)
2854 {
2855         struct inet6_dev *idev;
2856
2857         ASSERT_RTNL();
2858
2859         idev = addrconf_add_dev(dev);
2860         if (IS_ERR(idev)) {
2861                 pr_debug("init ip6-ip6: add_dev failed\n");
2862                 return;
2863         }
2864         ip6_tnl_add_linklocal(idev);
2865 }
2866
2867 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2868                            void *data)
2869 {
2870         struct net_device *dev = (struct net_device *) data;
2871         struct inet6_dev *idev = __in6_dev_get(dev);
2872         int run_pending = 0;
2873         int err;
2874
2875         switch (event) {
2876         case NETDEV_REGISTER:
2877                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2878                         idev = ipv6_add_dev(dev);
2879                         if (!idev)
2880                                 return notifier_from_errno(-ENOMEM);
2881                 }
2882                 break;
2883
2884         case NETDEV_UP:
2885         case NETDEV_CHANGE:
2886                 if (dev->flags & IFF_SLAVE)
2887                         break;
2888
2889                 if (event == NETDEV_UP) {
2890                         if (!addrconf_qdisc_ok(dev)) {
2891                                 /* device is not ready yet. */
2892                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2893                                         dev->name);
2894                                 break;
2895                         }
2896
2897                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2898                                 idev = ipv6_add_dev(dev);
2899
2900                         if (idev) {
2901                                 idev->if_flags |= IF_READY;
2902                                 run_pending = 1;
2903                         }
2904                 } else {
2905                         if (!addrconf_qdisc_ok(dev)) {
2906                                 /* device is still not ready. */
2907                                 break;
2908                         }
2909
2910                         if (idev) {
2911                                 if (idev->if_flags & IF_READY)
2912                                         /* device is already configured. */
2913                                         break;
2914                                 idev->if_flags |= IF_READY;
2915                         }
2916
2917                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2918                                 dev->name);
2919
2920                         run_pending = 1;
2921                 }
2922
2923                 switch (dev->type) {
2924 #if IS_ENABLED(CONFIG_IPV6_SIT)
2925                 case ARPHRD_SIT:
2926                         addrconf_sit_config(dev);
2927                         break;
2928 #endif
2929 #if IS_ENABLED(CONFIG_NET_IPGRE)
2930                 case ARPHRD_IPGRE:
2931                         addrconf_gre_config(dev);
2932                         break;
2933 #endif
2934                 case ARPHRD_TUNNEL6:
2935                         addrconf_ip6_tnl_config(dev);
2936                         break;
2937                 case ARPHRD_LOOPBACK:
2938                         init_loopback(dev);
2939                         break;
2940
2941                 default:
2942                         addrconf_dev_config(dev);
2943                         break;
2944                 }
2945
2946                 if (idev) {
2947                         if (run_pending)
2948                                 addrconf_dad_run(idev);
2949
2950                         /*
2951                          * If the MTU changed during the interface down,
2952                          * when the interface up, the changed MTU must be
2953                          * reflected in the idev as well as routers.
2954                          */
2955                         if (idev->cnf.mtu6 != dev->mtu &&
2956                             dev->mtu >= IPV6_MIN_MTU) {
2957                                 rt6_mtu_change(dev, dev->mtu);
2958                                 idev->cnf.mtu6 = dev->mtu;
2959                         }
2960                         idev->tstamp = jiffies;
2961                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2962
2963                         /*
2964                          * If the changed mtu during down is lower than
2965                          * IPV6_MIN_MTU stop IPv6 on this interface.
2966                          */
2967                         if (dev->mtu < IPV6_MIN_MTU)
2968                                 addrconf_ifdown(dev, 1);
2969                 }
2970                 break;
2971
2972         case NETDEV_CHANGEMTU:
2973                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2974                         rt6_mtu_change(dev, dev->mtu);
2975                         idev->cnf.mtu6 = dev->mtu;
2976                         break;
2977                 }
2978
2979                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2980                         idev = ipv6_add_dev(dev);
2981                         if (idev)
2982                                 break;
2983                 }
2984
2985                 /*
2986                  * MTU falled under IPV6_MIN_MTU.
2987                  * Stop IPv6 on this interface.
2988                  */
2989
2990         case NETDEV_DOWN:
2991         case NETDEV_UNREGISTER:
2992                 /*
2993                  *      Remove all addresses from this interface.
2994                  */
2995                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2996                 break;
2997
2998         case NETDEV_CHANGENAME:
2999                 if (idev) {
3000                         snmp6_unregister_dev(idev);
3001                         addrconf_sysctl_unregister(idev);
3002                         addrconf_sysctl_register(idev);
3003                         err = snmp6_register_dev(idev);
3004                         if (err)
3005                                 return notifier_from_errno(err);
3006                 }
3007                 break;
3008
3009         case NETDEV_PRE_TYPE_CHANGE:
3010         case NETDEV_POST_TYPE_CHANGE:
3011                 addrconf_type_change(dev, event);
3012                 break;
3013         }
3014
3015         return NOTIFY_OK;
3016 }
3017
3018 /*
3019  *      addrconf module should be notified of a device going up
3020  */
3021 static struct notifier_block ipv6_dev_notf = {
3022         .notifier_call = addrconf_notify,
3023 };
3024
3025 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3026 {
3027         struct inet6_dev *idev;
3028         ASSERT_RTNL();
3029
3030         idev = __in6_dev_get(dev);
3031
3032         if (event == NETDEV_POST_TYPE_CHANGE)
3033                 ipv6_mc_remap(idev);
3034         else if (event == NETDEV_PRE_TYPE_CHANGE)
3035                 ipv6_mc_unmap(idev);
3036 }
3037
3038 static int addrconf_ifdown(struct net_device *dev, int how)
3039 {
3040         struct net *net = dev_net(dev);
3041         struct inet6_dev *idev;
3042         struct inet6_ifaddr *ifa;
3043         int state, i;
3044
3045         ASSERT_RTNL();
3046
3047         rt6_ifdown(net, dev);
3048         neigh_ifdown(&nd_tbl, dev);
3049
3050         idev = __in6_dev_get(dev);
3051         if (idev == NULL)
3052                 return -ENODEV;
3053
3054         /*
3055          * Step 1: remove reference to ipv6 device from parent device.
3056          *         Do not dev_put!
3057          */
3058         if (how) {
3059                 idev->dead = 1;
3060
3061                 /* protected by rtnl_lock */
3062                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3063
3064                 /* Step 1.5: remove snmp6 entry */
3065                 snmp6_unregister_dev(idev);
3066
3067         }
3068
3069         /* Step 2: clear hash table */
3070         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3071                 struct hlist_head *h = &inet6_addr_lst[i];
3072
3073                 spin_lock_bh(&addrconf_hash_lock);
3074         restart:
3075                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3076                         if (ifa->idev == idev) {
3077                                 hlist_del_init_rcu(&ifa->addr_lst);
3078                                 addrconf_del_timer(ifa);
3079                                 goto restart;
3080                         }
3081                 }
3082                 spin_unlock_bh(&addrconf_hash_lock);
3083         }
3084
3085         write_lock_bh(&idev->lock);
3086
3087         /* Step 2: clear flags for stateless addrconf */
3088         if (!how)
3089                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3090
3091 #ifdef CONFIG_IPV6_PRIVACY
3092         if (how && del_timer(&idev->regen_timer))
3093                 in6_dev_put(idev);
3094
3095         /* Step 3: clear tempaddr list */
3096         while (!list_empty(&idev->tempaddr_list)) {
3097                 ifa = list_first_entry(&idev->tempaddr_list,
3098                                        struct inet6_ifaddr, tmp_list);
3099                 list_del(&ifa->tmp_list);
3100                 write_unlock_bh(&idev->lock);
3101                 spin_lock_bh(&ifa->lock);
3102
3103                 if (ifa->ifpub) {
3104                         in6_ifa_put(ifa->ifpub);
3105                         ifa->ifpub = NULL;
3106                 }
3107                 spin_unlock_bh(&ifa->lock);
3108                 in6_ifa_put(ifa);
3109                 write_lock_bh(&idev->lock);
3110         }
3111 #endif
3112
3113         while (!list_empty(&idev->addr_list)) {
3114                 ifa = list_first_entry(&idev->addr_list,
3115                                        struct inet6_ifaddr, if_list);
3116                 addrconf_del_timer(ifa);
3117
3118                 list_del(&ifa->if_list);
3119
3120                 write_unlock_bh(&idev->lock);
3121
3122                 spin_lock_bh(&ifa->state_lock);
3123                 state = ifa->state;
3124                 ifa->state = INET6_IFADDR_STATE_DEAD;
3125                 spin_unlock_bh(&ifa->state_lock);
3126
3127                 if (state != INET6_IFADDR_STATE_DEAD) {
3128                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3129                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3130                 }
3131                 in6_ifa_put(ifa);
3132
3133                 write_lock_bh(&idev->lock);
3134         }
3135
3136         write_unlock_bh(&idev->lock);
3137
3138         /* Step 5: Discard multicast list */
3139         if (how)
3140                 ipv6_mc_destroy_dev(idev);
3141         else
3142                 ipv6_mc_down(idev);
3143
3144         idev->tstamp = jiffies;
3145
3146         /* Last: Shot the device (if unregistered) */
3147         if (how) {
3148                 addrconf_sysctl_unregister(idev);
3149                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3150                 neigh_ifdown(&nd_tbl, dev);
3151                 in6_dev_put(idev);
3152         }
3153         return 0;
3154 }
3155
3156 static void addrconf_rs_timer(unsigned long data)
3157 {
3158         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3159         struct inet6_dev *idev = ifp->idev;
3160
3161         read_lock(&idev->lock);
3162         if (idev->dead || !(idev->if_flags & IF_READY))
3163                 goto out;
3164
3165         if (!ipv6_accept_ra(idev))
3166                 goto out;
3167
3168         /* Announcement received after solicitation was sent */
3169         if (idev->if_flags & IF_RA_RCVD)
3170                 goto out;
3171
3172         spin_lock(&ifp->lock);
3173         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3174                 /* The wait after the last probe can be shorter */
3175                 addrconf_mod_timer(ifp, AC_RS,
3176                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3177                                    idev->cnf.rtr_solicit_delay :
3178                                    idev->cnf.rtr_solicit_interval);
3179                 spin_unlock(&ifp->lock);
3180
3181                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3182         } else {
3183                 spin_unlock(&ifp->lock);
3184                 /*
3185                  * Note: we do not support deprecated "all on-link"
3186                  * assumption any longer.
