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         .accept_ra_rt_table     = 0,
 202         .proxy_ndp              = 0,
 203         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 204         .disable_ipv6           = 0,
 205         .accept_dad             = 1,
 206 };
 207
 208 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 209         .forwarding             = 0,
 210         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 211         .mtu6                   = IPV6_MIN_MTU,
 212         .accept_ra              = 1,
 213         .accept_redirects       = 1,
 214         .autoconf               = 1,
 215         .dad_transmits          = 1,
 216         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 217         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 218         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 219 #ifdef CONFIG_IPV6_PRIVACY
 220         .use_tempaddr           = 0,
 221         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 222         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 223         .regen_max_retry        = REGEN_MAX_RETRY,
 224         .max_desync_factor      = MAX_DESYNC_FACTOR,
 225 #endif
 226         .max_addresses          = IPV6_MAX_ADDRESSES,
 227         .accept_ra_defrtr       = 1,
 228         .accept_ra_pinfo        = 1,
 229 #ifdef CONFIG_IPV6_ROUTER_PREF
 230         .accept_ra_rtr_pref     = 1,
 231         .rtr_probe_interval     = 60 * HZ,
 232 #ifdef CONFIG_IPV6_ROUTE_INFO
 233         .accept_ra_rt_info_max_plen = 0,
 234 #endif
 235 #endif
 236         .accept_ra_rt_table     = 0,
 237         .proxy_ndp              = 0,
 238         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 239         .disable_ipv6           = 0,
 240         .accept_dad             = 1,
 241 };
 242
 243 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 244 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 245 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 246 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 247 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 248 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 249 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 250 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 251
 252 /* Check if a valid qdisc is available */
 253 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 254 {
 255         return !qdisc_tx_is_noop(dev);
 256 }
 257
 258 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 259 {
 260         if (del_timer(&ifp->timer))
 261                 __in6_ifa_put(ifp);
 262 }
 263
 264 enum addrconf_timer_t {
 265         AC_NONE,
 266         AC_DAD,
 267         AC_RS,
 268 };
 269
 270 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 271                                enum addrconf_timer_t what,
 272                                unsigned long when)
 273 {
 274         if (!del_timer(&ifp->timer))
 275                 in6_ifa_hold(ifp);
 276
 277         switch (what) {
 278         case AC_DAD:
 279                 ifp->timer.function = addrconf_dad_timer;
 280                 break;
 281         case AC_RS:
 282                 ifp->timer.function = addrconf_rs_timer;
 283                 break;
 284         default:
 285                 break;
 286         }
 287         ifp->timer.expires = jiffies + when;
 288         add_timer(&ifp->timer);
 289 }
 290
 291 static int snmp6_alloc_dev(struct inet6_dev *idev)
 292 {
 293         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 294                           sizeof(struct ipstats_mib),
 295                           __alignof__(struct ipstats_mib)) < 0)
 296                 goto err_ip;
 297         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 298                                         GFP_KERNEL);
 299         if (!idev->stats.icmpv6dev)
 300                 goto err_icmp;
 301         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 302                                            GFP_KERNEL);
 303         if (!idev->stats.icmpv6msgdev)
 304                 goto err_icmpmsg;
 305
 306         return 0;
 307
 308 err_icmpmsg:
 309         kfree(idev->stats.icmpv6dev);
 310 err_icmp:
 311         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 312 err_ip:
 313         return -ENOMEM;
 314 }
 315
 316 static void snmp6_free_dev(struct inet6_dev *idev)
 317 {
 318         kfree(idev->stats.icmpv6msgdev);
 319         kfree(idev->stats.icmpv6dev);
 320         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 321 }
 322
 323 /* Nobody refers to this device, we may destroy it. */
 324
 325 void in6_dev_finish_destroy(struct inet6_dev *idev)
 326 {
 327         struct net_device *dev = idev->dev;
 328
 329         WARN_ON(!list_empty(&idev->addr_list));
 330         WARN_ON(idev->mc_list != NULL);
 331
 332 #ifdef NET_REFCNT_DEBUG
 333         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 334 #endif
 335         dev_put(dev);
 336         if (!idev->dead) {
 337                 pr_warn("Freeing alive inet6 device %p\n", idev);
 338                 return;
 339         }
 340         snmp6_free_dev(idev);
 341         kfree_rcu(idev, rcu);
 342 }
 343 EXPORT_SYMBOL(in6_dev_finish_destroy);
 344
 345 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 346 {
 347         struct inet6_dev *ndev;
 348
 349         ASSERT_RTNL();
 350
 351         if (dev->mtu < IPV6_MIN_MTU)
 352                 return NULL;
 353
 354         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 355
 356         if (ndev == NULL)
 357                 return NULL;
 358
 359         rwlock_init(&ndev->lock);
 360         ndev->dev = dev;
 361         INIT_LIST_HEAD(&ndev->addr_list);
 362
 363         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 364         ndev->cnf.mtu6 = dev->mtu;
 365         ndev->cnf.sysctl = NULL;
 366         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 367         if (ndev->nd_parms == NULL) {
 368                 kfree(ndev);
 369                 return NULL;
 370         }
 371         if (ndev->cnf.forwarding)
 372                 dev_disable_lro(dev);
 373         /* We refer to the device */
 374         dev_hold(dev);
 375
 376         if (snmp6_alloc_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 dev_put(dev);
 382                 kfree(ndev);
 383                 return NULL;
 384         }
 385
 386         if (snmp6_register_dev(ndev) < 0) {
 387                 ADBG((KERN_WARNING
 388                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 389                         __func__, dev->name));
 390                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 391                 ndev->dead = 1;
 392                 in6_dev_finish_destroy(ndev);
 393                 return NULL;
 394         }
 395
 396         /* One reference from device.  We must do this before
 397          * we invoke __ipv6_regen_rndid().
 398          */
 399         in6_dev_hold(ndev);
 400
 401         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 402                 ndev->cnf.accept_dad = -1;
 403
 404 #if IS_ENABLED(CONFIG_IPV6_SIT)
 405         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 406                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 407                 ndev->cnf.rtr_solicits = 0;
 408         }
 409 #endif
 410
 411 #ifdef CONFIG_IPV6_PRIVACY
 412         INIT_LIST_HEAD(&ndev->tempaddr_list);
 413         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 414         if ((dev->flags&IFF_LOOPBACK) ||
 415             dev->type == ARPHRD_TUNNEL ||
 416             dev->type == ARPHRD_TUNNEL6 ||
 417             dev->type == ARPHRD_SIT ||
 418             dev->type == ARPHRD_NONE) {
 419                 ndev->cnf.use_tempaddr = -1;
 420         } else {
 421                 in6_dev_hold(ndev);
 422                 ipv6_regen_rndid((unsigned long) ndev);
 423         }
 424 #endif
 425         ndev->token = in6addr_any;
 426
 427         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 428                 ndev->if_flags |= IF_READY;
 429
 430         ipv6_mc_init_dev(ndev);
 431         ndev->tstamp = jiffies;
 432         addrconf_sysctl_register(ndev);
 433         /* protected by rtnl_lock */
 434         rcu_assign_pointer(dev->ip6_ptr, ndev);
 435
 436         /* Join interface-local all-node multicast group */
 437         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 438
 439         /* Join all-node multicast group */
 440         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 441
 442         /* Join all-router multicast group if forwarding is set */
 443         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 444                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 445
 446         return ndev;
 447 }
 448
 449 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 450 {
 451         struct inet6_dev *idev;
 452
 453         ASSERT_RTNL();
 454
 455         idev = __in6_dev_get(dev);
 456         if (!idev) {
 457                 idev = ipv6_add_dev(dev);
 458                 if (!idev)
 459                         return NULL;
 460         }
 461
 462         if (dev->flags&IFF_UP)
 463                 ipv6_mc_up(idev);
 464         return idev;
 465 }
 466
 467 static int inet6_netconf_msgsize_devconf(int type)
 468 {
 469         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 470                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 471
 472         /* type -1 is used for ALL */
 473         if (type == -1 || type == NETCONFA_FORWARDING)
 474                 size += nla_total_size(4);
 475 #ifdef CONFIG_IPV6_MROUTE
 476         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 477                 size += nla_total_size(4);
 478 #endif
 479
 480         return size;
 481 }
 482
 483 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 484                                       struct ipv6_devconf *devconf, u32 portid,
 485                                       u32 seq, int event, unsigned int flags,
 486                                       int type)
 487 {
 488         struct nlmsghdr  *nlh;
 489         struct netconfmsg *ncm;
 490
 491         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 492                         flags);
 493         if (nlh == NULL)
 494                 return -EMSGSIZE;
 495
 496         ncm = nlmsg_data(nlh);
 497         ncm->ncm_family = AF_INET6;
 498
 499         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 500                 goto nla_put_failure;
 501
 502         /* type -1 is used for ALL */
 503         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 504             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 505                 goto nla_put_failure;
 506 #ifdef CONFIG_IPV6_MROUTE
 507         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 508             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 509                         devconf->mc_forwarding) < 0)
 510                 goto nla_put_failure;
 511 #endif
 512         return nlmsg_end(skb, nlh);
 513
 514 nla_put_failure:
 515         nlmsg_cancel(skb, nlh);
 516         return -EMSGSIZE;
 517 }
 518
 519 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 520                                   struct ipv6_devconf *devconf)
 521 {
 522         struct sk_buff *skb;
 523         int err = -ENOBUFS;
 524
 525         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 526         if (skb == NULL)
 527                 goto errout;
 528
 529         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 530                                          RTM_NEWNETCONF, 0, type);
 531         if (err < 0) {
 532                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 533                 WARN_ON(err == -EMSGSIZE);
 534                 kfree_skb(skb);
 535                 goto errout;
 536         }
 537         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 538         return;
 539 errout:
 540         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 541 }
 542
 543 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 544         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 545         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 546 };
 547
 548 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 549                                      struct nlmsghdr *nlh)
 550 {
 551         struct net *net = sock_net(in_skb->sk);
 552         struct nlattr *tb[NETCONFA_MAX+1];
 553         struct netconfmsg *ncm;
 554         struct sk_buff *skb;
 555         struct ipv6_devconf *devconf;
 556         struct inet6_dev *in6_dev;
 557         struct net_device *dev;
 558         int ifindex;
 559         int err;
 560
 561         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 562                           devconf_ipv6_policy);
 563         if (err < 0)
 564                 goto errout;
 565
 566         err = EINVAL;
 567         if (!tb[NETCONFA_IFINDEX])
 568                 goto errout;
 569
 570         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 571         switch (ifindex) {
 572         case NETCONFA_IFINDEX_ALL:
 573                 devconf = net->ipv6.devconf_all;
 574                 break;
 575         case NETCONFA_IFINDEX_DEFAULT:
 576                 devconf = net->ipv6.devconf_dflt;
 577                 break;
 578         default:
 579                 dev = __dev_get_by_index(net, ifindex);
 580                 if (dev == NULL)
 581                         goto errout;
 582                 in6_dev = __in6_dev_get(dev);
 583                 if (in6_dev == NULL)
 584                         goto errout;
 585                 devconf = &in6_dev->cnf;
 586                 break;
 587         }
 588
 589         err = -ENOBUFS;
 590         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 591         if (skb == NULL)
 592                 goto errout;
 593
 594         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 595                                          NETLINK_CB(in_skb).portid,
 596                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 597                                          -1);
 598         if (err < 0) {
 599                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 600                 WARN_ON(err == -EMSGSIZE);
 601                 kfree_skb(skb);
 602                 goto errout;
 603         }
 604         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 605 errout:
 606         return err;
 607 }
 608
 609 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 610                                       struct netlink_callback *cb)
 611 {
 612         struct net *net = sock_net(skb->sk);
 613         int h, s_h;
 614         int idx, s_idx;
 615         struct net_device *dev;
 616         struct inet6_dev *idev;
 617         struct hlist_head *head;
 618
 619         s_h = cb->args[0];
 620         s_idx = idx = cb->args[1];
 621
 622         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 623                 idx = 0;
 624                 head = &net->dev_index_head[h];
 625                 rcu_read_lock();
 626                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 627                           net->dev_base_seq;
 628                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 629                         if (idx < s_idx)
 630                                 goto cont;
 631                         idev = __in6_dev_get(dev);
 632                         if (!idev)
 633                                 goto cont;
 634
 635                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 636                                                        &idev->cnf,
 637                                                        NETLINK_CB(cb->skb).portid,
 638                                                        cb->nlh->nlmsg_seq,
 639                                                        RTM_NEWNETCONF,
 640                                                        NLM_F_MULTI,
 641                                                        -1) <= 0) {
 642                                 rcu_read_unlock();
 643                                 goto done;
 644                         }
 645                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 646 cont:
 647                         idx++;
 648                 }
 649                 rcu_read_unlock();
 650         }
 651         if (h == NETDEV_HASHENTRIES) {
 652                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 653                                                net->ipv6.devconf_all,
 654                                                NETLINK_CB(cb->skb).portid,
 655                                                cb->nlh->nlmsg_seq,
 656                                                RTM_NEWNETCONF, NLM_F_MULTI,
 657                                                -1) <= 0)
 658                         goto done;
 659                 else
 660                         h++;
 661         }
 662         if (h == NETDEV_HASHENTRIES + 1) {
 663                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 664                                                net->ipv6.devconf_dflt,
 665                                                NETLINK_CB(cb->skb).portid,
 666                                                cb->nlh->nlmsg_seq,
 667                                                RTM_NEWNETCONF, NLM_F_MULTI,
 668                                                -1) <= 0)
 669                         goto done;
 670                 else
 671                         h++;
 672         }
 673 done:
 674         cb->args[0] = h;
 675         cb->args[1] = idx;
 676
 677         return skb->len;
 678 }
 679
 680 #ifdef CONFIG_SYSCTL
 681 static void dev_forward_change(struct inet6_dev *idev)
 682 {
 683         struct net_device *dev;
 684         struct inet6_ifaddr *ifa;
 685
 686         if (!idev)
 687                 return;
 688         dev = idev->dev;
 689         if (idev->cnf.forwarding)
 690                 dev_disable_lro(dev);
 691         if (dev->flags & IFF_MULTICAST) {
 692                 if (idev->cnf.forwarding) {
 693                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 694                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 695                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 696                 } else {
 697                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 698                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 699                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 700                 }
 701         }
 702
 703         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 704                 if (ifa->flags&IFA_F_TENTATIVE)
 705                         continue;
 706                 if (idev->cnf.forwarding)
 707                         addrconf_join_anycast(ifa);
 708                 else
 709                         addrconf_leave_anycast(ifa);
 710         }
 711         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 712                                      dev->ifindex, &idev->cnf);
 713 }
 714
 715
 716 static void addrconf_forward_change(struct net *net, __s32 newf)
 717 {
 718         struct net_device *dev;
 719         struct inet6_dev *idev;
 720
 721         for_each_netdev(net, dev) {
 722                 idev = __in6_dev_get(dev);
 723                 if (idev) {
 724                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 725                         idev->cnf.forwarding = newf;
 726                         if (changed)
 727                                 dev_forward_change(idev);
 728                 }
 729         }
 730 }
 731
 732 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 733 {
 734         struct net *net;
 735         int old;
 736
 737         if (!rtnl_trylock())
 738                 return restart_syscall();
 739
 740         net = (struct net *)table->extra2;
 741         old = *p;
 742         *p = newf;
 743
 744         if (p == &net->ipv6.devconf_dflt->forwarding) {
 745                 if ((!newf) ^ (!old))
 746                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 747                                                      NETCONFA_IFINDEX_DEFAULT,
 748                                                      net->ipv6.devconf_dflt);
 749                 rtnl_unlock();
 750                 return 0;
 751         }
 752
 753         if (p == &net->ipv6.devconf_all->forwarding) {
 754                 net->ipv6.devconf_dflt->forwarding = newf;
 755                 addrconf_forward_change(net, newf);
 756                 if ((!newf) ^ (!old))
 757                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 758                                                      NETCONFA_IFINDEX_ALL,
 759                                                      net->ipv6.devconf_all);
 760         } else if ((!newf) ^ (!old))
 761                 dev_forward_change((struct inet6_dev *)table->extra1);
 762         rtnl_unlock();
 763
 764         if (newf)
 765                 rt6_purge_dflt_routers(net);
 766         return 1;
 767 }
 768 #endif
 769
 770 /* Nobody refers to this ifaddr, destroy it */
 771 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 772 {
 773         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 774
 775 #ifdef NET_REFCNT_DEBUG
 776         pr_debug("%s\n", __func__);
 777 #endif
 778
 779         in6_dev_put(ifp->idev);
 780
 781         if (del_timer(&ifp->timer))
 782                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 783
 784         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 785                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 786                 return;
 787         }
 788         ip6_rt_put(ifp->rt);
 789
 790         kfree_rcu(ifp, rcu);
 791 }
 792
 793 static void
 794 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 795 {
 796         struct list_head *p;
 797         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 798
 799         /*
 800          * Each device address list is sorted in order of scope -
 801          * global before linklocal.