3187                  */
3188                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3189         }
3190
3191 out:
3192         read_unlock(&idev->lock);
3193         in6_ifa_put(ifp);
3194 }
3195
3196 /*
3197  *      Duplicate Address Detection
3198  */
3199 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3200 {
3201         unsigned long rand_num;
3202         struct inet6_dev *idev = ifp->idev;
3203
3204         if (ifp->flags & IFA_F_OPTIMISTIC)
3205                 rand_num = 0;
3206         else
3207                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3208
3209         ifp->probes = idev->cnf.dad_transmits;
3210         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3211 }
3212
3213 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3214 {
3215         struct inet6_dev *idev = ifp->idev;
3216         struct net_device *dev = idev->dev;
3217
3218         addrconf_join_solict(dev, &ifp->addr);
3219
3220         net_srandom(ifp->addr.s6_addr32[3]);
3221
3222         read_lock_bh(&idev->lock);
3223         spin_lock(&ifp->lock);
3224         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3225                 goto out;
3226
3227         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3228             idev->cnf.accept_dad < 1 ||
3229             !(ifp->flags&IFA_F_TENTATIVE) ||
3230             ifp->flags & IFA_F_NODAD) {
3231                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3232                 spin_unlock(&ifp->lock);
3233                 read_unlock_bh(&idev->lock);
3234
3235                 addrconf_dad_completed(ifp);
3236                 return;
3237         }
3238
3239         if (!(idev->if_flags & IF_READY)) {
3240                 spin_unlock(&ifp->lock);
3241                 read_unlock_bh(&idev->lock);
3242                 /*
3243                  * If the device is not ready:
3244                  * - keep it tentative if it is a permanent address.
3245                  * - otherwise, kill it.
3246                  */
3247                 in6_ifa_hold(ifp);
3248                 addrconf_dad_stop(ifp, 0);
3249                 return;
3250         }
3251
3252         /*
3253          * Optimistic nodes can start receiving
3254          * Frames right away
3255          */
3256         if (ifp->flags & IFA_F_OPTIMISTIC)
3257                 ip6_ins_rt(ifp->rt);
3258
3259         addrconf_dad_kick(ifp);
3260 out:
3261         spin_unlock(&ifp->lock);
3262         read_unlock_bh(&idev->lock);
3263 }
3264
3265 static void addrconf_dad_timer(unsigned long data)
3266 {
3267         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3268         struct inet6_dev *idev = ifp->idev;
3269         struct in6_addr mcaddr;
3270
3271         if (!ifp->probes && addrconf_dad_end(ifp))
3272                 goto out;
3273
3274         read_lock(&idev->lock);
3275         if (idev->dead || !(idev->if_flags & IF_READY)) {
3276                 read_unlock(&idev->lock);
3277                 goto out;
3278         }
3279
3280         spin_lock(&ifp->lock);
3281         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3282                 spin_unlock(&ifp->lock);
3283                 read_unlock(&idev->lock);
3284                 goto out;
3285         }
3286
3287         if (ifp->probes == 0) {
3288                 /*
3289                  * DAD was successful
3290                  */
3291
3292                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3293                 spin_unlock(&ifp->lock);
3294                 read_unlock(&idev->lock);
3295
3296                 addrconf_dad_completed(ifp);
3297
3298                 goto out;
3299         }
3300
3301         ifp->probes--;
3302         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3303         spin_unlock(&ifp->lock);
3304         read_unlock(&idev->lock);
3305
3306         /* send a neighbour solicitation for our addr */
3307         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3308         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3309 out:
3310         in6_ifa_put(ifp);
3311 }
3312
3313 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3314 {
3315         struct net_device *dev = ifp->idev->dev;
3316
3317         /*
3318          *      Configure the address for reception. Now it is valid.
3319          */
3320
3321         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3322
3323         /* If added prefix is link local and we are prepared to process
3324            router advertisements, start sending router solicitations.
3325          */
3326
3327         if (ipv6_accept_ra(ifp->idev) &&
3328             ifp->idev->cnf.rtr_solicits > 0 &&
3329             (dev->flags&IFF_LOOPBACK) == 0 &&
3330             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3331                 /*
3332                  *      If a host as already performed a random delay
3333                  *      [...] as part of DAD [...] there is no need
3334                  *      to delay again before sending the first RS
3335                  */
3336                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3337
3338                 spin_lock_bh(&ifp->lock);
3339                 ifp->probes = 1;
3340                 ifp->idev->if_flags |= IF_RS_SENT;
3341                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3342                 spin_unlock_bh(&ifp->lock);
3343         }
3344 }
3345
3346 static void addrconf_dad_run(struct inet6_dev *idev)
3347 {
3348         struct inet6_ifaddr *ifp;
3349
3350         read_lock_bh(&idev->lock);
3351         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3352                 spin_lock(&ifp->lock);
3353                 if (ifp->flags & IFA_F_TENTATIVE &&
3354                     ifp->state == INET6_IFADDR_STATE_DAD)
3355                         addrconf_dad_kick(ifp);
3356                 spin_unlock(&ifp->lock);
3357         }
3358         read_unlock_bh(&idev->lock);
3359 }
3360
3361 #ifdef CONFIG_PROC_FS
3362 struct if6_iter_state {
3363         struct seq_net_private p;
3364         int bucket;
3365         int offset;
3366 };
3367
3368 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3369 {
3370         struct inet6_ifaddr *ifa = NULL;
3371         struct if6_iter_state *state = seq->private;
3372         struct net *net = seq_file_net(seq);
3373         int p = 0;
3374
3375         /* initial bucket if pos is 0 */
3376         if (pos == 0) {
3377                 state->bucket = 0;
3378                 state->offset = 0;
3379         }
3380
3381         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3382                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3383                                          addr_lst) {
3384                         if (!net_eq(dev_net(ifa->idev->dev), net))
3385                                 continue;
3386                         /* sync with offset */
3387                         if (p < state->offset) {
3388                                 p++;
3389                                 continue;
3390                         }
3391                         state->offset++;
3392                         return ifa;
3393                 }
3394
3395                 /* prepare for next bucket */
3396                 state->offset = 0;
3397                 p = 0;
3398         }
3399         return NULL;
3400 }
3401
3402 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3403                                          struct inet6_ifaddr *ifa)
3404 {
3405         struct if6_iter_state *state = seq->private;
3406         struct net *net = seq_file_net(seq);
3407
3408         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3409                 if (!net_eq(dev_net(ifa->idev->dev), net))
3410                         continue;
3411                 state->offset++;
3412                 return ifa;
3413         }
3414
3415         while (++state->bucket < IN6_ADDR_HSIZE) {
3416                 state->offset = 0;
3417                 hlist_for_each_entry_rcu_bh(ifa,
3418                                      &inet6_addr_lst[state->bucket], addr_lst) {
3419                         if (!net_eq(dev_net(ifa->idev->dev), net))
3420                                 continue;
3421                         state->offset++;
3422                         return ifa;
3423                 }
3424         }
3425
3426         return NULL;
3427 }
3428
3429 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3430         __acquires(rcu_bh)
3431 {
3432         rcu_read_lock_bh();
3433         return if6_get_first(seq, *pos);
3434 }
3435
3436 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3437 {
3438         struct inet6_ifaddr *ifa;
3439
3440         ifa = if6_get_next(seq, v);
3441         ++*pos;
3442         return ifa;
3443 }
3444
3445 static void if6_seq_stop(struct seq_file *seq, void *v)
3446         __releases(rcu_bh)
3447 {
3448         rcu_read_unlock_bh();
3449 }
3450
3451 static int if6_seq_show(struct seq_file *seq, void *v)
3452 {
3453         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3454         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3455                    &ifp->addr,
3456                    ifp->idev->dev->ifindex,
3457                    ifp->prefix_len,
3458                    ifp->scope,
3459                    ifp->flags,
3460                    ifp->idev->dev->name);
3461         return 0;
3462 }
3463
3464 static const struct seq_operations if6_seq_ops = {
3465         .start  = if6_seq_start,
3466         .next   = if6_seq_next,
3467         .show   = if6_seq_show,
3468         .stop   = if6_seq_stop,
3469 };
3470
3471 static int if6_seq_open(struct inode *inode, struct file *file)
3472 {
3473         return seq_open_net(inode, file, &if6_seq_ops,
3474                             sizeof(struct if6_iter_state));
3475 }
3476
3477 static const struct file_operations if6_fops = {
3478         .owner          = THIS_MODULE,
3479         .open           = if6_seq_open,
3480         .read           = seq_read,
3481         .llseek         = seq_lseek,
3482         .release        = seq_release_net,
3483 };
3484
3485 static int __net_init if6_proc_net_init(struct net *net)
3486 {
3487         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3488                 return -ENOMEM;
3489         return 0;
3490 }
3491
3492 static void __net_exit if6_proc_net_exit(struct net *net)
3493 {
3494         remove_proc_entry("if_inet6", net->proc_net);
3495 }
3496
3497 static struct pernet_operations if6_proc_net_ops = {
3498        .init = if6_proc_net_init,
3499        .exit = if6_proc_net_exit,
3500 };
3501
3502 int __init if6_proc_init(void)
3503 {
3504         return register_pernet_subsys(&if6_proc_net_ops);
3505 }
3506
3507 void if6_proc_exit(void)
3508 {
3509         unregister_pernet_subsys(&if6_proc_net_ops);
3510 }
3511 #endif  /* CONFIG_PROC_FS */
3512
3513 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3514 /* Check if address is a home address configured on any interface. */
3515 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3516 {
3517         int ret = 0;
3518         struct inet6_ifaddr *ifp = NULL;
3519         unsigned int hash = inet6_addr_hash(addr);
3520
3521         rcu_read_lock_bh();
3522         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3523                 if (!net_eq(dev_net(ifp->idev->dev), net))
3524                         continue;
3525                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3526                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3527                         ret = 1;
3528                         break;
3529                 }
3530         }
3531         rcu_read_unlock_bh();
3532         return ret;
3533 }
3534 #endif
3535
3536 /*
3537  *      Periodic address status verification
3538  */
3539
3540 static void addrconf_verify(unsigned long foo)
3541 {
3542         unsigned long now, next, next_sec, next_sched;
3543         struct inet6_ifaddr *ifp;
3544         int i;
3545
3546         rcu_read_lock_bh();
3547         spin_lock(&addrconf_verify_lock);
3548         now = jiffies;
3549         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3550
3551         del_timer(&addr_chk_timer);
3552
3553         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3554 restart:
3555                 hlist_for_each_entry_rcu_bh(ifp,
3556                                          &inet6_addr_lst[i], addr_lst) {
3557                         unsigned long age;
3558
3559                         if (ifp->flags & IFA_F_PERMANENT)
3560                                 continue;
3561
3562                         spin_lock(&ifp->lock);
3563                         /* We try to batch several events at once. */
3564                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3565
3566                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3567                             age >= ifp->valid_lft) {
3568                                 spin_unlock(&ifp->lock);
3569                                 in6_ifa_hold(ifp);
3570                                 ipv6_del_addr(ifp);
3571                                 goto restart;
3572                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3573                                 spin_unlock(&ifp->lock);
3574                                 continue;
3575                         } else if (age >= ifp->prefered_lft) {
3576                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3577                                 int deprecate = 0;
3578
3579                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3580                                         deprecate = 1;
3581                                         ifp->flags |= IFA_F_DEPRECATED;
3582                                 }
3583
3584                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3585                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3586
3587                                 spin_unlock(&ifp->lock);
3588
3589                                 if (deprecate) {
3590                                         in6_ifa_hold(ifp);
3591
3592                                         ipv6_ifa_notify(0, ifp);
3593                                         in6_ifa_put(ifp);
3594                                         goto restart;
3595                                 }
3596 #ifdef CONFIG_IPV6_PRIVACY
3597                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3598                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3599                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3600                                         ifp->idev->cnf.dad_transmits *
3601                                         ifp->idev->nd_parms->retrans_time / HZ;
3602
3603                                 if (age >= ifp->prefered_lft - regen_advance) {
3604                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3605                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3606                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3607                                         if (!ifp->regen_count && ifpub) {
3608                                                 ifp->regen_count++;
3609                                                 in6_ifa_hold(ifp);
3610                                                 in6_ifa_hold(ifpub);
3611                                                 spin_unlock(&ifp->lock);
3612
3613                                                 spin_lock(&ifpub->lock);
3614                                                 ifpub->regen_count = 0;
3615                                                 spin_unlock(&ifpub->lock);
3616                                                 ipv6_create_tempaddr(ifpub, ifp);
3617                                                 in6_ifa_put(ifpub);
3618                                                 in6_ifa_put(ifp);
3619                                                 goto restart;
3620                                         }
3621                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3622                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3623                                 spin_unlock(&ifp->lock);
3624 #endif
3625                         } else {
3626                                 /* ifp->prefered_lft <= ifp->valid_lft */
3627                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3628                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3629                                 spin_unlock(&ifp->lock);
3630                         }
3631                 }
3632         }
3633
3634         next_sec = round_jiffies_up(next);
3635         next_sched = next;
3636
3637         /* If rounded timeout is accurate enough, accept it. */
3638         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3639                 next_sched = next_sec;
3640
3641         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3642         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3643                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3644
3645         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3646               now, next, next_sec, next_sched));
3647
3648         addr_chk_timer.expires = next_sched;
3649         add_timer(&addr_chk_timer);
3650         spin_unlock(&addrconf_verify_lock);
3651         rcu_read_unlock_bh();
3652 }
3653
3654 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3655 {
3656         struct in6_addr *pfx = NULL;
3657
3658         if (addr)
3659                 pfx = nla_data(addr);
3660
3661         if (local) {
3662                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3663                         pfx = NULL;
3664                 else
3665                         pfx = nla_data(local);
3666         }
3667
3668         return pfx;
3669 }
3670
3671 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3672         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3673         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3674         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3675 };
3676
3677 static int
3678 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3679 {
3680         struct net *net = sock_net(skb->sk);
3681         struct ifaddrmsg *ifm;
3682         struct nlattr *tb[IFA_MAX+1];
3683         struct in6_addr *pfx;
3684         int err;
3685
3686         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3687         if (err < 0)
3688                 return err;
3689
3690         ifm = nlmsg_data(nlh);
3691         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3692         if (pfx == NULL)
3693                 return -EINVAL;
3694
3695         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3696 }
3697
3698 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3699                              u32 prefered_lft, u32 valid_lft)
3700 {
3701         u32 flags;
3702         clock_t expires;
3703         unsigned long timeout;
3704
3705         if (!valid_lft || (prefered_lft > valid_lft))
3706                 return -EINVAL;
3707
3708         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3709         if (addrconf_finite_timeout(timeout)) {
3710                 expires = jiffies_to_clock_t(timeout * HZ);
3711                 valid_lft = timeout;
3712                 flags = RTF_EXPIRES;
3713         } else {
3714                 expires = 0;
3715                 flags = 0;
3716                 ifa_flags |= IFA_F_PERMANENT;
3717         }
3718
3719         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3720         if (addrconf_finite_timeout(timeout)) {
3721                 if (timeout == 0)
3722                         ifa_flags |= IFA_F_DEPRECATED;
3723                 prefered_lft = timeout;
3724         }
3725
3726         spin_lock_bh(&ifp->lock);
3727         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3728         ifp->tstamp = jiffies;
3729         ifp->valid_lft = valid_lft;
3730         ifp->prefered_lft = prefered_lft;
3731
3732         spin_unlock_bh(&ifp->lock);
3733         if (!(ifp->flags&IFA_F_TENTATIVE))
3734                 ipv6_ifa_notify(0, ifp);
3735
3736         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3737                               expires, flags);
3738         addrconf_verify(0);
3739
3740         return 0;
3741 }
3742
3743 static int
3744 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3745 {
3746         struct net *net = sock_net(skb->sk);
3747         struct ifaddrmsg *ifm;
3748         struct nlattr *tb[IFA_MAX+1];
3749         struct in6_addr *pfx;
3750         struct inet6_ifaddr *ifa;
3751         struct net_device *dev;
3752         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3753         u8 ifa_flags;
3754         int err;
3755
3756         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3757         if (err < 0)
3758                 return err;
3759
3760         ifm = nlmsg_data(nlh);
3761         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3762         if (pfx == NULL)
3763                 return -EINVAL;
3764
3765         if (tb[IFA_CACHEINFO]) {
3766                 struct ifa_cacheinfo *ci;
3767
3768                 ci = nla_data(tb[IFA_CACHEINFO]);
3769                 valid_lft = ci->ifa_valid;
3770                 preferred_lft = ci->ifa_prefered;
3771         } else {
3772                 preferred_lft = INFINITY_LIFE_TIME;
3773                 valid_lft = INFINITY_LIFE_TIME;
3774         }
3775
3776         dev =  __dev_get_by_index(net, ifm->ifa_index);
3777         if (dev == NULL)
3778                 return -ENODEV;
3779
3780         /* We ignore other flags so far. */
3781         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3782
3783         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3784         if (ifa == NULL) {
3785                 /*
3786                  * It would be best to check for !NLM_F_CREATE here but
3787                  * userspace alreay relies on not having to provide this.