 802          */
 803         list_for_each(p, &idev->addr_list) {
 804                 struct inet6_ifaddr *ifa
 805                         = list_entry(p, struct inet6_ifaddr, if_list);
 806                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 807                         break;
 808         }
 809
 810         list_add_tail(&ifp->if_list, p);
 811 }
 812
 813 static u32 inet6_addr_hash(const struct in6_addr *addr)
 814 {
 815         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 816 }
 817
 818 /* On success it returns ifp with increased reference count */
 819
 820 static struct inet6_ifaddr *
 821 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 822               int scope, u32 flags)
 823 {
 824         struct inet6_ifaddr *ifa = NULL;
 825         struct rt6_info *rt;
 826         unsigned int hash;
 827         int err = 0;
 828         int addr_type = ipv6_addr_type(addr);
 829
 830         if (addr_type == IPV6_ADDR_ANY ||
 831             addr_type & IPV6_ADDR_MULTICAST ||
 832             (!(idev->dev->flags & IFF_LOOPBACK) &&
 833              addr_type & IPV6_ADDR_LOOPBACK))
 834                 return ERR_PTR(-EADDRNOTAVAIL);
 835
 836         rcu_read_lock_bh();
 837         if (idev->dead) {
 838                 err = -ENODEV;                  /*XXX*/
 839                 goto out2;
 840         }
 841
 842         if (idev->cnf.disable_ipv6) {
 843                 err = -EACCES;
 844                 goto out2;
 845         }
 846
 847         spin_lock(&addrconf_hash_lock);
 848
 849         /* Ignore adding duplicate addresses on an interface */
 850         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 851                 ADBG(("ipv6_add_addr: already assigned\n"));
 852                 err = -EEXIST;
 853                 goto out;
 854         }
 855
 856         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 857
 858         if (ifa == NULL) {
 859                 ADBG(("ipv6_add_addr: malloc failed\n"));
 860                 err = -ENOBUFS;
 861                 goto out;
 862         }
 863
 864         rt = addrconf_dst_alloc(idev, addr, false);
 865         if (IS_ERR(rt)) {
 866                 err = PTR_ERR(rt);
 867                 goto out;
 868         }
 869
 870         ifa->addr = *addr;
 871
 872         spin_lock_init(&ifa->lock);
 873         spin_lock_init(&ifa->state_lock);
 874         init_timer(&ifa->timer);
 875         INIT_HLIST_NODE(&ifa->addr_lst);
 876         ifa->timer.data = (unsigned long) ifa;
 877         ifa->scope = scope;
 878         ifa->prefix_len = pfxlen;
 879         ifa->flags = flags | IFA_F_TENTATIVE;
 880         ifa->cstamp = ifa->tstamp = jiffies;
 881         ifa->tokenized = false;
 882
 883         ifa->rt = rt;
 884
 885         ifa->idev = idev;
 886         in6_dev_hold(idev);
 887         /* For caller */
 888         in6_ifa_hold(ifa);
 889
 890         /* Add to big hash table */
 891         hash = inet6_addr_hash(addr);
 892
 893         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 894         spin_unlock(&addrconf_hash_lock);
 895
 896         write_lock(&idev->lock);
 897         /* Add to inet6_dev unicast addr list. */
 898         ipv6_link_dev_addr(idev, ifa);
 899
 900 #ifdef CONFIG_IPV6_PRIVACY
 901         if (ifa->flags&IFA_F_TEMPORARY) {
 902                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 903                 in6_ifa_hold(ifa);
 904         }
 905 #endif
 906
 907         in6_ifa_hold(ifa);
 908         write_unlock(&idev->lock);
 909 out2:
 910         rcu_read_unlock_bh();
 911
 912         if (likely(err == 0))
 913                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 914         else {
 915                 kfree(ifa);
 916                 ifa = ERR_PTR(err);
 917         }
 918
 919         return ifa;
 920 out:
 921         spin_unlock(&addrconf_hash_lock);
 922         goto out2;
 923 }
 924
 925 /* This function wants to get referenced ifp and releases it before return */
 926
 927 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 928 {
 929         struct inet6_ifaddr *ifa, *ifn;
 930         struct inet6_dev *idev = ifp->idev;
 931         int state;
 932         int deleted = 0, onlink = 0;
 933         unsigned long expires = jiffies;
 934
 935         spin_lock_bh(&ifp->state_lock);
 936         state = ifp->state;
 937         ifp->state = INET6_IFADDR_STATE_DEAD;
 938         spin_unlock_bh(&ifp->state_lock);
 939
 940         if (state == INET6_IFADDR_STATE_DEAD)
 941                 goto out;
 942
 943         spin_lock_bh(&addrconf_hash_lock);
 944         hlist_del_init_rcu(&ifp->addr_lst);
 945         spin_unlock_bh(&addrconf_hash_lock);
 946
 947         write_lock_bh(&idev->lock);
 948 #ifdef CONFIG_IPV6_PRIVACY
 949         if (ifp->flags&IFA_F_TEMPORARY) {
 950                 list_del(&ifp->tmp_list);
 951                 if (ifp->ifpub) {
 952                         in6_ifa_put(ifp->ifpub);
 953                         ifp->ifpub = NULL;
 954                 }
 955                 __in6_ifa_put(ifp);
 956         }
 957 #endif
 958
 959         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 960                 if (ifa == ifp) {
 961                         list_del_init(&ifp->if_list);
 962                         __in6_ifa_put(ifp);
 963
 964                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 965                                 break;
 966                         deleted = 1;
 967                         continue;
 968                 } else if (ifp->flags & IFA_F_PERMANENT) {
 969                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 970                                               ifp->prefix_len)) {
 971                                 if (ifa->flags & IFA_F_PERMANENT) {
 972                                         onlink = 1;
 973                                         if (deleted)
 974                                                 break;
 975                                 } else {
 976                                         unsigned long lifetime;
 977
 978                                         if (!onlink)
 979                                                 onlink = -1;
 980
 981                                         spin_lock(&ifa->lock);
 982
 983                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 984                                         /*
 985                                          * Note: Because this address is
 986                                          * not permanent, lifetime <
 987                                          * LONG_MAX / HZ here.
 988                                          */
 989                                         if (time_before(expires,
 990                                                         ifa->tstamp + lifetime * HZ))
 991                                                 expires = ifa->tstamp + lifetime * HZ;
 992                                         spin_unlock(&ifa->lock);
 993                                 }
 994                         }
 995                 }
 996         }
 997         write_unlock_bh(&idev->lock);
 998
 999         addrconf_del_timer(ifp);
1000
1001         ipv6_ifa_notify(RTM_DELADDR, ifp);
1002
1003         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1004
1005         /*
1006          * Purge or update corresponding prefix
1007          *
1008          * 1) we don't purge prefix here if address was not permanent.
1009          *    prefix is managed by its own lifetime.
1010          * 2) if there're no addresses, delete prefix.
1011          * 3) if there're still other permanent address(es),
1012          *    corresponding prefix is still permanent.
1013          * 4) otherwise, update prefix lifetime to the
1014          *    longest valid lifetime among the corresponding
1015          *    addresses on the device.
1016          *    Note: subsequent RA will update lifetime.
1017          *
1018          * --yoshfuji
1019          */
1020         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1021                 struct in6_addr prefix;
1022                 struct rt6_info *rt;
1023
1024                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1025
1026                 rt = addrconf_get_prefix_route(&prefix,
1027                                                ifp->prefix_len,
1028                                                ifp->idev->dev,
1029                                                0, RTF_GATEWAY | RTF_DEFAULT);
1030
1031                 if (rt) {
1032                         if (onlink == 0) {
1033                                 ip6_del_rt(rt);
1034                                 rt = NULL;
1035                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1036                                 rt6_set_expires(rt, expires);
1037                         }
1038                 }
1039                 ip6_rt_put(rt);
1040         }
1041
1042         /* clean up prefsrc entries */
1043         rt6_remove_prefsrc(ifp);
1044 out:
1045         in6_ifa_put(ifp);
1046 }
1047
1048 #ifdef CONFIG_IPV6_PRIVACY
1049 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1050 {
1051         struct inet6_dev *idev = ifp->idev;
1052         struct in6_addr addr, *tmpaddr;
1053         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1054         unsigned long regen_advance;
1055         int tmp_plen;
1056         int ret = 0;
1057         int max_addresses;
1058         u32 addr_flags;
1059         unsigned long now = jiffies;
1060
1061         write_lock(&idev->lock);
1062         if (ift) {
1063                 spin_lock_bh(&ift->lock);
1064                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1065                 spin_unlock_bh(&ift->lock);
1066                 tmpaddr = &addr;
1067         } else {
1068                 tmpaddr = NULL;
1069         }
1070 retry:
1071         in6_dev_hold(idev);
1072         if (idev->cnf.use_tempaddr <= 0) {
1073                 write_unlock(&idev->lock);
1074                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1075                 in6_dev_put(idev);
1076                 ret = -1;
1077                 goto out;
1078         }
1079         spin_lock_bh(&ifp->lock);
1080         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1081                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1082                 spin_unlock_bh(&ifp->lock);
1083                 write_unlock(&idev->lock);
1084                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1085                         __func__);
1086                 in6_dev_put(idev);
1087                 ret = -1;
1088                 goto out;
1089         }
1090         in6_ifa_hold(ifp);
1091         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1092         __ipv6_try_regen_rndid(idev, tmpaddr);
1093         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1094         age = (now - ifp->tstamp) / HZ;
1095         tmp_valid_lft = min_t(__u32,
1096                               ifp->valid_lft,
1097                               idev->cnf.temp_valid_lft + age);
1098         tmp_prefered_lft = min_t(__u32,
1099                                  ifp->prefered_lft,
1100                                  idev->cnf.temp_prefered_lft + age -
1101                                  idev->cnf.max_desync_factor);
1102         tmp_plen = ifp->prefix_len;
1103         max_addresses = idev->cnf.max_addresses;
1104         tmp_tstamp = ifp->tstamp;
1105         spin_unlock_bh(&ifp->lock);
1106
1107         regen_advance = idev->cnf.regen_max_retry *
1108                         idev->cnf.dad_transmits *
1109                         idev->nd_parms->retrans_time / HZ;
1110         write_unlock(&idev->lock);
1111
1112         /* A temporary address is created only if this calculated Preferred
1113          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1114          * an implementation must not create a temporary address with a zero
1115          * Preferred Lifetime.
1116          * Use age calculation as in addrconf_verify to avoid unnecessary
1117          * temporary addresses being generated.
1118          */
1119         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1120         if (tmp_prefered_lft <= regen_advance + age) {
1121                 in6_ifa_put(ifp);
1122                 in6_dev_put(idev);
1123                 ret = -1;
1124                 goto out;
1125         }
1126
1127         addr_flags = IFA_F_TEMPORARY;
1128         /* set in addrconf_prefix_rcv() */
1129         if (ifp->flags & IFA_F_OPTIMISTIC)
1130                 addr_flags |= IFA_F_OPTIMISTIC;
1131
1132         ift = ipv6_add_addr(idev, &addr, tmp_plen,
1133                             ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1134                             addr_flags);
1135         if (IS_ERR(ift)) {
1136                 in6_ifa_put(ifp);
1137                 in6_dev_put(idev);
1138                 pr_info("%s: retry temporary address regeneration\n", __func__);
1139                 tmpaddr = &addr;
1140                 write_lock(&idev->lock);
1141                 goto retry;
1142         }
1143
1144         spin_lock_bh(&ift->lock);
1145         ift->ifpub = ifp;
1146         ift->valid_lft = tmp_valid_lft;
1147         ift->prefered_lft = tmp_prefered_lft;
1148         ift->cstamp = now;
1149         ift->tstamp = tmp_tstamp;
1150         spin_unlock_bh(&ift->lock);
1151
1152         addrconf_dad_start(ift);
1153         in6_ifa_put(ift);
1154         in6_dev_put(idev);
1155 out:
1156         return ret;
1157 }
1158 #endif
1159
1160 /*
1161  *      Choose an appropriate source address (RFC3484)
1162  */
1163 enum {
1164         IPV6_SADDR_RULE_INIT = 0,
1165         IPV6_SADDR_RULE_LOCAL,
1166         IPV6_SADDR_RULE_SCOPE,
1167         IPV6_SADDR_RULE_PREFERRED,
1168 #ifdef CONFIG_IPV6_MIP6
1169         IPV6_SADDR_RULE_HOA,
1170 #endif
1171         IPV6_SADDR_RULE_OIF,
1172         IPV6_SADDR_RULE_LABEL,
1173 #ifdef CONFIG_IPV6_PRIVACY
1174         IPV6_SADDR_RULE_PRIVACY,
1175 #endif
1176         IPV6_SADDR_RULE_ORCHID,
1177         IPV6_SADDR_RULE_PREFIX,
1178         IPV6_SADDR_RULE_MAX
1179 };
1180
1181 struct ipv6_saddr_score {
1182         int                     rule;
1183         int                     addr_type;
1184         struct inet6_ifaddr     *ifa;
1185         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1186         int                     scopedist;
1187         int                     matchlen;
1188 };
1189
1190 struct ipv6_saddr_dst {
1191         const struct in6_addr *addr;
1192         int ifindex;
1193         int scope;
1194         int label;
1195         unsigned int prefs;
1196 };
1197
1198 static inline int ipv6_saddr_preferred(int type)
1199 {
1200         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1201                 return 1;
1202         return 0;
1203 }
1204
1205 static int ipv6_get_saddr_eval(struct net *net,
1206                                struct ipv6_saddr_score *score,
1207                                struct ipv6_saddr_dst *dst,
1208                                int i)
1209 {
1210         int ret;
1211
1212         if (i <= score->rule) {
1213                 switch (i) {
1214                 case IPV6_SADDR_RULE_SCOPE:
1215                         ret = score->scopedist;
1216                         break;
1217                 case IPV6_SADDR_RULE_PREFIX:
1218                         ret = score->matchlen;
1219                         break;
1220                 default:
1221                         ret = !!test_bit(i, score->scorebits);
1222                 }
1223                 goto out;
1224         }
1225
1226         switch (i) {
1227         case IPV6_SADDR_RULE_INIT:
1228                 /* Rule 0: remember if hiscore is not ready yet */
1229                 ret = !!score->ifa;
1230                 break;
1231         case IPV6_SADDR_RULE_LOCAL:
1232                 /* Rule 1: Prefer same address */
1233                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1234                 break;
1235         case IPV6_SADDR_RULE_SCOPE:
1236                 /* Rule 2: Prefer appropriate scope
1237                  *
1238                  *      ret
1239                  *       ^
1240                  *    -1 |  d 15
1241                  *    ---+--+-+---> scope
1242                  *       |
1243                  *       |             d is scope of the destination.