3788                  */
3789                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3790                                       ifm->ifa_prefixlen, ifa_flags,
3791                                       preferred_lft, valid_lft);
3792         }
3793
3794         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3795             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3796                 err = -EEXIST;
3797         else
3798                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3799
3800         in6_ifa_put(ifa);
3801
3802         return err;
3803 }
3804
3805 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3806                           u8 scope, int ifindex)
3807 {
3808         struct ifaddrmsg *ifm;
3809
3810         ifm = nlmsg_data(nlh);
3811         ifm->ifa_family = AF_INET6;
3812         ifm->ifa_prefixlen = prefixlen;
3813         ifm->ifa_flags = flags;
3814         ifm->ifa_scope = scope;
3815         ifm->ifa_index = ifindex;
3816 }
3817
3818 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3819                          unsigned long tstamp, u32 preferred, u32 valid)
3820 {
3821         struct ifa_cacheinfo ci;
3822
3823         ci.cstamp = cstamp_delta(cstamp);
3824         ci.tstamp = cstamp_delta(tstamp);
3825         ci.ifa_prefered = preferred;
3826         ci.ifa_valid = valid;
3827
3828         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3829 }
3830
3831 static inline int rt_scope(int ifa_scope)
3832 {
3833         if (ifa_scope & IFA_HOST)
3834                 return RT_SCOPE_HOST;
3835         else if (ifa_scope & IFA_LINK)
3836                 return RT_SCOPE_LINK;
3837         else if (ifa_scope & IFA_SITE)
3838                 return RT_SCOPE_SITE;
3839         else
3840                 return RT_SCOPE_UNIVERSE;
3841 }
3842
3843 static inline int inet6_ifaddr_msgsize(void)
3844 {
3845         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3846                + nla_total_size(16) /* IFA_ADDRESS */
3847                + nla_total_size(sizeof(struct ifa_cacheinfo));
3848 }
3849
3850 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3851                              u32 portid, u32 seq, int event, unsigned int flags)
3852 {
3853         struct nlmsghdr  *nlh;
3854         u32 preferred, valid;
3855
3856         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3857         if (nlh == NULL)
3858                 return -EMSGSIZE;
3859
3860         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3861                       ifa->idev->dev->ifindex);
3862
3863         if (!(ifa->flags&IFA_F_PERMANENT)) {
3864                 preferred = ifa->prefered_lft;
3865                 valid = ifa->valid_lft;
3866                 if (preferred != INFINITY_LIFE_TIME) {
3867                         long tval = (jiffies - ifa->tstamp)/HZ;
3868                         if (preferred > tval)
3869                                 preferred -= tval;
3870                         else
3871                                 preferred = 0;
3872                         if (valid != INFINITY_LIFE_TIME) {
3873                                 if (valid > tval)
3874                                         valid -= tval;
3875                                 else
3876                                         valid = 0;
3877                         }
3878                 }
3879         } else {
3880                 preferred = INFINITY_LIFE_TIME;
3881                 valid = INFINITY_LIFE_TIME;
3882         }
3883
3884         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3885             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3886                 nlmsg_cancel(skb, nlh);
3887                 return -EMSGSIZE;
3888         }
3889
3890         return nlmsg_end(skb, nlh);
3891 }
3892
3893 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3894                                 u32 portid, u32 seq, int event, u16 flags)
3895 {
3896         struct nlmsghdr  *nlh;
3897         u8 scope = RT_SCOPE_UNIVERSE;
3898         int ifindex = ifmca->idev->dev->ifindex;
3899
3900         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3901                 scope = RT_SCOPE_SITE;
3902
3903         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3904         if (nlh == NULL)
3905                 return -EMSGSIZE;
3906
3907         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3908         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3909             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3910                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3911                 nlmsg_cancel(skb, nlh);
3912                 return -EMSGSIZE;
3913         }
3914
3915         return nlmsg_end(skb, nlh);
3916 }
3917
3918 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3919                                 u32 portid, u32 seq, int event, unsigned int flags)
3920 {
3921         struct nlmsghdr  *nlh;
3922         u8 scope = RT_SCOPE_UNIVERSE;
3923         int ifindex = ifaca->aca_idev->dev->ifindex;
3924
3925         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3926                 scope = RT_SCOPE_SITE;
3927
3928         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3929         if (nlh == NULL)
3930                 return -EMSGSIZE;
3931
3932         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3933         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3934             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3935                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3936                 nlmsg_cancel(skb, nlh);
3937                 return -EMSGSIZE;
3938         }
3939
3940         return nlmsg_end(skb, nlh);
3941 }
3942
3943 enum addr_type_t {
3944         UNICAST_ADDR,
3945         MULTICAST_ADDR,
3946         ANYCAST_ADDR,
3947 };
3948
3949 /* called with rcu_read_lock() */
3950 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3951                           struct netlink_callback *cb, enum addr_type_t type,
3952                           int s_ip_idx, int *p_ip_idx)
3953 {
3954         struct ifmcaddr6 *ifmca;
3955         struct ifacaddr6 *ifaca;
3956         int err = 1;
3957         int ip_idx = *p_ip_idx;
3958
3959         read_lock_bh(&idev->lock);
3960         switch (type) {
3961         case UNICAST_ADDR: {
3962                 struct inet6_ifaddr *ifa;
3963
3964                 /* unicast address incl. temp addr */
3965                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3966                         if (++ip_idx < s_ip_idx)
3967                                 continue;
3968                         err = inet6_fill_ifaddr(skb, ifa,
3969                                                 NETLINK_CB(cb->skb).portid,
3970                                                 cb->nlh->nlmsg_seq,
3971                                                 RTM_NEWADDR,
3972                                                 NLM_F_MULTI);
3973                         if (err <= 0)
3974                                 break;
3975                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3976                 }
3977                 break;
3978         }
3979         case MULTICAST_ADDR:
3980                 /* multicast address */
3981                 for (ifmca = idev->mc_list; ifmca;
3982                      ifmca = ifmca->next, ip_idx++) {
3983                         if (ip_idx < s_ip_idx)
3984                                 continue;
3985                         err = inet6_fill_ifmcaddr(skb, ifmca,
3986                                                   NETLINK_CB(cb->skb).portid,
3987                                                   cb->nlh->nlmsg_seq,
3988                                                   RTM_GETMULTICAST,
3989                                                   NLM_F_MULTI);
3990                         if (err <= 0)
3991                                 break;
3992                 }
3993                 break;
3994         case ANYCAST_ADDR:
3995                 /* anycast address */
3996                 for (ifaca = idev->ac_list; ifaca;
3997                      ifaca = ifaca->aca_next, ip_idx++) {
3998                         if (ip_idx < s_ip_idx)
3999                                 continue;
4000                         err = inet6_fill_ifacaddr(skb, ifaca,
4001                                                   NETLINK_CB(cb->skb).portid,
4002                                                   cb->nlh->nlmsg_seq,
4003                                                   RTM_GETANYCAST,
4004                                                   NLM_F_MULTI);
4005                         if (err <= 0)
4006                                 break;
4007                 }
4008                 break;
4009         default:
4010                 break;
4011         }
4012         read_unlock_bh(&idev->lock);
4013         *p_ip_idx = ip_idx;
4014         return err;
4015 }
4016
4017 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4018                            enum addr_type_t type)
4019 {
4020         struct net *net = sock_net(skb->sk);
4021         int h, s_h;
4022         int idx, ip_idx;
4023         int s_idx, s_ip_idx;
4024         struct net_device *dev;
4025         struct inet6_dev *idev;
4026         struct hlist_head *head;
4027
4028         s_h = cb->args[0];
4029         s_idx = idx = cb->args[1];
4030         s_ip_idx = ip_idx = cb->args[2];
4031
4032         rcu_read_lock();
4033         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4034         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4035                 idx = 0;
4036                 head = &net->dev_index_head[h];
4037                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4038                         if (idx < s_idx)
4039                                 goto cont;
4040                         if (h > s_h || idx > s_idx)
4041                                 s_ip_idx = 0;
4042                         ip_idx = 0;
4043                         idev = __in6_dev_get(dev);
4044                         if (!idev)
4045                                 goto cont;
4046
4047                         if (in6_dump_addrs(idev, skb, cb, type,
4048                                            s_ip_idx, &ip_idx) <= 0)
4049                                 goto done;
4050 cont:
4051                         idx++;
4052                 }
4053         }
4054 done:
4055         rcu_read_unlock();
4056         cb->args[0] = h;
4057         cb->args[1] = idx;
4058         cb->args[2] = ip_idx;
4059
4060         return skb->len;
4061 }
4062
4063 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4064 {
4065         enum addr_type_t type = UNICAST_ADDR;
4066
4067         return inet6_dump_addr(skb, cb, type);
4068 }
4069
4070 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4071 {
4072         enum addr_type_t type = MULTICAST_ADDR;
4073
4074         return inet6_dump_addr(skb, cb, type);
4075 }
4076
4077
4078 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4079 {
4080         enum addr_type_t type = ANYCAST_ADDR;
4081
4082         return inet6_dump_addr(skb, cb, type);
4083 }
4084
4085 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4086 {
4087         struct net *net = sock_net(in_skb->sk);
4088         struct ifaddrmsg *ifm;
4089         struct nlattr *tb[IFA_MAX+1];
4090         struct in6_addr *addr = NULL;
4091         struct net_device *dev = NULL;
4092         struct inet6_ifaddr *ifa;
4093         struct sk_buff *skb;
4094         int err;
4095
4096         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4097         if (err < 0)
4098                 goto errout;
4099
4100         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4101         if (addr == NULL) {
4102                 err = -EINVAL;
4103                 goto errout;
4104         }
4105
4106         ifm = nlmsg_data(nlh);
4107         if (ifm->ifa_index)
4108                 dev = __dev_get_by_index(net, ifm->ifa_index);
4109
4110         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4111         if (!ifa) {
4112                 err = -EADDRNOTAVAIL;
4113                 goto errout;
4114         }
4115
4116         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4117         if (!skb) {
4118                 err = -ENOBUFS;
4119                 goto errout_ifa;
4120         }
4121
4122         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4123                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4124         if (err < 0) {
4125                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4126                 WARN_ON(err == -EMSGSIZE);
4127                 kfree_skb(skb);
4128                 goto errout_ifa;
4129         }
4130         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4131 errout_ifa:
4132         in6_ifa_put(ifa);
4133 errout:
4134         return err;
4135 }
4136
4137 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4138 {
4139         struct sk_buff *skb;
4140         struct net *net = dev_net(ifa->idev->dev);
4141         int err = -ENOBUFS;
4142
4143         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4144         if (skb == NULL)
4145                 goto errout;
4146
4147         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4148         if (err < 0) {
4149                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4150                 WARN_ON(err == -EMSGSIZE);
4151                 kfree_skb(skb);
4152                 goto errout;
4153         }
4154         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4155         return;
4156 errout:
4157         if (err < 0)
4158                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4159 }
4160
4161 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4162                                 __s32 *array, int bytes)
4163 {
4164         BUG_ON(bytes < (DEVCONF_MAX * 4));
4165
4166         memset(array, 0, bytes);
4167         array[DEVCONF_FORWARDING] = cnf->forwarding;
4168         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4169         array[DEVCONF_MTU6] = cnf->mtu6;
4170         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4171         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4172         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4173         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4174         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4175         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4176                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4177         array[DEVCONF_RTR_SOLICIT_DELAY] =
4178                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4179         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4180 #ifdef CONFIG_IPV6_PRIVACY
4181         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4182         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4183         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4184         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4185         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4186 #endif
4187         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4188         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4189         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4190 #ifdef CONFIG_IPV6_ROUTER_PREF
4191         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4192         array[DEVCONF_RTR_PROBE_INTERVAL] =
4193                 jiffies_to_msecs(cnf->rtr_probe_interval);
4194 #ifdef CONFIG_IPV6_ROUTE_INFO
4195         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4196 #endif
4197 #endif
4198         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4199         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4200 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4201         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4202 #endif
4203 #ifdef CONFIG_IPV6_MROUTE
4204         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4205 #endif
4206         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4207         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4208         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4209         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4210 }
4211
4212 static inline size_t inet6_ifla6_size(void)
4213 {
4214         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4215              + nla_total_size(sizeof(struct ifla_cacheinfo))
4216              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4217              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4218              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4219              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4220 }
4221
4222 static inline size_t inet6_if_nlmsg_size(void)
4223 {
4224         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4225                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4226                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4227                + nla_total_size(4) /* IFLA_MTU */
4228                + nla_total_size(4) /* IFLA_LINK */
4229                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4230 }
4231
4232 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4233                                       int items, int bytes)
4234 {
4235         int i;
4236         int pad = bytes - sizeof(u64) * items;
4237         BUG_ON(pad < 0);
4238
4239         /* Use put_unaligned() because stats may not be aligned for u64. */
4240         put_unaligned(items, &stats[0]);
4241         for (i = 1; i < items; i++)
4242                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4243
4244         memset(&stats[items], 0, pad);
4245 }
4246
4247 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4248                                       int items, int bytes, size_t syncpoff)
4249 {
4250         int i;
4251         int pad = bytes - sizeof(u64) * items;
4252         BUG_ON(pad < 0);
4253
4254         /* Use put_unaligned() because stats may not be aligned for u64. */
4255         put_unaligned(items, &stats[0]);
4256         for (i = 1; i < items; i++)
4257                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4258
4259         memset(&stats[items], 0, pad);
4260 }
4261
4262 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4263                              int bytes)
4264 {
4265         switch (attrtype) {
4266         case IFLA_INET6_STATS:
4267                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4268                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4269                 break;
4270         case IFLA_INET6_ICMP6STATS:
4271                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4272                 break;
4273         }
4274 }
4275
4276 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4277 {
4278         struct nlattr *nla;
4279         struct ifla_cacheinfo ci;
4280
4281         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4282                 goto nla_put_failure;
4283         ci.max_reasm_len = IPV6_MAXPLEN;
4284         ci.tstamp = cstamp_delta(idev->tstamp);
4285         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4286         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4287         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4288                 goto nla_put_failure;
4289         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4290         if (nla == NULL)
4291                 goto nla_put_failure;
4292         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4293
4294         /* XXX - MC not implemented */
4295
4296         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4297         if (nla == NULL)
4298                 goto nla_put_failure;
4299         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4300
4301         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4302         if (nla == NULL)
4303                 goto nla_put_failure;
4304         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4305
4306         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4307         if (nla == NULL)
4308                 goto nla_put_failure;
4309         read_lock_bh(&idev->lock);
4310         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4311         read_unlock_bh(&idev->lock);
4312
4313         return 0;
4314
4315 nla_put_failure:
4316         return -EMSGSIZE;
4317 }
4318
4319 static size_t inet6_get_link_af_size(const struct net_device *dev)
4320 {
4321         if (!__in6_dev_get(dev))
4322                 return 0;
4323
4324         return inet6_ifla6_size();
4325 }
4326
4327 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4328 {
4329         struct inet6_dev *idev = __in6_dev_get(dev);
4330
4331         if (!idev)
4332                 return -ENODATA;
4333
4334         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4335                 return -EMSGSIZE;
4336
4337         return 0;
4338 }
4339
4340 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4341 {
4342         struct inet6_ifaddr *ifp;
4343         struct net_device *dev = idev->dev;
4344         bool update_rs = false;
4345         struct in6_addr ll_addr;
4346
4347         if (token == NULL)
4348                 return -EINVAL;
4349         if (ipv6_addr_any(token))
4350                 return -EINVAL;
4351         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4352                 return -EINVAL;
4353         if (!ipv6_accept_ra(idev))
4354                 return -EINVAL;
4355         if (idev->cnf.rtr_solicits <= 0)
4356                 return -EINVAL;
4357
4358         write_lock_bh(&idev->lock);
4359
4360         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4361         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4362
4363         write_unlock_bh(&idev->lock);
4364
4365         if (!idev->dead && (idev->if_flags & IF_READY) &&
4366             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4367                              IFA_F_OPTIMISTIC)) {
4368
4369                 /* If we're not ready, then normal ifup will take care
4370                  * of this. Otherwise, we need to request our rs here.