1244                  *  B-d  |  \
1245                  *       |   \      <- smaller scope is better if
1246                  *  B-15 |    \        if scope is enough for destinaion.
1247                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1248                  * d-C-1 | /
1249                  *       |/         <- greater is better
1250                  *   -C  /             if scope is not enough for destination.
1251                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1252                  *
1253                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1254                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1255                  * Assume B = 0 and we get C > 29.
1256                  */
1257                 ret = __ipv6_addr_src_scope(score->addr_type);
1258                 if (ret >= dst->scope)
1259                         ret = -ret;
1260                 else
1261                         ret -= 128;     /* 30 is enough */
1262                 score->scopedist = ret;
1263                 break;
1264         case IPV6_SADDR_RULE_PREFERRED:
1265                 /* Rule 3: Avoid deprecated and optimistic addresses */
1266                 ret = ipv6_saddr_preferred(score->addr_type) ||
1267                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1268                 break;
1269 #ifdef CONFIG_IPV6_MIP6
1270         case IPV6_SADDR_RULE_HOA:
1271             {
1272                 /* Rule 4: Prefer home address */
1273                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1274                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1275                 break;
1276             }
1277 #endif
1278         case IPV6_SADDR_RULE_OIF:
1279                 /* Rule 5: Prefer outgoing interface */
1280                 ret = (!dst->ifindex ||
1281                        dst->ifindex == score->ifa->idev->dev->ifindex);
1282                 break;
1283         case IPV6_SADDR_RULE_LABEL:
1284                 /* Rule 6: Prefer matching label */
1285                 ret = ipv6_addr_label(net,
1286                                       &score->ifa->addr, score->addr_type,
1287                                       score->ifa->idev->dev->ifindex) == dst->label;
1288                 break;
1289 #ifdef CONFIG_IPV6_PRIVACY
1290         case IPV6_SADDR_RULE_PRIVACY:
1291             {
1292                 /* Rule 7: Prefer public address
1293                  * Note: prefer temporary address if use_tempaddr >= 2
1294                  */
1295                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1296                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1297                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1298                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1299                 break;
1300             }
1301 #endif
1302         case IPV6_SADDR_RULE_ORCHID:
1303                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1304                  *          non-ORCHID vs non-ORCHID
1305                  */
1306                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1307                         ipv6_addr_orchid(dst->addr));
1308                 break;
1309         case IPV6_SADDR_RULE_PREFIX:
1310                 /* Rule 8: Use longest matching prefix */
1311                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1312                 if (ret > score->ifa->prefix_len)
1313                         ret = score->ifa->prefix_len;
1314                 score->matchlen = ret;
1315                 break;
1316         default:
1317                 ret = 0;
1318         }
1319
1320         if (ret)
1321                 __set_bit(i, score->scorebits);
1322         score->rule = i;
1323 out:
1324         return ret;
1325 }
1326
1327 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1328                        const struct in6_addr *daddr, unsigned int prefs,
1329                        struct in6_addr *saddr)
1330 {
1331         struct ipv6_saddr_score scores[2],
1332                                 *score = &scores[0], *hiscore = &scores[1];
1333         struct ipv6_saddr_dst dst;
1334         struct net_device *dev;
1335         int dst_type;
1336
1337         dst_type = __ipv6_addr_type(daddr);
1338         dst.addr = daddr;
1339         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1340         dst.scope = __ipv6_addr_src_scope(dst_type);
1341         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1342         dst.prefs = prefs;
1343
1344         hiscore->rule = -1;
1345         hiscore->ifa = NULL;
1346
1347         rcu_read_lock();
1348
1349         for_each_netdev_rcu(net, dev) {
1350                 struct inet6_dev *idev;
1351
1352                 /* Candidate Source Address (section 4)
1353                  *  - multicast and link-local destination address,
1354                  *    the set of candidate source address MUST only
1355                  *    include addresses assigned to interfaces
1356                  *    belonging to the same link as the outgoing
1357                  *    interface.
1358                  * (- For site-local destination addresses, the
1359                  *    set of candidate source addresses MUST only
1360                  *    include addresses assigned to interfaces
1361                  *    belonging to the same site as the outgoing
1362                  *    interface.)
1363                  */
1364                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1365                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1366                     dst.ifindex && dev->ifindex != dst.ifindex)
1367                         continue;
1368
1369                 idev = __in6_dev_get(dev);
1370                 if (!idev)
1371                         continue;
1372
1373                 read_lock_bh(&idev->lock);
1374                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1375                         int i;
1376
1377                         /*
1378                          * - Tentative Address (RFC2462 section 5.4)
1379                          *  - A tentative address is not considered
1380                          *    "assigned to an interface" in the traditional
1381                          *    sense, unless it is also flagged as optimistic.
1382                          * - Candidate Source Address (section 4)
1383                          *  - In any case, anycast addresses, multicast
1384                          *    addresses, and the unspecified address MUST
1385                          *    NOT be included in a candidate set.
1386                          */
1387                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1388                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1389                                 continue;
1390
1391                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1392
1393                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1394                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1395                                 LIMIT_NETDEBUG(KERN_DEBUG
1396                                                "ADDRCONF: unspecified / multicast address "
1397                                                "assigned as unicast address on %s",
1398                                                dev->name);
1399                                 continue;
1400                         }
1401
1402                         score->rule = -1;
1403                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1404
1405                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1406                                 int minihiscore, miniscore;
1407
1408                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1409                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1410
1411                                 if (minihiscore > miniscore) {
1412                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1413                                             score->scopedist > 0) {
1414                                                 /*
1415                                                  * special case:
1416                                                  * each remaining entry
1417                                                  * has too small (not enough)
1418                                                  * scope, because ifa entries
1419                                                  * are sorted by their scope
1420                                                  * values.
1421                                                  */
1422                                                 goto try_nextdev;
1423                                         }
1424                                         break;
1425                                 } else if (minihiscore < miniscore) {
1426                                         if (hiscore->ifa)
1427                                                 in6_ifa_put(hiscore->ifa);
1428
1429                                         in6_ifa_hold(score->ifa);
1430
1431                                         swap(hiscore, score);
1432
1433                                         /* restore our iterator */
1434                                         score->ifa = hiscore->ifa;
1435
1436                                         break;
1437                                 }
1438                         }
1439                 }
1440 try_nextdev:
1441                 read_unlock_bh(&idev->lock);
1442         }
1443         rcu_read_unlock();
1444
1445         if (!hiscore->ifa)
1446                 return -EADDRNOTAVAIL;
1447
1448         *saddr = hiscore->ifa->addr;
1449         in6_ifa_put(hiscore->ifa);
1450         return 0;
1451 }
1452 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1453
1454 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1455                       unsigned char banned_flags)
1456 {
1457         struct inet6_ifaddr *ifp;
1458         int err = -EADDRNOTAVAIL;
1459
1460         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1461                 if (ifp->scope == IFA_LINK &&
1462                     !(ifp->flags & banned_flags)) {
1463                         *addr = ifp->addr;
1464                         err = 0;
1465                         break;
1466                 }
1467         }
1468         return err;
1469 }
1470
1471 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1472                     unsigned char banned_flags)
1473 {
1474         struct inet6_dev *idev;
1475         int err = -EADDRNOTAVAIL;
1476
1477         rcu_read_lock();
1478         idev = __in6_dev_get(dev);
1479         if (idev) {
1480                 read_lock_bh(&idev->lock);
1481                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1482                 read_unlock_bh(&idev->lock);
1483         }
1484         rcu_read_unlock();
1485         return err;
1486 }
1487
1488 static int ipv6_count_addresses(struct inet6_dev *idev)
1489 {
1490         int cnt = 0;
1491         struct inet6_ifaddr *ifp;
1492
1493         read_lock_bh(&idev->lock);
1494         list_for_each_entry(ifp, &idev->addr_list, if_list)
1495                 cnt++;
1496         read_unlock_bh(&idev->lock);
1497         return cnt;
1498 }
1499
1500 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1501                   const struct net_device *dev, int strict)
1502 {
1503         struct inet6_ifaddr *ifp;
1504         unsigned int hash = inet6_addr_hash(addr);
1505
1506         rcu_read_lock_bh();
1507         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1508                 if (!net_eq(dev_net(ifp->idev->dev), net))
1509                         continue;
1510                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1511                     !(ifp->flags&IFA_F_TENTATIVE) &&
1512                     (dev == NULL || ifp->idev->dev == dev ||
1513                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1514                         rcu_read_unlock_bh();
1515                         return 1;
1516                 }
1517         }
1518
1519         rcu_read_unlock_bh();
1520         return 0;
1521 }
1522 EXPORT_SYMBOL(ipv6_chk_addr);
1523
1524 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1525                                struct net_device *dev)
1526 {
1527         unsigned int hash = inet6_addr_hash(addr);
1528         struct inet6_ifaddr *ifp;
1529
1530         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1531                 if (!net_eq(dev_net(ifp->idev->dev), net))
1532                         continue;
1533                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1534                         if (dev == NULL || ifp->idev->dev == dev)
1535                                 return true;
1536                 }
1537         }
1538         return false;
1539 }
1540
1541 /* Compares an address/prefix_len with addresses on device @dev.
1542  * If one is found it returns true.
1543  */
1544 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1545         const unsigned int prefix_len, struct net_device *dev)
1546 {
1547         struct inet6_dev *idev;
1548         struct inet6_ifaddr *ifa;
1549         bool ret = false;
1550
1551         rcu_read_lock();
1552         idev = __in6_dev_get(dev);
1553         if (idev) {
1554                 read_lock_bh(&idev->lock);
1555                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1556                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1557                         if (ret)
1558                                 break;
1559                 }
1560                 read_unlock_bh(&idev->lock);
1561         }
1562         rcu_read_unlock();
1563
1564         return ret;
1565 }
1566 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1567
1568 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1569 {
1570         struct inet6_dev *idev;
1571         struct inet6_ifaddr *ifa;
1572         int     onlink;
1573
1574         onlink = 0;
1575         rcu_read_lock();
1576         idev = __in6_dev_get(dev);
1577         if (idev) {
1578                 read_lock_bh(&idev->lock);
1579                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1580                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1581                                                    ifa->prefix_len);
1582                         if (onlink)
1583                                 break;
1584                 }
1585                 read_unlock_bh(&idev->lock);
1586         }
1587         rcu_read_unlock();
1588         return onlink;
1589 }
1590 EXPORT_SYMBOL(ipv6_chk_prefix);
1591
1592 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1593                                      struct net_device *dev, int strict)
1594 {
1595         struct inet6_ifaddr *ifp, *result = NULL;
1596         unsigned int hash = inet6_addr_hash(addr);
1597
1598         rcu_read_lock_bh();
1599         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1600                 if (!net_eq(dev_net(ifp->idev->dev), net))
1601                         continue;
1602                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1603                         if (dev == NULL || ifp->idev->dev == dev ||
1604                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1605                                 result = ifp;
1606                                 in6_ifa_hold(ifp);
1607                                 break;
1608                         }
1609                 }
1610         }
1611         rcu_read_unlock_bh();
1612
1613         return result;
1614 }
1615
1616 /* Gets referenced address, destroys ifaddr */
1617
1618 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1619 {
1620         if (ifp->flags&IFA_F_PERMANENT) {
1621                 spin_lock_bh(&ifp->lock);
1622                 addrconf_del_timer(ifp);
1623                 ifp->flags |= IFA_F_TENTATIVE;
1624                 if (dad_failed)
1625                         ifp->flags |= IFA_F_DADFAILED;
1626                 spin_unlock_bh(&ifp->lock);
1627                 if (dad_failed)
1628                         ipv6_ifa_notify(0, ifp);
1629                 in6_ifa_put(ifp);
1630 #ifdef CONFIG_IPV6_PRIVACY
1631         } else if (ifp->flags&IFA_F_TEMPORARY) {
1632                 struct inet6_ifaddr *ifpub;
1633                 spin_lock_bh(&ifp->lock);
1634                 ifpub = ifp->ifpub;
1635                 if (ifpub) {
1636                         in6_ifa_hold(ifpub);
1637                         spin_unlock_bh(&ifp->lock);
1638                         ipv6_create_tempaddr(ifpub, ifp);
1639                         in6_ifa_put(ifpub);
1640                 } else {
1641                         spin_unlock_bh(&ifp->lock);
1642                 }
1643                 ipv6_del_addr(ifp);
1644 #endif
1645         } else
1646                 ipv6_del_addr(ifp);
1647 }
1648
1649 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1650 {
1651         int err = -ENOENT;
1652
1653         spin_lock(&ifp->state_lock);
1654         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1655                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1656                 err = 0;
1657         }
1658         spin_unlock(&ifp->state_lock);
1659
1660         return err;
1661 }
1662
1663 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1664 {
1665         struct inet6_dev *idev = ifp->idev;
1666
1667         if (addrconf_dad_end(ifp)) {
1668                 in6_ifa_put(ifp);
1669                 return;
1670         }
1671
1672         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1673                              ifp->idev->dev->name, &ifp->addr);
1674
1675         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1676                 struct in6_addr addr;
1677
1678                 addr.s6_addr32[0] = htonl(0xfe800000);
1679                 addr.s6_addr32[1] = 0;
1680
1681                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1682                     ipv6_addr_equal(&ifp->addr, &addr)) {
1683                         /* DAD failed for link-local based on MAC address */
1684                         idev->cnf.disable_ipv6 = 1;
1685
1686                         pr_info("%s: IPv6 being disabled!\n",
1687                                 ifp->idev->dev->name);
1688                 }
1689         }
1690
1691         addrconf_dad_stop(ifp, 1);
1692 }
1693
1694 /* Join to solicited addr multicast group. */
1695
1696 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1697 {
1698         struct in6_addr maddr;
1699
1700         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1701                 return;
1702
1703         addrconf_addr_solict_mult(addr, &maddr);
1704         ipv6_dev_mc_inc(dev, &maddr);
1705 }
1706
1707 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1708 {
1709         struct in6_addr maddr;
1710
1711         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1712                 return;
1713
1714         addrconf_addr_solict_mult(addr, &maddr);
1715         __ipv6_dev_mc_dec(idev, &maddr);
1716 }
1717
1718 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1719 {
1720         struct in6_addr addr;
1721         if (ifp->prefix_len == 127) /* RFC 6164 */
1722                 return;
1723         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1724         if (ipv6_addr_any(&addr))
1725                 return;
1726         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1727 }
1728
1729 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1730 {
1731         struct in6_addr addr;
1732         if (ifp->prefix_len == 127) /* RFC 6164 */
1733                 return;
1734         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1735         if (ipv6_addr_any(&addr))
1736                 return;
1737         __ipv6_dev_ac_dec(ifp->idev, &addr);
1738 }
1739
1740 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1741 {
1742         if (dev->addr_len != ETH_ALEN)
1743                 return -1;
1744         memcpy(eui, dev->dev_addr, 3);
1745         memcpy(eui + 5, dev->dev_addr + 3, 3);
1746
1747         /*
1748          * The zSeries OSA network cards can be shared among various
1749          * OS instances, but the OSA cards have only one MAC address.