4371                  */
4372                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4373                 update_rs = true;
4374         }
4375
4376         write_lock_bh(&idev->lock);
4377
4378         if (update_rs)
4379                 idev->if_flags |= IF_RS_SENT;
4380
4381         /* Well, that's kinda nasty ... */
4382         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4383                 spin_lock(&ifp->lock);
4384                 if (ifp->tokenized) {
4385                         ifp->valid_lft = 0;
4386                         ifp->prefered_lft = 0;
4387                 }
4388                 spin_unlock(&ifp->lock);
4389         }
4390
4391         write_unlock_bh(&idev->lock);
4392         return 0;
4393 }
4394
4395 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4396 {
4397         int err = -EINVAL;
4398         struct inet6_dev *idev = __in6_dev_get(dev);
4399         struct nlattr *tb[IFLA_INET6_MAX + 1];
4400
4401         if (!idev)
4402                 return -EAFNOSUPPORT;
4403
4404         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4405                 BUG();
4406
4407         if (tb[IFLA_INET6_TOKEN])
4408                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4409
4410         return err;
4411 }
4412
4413 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4414                              u32 portid, u32 seq, int event, unsigned int flags)
4415 {
4416         struct net_device *dev = idev->dev;
4417         struct ifinfomsg *hdr;
4418         struct nlmsghdr *nlh;
4419         void *protoinfo;
4420
4421         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4422         if (nlh == NULL)
4423                 return -EMSGSIZE;
4424
4425         hdr = nlmsg_data(nlh);
4426         hdr->ifi_family = AF_INET6;
4427         hdr->__ifi_pad = 0;
4428         hdr->ifi_type = dev->type;
4429         hdr->ifi_index = dev->ifindex;
4430         hdr->ifi_flags = dev_get_flags(dev);
4431         hdr->ifi_change = 0;
4432
4433         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4434             (dev->addr_len &&
4435              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4436             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4437             (dev->ifindex != dev->iflink &&
4438              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4439                 goto nla_put_failure;
4440         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4441         if (protoinfo == NULL)
4442                 goto nla_put_failure;
4443
4444         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4445                 goto nla_put_failure;
4446
4447         nla_nest_end(skb, protoinfo);
4448         return nlmsg_end(skb, nlh);
4449
4450 nla_put_failure:
4451         nlmsg_cancel(skb, nlh);
4452         return -EMSGSIZE;
4453 }
4454
4455 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4456 {
4457         struct net *net = sock_net(skb->sk);
4458         int h, s_h;
4459         int idx = 0, s_idx;
4460         struct net_device *dev;
4461         struct inet6_dev *idev;
4462         struct hlist_head *head;
4463
4464         s_h = cb->args[0];
4465         s_idx = cb->args[1];
4466
4467         rcu_read_lock();
4468         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4469                 idx = 0;
4470                 head = &net->dev_index_head[h];
4471                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4472                         if (idx < s_idx)
4473                                 goto cont;
4474                         idev = __in6_dev_get(dev);
4475                         if (!idev)
4476                                 goto cont;
4477                         if (inet6_fill_ifinfo(skb, idev,
4478                                               NETLINK_CB(cb->skb).portid,
4479                                               cb->nlh->nlmsg_seq,
4480                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4481                                 goto out;
4482 cont:
4483                         idx++;
4484                 }
4485         }
4486 out:
4487         rcu_read_unlock();
4488         cb->args[1] = idx;
4489         cb->args[0] = h;
4490
4491         return skb->len;
4492 }
4493
4494 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4495 {
4496         struct sk_buff *skb;
4497         struct net *net = dev_net(idev->dev);
4498         int err = -ENOBUFS;
4499
4500         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4501         if (skb == NULL)
4502                 goto errout;
4503
4504         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4505         if (err < 0) {
4506                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4507                 WARN_ON(err == -EMSGSIZE);
4508                 kfree_skb(skb);
4509                 goto errout;
4510         }
4511         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4512         return;
4513 errout:
4514         if (err < 0)
4515                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4516 }
4517
4518 static inline size_t inet6_prefix_nlmsg_size(void)
4519 {
4520         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4521                + nla_total_size(sizeof(struct in6_addr))
4522                + nla_total_size(sizeof(struct prefix_cacheinfo));
4523 }
4524
4525 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4526                              struct prefix_info *pinfo, u32 portid, u32 seq,
4527                              int event, unsigned int flags)
4528 {
4529         struct prefixmsg *pmsg;
4530         struct nlmsghdr *nlh;
4531         struct prefix_cacheinfo ci;
4532
4533         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4534         if (nlh == NULL)
4535                 return -EMSGSIZE;
4536
4537         pmsg = nlmsg_data(nlh);
4538         pmsg->prefix_family = AF_INET6;
4539         pmsg->prefix_pad1 = 0;
4540         pmsg->prefix_pad2 = 0;
4541         pmsg->prefix_ifindex = idev->dev->ifindex;
4542         pmsg->prefix_len = pinfo->prefix_len;
4543         pmsg->prefix_type = pinfo->type;
4544         pmsg->prefix_pad3 = 0;
4545         pmsg->prefix_flags = 0;
4546         if (pinfo->onlink)
4547                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4548         if (pinfo->autoconf)
4549                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4550
4551         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4552                 goto nla_put_failure;
4553         ci.preferred_time = ntohl(pinfo->prefered);
4554         ci.valid_time = ntohl(pinfo->valid);
4555         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4556                 goto nla_put_failure;
4557         return nlmsg_end(skb, nlh);
4558
4559 nla_put_failure:
4560         nlmsg_cancel(skb, nlh);
4561         return -EMSGSIZE;
4562 }
4563
4564 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4565                          struct prefix_info *pinfo)
4566 {
4567         struct sk_buff *skb;
4568         struct net *net = dev_net(idev->dev);
4569         int err = -ENOBUFS;
4570
4571         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4572         if (skb == NULL)
4573                 goto errout;
4574
4575         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4576         if (err < 0) {
4577                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4578                 WARN_ON(err == -EMSGSIZE);
4579                 kfree_skb(skb);
4580                 goto errout;
4581         }
4582         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4583         return;
4584 errout:
4585         if (err < 0)
4586                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4587 }
4588
4589 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4590 {
4591         struct net *net = dev_net(ifp->idev->dev);
4592
4593         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4594
4595         switch (event) {
4596         case RTM_NEWADDR:
4597                 /*
4598                  * If the address was optimistic
4599                  * we inserted the route at the start of
4600                  * our DAD process, so we don't need
4601                  * to do it again
4602                  */
4603                 if (!(ifp->rt->rt6i_node))
4604                         ip6_ins_rt(ifp->rt);
4605                 if (ifp->idev->cnf.forwarding)
4606                         addrconf_join_anycast(ifp);
4607                 break;
4608         case RTM_DELADDR:
4609                 if (ifp->idev->cnf.forwarding)
4610                         addrconf_leave_anycast(ifp);
4611                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4612                 dst_hold(&ifp->rt->dst);
4613
4614                 if (ip6_del_rt(ifp->rt))
4615                         dst_free(&ifp->rt->dst);
4616                 break;
4617         }
4618         atomic_inc(&net->ipv6.dev_addr_genid);
4619 }
4620
4621 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4622 {
4623         rcu_read_lock_bh();
4624         if (likely(ifp->idev->dead == 0))
4625                 __ipv6_ifa_notify(event, ifp);
4626         rcu_read_unlock_bh();
4627 }
4628
4629 #ifdef CONFIG_SYSCTL
4630
4631 static
4632 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4633                            void __user *buffer, size_t *lenp, loff_t *ppos)
4634 {
4635         int *valp = ctl->data;
4636         int val = *valp;
4637         loff_t pos = *ppos;
4638         ctl_table lctl;
4639         int ret;
4640
4641         /*
4642          * ctl->data points to idev->cnf.forwarding, we should
4643          * not modify it until we get the rtnl lock.