1750          * This leads to duplicate address conflicts in conjunction
1751          * with IPv6 if more than one instance uses the same card.
1752          *
1753          * The driver for these cards can deliver a unique 16-bit
1754          * identifier for each instance sharing the same card.  It is
1755          * placed instead of 0xFFFE in the interface identifier.  The
1756          * "u" bit of the interface identifier is not inverted in this
1757          * case.  Hence the resulting interface identifier has local
1758          * scope according to RFC2373.
1759          */
1760         if (dev->dev_id) {
1761                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1762                 eui[4] = dev->dev_id & 0xFF;
1763         } else {
1764                 eui[3] = 0xFF;
1765                 eui[4] = 0xFE;
1766                 eui[0] ^= 2;
1767         }
1768         return 0;
1769 }
1770
1771 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1772 {
1773         if (dev->addr_len != IEEE802154_ADDR_LEN)
1774                 return -1;
1775         memcpy(eui, dev->dev_addr, 8);
1776         eui[0] ^= 2;
1777         return 0;
1778 }
1779
1780 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1781 {
1782         union fwnet_hwaddr *ha;
1783
1784         if (dev->addr_len != FWNET_ALEN)
1785                 return -1;
1786
1787         ha = (union fwnet_hwaddr *)dev->dev_addr;
1788
1789         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1790         eui[0] ^= 2;
1791         return 0;
1792 }
1793
1794 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1795 {
1796         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1797         if (dev->addr_len != ARCNET_ALEN)
1798                 return -1;
1799         memset(eui, 0, 7);
1800         eui[7] = *(u8 *)dev->dev_addr;
1801         return 0;
1802 }
1803
1804 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1805 {
1806         if (dev->addr_len != INFINIBAND_ALEN)
1807                 return -1;
1808         memcpy(eui, dev->dev_addr + 12, 8);
1809         eui[0] |= 2;
1810         return 0;
1811 }
1812
1813 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1814 {
1815         if (addr == 0)
1816                 return -1;
1817         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1818                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1819                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1820                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1821                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1822                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1823         eui[1] = 0;
1824         eui[2] = 0x5E;
1825         eui[3] = 0xFE;
1826         memcpy(eui + 4, &addr, 4);
1827         return 0;
1828 }
1829
1830 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1831 {
1832         if (dev->priv_flags & IFF_ISATAP)
1833                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1834         return -1;
1835 }
1836
1837 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1838 {
1839         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1840 }
1841
1842 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1843 {
1844         switch (dev->type) {
1845         case ARPHRD_ETHER:
1846         case ARPHRD_FDDI:
1847                 return addrconf_ifid_eui48(eui, dev);
1848         case ARPHRD_ARCNET:
1849                 return addrconf_ifid_arcnet(eui, dev);
1850         case ARPHRD_INFINIBAND:
1851                 return addrconf_ifid_infiniband(eui, dev);
1852         case ARPHRD_SIT:
1853                 return addrconf_ifid_sit(eui, dev);
1854         case ARPHRD_IPGRE:
1855                 return addrconf_ifid_gre(eui, dev);
1856         case ARPHRD_IEEE802154:
1857                 return addrconf_ifid_eui64(eui, dev);
1858         case ARPHRD_IEEE1394:
1859                 return addrconf_ifid_ieee1394(eui, dev);
1860         }
1861         return -1;
1862 }
1863
1864 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1865 {
1866         int err = -1;
1867         struct inet6_ifaddr *ifp;
1868
1869         read_lock_bh(&idev->lock);
1870         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1871                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1872                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1873                         err = 0;
1874                         break;
1875                 }
1876         }
1877         read_unlock_bh(&idev->lock);
1878         return err;
1879 }
1880
1881 #ifdef CONFIG_IPV6_PRIVACY
1882 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1883 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1884 {
1885 regen:
1886         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1887         idev->rndid[0] &= ~0x02;
1888
1889         /*
1890          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1891          * check if generated address is not inappropriate
1892          *
1893          *  - Reserved subnet anycast (RFC 2526)
1894          *      11111101 11....11 1xxxxxxx
1895          *  - ISATAP (RFC4214) 6.1
1896          *      00-00-5E-FE-xx-xx-xx-xx
1897          *  - value 0
1898          *  - XXX: already assigned to an address on the device
1899          */
1900         if (idev->rndid[0] == 0xfd &&
1901             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1902             (idev->rndid[7]&0x80))
1903                 goto regen;
1904         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1905                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1906                         goto regen;
1907                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1908                         goto regen;
1909         }
1910 }
1911
1912 static void ipv6_regen_rndid(unsigned long data)
1913 {
1914         struct inet6_dev *idev = (struct inet6_dev *) data;
1915         unsigned long expires;
1916
1917         rcu_read_lock_bh();
1918         write_lock_bh(&idev->lock);
1919
1920         if (idev->dead)
1921                 goto out;
1922
1923         __ipv6_regen_rndid(idev);
1924
1925         expires = jiffies +
1926                 idev->cnf.temp_prefered_lft * HZ -
1927                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1928                 idev->cnf.max_desync_factor * HZ;
1929         if (time_before(expires, jiffies)) {
1930                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1931                         __func__, idev->dev->name);
1932                 goto out;
1933         }
1934
1935         if (!mod_timer(&idev->regen_timer, expires))
1936                 in6_dev_hold(idev);
1937
1938 out:
1939         write_unlock_bh(&idev->lock);
1940         rcu_read_unlock_bh();
1941         in6_dev_put(idev);
1942 }
1943
1944 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1945 {
1946         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1947                 __ipv6_regen_rndid(idev);
1948 }
1949 #endif
1950
1951 u32 addrconf_rt_table(const struct net_device *dev, u32 default_table) {
1952         /* Determines into what table to put autoconf PIO/RIO/default routes
1953          * learned on this device.
1954          *
1955          * - If 0, use the same table for every device. This puts routes into
1956          *   one of RT_TABLE_{PREFIX,INFO,DFLT} depending on the type of route
1957          *   (but note that these three are currently all equal to
1958          *   RT6_TABLE_MAIN).
1959          * - If > 0, use the specified table.
1960          * - If < 0, put routes into table dev->ifindex + (-rt_table).
1961          */
1962         struct inet6_dev *idev = in6_dev_get(dev);
1963         u32 table;
1964         int sysctl = idev->cnf.accept_ra_rt_table;
1965         if (sysctl == 0) {
1966                 table = default_table;
1967         } else if (sysctl > 0) {
1968                 table = (u32) sysctl;
1969         } else {
1970                 table = (unsigned) dev->ifindex + (-sysctl);
1971         }
1972         in6_dev_put(idev);
1973         return table;
1974 }
1975
1976 /*
1977  *      Add prefix route.
1978  */
1979
1980 static void
1981 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1982                       unsigned long expires, u32 flags)
1983 {
1984         struct fib6_config cfg = {
1985                 .fc_table = addrconf_rt_table(dev, RT6_TABLE_PREFIX),
1986                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1987                 .fc_ifindex = dev->ifindex,
1988                 .fc_expires = expires,
1989                 .fc_dst_len = plen,
1990                 .fc_flags = RTF_UP | flags,
1991                 .fc_nlinfo.nl_net = dev_net(dev),
1992                 .fc_protocol = RTPROT_KERNEL,
1993         };
1994
1995         cfg.fc_dst = *pfx;
1996
1997         /* Prevent useless cloning on PtP SIT.
1998            This thing is done here expecting that the whole
1999            class of non-broadcast devices need not cloning.
2000          */
2001 #if IS_ENABLED(CONFIG_IPV6_SIT)
2002         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2003                 cfg.fc_flags |= RTF_NONEXTHOP;
2004 #endif
2005
2006         ip6_route_add(&cfg);
2007 }
2008
2009
2010 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2011                                                   int plen,
2012                                                   const struct net_device *dev,
2013                                                   u32 flags, u32 noflags)
2014 {
2015         struct fib6_node *fn;
2016         struct rt6_info *rt = NULL;
2017         struct fib6_table *table;
2018
2019         table = fib6_get_table(dev_net(dev),
2020                                addrconf_rt_table(dev, RT6_TABLE_PREFIX));
2021         if (table == NULL)
2022                 return NULL;
2023
2024         read_lock_bh(&table->tb6_lock);
2025         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2026         if (!fn)
2027                 goto out;
2028         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2029                 if (rt->dst.dev->ifindex != dev->ifindex)
2030                         continue;
2031                 if ((rt->rt6i_flags & flags) != flags)
2032                         continue;
2033                 if ((rt->rt6i_flags & noflags) != 0)
2034                         continue;
2035                 dst_hold(&rt->dst);
2036                 break;
2037         }
2038 out:
2039         read_unlock_bh(&table->tb6_lock);
2040         return rt;
2041 }
2042
2043
2044 /* Create "default" multicast route to the interface */
2045
2046 static void addrconf_add_mroute(struct net_device *dev)
2047 {
2048         struct fib6_config cfg = {
2049                 .fc_table = RT6_TABLE_LOCAL,
2050                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2051                 .fc_ifindex = dev->ifindex,
2052                 .fc_dst_len = 8,
2053                 .fc_flags = RTF_UP,
2054                 .fc_nlinfo.nl_net = dev_net(dev),
2055         };
2056
2057         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2058
2059         ip6_route_add(&cfg);
2060 }
2061
2062 #if IS_ENABLED(CONFIG_IPV6_SIT)
2063 static void sit_route_add(struct net_device *dev)
2064 {
2065         struct fib6_config cfg = {
2066                 .fc_table = RT6_TABLE_MAIN,
2067                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2068                 .fc_ifindex = dev->ifindex,
2069                 .fc_dst_len = 96,
2070                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2071                 .fc_nlinfo.nl_net = dev_net(dev),
2072         };
2073
2074         /* prefix length - 96 bits "::d.d.d.d" */
2075         ip6_route_add(&cfg);
2076 }
2077 #endif
2078
2079 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2080 {
2081         struct inet6_dev *idev;
2082
2083         ASSERT_RTNL();
2084
2085         idev = ipv6_find_idev(dev);
2086         if (!idev)
2087                 return ERR_PTR(-ENOBUFS);
2088
2089         if (idev->cnf.disable_ipv6)
2090                 return ERR_PTR(-EACCES);
2091
2092         /* Add default multicast route */
2093         if (!(dev->flags & IFF_LOOPBACK))
2094                 addrconf_add_mroute(dev);
2095
2096         return idev;
2097 }
2098
2099 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2100 {
2101         struct prefix_info *pinfo;
2102         __u32 valid_lft;
2103         __u32 prefered_lft;
2104         int addr_type;
2105         struct inet6_dev *in6_dev;
2106         struct net *net = dev_net(dev);
2107
2108         pinfo = (struct prefix_info *) opt;
2109
2110         if (len < sizeof(struct prefix_info)) {
2111                 ADBG(("addrconf: prefix option too short\n"));
2112                 return;
2113         }
2114
2115         /*
2116          *      Validation checks ([ADDRCONF], page 19)
2117          */
2118
2119         addr_type = ipv6_addr_type(&pinfo->prefix);
2120
2121         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2122                 return;
2123
2124         valid_lft = ntohl(pinfo->valid);
2125         prefered_lft = ntohl(pinfo->prefered);
2126
2127         if (prefered_lft > valid_lft) {
2128                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2129                 return;
2130         }
2131
2132         in6_dev = in6_dev_get(dev);
2133
2134         if (in6_dev == NULL) {
2135                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2136                                     dev->name);
2137                 return;
2138         }
2139
2140         /*
2141          *      Two things going on here:
2142          *      1) Add routes for on-link prefixes
2143          *      2) Configure prefixes with the auto flag set
2144          */
2145
2146         if (pinfo->onlink) {
2147                 struct rt6_info *rt;
2148                 unsigned long rt_expires;
2149
2150                 /* Avoid arithmetic overflow. Really, we could
2151                  * save rt_expires in seconds, likely valid_lft,
2152                  * but it would require division in fib gc, that it
2153                  * not good.
2154                  */
2155                 if (HZ > USER_HZ)
2156                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2157                 else
2158                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2159
2160                 if (addrconf_finite_timeout(rt_expires))
2161                         rt_expires *= HZ;
2162
2163                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2164                                                pinfo->prefix_len,
2165                                                dev,
2166                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2167                                                RTF_GATEWAY | RTF_DEFAULT);
2168
2169                 if (rt) {
2170                         /* Autoconf prefix route */
2171                         if (valid_lft == 0) {
2172                                 ip6_del_rt(rt);
2173                                 rt = NULL;
2174                         } else if (addrconf_finite_timeout(rt_expires)) {
2175                                 /* not infinity */
2176                                 rt6_set_expires(rt, jiffies + rt_expires);
2177                         } else {
2178                                 rt6_clean_expires(rt);
2179                         }
2180                 } else if (valid_lft) {
2181                         clock_t expires = 0;
2182                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2183                         if (addrconf_finite_timeout(rt_expires)) {
2184                                 /* not infinity */
2185                                 flags |= RTF_EXPIRES;
2186                                 expires = jiffies_to_clock_t(rt_expires);
2187                         }
2188                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2189                                               dev, expires, flags);
2190                 }
2191                 ip6_rt_put(rt);
2192         }
2193
2194         /* Try to figure out our local address for this prefix */
2195
2196         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2197                 struct inet6_ifaddr *ifp;
2198                 struct in6_addr addr;
2199                 int create = 0, update_lft = 0;
2200                 bool tokenized = false;
2201
2202                 if (pinfo->prefix_len == 64) {
2203                         memcpy(&addr, &pinfo->prefix, 8);
2204
2205                         if (!ipv6_addr_any(&in6_dev->token)) {
2206                                 read_lock_bh(&in6_dev->lock);
2207                                 memcpy(addr.s6_addr + 8,
2208                                        in6_dev->token.s6_addr + 8, 8);
2209                                 read_unlock_bh(&in6_dev->lock);
2210                                 tokenized = true;
2211                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2212                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2213                                 in6_dev_put(in6_dev);
2214                                 return;
2215                         }
2216                         goto ok;
2217                 }
2218                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2219                                     pinfo->prefix_len);
2220                 in6_dev_put(in6_dev);
2221                 return;
2222
2223 ok:
2224
2225                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2226
2227                 if (ifp == NULL && valid_lft) {
2228                         int max_addresses = in6_dev->cnf.max_addresses;
2229                         u32 addr_flags = 0;
2230
2231 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2232                         if (in6_dev->cnf.optimistic_dad &&
2233                             !net->ipv6.devconf_all->forwarding && sllao)
2234                                 addr_flags = IFA_F_OPTIMISTIC;
2235 #endif
2236
2237                         /* Do not allow to create too much of autoconfigured
2238                          * addresses; this would be too easy way to crash kernel.
2239                          */
2240                         if (!max_addresses ||
2241                             ipv6_count_addresses(in6_dev) < max_addresses)
2242                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2243                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2244                                                     addr_flags);
2245
2246                         if (IS_ERR_OR_NULL(ifp)) {
2247                                 in6_dev_put(in6_dev);
2248                                 return;
2249                         }
2250
2251                         update_lft = create = 1;
2252                         ifp->cstamp = jiffies;
2253                         ifp->tokenized = tokenized;
2254                         addrconf_dad_start(ifp);
2255                 }
2256
2257                 if (ifp) {
2258                         int flags;
2259                         unsigned long now;
2260 #ifdef CONFIG_IPV6_PRIVACY
2261                         struct inet6_ifaddr *ift;
2262 #endif
2263                         u32 stored_lft;
2264
2265                         /* update lifetime (RFC2462 5.5.3 e) */
2266                         spin_lock(&ifp->lock);
2267                         now = jiffies;
2268                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2269                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2270                         else
2271                                 stored_lft = 0;
2272                         if (!update_lft && stored_lft) {
2273                                 if (valid_lft > MIN_VALID_LIFETIME ||
2274                                     valid_lft > stored_lft)
2275                                         update_lft = 1;
2276                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2277                                         /* valid_lft <= stored_lft is always true */
2278                                         /*
2279                                          * RFC 4862 Section 5.5.3e:
2280                                          * "Note that the preferred lifetime of
2281                                          *  the corresponding address is always
2282                                          *  reset to the Preferred Lifetime in
2283                                          *  the received Prefix Information
2284                                          *  option, regardless of whether the
2285                                          *  valid lifetime is also reset or
2286                                          *  ignored."
2287                                          *
2288                                          *  So if the preferred lifetime in
2289                                          *  this advertisement is different
2290                                          *  than what we have stored, but the
2291                                          *  valid lifetime is invalid, just
2292                                          *  reset prefered_lft.
2293                                          *
2294                                          *  We must set the valid lifetime
2295                                          *  to the stored lifetime since we'll
2296                                          *  be updating the timestamp below,
2297                                          *  else we'll set it back to the
2298                                          *  minimum.
2299                                          */
2300                                         if (prefered_lft != ifp->prefered_lft) {
2301                                                 valid_lft = stored_lft;
2302                                                 update_lft = 1;
2303                                         }
2304                                 } else {
2305                                         valid_lft = MIN_VALID_LIFETIME;
2306                                         if (valid_lft < prefered_lft)
2307                                                 prefered_lft = valid_lft;
2308                                         update_lft = 1;
2309                                 }
2310                         }
2311
2312                         if (update_lft) {
2313                                 ifp->valid_lft = valid_lft;
2314                                 ifp->prefered_lft = prefered_lft;
2315                                 ifp->tstamp = now;
2316                                 flags = ifp->flags;
2317                                 ifp->flags &= ~IFA_F_DEPRECATED;
2318                                 spin_unlock(&ifp->lock);
2319
2320                                 if (!(flags&IFA_F_TENTATIVE))
2321                                         ipv6_ifa_notify(0, ifp);
2322                         } else
2323                                 spin_unlock(&ifp->lock);
2324
2325 #ifdef CONFIG_IPV6_PRIVACY
2326                         read_lock_bh(&in6_dev->lock);
2327                         /* update all temporary addresses in the list */
2328                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2329                                             tmp_list) {
2330                                 int age, max_valid, max_prefered;
2331
2332                                 if (ifp != ift->ifpub)
2333                                         continue;
2334
2335                                 /*
2336                                  * RFC 4941 section 3.3:
2337                                  * If a received option will extend the lifetime
2338                                  * of a public address, the lifetimes of
2339                                  * temporary addresses should be extended,
2340                                  * subject to the overall constraint that no
2341                                  * temporary addresses should ever remain
2342                                  * "valid" or "preferred" for a time longer than
2343                                  * (TEMP_VALID_LIFETIME) or
2344                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2345                                  * respectively.
2346                                  */
2347                                 age = (now - ift->cstamp) / HZ;
2348                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2349                                 if (max_valid < 0)
2350                                         max_valid = 0;
2351
2352                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2353                                                in6_dev->cnf.max_desync_factor -
2354                                                age;
2355                                 if (max_prefered < 0)
2356                                         max_prefered = 0;
2357
2358                                 if (valid_lft > max_valid)
2359                                         valid_lft = max_valid;
2360
2361                                 if (prefered_lft > max_prefered)
2362                                         prefered_lft = max_prefered;
2363
2364                                 spin_lock(&ift->lock);
2365                                 flags = ift->flags;
2366                                 ift->valid_lft = valid_lft;
2367                                 ift->prefered_lft = prefered_lft;
2368                                 ift->tstamp = now;
2369                                 if (prefered_lft > 0)
2370                                         ift->flags &= ~IFA_F_DEPRECATED;
2371
2372                                 spin_unlock(&ift->lock);
2373                                 if (!(flags&IFA_F_TENTATIVE))
2374                                         ipv6_ifa_notify(0, ift);
2375                         }
2376
2377                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2378                                 /*
2379                                  * When a new public address is created as
2380                                  * described in [ADDRCONF], also create a new
2381                                  * temporary address. Also create a temporary
2382                                  * address if it's enabled but no temporary
2383                                  * address currently exists.
2384                                  */
2385                                 read_unlock_bh(&in6_dev->lock);
2386                                 ipv6_create_tempaddr(ifp, NULL);
2387                         } else {
2388                                 read_unlock_bh(&in6_dev->lock);
2389                         }
2390 #endif
2391                         in6_ifa_put(ifp);
2392                         addrconf_verify(0);
2393                 }
2394         }
2395         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2396         in6_dev_put(in6_dev);
2397 }
2398
2399 /*
2400  *      Set destination address.
2401  *      Special case for SIT interfaces where we create a new "virtual"
2402  *      device.
2403  */
2404 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2405 {
2406         struct in6_ifreq ireq;
2407         struct net_device *dev;
2408         int err = -EINVAL;
2409
2410         rtnl_lock();
2411
2412         err = -EFAULT;
2413         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2414                 goto err_exit;
2415
2416         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2417
2418         err = -ENODEV;
2419         if (dev == NULL)
2420                 goto err_exit;
2421
2422 #if IS_ENABLED(CONFIG_IPV6_SIT)
2423         if (dev->type == ARPHRD_SIT) {
2424                 const struct net_device_ops *ops = dev->netdev_ops;
2425                 struct ifreq ifr;
2426                 struct ip_tunnel_parm p;
2427
2428                 err = -EADDRNOTAVAIL;
2429                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2430                         goto err_exit;
2431
2432                 memset(&p, 0, sizeof(p));
2433                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2434                 p.iph.saddr = 0;
2435                 p.iph.version = 4;
2436                 p.iph.ihl = 5;
2437                 p.iph.protocol = IPPROTO_IPV6;
2438                 p.iph.ttl = 64;
2439                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2440
2441                 if (ops->ndo_do_ioctl) {
2442                         mm_segment_t oldfs = get_fs();
2443
2444                         set_fs(KERNEL_DS);
2445                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2446                         set_fs(oldfs);
2447                 } else
2448                         err = -EOPNOTSUPP;
2449
2450                 if (err == 0) {
2451                         err = -ENOBUFS;
2452                         dev = __dev_get_by_name(net, p.name);
2453                         if (!dev)
2454                                 goto err_exit;
2455                         err = dev_open(dev);
2456                 }
2457         }
2458 #endif
2459
2460 err_exit:
2461         rtnl_unlock();
2462         return err;
2463 }
2464
2465 /*
2466  *      Manual configuration of address on an interface
2467  */
2468 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2469                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2470                           __u32 valid_lft)
2471 {
2472         struct inet6_ifaddr *ifp;
2473         struct inet6_dev *idev;
2474         struct net_device *dev;
2475         int scope;
2476         u32 flags;
2477         clock_t expires;
2478         unsigned long timeout;
2479
2480         ASSERT_RTNL();
2481
2482         if (plen > 128)
2483                 return -EINVAL;
2484
2485         /* check the lifetime */
2486         if (!valid_lft || prefered_lft > valid_lft)
2487                 return -EINVAL;
2488
2489         dev = __dev_get_by_index(net, ifindex);
2490         if (!dev)
2491                 return -ENODEV;
2492
2493         idev = addrconf_add_dev(dev);
2494         if (IS_ERR(idev))
2495                 return PTR_ERR(idev);
2496
2497         scope = ipv6_addr_scope(pfx);
2498
2499         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2500         if (addrconf_finite_timeout(timeout)) {
2501                 expires = jiffies_to_clock_t(timeout * HZ);
2502                 valid_lft = timeout;
2503                 flags = RTF_EXPIRES;
2504         } else {
2505                 expires = 0;
2506                 flags = 0;
2507                 ifa_flags |= IFA_F_PERMANENT;
2508         }
2509
2510         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2511         if (addrconf_finite_timeout(timeout)) {
2512                 if (timeout == 0)
2513                         ifa_flags |= IFA_F_DEPRECATED;
2514                 prefered_lft = timeout;
2515         }
2516
2517         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2518
2519         if (!IS_ERR(ifp)) {
2520                 spin_lock_bh(&ifp->lock);
2521                 ifp->valid_lft = valid_lft;
2522                 ifp->prefered_lft = prefered_lft;
2523                 ifp->tstamp = jiffies;
2524                 spin_unlock_bh(&ifp->lock);
2525
2526                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2527                                       expires, flags);
2528                 /*
2529                  * Note that section 3.1 of RFC 4429 indicates
2530                  * that the Optimistic flag should not be set for
2531                  * manually configured addresses
2532                  */
2533                 addrconf_dad_start(ifp);
2534                 in6_ifa_put(ifp);
2535                 addrconf_verify(0);
2536                 return 0;
2537         }
2538
2539         return PTR_ERR(ifp);
2540 }
2541
2542 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2543                           unsigned int plen)
2544 {
2545         struct inet6_ifaddr *ifp;
2546         struct inet6_dev *idev;
2547         struct net_device *dev;
2548
2549         if (plen > 128)
2550                 return -EINVAL;
2551
2552         dev = __dev_get_by_index(net, ifindex);
2553         if (!dev)
2554                 return -ENODEV;
2555
2556         if ((idev = __in6_dev_get(dev)) == NULL)
2557                 return -ENXIO;
2558
2559         read_lock_bh(&idev->lock);
2560         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2561                 if (ifp->prefix_len == plen &&
2562                     ipv6_addr_equal(pfx, &ifp->addr)) {
2563                         in6_ifa_hold(ifp);
2564                         read_unlock_bh(&idev->lock);
2565
2566                         ipv6_del_addr(ifp);
2567
2568                         /* If the last address is deleted administratively,
2569                            disable IPv6 on this interface.