4644          */
4645         lctl = *ctl;
4646         lctl.data = &val;
4647
4648         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4649
4650         if (write)
4651                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4652         if (ret)
4653                 *ppos = pos;
4654         return ret;
4655 }
4656
4657 static void dev_disable_change(struct inet6_dev *idev)
4658 {
4659         if (!idev || !idev->dev)
4660                 return;
4661
4662         if (idev->cnf.disable_ipv6)
4663                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4664         else
4665                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4666 }
4667
4668 static void addrconf_disable_change(struct net *net, __s32 newf)
4669 {
4670         struct net_device *dev;
4671         struct inet6_dev *idev;
4672
4673         rcu_read_lock();
4674         for_each_netdev_rcu(net, dev) {
4675                 idev = __in6_dev_get(dev);
4676                 if (idev) {
4677                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4678                         idev->cnf.disable_ipv6 = newf;
4679                         if (changed)
4680                                 dev_disable_change(idev);
4681                 }
4682         }
4683         rcu_read_unlock();
4684 }
4685
4686 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4687 {
4688         struct net *net;
4689         int old;
4690
4691         if (!rtnl_trylock())
4692                 return restart_syscall();
4693
4694         net = (struct net *)table->extra2;
4695         old = *p;
4696         *p = newf;
4697
4698         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4699                 rtnl_unlock();
4700                 return 0;
4701         }
4702
4703         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4704                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4705                 addrconf_disable_change(net, newf);
4706         } else if ((!newf) ^ (!old))
4707                 dev_disable_change((struct inet6_dev *)table->extra1);
4708
4709         rtnl_unlock();
4710         return 0;
4711 }
4712
4713 static
4714 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4715                             void __user *buffer, size_t *lenp, loff_t *ppos)
4716 {
4717         int *valp = ctl->data;
4718         int val = *valp;
4719         loff_t pos = *ppos;
4720         ctl_table lctl;
4721         int ret;
4722
4723         /*
4724          * ctl->data points to idev->cnf.disable_ipv6, we should
4725          * not modify it until we get the rtnl lock.
4726          */
4727         lctl = *ctl;
4728         lctl.data = &val;
4729
4730         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4731
4732         if (write)
4733                 ret = addrconf_disable_ipv6(ctl, valp, val);
4734         if (ret)
4735                 *ppos = pos;
4736         return ret;
4737 }
4738
4739 static struct addrconf_sysctl_table
4740 {
4741         struct ctl_table_header *sysctl_header;
4742         ctl_table addrconf_vars[DEVCONF_MAX+1];
4743 } addrconf_sysctl __read_mostly = {
4744         .sysctl_header = NULL,
4745         .addrconf_vars = {
4746                 {
4747                         .procname       = "forwarding",
4748                         .data           = &ipv6_devconf.forwarding,
4749                         .maxlen         = sizeof(int),
4750                         .mode           = 0644,
4751                         .proc_handler   = addrconf_sysctl_forward,
4752                 },
4753                 {
4754                         .procname       = "hop_limit",
4755                         .data           = &ipv6_devconf.hop_limit,
4756                         .maxlen         = sizeof(int),
4757                         .mode           = 0644,
4758                         .proc_handler   = proc_dointvec,
4759                 },
4760                 {
4761                         .procname       = "mtu",
4762                         .data           = &ipv6_devconf.mtu6,
4763                         .maxlen         = sizeof(int),
4764                         .mode           = 0644,
4765                         .proc_handler   = proc_dointvec,
4766                 },
4767                 {
4768                         .procname       = "accept_ra",
4769                         .data           = &ipv6_devconf.accept_ra,
4770                         .maxlen         = sizeof(int),
4771                         .mode           = 0644,
4772                         .proc_handler   = proc_dointvec,
4773                 },
4774                 {
4775                         .procname       = "accept_redirects",
4776                         .data           = &ipv6_devconf.accept_redirects,
4777                         .maxlen         = sizeof(int),
4778                         .mode           = 0644,
4779                         .proc_handler   = proc_dointvec,
4780                 },
4781                 {
4782                         .procname       = "autoconf",
4783                         .data           = &ipv6_devconf.autoconf,
4784                         .maxlen         = sizeof(int),
4785                         .mode           = 0644,
4786                         .proc_handler   = proc_dointvec,
4787                 },
4788                 {
4789                         .procname       = "dad_transmits",
4790                         .data           = &ipv6_devconf.dad_transmits,
4791                         .maxlen         = sizeof(int),
4792                         .mode           = 0644,
4793                         .proc_handler   = proc_dointvec,
4794                 },
4795                 {
4796                         .procname       = "router_solicitations",
4797                         .data           = &ipv6_devconf.rtr_solicits,
4798                         .maxlen         = sizeof(int),
4799                         .mode           = 0644,
4800                         .proc_handler   = proc_dointvec,
4801                 },
4802                 {
4803                         .procname       = "router_solicitation_interval",
4804                         .data           = &ipv6_devconf.rtr_solicit_interval,
4805                         .maxlen         = sizeof(int),
4806                         .mode           = 0644,
4807                         .proc_handler   = proc_dointvec_jiffies,
4808                 },
4809                 {
4810                         .procname       = "router_solicitation_delay",
4811                         .data           = &ipv6_devconf.rtr_solicit_delay,
4812                         .maxlen         = sizeof(int),
4813                         .mode           = 0644,
4814                         .proc_handler   = proc_dointvec_jiffies,
4815                 },
4816                 {
4817                         .procname       = "force_mld_version",
4818                         .data           = &ipv6_devconf.force_mld_version,
4819                         .maxlen         = sizeof(int),
4820                         .mode           = 0644,
4821                         .proc_handler   = proc_dointvec,
4822                 },
4823 #ifdef CONFIG_IPV6_PRIVACY
4824                 {
4825                         .procname       = "use_tempaddr",
4826                         .data           = &ipv6_devconf.use_tempaddr,
4827                         .maxlen         = sizeof(int),
4828                         .mode           = 0644,
4829                         .proc_handler   = proc_dointvec,
4830                 },
4831                 {
4832                         .procname       = "temp_valid_lft",
4833                         .data           = &ipv6_devconf.temp_valid_lft,
4834                         .maxlen         = sizeof(int),
4835                         .mode           = 0644,
4836                         .proc_handler   = proc_dointvec,
4837                 },
4838                 {
4839                         .procname       = "temp_prefered_lft",
4840                         .data           = &ipv6_devconf.temp_prefered_lft,
4841                         .maxlen         = sizeof(int),
4842                         .mode           = 0644,
4843                         .proc_handler   = proc_dointvec,
4844                 },
4845                 {
4846                         .procname       = "regen_max_retry",
4847                         .data           = &ipv6_devconf.regen_max_retry,
4848                         .maxlen         = sizeof(int),
4849                         .mode           = 0644,
4850                         .proc_handler   = proc_dointvec,
4851                 },
4852                 {
4853                         .procname       = "max_desync_factor",
4854                         .data           = &ipv6_devconf.max_desync_factor,
4855                         .maxlen         = sizeof(int),
4856                         .mode           = 0644,
4857                         .proc_handler   = proc_dointvec,
4858                 },
4859 #endif
4860                 {
4861                         .procname       = "max_addresses",
4862                         .data           = &ipv6_devconf.max_addresses,
4863                         .maxlen         = sizeof(int),
4864                         .mode           = 0644,
4865                         .proc_handler   = proc_dointvec,
4866                 },
4867                 {
4868                         .procname       = "accept_ra_defrtr",
4869                         .data           = &ipv6_devconf.accept_ra_defrtr,
4870                         .maxlen         = sizeof(int),
4871                         .mode           = 0644,
4872                         .proc_handler   = proc_dointvec,
4873                 },
4874                 {
4875                         .procname       = "accept_ra_pinfo",
4876                         .data           = &ipv6_devconf.accept_ra_pinfo,
4877                         .maxlen         = sizeof(int),
4878                         .mode           = 0644,
4879                         .proc_handler   = proc_dointvec,
4880                 },
4881 #ifdef CONFIG_IPV6_ROUTER_PREF
4882                 {
4883                         .procname       = "accept_ra_rtr_pref",
4884                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4885                         .maxlen         = sizeof(int),
4886                         .mode           = 0644,
4887                         .proc_handler   = proc_dointvec,
4888                 },
4889                 {
4890                         .procname       = "router_probe_interval",
4891                         .data           = &ipv6_devconf.rtr_probe_interval,
4892                         .maxlen         = sizeof(int),
4893                         .mode           = 0644,
4894                         .proc_handler   = proc_dointvec_jiffies,
4895                 },
4896 #ifdef CONFIG_IPV6_ROUTE_INFO
4897                 {
4898                         .procname       = "accept_ra_rt_info_max_plen",
4899                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4900                         .maxlen         = sizeof(int),
4901                         .mode           = 0644,
4902                         .proc_handler   = proc_dointvec,
4903                 },
4904 #endif
4905 #endif
4906                 {
4907                         .procname       = "proxy_ndp",