2570                          */
2571                         if (list_empty(&idev->addr_list))
2572                                 addrconf_ifdown(idev->dev, 1);
2573                         return 0;
2574                 }
2575         }
2576         read_unlock_bh(&idev->lock);
2577         return -EADDRNOTAVAIL;
2578 }
2579
2580
2581 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2582 {
2583         struct in6_ifreq ireq;
2584         int err;
2585
2586         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2587                 return -EPERM;
2588
2589         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2590                 return -EFAULT;
2591
2592         rtnl_lock();
2593         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2594                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2595                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2596         rtnl_unlock();
2597         return err;
2598 }
2599
2600 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2601 {
2602         struct in6_ifreq ireq;
2603         int err;
2604
2605         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2606                 return -EPERM;
2607
2608         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2609                 return -EFAULT;
2610
2611         rtnl_lock();
2612         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2613                              ireq.ifr6_prefixlen);
2614         rtnl_unlock();
2615         return err;
2616 }
2617
2618 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2619                      int plen, int scope)
2620 {
2621         struct inet6_ifaddr *ifp;
2622
2623         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2624         if (!IS_ERR(ifp)) {
2625                 spin_lock_bh(&ifp->lock);
2626                 ifp->flags &= ~IFA_F_TENTATIVE;
2627                 spin_unlock_bh(&ifp->lock);
2628                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2629                 in6_ifa_put(ifp);
2630         }
2631 }
2632
2633 #if IS_ENABLED(CONFIG_IPV6_SIT)
2634 static void sit_add_v4_addrs(struct inet6_dev *idev)
2635 {
2636         struct in6_addr addr;
2637         struct net_device *dev;
2638         struct net *net = dev_net(idev->dev);
2639         int scope;
2640
2641         ASSERT_RTNL();
2642
2643         memset(&addr, 0, sizeof(struct in6_addr));
2644         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2645
2646         if (idev->dev->flags&IFF_POINTOPOINT) {
2647                 addr.s6_addr32[0] = htonl(0xfe800000);
2648                 scope = IFA_LINK;
2649         } else {
2650                 scope = IPV6_ADDR_COMPATv4;
2651         }
2652
2653         if (addr.s6_addr32[3]) {
2654                 add_addr(idev, &addr, 128, scope);
2655                 return;
2656         }
2657
2658         for_each_netdev(net, dev) {
2659                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2660                 if (in_dev && (dev->flags & IFF_UP)) {
2661                         struct in_ifaddr *ifa;
2662
2663                         int flag = scope;
2664
2665                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2666                                 int plen;
2667
2668                                 addr.s6_addr32[3] = ifa->ifa_local;
2669
2670                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2671                                         continue;
2672                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2673                                         if (idev->dev->flags&IFF_POINTOPOINT)
2674                                                 continue;
2675                                         flag |= IFA_HOST;
2676                                 }
2677                                 if (idev->dev->flags&IFF_POINTOPOINT)
2678                                         plen = 64;
2679                                 else
2680                                         plen = 96;
2681
2682                                 add_addr(idev, &addr, plen, flag);
2683                         }
2684                 }
2685         }
2686 }
2687 #endif
2688
2689 static void init_loopback(struct net_device *dev)
2690 {
2691         struct inet6_dev  *idev;
2692         struct net_device *sp_dev;
2693         struct inet6_ifaddr *sp_ifa;
2694         struct rt6_info *sp_rt;
2695
2696         /* ::1 */
2697
2698         ASSERT_RTNL();
2699
2700         if ((idev = ipv6_find_idev(dev)) == NULL) {
2701                 pr_debug("%s: add_dev failed\n", __func__);
2702                 return;
2703         }
2704
2705         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2706
2707         /* Add routes to other interface's IPv6 addresses */
2708         for_each_netdev(dev_net(dev), sp_dev) {
2709                 if (!strcmp(sp_dev->name, dev->name))
2710                         continue;
2711
2712                 idev = __in6_dev_get(sp_dev);
2713                 if (!idev)
2714                         continue;
2715
2716                 read_lock_bh(&idev->lock);
2717                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2718
2719                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2720                                 continue;
2721
2722                         if (sp_ifa->rt)
2723                                 continue;
2724
2725                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2726
2727                         /* Failure cases are ignored */
2728                         if (!IS_ERR(sp_rt)) {
2729                                 sp_ifa->rt = sp_rt;
2730                                 ip6_ins_rt(sp_rt);
2731                         }
2732                 }
2733                 read_unlock_bh(&idev->lock);
2734         }
2735 }
2736
2737 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2738 {
2739         struct inet6_ifaddr *ifp;
2740         u32 addr_flags = IFA_F_PERMANENT;
2741
2742 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2743         if (idev->cnf.optimistic_dad &&
2744             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2745                 addr_flags |= IFA_F_OPTIMISTIC;
2746 #endif
2747
2748
2749         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2750         if (!IS_ERR(ifp)) {
2751                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2752                 addrconf_dad_start(ifp);
2753                 in6_ifa_put(ifp);
2754         }
2755 }
2756
2757 static void addrconf_dev_config(struct net_device *dev)
2758 {
2759         struct in6_addr addr;
2760         struct inet6_dev *idev;
2761
2762         ASSERT_RTNL();
2763
2764         if ((dev->type != ARPHRD_ETHER) &&
2765             (dev->type != ARPHRD_FDDI) &&
2766             (dev->type != ARPHRD_ARCNET) &&
2767             (dev->type != ARPHRD_INFINIBAND) &&
2768             (dev->type != ARPHRD_IEEE802154) &&
2769             (dev->type != ARPHRD_IEEE1394)) {
2770                 /* Alas, we support only Ethernet autoconfiguration. */
2771                 return;
2772         }
2773
2774         idev = addrconf_add_dev(dev);
2775         if (IS_ERR(idev))
2776                 return;
2777
2778         memset(&addr, 0, sizeof(struct in6_addr));
2779         addr.s6_addr32[0] = htonl(0xFE800000);
2780
2781         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2782                 addrconf_add_linklocal(idev, &addr);
2783 }
2784
2785 #if IS_ENABLED(CONFIG_IPV6_SIT)
2786 static void addrconf_sit_config(struct net_device *dev)
2787 {
2788         struct inet6_dev *idev;
2789
2790         ASSERT_RTNL();
2791
2792         /*
2793          * Configure the tunnel with one of our IPv4
2794          * addresses... we should configure all of
2795          * our v4 addrs in the tunnel
2796          */
2797
2798         if ((idev = ipv6_find_idev(dev)) == NULL) {
2799                 pr_debug("%s: add_dev failed\n", __func__);
2800                 return;
2801         }
2802
2803         if (dev->priv_flags & IFF_ISATAP) {
2804                 struct in6_addr addr;
2805
2806                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2807                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2808                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2809                         addrconf_add_linklocal(idev, &addr);
2810                 return;
2811         }
2812
2813         sit_add_v4_addrs(idev);
2814
2815         if (dev->flags&IFF_POINTOPOINT)
2816                 addrconf_add_mroute(dev);
2817         else
2818                 sit_route_add(dev);
2819 }
2820 #endif
2821
2822 #if IS_ENABLED(CONFIG_NET_IPGRE)
2823 static void addrconf_gre_config(struct net_device *dev)
2824 {
2825         struct inet6_dev *idev;
2826         struct in6_addr addr;
2827
2828         pr_info("%s(%s)\n", __func__, dev->name);
2829
2830         ASSERT_RTNL();
2831
2832         if ((idev = ipv6_find_idev(dev)) == NULL) {
2833                 pr_debug("%s: add_dev failed\n", __func__);
2834                 return;
2835         }
2836
2837         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2838         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2839
2840         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2841                 addrconf_add_linklocal(idev, &addr);
2842 }
2843 #endif
2844
2845 static inline int
2846 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2847 {
2848         struct in6_addr lladdr;
2849
2850         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2851                 addrconf_add_linklocal(idev, &lladdr);
2852                 return 0;
2853         }
2854         return -1;
2855 }
2856
2857 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2858 {
2859         struct net_device *link_dev;
2860         struct net *net = dev_net(idev->dev);
2861
2862         /* first try to inherit the link-local address from the link device */
2863         if (idev->dev->iflink &&
2864             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2865                 if (!ipv6_inherit_linklocal(idev, link_dev))
2866                         return;
2867         }
2868         /* then try to inherit it from any device */
2869         for_each_netdev(net, link_dev) {
2870                 if (!ipv6_inherit_linklocal(idev, link_dev))
2871                         return;
2872         }
2873         pr_debug("init ip6-ip6: add_linklocal failed\n");
2874 }
2875
2876 /*
2877  * Autoconfigure tunnel with a link-local address so routing protocols,
2878  * DHCPv6, MLD etc. can be run over the virtual link
2879  */
2880
2881 static void addrconf_ip6_tnl_config(struct net_device *dev)
2882 {
2883         struct inet6_dev *idev;
2884
2885         ASSERT_RTNL();
2886
2887         idev = addrconf_add_dev(dev);
2888         if (IS_ERR(idev)) {
2889                 pr_debug("init ip6-ip6: add_dev failed\n");
2890                 return;
2891         }
2892         ip6_tnl_add_linklocal(idev);
2893 }
2894
2895 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2896                            void *data)
2897 {
2898         struct net_device *dev = (struct net_device *) data;
2899         struct inet6_dev *idev = __in6_dev_get(dev);
2900         int run_pending = 0;
2901         int err;
2902
2903         switch (event) {
2904         case NETDEV_REGISTER:
2905                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2906                         idev = ipv6_add_dev(dev);
2907                         if (!idev)
2908                                 return notifier_from_errno(-ENOMEM);
2909                 }
2910                 break;
2911
2912         case NETDEV_UP:
2913         case NETDEV_CHANGE:
2914                 if (dev->flags & IFF_SLAVE)
2915                         break;
2916
2917                 if (event == NETDEV_UP) {
2918                         if (!addrconf_qdisc_ok(dev)) {
2919                                 /* device is not ready yet. */
2920                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2921                                         dev->name);
2922                                 break;
2923                         }
2924
2925                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2926                                 idev = ipv6_add_dev(dev);
2927
2928                         if (idev) {
2929                                 idev->if_flags |= IF_READY;
2930                                 run_pending = 1;
2931                         }
2932                 } else {
2933                         if (!addrconf_qdisc_ok(dev)) {
2934                                 /* device is still not ready. */
2935                                 break;
2936                         }
2937
2938                         if (idev) {
2939                                 if (idev->if_flags & IF_READY)
2940                                         /* device is already configured. */
2941                                         break;
2942                                 idev->if_flags |= IF_READY;
2943                         }
2944
2945                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2946                                 dev->name);
2947
2948                         run_pending = 1;
2949                 }
2950
2951                 switch (dev->type) {
2952 #if IS_ENABLED(CONFIG_IPV6_SIT)
2953                 case ARPHRD_SIT:
2954                         addrconf_sit_config(dev);
2955                         break;
2956 #endif
2957 #if IS_ENABLED(CONFIG_NET_IPGRE)
2958                 case ARPHRD_IPGRE:
2959                         addrconf_gre_config(dev);
2960                         break;
2961 #endif
2962                 case ARPHRD_TUNNEL6:
2963                         addrconf_ip6_tnl_config(dev);
2964                         break;
2965                 case ARPHRD_LOOPBACK:
2966                         init_loopback(dev);
2967                         break;
2968
2969                 default:
2970                         addrconf_dev_config(dev);
2971                         break;
2972                 }
2973
2974                 if (idev) {
2975                         if (run_pending)
2976                                 addrconf_dad_run(idev);
2977
2978                         /*
2979                          * If the MTU changed during the interface down,
2980                          * when the interface up, the changed MTU must be
2981                          * reflected in the idev as well as routers.
2982                          */
2983                         if (idev->cnf.mtu6 != dev->mtu &&
2984                             dev->mtu >= IPV6_MIN_MTU) {
2985                                 rt6_mtu_change(dev, dev->mtu);
2986                                 idev->cnf.mtu6 = dev->mtu;
2987                         }
2988                         idev->tstamp = jiffies;
2989                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2990
2991                         /*
2992                          * If the changed mtu during down is lower than
2993                          * IPV6_MIN_MTU stop IPv6 on this interface.
2994                          */
2995                         if (dev->mtu < IPV6_MIN_MTU)
2996                                 addrconf_ifdown(dev, 1);
2997                 }
2998                 break;
2999
3000         case NETDEV_CHANGEMTU:
3001                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3002                         rt6_mtu_change(dev, dev->mtu);
3003                         idev->cnf.mtu6 = dev->mtu;
3004                         break;
3005                 }
3006
3007                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3008                         idev = ipv6_add_dev(dev);
3009                         if (idev)
3010                                 break;
3011                 }
3012
3013                 /*
3014                  * MTU falled under IPV6_MIN_MTU.
3015                  * Stop IPv6 on this interface.
3016                  */
3017
3018         case NETDEV_DOWN:
3019         case NETDEV_UNREGISTER:
3020                 /*
3021                  *      Remove all addresses from this interface.
3022                  */
3023                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3024                 break;
3025
3026         case NETDEV_CHANGENAME:
3027                 if (idev) {
3028                         snmp6_unregister_dev(idev);
3029                         addrconf_sysctl_unregister(idev);
3030                         addrconf_sysctl_register(idev);
3031                         err = snmp6_register_dev(idev);
3032                         if (err)
3033                                 return notifier_from_errno(err);
3034                 }
3035                 break;
3036
3037         case NETDEV_PRE_TYPE_CHANGE:
3038         case NETDEV_POST_TYPE_CHANGE:
3039                 addrconf_type_change(dev, event);
3040                 break;
3041         }
3042
3043         return NOTIFY_OK;
3044 }
3045
3046 /*
3047  *      addrconf module should be notified of a device going up
3048  */
3049 static struct notifier_block ipv6_dev_notf = {
3050         .notifier_call = addrconf_notify,
3051 };
3052
3053 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3054 {
3055         struct inet6_dev *idev;
3056         ASSERT_RTNL();
3057
3058         idev = __in6_dev_get(dev);
3059
3060         if (event == NETDEV_POST_TYPE_CHANGE)
3061                 ipv6_mc_remap(idev);
3062         else if (event == NETDEV_PRE_TYPE_CHANGE)
3063                 ipv6_mc_unmap(idev);
3064 }
3065
3066 static int addrconf_ifdown(struct net_device *dev, int how)
3067 {
3068         struct net *net = dev_net(dev);
3069         struct inet6_dev *idev;
3070         struct inet6_ifaddr *ifa;
3071         int state, i;
3072
3073         ASSERT_RTNL();
3074
3075         rt6_ifdown(net, dev);
3076         neigh_ifdown(&nd_tbl, dev);
3077
3078         idev = __in6_dev_get(dev);
3079         if (idev == NULL)
3080                 return -ENODEV;
3081
3082         /*
3083          * Step 1: remove reference to ipv6 device from parent device.
3084          *         Do not dev_put!
3085          */
3086         if (how) {
3087                 idev->dead = 1;
3088
3089                 /* protected by rtnl_lock */
3090                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3091
3092                 /* Step 1.5: remove snmp6 entry */
3093                 snmp6_unregister_dev(idev);
3094
3095         }
3096
3097         /* Step 2: clear hash table */
3098         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3099                 struct hlist_head *h = &inet6_addr_lst[i];
3100
3101                 spin_lock_bh(&addrconf_hash_lock);
3102         restart:
3103                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3104                         if (ifa->idev == idev) {
3105                                 hlist_del_init_rcu(&ifa->addr_lst);
3106                                 addrconf_del_timer(ifa);
3107                                 goto restart;
3108                         }
3109                 }
3110                 spin_unlock_bh(&addrconf_hash_lock);
3111         }
3112
3113         write_lock_bh(&idev->lock);
3114
3115         /* Step 2: clear flags for stateless addrconf */
3116         if (!how)
3117                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3118
3119 #ifdef CONFIG_IPV6_PRIVACY
3120         if (how && del_timer(&idev->regen_timer))
3121                 in6_dev_put(idev);
3122
3123         /* Step 3: clear tempaddr list */
3124         while (!list_empty(&idev->tempaddr_list)) {
3125                 ifa = list_first_entry(&idev->tempaddr_list,
3126                                        struct inet6_ifaddr, tmp_list);
3127                 list_del(&ifa->tmp_list);
3128                 write_unlock_bh(&idev->lock);
3129                 spin_lock_bh(&ifa->lock);
3130
3131                 if (ifa->ifpub) {
3132                         in6_ifa_put(ifa->ifpub);
3133                         ifa->ifpub = NULL;
3134                 }
3135                 spin_unlock_bh(&ifa->lock);
3136                 in6_ifa_put(ifa);
3137                 write_lock_bh(&idev->lock);
3138         }
3139 #endif
3140
3141         while (!list_empty(&idev->addr_list)) {
3142                 ifa = list_first_entry(&idev->addr_list,
3143                                        struct inet6_ifaddr, if_list);
3144                 addrconf_del_timer(ifa);
3145
3146                 list_del(&ifa->if_list);
3147
3148                 write_unlock_bh(&idev->lock);
3149
3150                 spin_lock_bh(&ifa->state_lock);
3151                 state = ifa->state;
3152                 ifa->state = INET6_IFADDR_STATE_DEAD;
3153                 spin_unlock_bh(&ifa->state_lock);
3154
3155                 if (state != INET6_IFADDR_STATE_DEAD) {
3156                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3157                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3158                 }
3159                 in6_ifa_put(ifa);
3160
3161                 write_lock_bh(&idev->lock);
3162         }
3163
3164         write_unlock_bh(&idev->lock);
3165
3166         /* Step 5: Discard multicast list */
3167         if (how)
3168                 ipv6_mc_destroy_dev(idev);
3169         else
3170                 ipv6_mc_down(idev);
3171
3172         idev->tstamp = jiffies;
3173
3174         /* Last: Shot the device (if unregistered) */
3175         if (how) {
3176                 addrconf_sysctl_unregister(idev);
3177                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3178                 neigh_ifdown(&nd_tbl, dev);
3179                 in6_dev_put(idev);
3180         }
3181         return 0;
3182 }
3183
3184 static void addrconf_rs_timer(unsigned long data)
3185 {
3186         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3187         struct inet6_dev *idev = ifp->idev;
3188
3189         read_lock(&idev->lock);
3190         if (idev->dead || !(idev->if_flags & IF_READY))
3191                 goto out;
3192
3193         if (!ipv6_accept_ra(idev))
3194                 goto out;
3195
3196         /* Announcement received after solicitation was sent */
3197         if (idev->if_flags & IF_RA_RCVD)
3198                 goto out;
3199
3200         spin_lock(&ifp->lock);
3201         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3202                 /* The wait after the last probe can be shorter */
3203                 addrconf_mod_timer(ifp, AC_RS,
3204                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3205                                    idev->cnf.rtr_solicit_delay :
3206                                    idev->cnf.rtr_solicit_interval);
3207                 spin_unlock(&ifp->lock);
3208
3209                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3210         } else {
3211                 spin_unlock(&ifp->lock);
3212                 /*
3213                  * Note: we do not support deprecated "all on-link"
3214                  * assumption any longer.
3215                  */
3216                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3217         }
3218
3219 out:
3220         read_unlock(&idev->lock);
3221         in6_ifa_put(ifp);
3222 }
3223
3224 /*
3225  *      Duplicate Address Detection
3226  */
3227 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3228 {
3229         unsigned long rand_num;
3230         struct inet6_dev *idev = ifp->idev;
3231
3232         if (ifp->flags & IFA_F_OPTIMISTIC)
3233                 rand_num = 0;
3234         else
3235                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3236
3237         ifp->probes = idev->cnf.dad_transmits;
3238         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3239 }
3240
3241 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3242 {
3243         struct inet6_dev *idev = ifp->idev;
3244         struct net_device *dev = idev->dev;
3245
3246         addrconf_join_solict(dev, &ifp->addr);
3247
3248         net_srandom(ifp->addr.s6_addr32[3]);
3249
3250         read_lock_bh(&idev->lock);
3251         spin_lock(&ifp->lock);
3252         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3253                 goto out;
3254
3255         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3256             idev->cnf.accept_dad < 1 ||
3257             !(ifp->flags&IFA_F_TENTATIVE) ||
3258             ifp->flags & IFA_F_NODAD) {
3259                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3260                 spin_unlock(&ifp->lock);
3261                 read_unlock_bh(&idev->lock);
3262
3263                 addrconf_dad_completed(ifp);
3264                 return;
3265         }
3266
3267         if (!(idev->if_flags & IF_READY)) {
3268                 spin_unlock(&ifp->lock);
3269                 read_unlock_bh(&idev->lock);
3270                 /*
3271                  * If the device is not ready:
3272                  * - keep it tentative if it is a permanent address.
3273                  * - otherwise, kill it.
3274                  */
3275                 in6_ifa_hold(ifp);
3276                 addrconf_dad_stop(ifp, 0);
3277                 return;
3278         }
3279
3280         /*
3281          * Optimistic nodes can start receiving
3282          * Frames right away
3283          */
3284         if (ifp->flags & IFA_F_OPTIMISTIC)
3285                 ip6_ins_rt(ifp->rt);
3286
3287         addrconf_dad_kick(ifp);
3288 out:
3289         spin_unlock(&ifp->lock);
3290         read_unlock_bh(&idev->lock);
3291 }
3292
3293 static void addrconf_dad_timer(unsigned long data)
3294 {
3295         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3296         struct inet6_dev *idev = ifp->idev;
3297         struct in6_addr mcaddr;
3298
3299         if (!ifp->probes && addrconf_dad_end(ifp))
3300                 goto out;
3301
3302         read_lock(&idev->lock);
3303         if (idev->dead || !(idev->if_flags & IF_READY)) {
3304                 read_unlock(&idev->lock);
3305                 goto out;
3306         }
3307
3308         spin_lock(&ifp->lock);
3309         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3310                 spin_unlock(&ifp->lock);
3311                 read_unlock(&idev->lock);
3312                 goto out;
3313         }
3314
3315         if (ifp->probes == 0) {
3316                 /*
3317                  * DAD was successful
3318                  */
3319
3320                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3321                 spin_unlock(&ifp->lock);
3322                 read_unlock(&idev->lock);
3323
3324                 addrconf_dad_completed(ifp);
3325
3326                 goto out;
3327         }
3328
3329         ifp->probes--;
3330         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3331         spin_unlock(&ifp->lock);
3332         read_unlock(&idev->lock);
3333
3334         /* send a neighbour solicitation for our addr */
3335         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3336         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3337 out:
3338         in6_ifa_put(ifp);
3339 }
3340
3341 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3342 {
3343         struct net_device *dev = ifp->idev->dev;
3344
3345         /*
3346          *      Configure the address for reception. Now it is valid.
3347          */
3348
3349         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3350
3351         /* If added prefix is link local and we are prepared to process
3352            router advertisements, start sending router solicitations.
3353          */
3354
3355         if (ipv6_accept_ra(ifp->idev) &&
3356             ifp->idev->cnf.rtr_solicits > 0 &&
3357             (dev->flags&IFF_LOOPBACK) == 0 &&
3358             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3359                 /*
3360                  *      If a host as already performed a random delay
3361                  *      [...] as part of DAD [...] there is no need
3362                  *      to delay again before sending the first RS
3363                  */
3364                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3365
3366                 spin_lock_bh(&ifp->lock);
3367                 ifp->probes = 1;
3368                 ifp->idev->if_flags |= IF_RS_SENT;
3369                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3370                 spin_unlock_bh(&ifp->lock);
3371         }
3372 }
3373
3374 static void addrconf_dad_run(struct inet6_dev *idev)
3375 {
3376         struct inet6_ifaddr *ifp;
3377
3378         read_lock_bh(&idev->lock);
3379         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3380                 spin_lock(&ifp->lock);
3381                 if (ifp->flags & IFA_F_TENTATIVE &&
3382                     ifp->state == INET6_IFADDR_STATE_DAD)
3383                         addrconf_dad_kick(ifp);
3384                 spin_unlock(&ifp->lock);
3385         }
3386         read_unlock_bh(&idev->lock);
3387 }
3388
3389 #ifdef CONFIG_PROC_FS
3390 struct if6_iter_state {
3391         struct seq_net_private p;
3392         int bucket;
3393         int offset;
3394 };
3395
3396 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3397 {
3398         struct inet6_ifaddr *ifa = NULL;
3399         struct if6_iter_state *state = seq->private;
3400         struct net *net = seq_file_net(seq);
3401         int p = 0;
3402
3403         /* initial bucket if pos is 0 */
3404         if (pos == 0) {
3405                 state->bucket = 0;
3406                 state->offset = 0;
3407         }
3408
3409         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3410                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3411                                          addr_lst) {
3412                         if (!net_eq(dev_net(ifa->idev->dev), net))
3413                                 continue;
3414                         /* sync with offset */
3415                         if (p < state->offset) {
3416                                 p++;
3417                                 continue;
3418                         }
3419                         state->offset++;
3420                         return ifa;
3421                 }
3422
3423                 /* prepare for next bucket */
3424                 state->offset = 0;
3425                 p = 0;
3426         }
3427         return NULL;
3428 }
3429
3430 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3431                                          struct inet6_ifaddr *ifa)
3432 {
3433         struct if6_iter_state *state = seq->private;
3434         struct net *net = seq_file_net(seq);
3435
3436         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3437                 if (!net_eq(dev_net(ifa->idev->dev), net))
3438                         continue;
3439                 state->offset++;
3440                 return ifa;
3441         }
3442
3443         while (++state->bucket < IN6_ADDR_HSIZE) {
3444                 state->offset = 0;
3445                 hlist_for_each_entry_rcu_bh(ifa,
3446                                      &inet6_addr_lst[state->bucket], addr_lst) {
3447                         if (!net_eq(dev_net(ifa->idev->dev), net))
3448                                 continue;
3449                         state->offset++;
3450                         return ifa;
3451                 }
3452         }
3453
3454         return NULL;
3455 }
3456
3457 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3458         __acquires(rcu_bh)
3459 {
3460         rcu_read_lock_bh();
3461         return if6_get_first(seq, *pos);
3462 }
3463
3464 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3465 {
3466         struct inet6_ifaddr *ifa;
3467
3468         ifa = if6_get_next(seq, v);
3469         ++*pos;
3470         return ifa;
3471 }
3472
3473 static void if6_seq_stop(struct seq_file *seq, void *v)
3474         __releases(rcu_bh)
3475 {
3476         rcu_read_unlock_bh();
3477 }
3478
3479 static int if6_seq_show(struct seq_file *seq, void *v)
3480 {
3481         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3482         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3483                    &ifp->addr,
3484                    ifp->idev->dev->ifindex,
3485                    ifp->prefix_len,
3486                    ifp->scope,
3487                    ifp->flags,
3488                    ifp->idev->dev->name);
3489         return 0;
3490 }
3491
3492 static const struct seq_operations if6_seq_ops = {
3493         .start  = if6_seq_start,
3494         .next   = if6_seq_next,
3495         .show   = if6_seq_show,
3496         .stop   = if6_seq_stop,
3497 };
3498
3499 static int if6_seq_open(struct inode *inode, struct file *file)
3500 {
3501         return seq_open_net(inode, file, &if6_seq_ops,
3502                             sizeof(struct if6_iter_state));
3503 }
3504
3505 static const struct file_operations if6_fops = {
3506         .owner          = THIS_MODULE,
3507         .open           = if6_seq_open,
3508         .read           = seq_read,
3509         .llseek         = seq_lseek,
3510         .release        = seq_release_net,
3511 };
3512
3513 static int __net_init if6_proc_net_init(struct net *net)
3514 {
3515         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3516                 return -ENOMEM;
3517         return 0;
3518 }
3519
3520 static void __net_exit if6_proc_net_exit(struct net *net)
3521 {
3522         remove_proc_entry("if_inet6", net->proc_net);
3523 }
3524
3525 static struct pernet_operations if6_proc_net_ops = {
3526        .init = if6_proc_net_init,
3527        .exit = if6_proc_net_exit,
3528 };
3529
3530 int __init if6_proc_init(void)
3531 {
3532         return register_pernet_subsys(&if6_proc_net_ops);
3533 }
3534
3535 void if6_proc_exit(void)
3536 {
3537         unregister_pernet_subsys(&if6_proc_net_ops);
3538 }
3539 #endif  /* CONFIG_PROC_FS */
3540
3541 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3542 /* Check if address is a home address configured on any interface. */
3543 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3544 {
3545         int ret = 0;
3546         struct inet6_ifaddr *ifp = NULL;
3547         unsigned int hash = inet6_addr_hash(addr);
3548
3549         rcu_read_lock_bh();
3550         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3551                 if (!net_eq(dev_net(ifp->idev->dev), net))
3552                         continue;
3553                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3554                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3555                         ret = 1;
3556                         break;
3557                 }
3558         }
3559         rcu_read_unlock_bh();
3560         return ret;
3561 }
3562 #endif
3563
3564 /*
3565  *      Periodic address status verification
3566  */
3567
3568 static void addrconf_verify(unsigned long foo)
3569 {
3570         unsigned long now, next, next_sec, next_sched;
3571         struct inet6_ifaddr *ifp;
3572         int i;
3573
3574         rcu_read_lock_bh();
3575         spin_lock(&addrconf_verify_lock);
3576         now = jiffies;
3577         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3578
3579         del_timer(&addr_chk_timer);
3580
3581         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3582 restart:
3583                 hlist_for_each_entry_rcu_bh(ifp,
3584                                          &inet6_addr_lst[i], addr_lst) {
3585                         unsigned long age;
3586
3587                         if (ifp->flags & IFA_F_PERMANENT)
3588                                 continue;
3589
3590                         spin_lock(&ifp->lock);
3591                         /* We try to batch several events at once. */
3592                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3593
3594                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3595                             age >= ifp->valid_lft) {
3596                                 spin_unlock(&ifp->lock);
3597                                 in6_ifa_hold(ifp);
3598                                 ipv6_del_addr(ifp);
3599                                 goto restart;
3600                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3601                                 spin_unlock(&ifp->lock);
3602                                 continue;
3603                         } else if (age >= ifp->prefered_lft) {
3604                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3605                                 int deprecate = 0;
3606
3607                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3608                                         deprecate = 1;
3609                                         ifp->flags |= IFA_F_DEPRECATED;
3610                                 }
3611
3612                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3613                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3614
3615                                 spin_unlock(&ifp->lock);
3616
3617                                 if (deprecate) {
3618                                         in6_ifa_hold(ifp);
3619
3620                                         ipv6_ifa_notify(0, ifp);
3621                                         in6_ifa_put(ifp);
3622                                         goto restart;
3623                                 }
3624 #ifdef CONFIG_IPV6_PRIVACY
3625                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3626                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3627                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3628                                         ifp->idev->cnf.dad_transmits *
3629                                         ifp->idev->nd_parms->retrans_time / HZ;
3630
3631                                 if (age >= ifp->prefered_lft - regen_advance) {
3632                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3633                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3634                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3635                                         if (!ifp->regen_count && ifpub) {
3636                                                 ifp->regen_count++;
3637                                                 in6_ifa_hold(ifp);
3638                                                 in6_ifa_hold(ifpub);
3639                                                 spin_unlock(&ifp->lock);
3640
3641                                                 spin_lock(&ifpub->lock);
3642                                                 ifpub->regen_count = 0;
3643                                                 spin_unlock(&ifpub->lock);
3644                                                 ipv6_create_tempaddr(ifpub, ifp);
3645                                                 in6_ifa_put(ifpub);
3646                                                 in6_ifa_put(ifp);
3647                                                 goto restart;
3648                                         }
3649                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3650                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3651                                 spin_unlock(&ifp->lock);
3652 #endif
3653                         } else {
3654                                 /* ifp->prefered_lft <= ifp->valid_lft */
3655                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3656                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3657                                 spin_unlock(&ifp->lock);
3658                         }
3659                 }
3660         }
3661
3662         next_sec = round_jiffies_up(next);
3663         next_sched = next;
3664
3665         /* If rounded timeout is accurate enough, accept it. */
3666         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3667                 next_sched = next_sec;
3668
3669         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3670         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3671                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3672
3673         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3674               now, next, next_sec, next_sched));
3675
3676         addr_chk_timer.expires = next_sched;
3677         add_timer(&addr_chk_timer);
3678         spin_unlock(&addrconf_verify_lock);
3679         rcu_read_unlock_bh();
3680 }
3681
3682 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3683 {
3684         struct in6_addr *pfx = NULL;
3685
3686         if (addr)
3687                 pfx = nla_data(addr);
3688
3689         if (local) {
3690                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3691                         pfx = NULL;
3692                 else
3693                         pfx = nla_data(local);
3694         }
3695
3696         return pfx;
3697 }
3698
3699 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3700         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3701         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3702         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3703 };
3704
3705 static int
3706 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3707 {
3708         struct net *net = sock_net(skb->sk);
3709         struct ifaddrmsg *ifm;
3710         struct nlattr *tb[IFA_MAX+1];
3711         struct in6_addr *pfx;
3712         int err;
3713
3714         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3715         if (err < 0)
3716                 return err;
3717
3718         ifm = nlmsg_data(nlh);
3719         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3720         if (pfx == NULL)
3721                 return -EINVAL;
3722
3723         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3724 }
3725
3726 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3727                              u32 prefered_lft, u32 valid_lft)
3728 {
3729         u32 flags;
3730         clock_t expires;
3731         unsigned long timeout;
3732
3733         if (!valid_lft || (prefered_lft > valid_lft))
3734                 return -EINVAL;
3735
3736         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3737         if (addrconf_finite_timeout(timeout)) {
3738                 expires = jiffies_to_clock_t(timeout * HZ);
3739                 valid_lft = timeout;
3740                 flags = RTF_EXPIRES;
3741         } else {
3742                 expires = 0;
3743                 flags = 0;
3744                 ifa_flags |= IFA_F_PERMANENT;
3745         }
3746
3747         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3748         if (addrconf_finite_timeout(timeout)) {
3749                 if (timeout == 0)
3750                         ifa_flags |= IFA_F_DEPRECATED;
3751                 prefered_lft = timeout;
3752         }
3753
3754         spin_lock_bh(&ifp->lock);
3755         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3756         ifp->tstamp = jiffies;
3757         ifp->valid_lft = valid_lft;
3758         ifp->prefered_lft = prefered_lft;
3759
3760         spin_unlock_bh(&ifp->lock);
3761         if (!(ifp->flags&IFA_F_TENTATIVE))
3762                 ipv6_ifa_notify(0, ifp);
3763
3764         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3765                               expires, flags);
3766         addrconf_verify(0);
3767
3768         return 0;
3769 }
3770
3771 static int
3772 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3773 {
3774         struct net *net = sock_net(skb->sk);
3775         struct ifaddrmsg *ifm;
3776         struct nlattr *tb[IFA_MAX+1];
3777         struct in6_addr *pfx;
3778         struct inet6_ifaddr *ifa;
3779         struct net_device *dev;
3780         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3781         u8 ifa_flags;
3782         int err;
3783
3784         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3785         if (err < 0)
3786                 return err;
3787
3788         ifm = nlmsg_data(nlh);
3789         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3790         if (pfx == NULL)
3791                 return -EINVAL;
3792
3793         if (tb[IFA_CACHEINFO]) {
3794                 struct ifa_cacheinfo *ci;
3795
3796                 ci = nla_data(tb[IFA_CACHEINFO]);
3797                 valid_lft = ci->ifa_valid;
3798                 preferred_lft = ci->ifa_prefered;
3799         } else {
3800                 preferred_lft = INFINITY_LIFE_TIME;
3801                 valid_lft = INFINITY_LIFE_TIME;
3802         }
3803
3804         dev =  __dev_get_by_index(net, ifm->ifa_index);
3805         if (dev == NULL)
3806                 return -ENODEV;
3807
3808         /* We ignore other flags so far. */
3809         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3810
3811         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3812         if (ifa == NULL) {
3813                 /*
3814                  * It would be best to check for !NLM_F_CREATE here but
3815                  * userspace alreay relies on not having to provide this.