4908                         .data           = &ipv6_devconf.proxy_ndp,
4909                         .maxlen         = sizeof(int),
4910                         .mode           = 0644,
4911                         .proc_handler   = proc_dointvec,
4912                 },
4913                 {
4914                         .procname       = "accept_source_route",
4915                         .data           = &ipv6_devconf.accept_source_route,
4916                         .maxlen         = sizeof(int),
4917                         .mode           = 0644,
4918                         .proc_handler   = proc_dointvec,
4919                 },
4920 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4921                 {
4922                         .procname       = "optimistic_dad",
4923                         .data           = &ipv6_devconf.optimistic_dad,
4924                         .maxlen         = sizeof(int),
4925                         .mode           = 0644,
4926                         .proc_handler   = proc_dointvec,
4927
4928                 },
4929 #endif
4930 #ifdef CONFIG_IPV6_MROUTE
4931                 {
4932                         .procname       = "mc_forwarding",
4933                         .data           = &ipv6_devconf.mc_forwarding,
4934                         .maxlen         = sizeof(int),
4935                         .mode           = 0444,
4936                         .proc_handler   = proc_dointvec,
4937                 },
4938 #endif
4939                 {
4940                         .procname       = "disable_ipv6",
4941                         .data           = &ipv6_devconf.disable_ipv6,
4942                         .maxlen         = sizeof(int),
4943                         .mode           = 0644,
4944                         .proc_handler   = addrconf_sysctl_disable,
4945                 },
4946                 {
4947                         .procname       = "accept_dad",
4948                         .data           = &ipv6_devconf.accept_dad,
4949                         .maxlen         = sizeof(int),
4950                         .mode           = 0644,
4951                         .proc_handler   = proc_dointvec,
4952                 },
4953                 {
4954                         .procname       = "force_tllao",
4955                         .data           = &ipv6_devconf.force_tllao,
4956                         .maxlen         = sizeof(int),
4957                         .mode           = 0644,
4958                         .proc_handler   = proc_dointvec
4959                 },
4960                 {
4961                         .procname       = "ndisc_notify",
4962                         .data           = &ipv6_devconf.ndisc_notify,
4963                         .maxlen         = sizeof(int),
4964                         .mode           = 0644,
4965                         .proc_handler   = proc_dointvec
4966                 },
4967                 {
4968                         /* sentinel */
4969                 }
4970         },
4971 };
4972
4973 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4974                 struct inet6_dev *idev, struct ipv6_devconf *p)
4975 {
4976         int i;
4977         struct addrconf_sysctl_table *t;
4978         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4979
4980         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4981         if (t == NULL)
4982                 goto out;
4983
4984         for (i = 0; t->addrconf_vars[i].data; i++) {
4985                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4986                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4987                 t->addrconf_vars[i].extra2 = net;
4988         }
4989
4990         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4991
4992         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4993         if (t->sysctl_header == NULL)
4994                 goto free;
4995
4996         p->sysctl = t;
4997         return 0;
4998
4999 free:
5000         kfree(t);
5001 out:
5002         return -ENOBUFS;
5003 }
5004
5005 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5006 {
5007         struct addrconf_sysctl_table *t;
5008
5009         if (p->sysctl == NULL)
5010                 return;
5011
5012         t = p->sysctl;
5013         p->sysctl = NULL;
5014         unregister_net_sysctl_table(t->sysctl_header);
5015         kfree(t);
5016 }
5017
5018 static void addrconf_sysctl_register(struct inet6_dev *idev)
5019 {
5020         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5021                               &ndisc_ifinfo_sysctl_change);
5022         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5023                                         idev, &idev->cnf);
5024 }
5025
5026 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5027 {
5028         __addrconf_sysctl_unregister(&idev->cnf);
5029         neigh_sysctl_unregister(idev->nd_parms);
5030 }
5031
5032
5033 #endif
5034
5035 static int __net_init addrconf_init_net(struct net *net)
5036 {
5037         int err = -ENOMEM;
5038         struct ipv6_devconf *all, *dflt;
5039
5040         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5041         if (all == NULL)
5042                 goto err_alloc_all;
5043
5044         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5045         if (dflt == NULL)
5046                 goto err_alloc_dflt;
5047
5048         /* these will be inherited by all namespaces */
5049         dflt->autoconf = ipv6_defaults.autoconf;
5050         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5051
5052         net->ipv6.devconf_all = all;
5053         net->ipv6.devconf_dflt = dflt;
5054
5055 #ifdef CONFIG_SYSCTL
5056         err = __addrconf_sysctl_register(net, "all", NULL, all);
5057         if (err < 0)
5058                 goto err_reg_all;
5059
5060         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5061         if (err < 0)
5062                 goto err_reg_dflt;
5063 #endif
5064         return 0;
5065
5066 #ifdef CONFIG_SYSCTL
5067 err_reg_dflt:
5068         __addrconf_sysctl_unregister(all);
5069 err_reg_all:
5070         kfree(dflt);
5071 #endif
5072 err_alloc_dflt:
5073         kfree(all);
5074 err_alloc_all:
5075         return err;
5076 }
5077
5078 static void __net_exit addrconf_exit_net(struct net *net)
5079 {
5080 #ifdef CONFIG_SYSCTL
5081         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5082         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5083 #endif
5084         if (!net_eq(net, &init_net)) {
5085                 kfree(net->ipv6.devconf_dflt);
5086                 kfree(net->ipv6.devconf_all);
5087         }
5088 }
5089
5090 static struct pernet_operations addrconf_ops = {
5091         .init = addrconf_init_net,
5092         .exit = addrconf_exit_net,
5093 };
5094
5095 static struct rtnl_af_ops inet6_ops = {
5096         .family           = AF_INET6,
5097         .fill_link_af     = inet6_fill_link_af,
5098         .get_link_af_size = inet6_get_link_af_size,
5099         .set_link_af      = inet6_set_link_af,
5100 };
5101
5102 /*
5103  *      Init / cleanup code
5104  */
5105
5106 int __init addrconf_init(void)
5107 {
5108         int i, err;
5109
5110         err = ipv6_addr_label_init();
5111         if (err < 0) {
5112                 pr_crit("%s: cannot initialize default policy table: %d\n",
5113                         __func__, err);
5114                 goto out;
5115         }
5116
5117         err = register_pernet_subsys(&addrconf_ops);
5118         if (err < 0)
5119                 goto out_addrlabel;
5120
5121         /* The addrconf netdev notifier requires that loopback_dev
5122          * has it's ipv6 private information allocated and setup
5123          * before it can bring up and give link-local addresses
5124          * to other devices which are up.
5125          *
5126          * Unfortunately, loopback_dev is not necessarily the first
5127          * entry in the global dev_base list of net devices.  In fact,
5128          * it is likely to be the very last entry on that list.
5129          * So this causes the notifier registry below to try and
5130          * give link-local addresses to all devices besides loopback_dev
5131          * first, then loopback_dev, which cases all the non-loopback_dev
5132          * devices to fail to get a link-local address.
5133          *
5134          * So, as a temporary fix, allocate the ipv6 structure for
5135          * loopback_dev first by hand.
5136          * Longer term, all of the dependencies ipv6 has upon the loopback
5137          * device and it being up should be removed.
5138          */
5139         rtnl_lock();
5140         if (!ipv6_add_dev(init_net.loopback_dev))
5141                 err = -ENOMEM;
5142         rtnl_unlock();
5143         if (err)
5144                 goto errlo;
5145
5146         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5147                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5148
5149         register_netdevice_notifier(&ipv6_dev_notf);
5150
5151         addrconf_verify(0);
5152
5153         err = rtnl_af_register(&inet6_ops);
5154         if (err < 0)
5155                 goto errout_af;
5156
5157         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5158                               NULL);
5159         if (err < 0)
5160                 goto errout;
5161
5162         /* Only the first call to __rtnl_register can fail */
5163         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5164         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5165         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5166                         inet6_dump_ifaddr, NULL);
5167         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5168                         inet6_dump_ifmcaddr, NULL);
5169         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5170                         inet6_dump_ifacaddr, NULL);
5171         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5172                         inet6_netconf_dump_devconf, NULL);
5173
5174         ipv6_addr_label_rtnl_register();
5175
5176         return 0;
5177 errout:
5178         rtnl_af_unregister(&inet6_ops);
5179 errout_af:
5180         unregister_netdevice_notifier(&ipv6_dev_notf);
5181 errlo:
5182         unregister_pernet_subsys(&addrconf_ops);
5183 out_addrlabel:
5184         ipv6_addr_label_cleanup();
5185 out:
5186         return err;
5187 }
5188
5189 void addrconf_cleanup(void)
5190 {
5191         struct net_device *dev;
5192         int i;
5193
5194         unregister_netdevice_notifier(&ipv6_dev_notf);
5195         unregister_pernet_subsys(&addrconf_ops);
5196         ipv6_addr_label_cleanup();
5197
5198         rtnl_lock();
5199
5200         __rtnl_af_unregister(&inet6_ops);
5201
5202         /* clean dev list */
5203         for_each_netdev(&init_net, dev) {
5204                 if (__in6_dev_get(dev) == NULL)
5205                         continue;
5206                 addrconf_ifdown(dev, 1);
5207         }
5208         addrconf_ifdown(init_net.loopback_dev, 2);
5209
5210         /*
5211          *      Check hash table.
5212          */
5213         spin_lock_bh(&addrconf_hash_lock);
5214         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5215                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5216         spin_unlock_bh(&addrconf_hash_lock);
5217
5218         del_timer(&addr_chk_timer);
5219         rtnl_unlock();
5220 }