3816                  */
3817                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3818                                       ifm->ifa_prefixlen, ifa_flags,
3819                                       preferred_lft, valid_lft);
3820         }
3821
3822         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3823             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3824                 err = -EEXIST;
3825         else
3826                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3827
3828         in6_ifa_put(ifa);
3829
3830         return err;
3831 }
3832
3833 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3834                           u8 scope, int ifindex)
3835 {
3836         struct ifaddrmsg *ifm;
3837
3838         ifm = nlmsg_data(nlh);
3839         ifm->ifa_family = AF_INET6;
3840         ifm->ifa_prefixlen = prefixlen;
3841         ifm->ifa_flags = flags;
3842         ifm->ifa_scope = scope;
3843         ifm->ifa_index = ifindex;
3844 }
3845
3846 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3847                          unsigned long tstamp, u32 preferred, u32 valid)
3848 {
3849         struct ifa_cacheinfo ci;
3850
3851         ci.cstamp = cstamp_delta(cstamp);
3852         ci.tstamp = cstamp_delta(tstamp);
3853         ci.ifa_prefered = preferred;
3854         ci.ifa_valid = valid;
3855
3856         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3857 }
3858
3859 static inline int rt_scope(int ifa_scope)
3860 {
3861         if (ifa_scope & IFA_HOST)
3862                 return RT_SCOPE_HOST;
3863         else if (ifa_scope & IFA_LINK)
3864                 return RT_SCOPE_LINK;
3865         else if (ifa_scope & IFA_SITE)
3866                 return RT_SCOPE_SITE;
3867         else
3868                 return RT_SCOPE_UNIVERSE;
3869 }
3870
3871 static inline int inet6_ifaddr_msgsize(void)
3872 {
3873         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3874                + nla_total_size(16) /* IFA_ADDRESS */
3875                + nla_total_size(sizeof(struct ifa_cacheinfo));
3876 }
3877
3878 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3879                              u32 portid, u32 seq, int event, unsigned int flags)
3880 {
3881         struct nlmsghdr  *nlh;
3882         u32 preferred, valid;
3883
3884         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3885         if (nlh == NULL)
3886                 return -EMSGSIZE;
3887
3888         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3889                       ifa->idev->dev->ifindex);
3890
3891         if (!(ifa->flags&IFA_F_PERMANENT)) {
3892                 preferred = ifa->prefered_lft;
3893                 valid = ifa->valid_lft;
3894                 if (preferred != INFINITY_LIFE_TIME) {
3895                         long tval = (jiffies - ifa->tstamp)/HZ;
3896                         if (preferred > tval)
3897                                 preferred -= tval;
3898                         else
3899                                 preferred = 0;
3900                         if (valid != INFINITY_LIFE_TIME) {
3901                                 if (valid > tval)
3902                                         valid -= tval;
3903                                 else
3904                                         valid = 0;
3905                         }
3906                 }
3907         } else {
3908                 preferred = INFINITY_LIFE_TIME;
3909                 valid = INFINITY_LIFE_TIME;
3910         }
3911
3912         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3913             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3914                 nlmsg_cancel(skb, nlh);
3915                 return -EMSGSIZE;
3916         }
3917
3918         return nlmsg_end(skb, nlh);
3919 }
3920
3921 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3922                                 u32 portid, u32 seq, int event, u16 flags)
3923 {
3924         struct nlmsghdr  *nlh;
3925         u8 scope = RT_SCOPE_UNIVERSE;
3926         int ifindex = ifmca->idev->dev->ifindex;
3927
3928         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3929                 scope = RT_SCOPE_SITE;
3930
3931         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3932         if (nlh == NULL)
3933                 return -EMSGSIZE;
3934
3935         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3936         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3937             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3938                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3939                 nlmsg_cancel(skb, nlh);
3940                 return -EMSGSIZE;
3941         }
3942
3943         return nlmsg_end(skb, nlh);
3944 }
3945
3946 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3947                                 u32 portid, u32 seq, int event, unsigned int flags)
3948 {
3949         struct nlmsghdr  *nlh;
3950         u8 scope = RT_SCOPE_UNIVERSE;
3951         int ifindex = ifaca->aca_idev->dev->ifindex;
3952
3953         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3954                 scope = RT_SCOPE_SITE;
3955
3956         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3957         if (nlh == NULL)
3958                 return -EMSGSIZE;
3959
3960         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3961         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3962             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3963                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3964                 nlmsg_cancel(skb, nlh);
3965                 return -EMSGSIZE;
3966         }
3967
3968         return nlmsg_end(skb, nlh);
3969 }
3970
3971 enum addr_type_t {
3972         UNICAST_ADDR,
3973         MULTICAST_ADDR,
3974         ANYCAST_ADDR,
3975 };
3976
3977 /* called with rcu_read_lock() */
3978 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3979                           struct netlink_callback *cb, enum addr_type_t type,
3980                           int s_ip_idx, int *p_ip_idx)
3981 {
3982         struct ifmcaddr6 *ifmca;
3983         struct ifacaddr6 *ifaca;
3984         int err = 1;
3985         int ip_idx = *p_ip_idx;
3986
3987         read_lock_bh(&idev->lock);
3988         switch (type) {
3989         case UNICAST_ADDR: {
3990                 struct inet6_ifaddr *ifa;
3991
3992                 /* unicast address incl. temp addr */
3993                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3994                         if (++ip_idx < s_ip_idx)
3995                                 continue;
3996                         err = inet6_fill_ifaddr(skb, ifa,
3997                                                 NETLINK_CB(cb->skb).portid,
3998                                                 cb->nlh->nlmsg_seq,
3999                                                 RTM_NEWADDR,
4000                                                 NLM_F_MULTI);
4001                         if (err <= 0)
4002                                 break;
4003                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4004                 }
4005                 break;
4006         }
4007         case MULTICAST_ADDR:
4008                 /* multicast address */
4009                 for (ifmca = idev->mc_list; ifmca;
4010                      ifmca = ifmca->next, ip_idx++) {
4011                         if (ip_idx < s_ip_idx)
4012                                 continue;
4013                         err = inet6_fill_ifmcaddr(skb, ifmca,
4014                                                   NETLINK_CB(cb->skb).portid,
4015                                                   cb->nlh->nlmsg_seq,
4016                                                   RTM_GETMULTICAST,
4017                                                   NLM_F_MULTI);
4018                         if (err <= 0)
4019                                 break;
4020                 }
4021                 break;
4022         case ANYCAST_ADDR:
4023                 /* anycast address */
4024                 for (ifaca = idev->ac_list; ifaca;
4025                      ifaca = ifaca->aca_next, ip_idx++) {
4026                         if (ip_idx < s_ip_idx)
4027                                 continue;
4028                         err = inet6_fill_ifacaddr(skb, ifaca,
4029                                                   NETLINK_CB(cb->skb).portid,
4030                                                   cb->nlh->nlmsg_seq,
4031                                                   RTM_GETANYCAST,
4032                                                   NLM_F_MULTI);
4033                         if (err <= 0)
4034                                 break;
4035                 }
4036                 break;
4037         default:
4038                 break;
4039         }
4040         read_unlock_bh(&idev->lock);
4041         *p_ip_idx = ip_idx;
4042         return err;
4043 }
4044
4045 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4046                            enum addr_type_t type)
4047 {
4048         struct net *net = sock_net(skb->sk);
4049         int h, s_h;
4050         int idx, ip_idx;
4051         int s_idx, s_ip_idx;
4052         struct net_device *dev;
4053         struct inet6_dev *idev;
4054         struct hlist_head *head;
4055
4056         s_h = cb->args[0];
4057         s_idx = idx = cb->args[1];
4058         s_ip_idx = ip_idx = cb->args[2];
4059
4060         rcu_read_lock();
4061         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4062         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4063                 idx = 0;
4064                 head = &net->dev_index_head[h];
4065                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4066                         if (idx < s_idx)
4067                                 goto cont;
4068                         if (h > s_h || idx > s_idx)
4069                                 s_ip_idx = 0;
4070                         ip_idx = 0;
4071                         idev = __in6_dev_get(dev);
4072                         if (!idev)
4073                                 goto cont;
4074
4075                         if (in6_dump_addrs(idev, skb, cb, type,
4076                                            s_ip_idx, &ip_idx) <= 0)
4077                                 goto done;
4078 cont:
4079                         idx++;
4080                 }
4081         }
4082 done:
4083         rcu_read_unlock();
4084         cb->args[0] = h;
4085         cb->args[1] = idx;
4086         cb->args[2] = ip_idx;
4087
4088         return skb->len;
4089 }
4090
4091 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4092 {
4093         enum addr_type_t type = UNICAST_ADDR;
4094
4095         return inet6_dump_addr(skb, cb, type);
4096 }
4097
4098 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4099 {
4100         enum addr_type_t type = MULTICAST_ADDR;
4101
4102         return inet6_dump_addr(skb, cb, type);
4103 }
4104
4105
4106 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4107 {
4108         enum addr_type_t type = ANYCAST_ADDR;
4109
4110         return inet6_dump_addr(skb, cb, type);
4111 }
4112
4113 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4114 {
4115         struct net *net = sock_net(in_skb->sk);
4116         struct ifaddrmsg *ifm;
4117         struct nlattr *tb[IFA_MAX+1];
4118         struct in6_addr *addr = NULL;
4119         struct net_device *dev = NULL;
4120         struct inet6_ifaddr *ifa;
4121         struct sk_buff *skb;
4122         int err;
4123
4124         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4125         if (err < 0)
4126                 goto errout;
4127
4128         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4129         if (addr == NULL) {
4130                 err = -EINVAL;
4131                 goto errout;
4132         }
4133
4134         ifm = nlmsg_data(nlh);
4135         if (ifm->ifa_index)
4136                 dev = __dev_get_by_index(net, ifm->ifa_index);
4137
4138         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4139         if (!ifa) {
4140                 err = -EADDRNOTAVAIL;
4141                 goto errout;
4142         }
4143
4144         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4145         if (!skb) {
4146                 err = -ENOBUFS;
4147                 goto errout_ifa;
4148         }
4149
4150         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4151                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4152         if (err < 0) {
4153                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4154                 WARN_ON(err == -EMSGSIZE);
4155                 kfree_skb(skb);
4156                 goto errout_ifa;
4157         }
4158         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4159 errout_ifa:
4160         in6_ifa_put(ifa);
4161 errout:
4162         return err;
4163 }
4164
4165 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4166 {
4167         struct sk_buff *skb;
4168         struct net *net = dev_net(ifa->idev->dev);
4169         int err = -ENOBUFS;
4170
4171         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4172         if (skb == NULL)
4173                 goto errout;
4174
4175         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4176         if (err < 0) {
4177                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4178                 WARN_ON(err == -EMSGSIZE);
4179                 kfree_skb(skb);
4180                 goto errout;
4181         }
4182         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4183         return;
4184 errout:
4185         if (err < 0)
4186                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4187 }
4188
4189 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4190                                 __s32 *array, int bytes)
4191 {
4192         BUG_ON(bytes < (DEVCONF_MAX * 4));
4193
4194         memset(array, 0, bytes);
4195         array[DEVCONF_FORWARDING] = cnf->forwarding;
4196         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4197         array[DEVCONF_MTU6] = cnf->mtu6;
4198         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4199         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4200         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4201         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4202         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4203         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4204                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4205         array[DEVCONF_RTR_SOLICIT_DELAY] =
4206                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4207         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4208 #ifdef CONFIG_IPV6_PRIVACY
4209         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4210         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4211         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4212         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4213         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4214 #endif
4215         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4216         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4217         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4218 #ifdef CONFIG_IPV6_ROUTER_PREF
4219         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4220         array[DEVCONF_RTR_PROBE_INTERVAL] =
4221                 jiffies_to_msecs(cnf->rtr_probe_interval);
4222 #ifdef CONFIG_IPV6_ROUTE_INFO
4223         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4224 #endif
4225 #endif
4226         array[DEVCONF_ACCEPT_RA_RT_TABLE] = cnf->accept_ra_rt_table;
4227         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4228         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4229 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4230         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4231 #endif
4232 #ifdef CONFIG_IPV6_MROUTE
4233         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4234 #endif
4235         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4236         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4237         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4238         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4239 }
4240
4241 static inline size_t inet6_ifla6_size(void)
4242 {
4243         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4244              + nla_total_size(sizeof(struct ifla_cacheinfo))
4245              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4246              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4247              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4248              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4249 }
4250
4251 static inline size_t inet6_if_nlmsg_size(void)
4252 {
4253         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4254                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4255                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4256                + nla_total_size(4) /* IFLA_MTU */
4257                + nla_total_size(4) /* IFLA_LINK */
4258                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4259 }
4260
4261 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4262                                       int items, int bytes)
4263 {
4264         int i;
4265         int pad = bytes - sizeof(u64) * items;
4266         BUG_ON(pad < 0);
4267
4268         /* Use put_unaligned() because stats may not be aligned for u64. */
4269         put_unaligned(items, &stats[0]);
4270         for (i = 1; i < items; i++)
4271                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4272
4273         memset(&stats[items], 0, pad);
4274 }
4275
4276 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4277                                       int items, int bytes, size_t syncpoff)
4278 {
4279         int i;
4280         int pad = bytes - sizeof(u64) * items;
4281         BUG_ON(pad < 0);
4282
4283         /* Use put_unaligned() because stats may not be aligned for u64. */
4284         put_unaligned(items, &stats[0]);
4285         for (i = 1; i < items; i++)
4286                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4287
4288         memset(&stats[items], 0, pad);
4289 }
4290
4291 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4292                              int bytes)
4293 {
4294         switch (attrtype) {
4295         case IFLA_INET6_STATS:
4296                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4297                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4298                 break;
4299         case IFLA_INET6_ICMP6STATS:
4300                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4301                 break;
4302         }
4303 }
4304
4305 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4306 {
4307         struct nlattr *nla;
4308         struct ifla_cacheinfo ci;
4309
4310         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4311                 goto nla_put_failure;
4312         ci.max_reasm_len = IPV6_MAXPLEN;
4313         ci.tstamp = cstamp_delta(idev->tstamp);
4314         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4315         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4316         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4317                 goto nla_put_failure;
4318         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4319         if (nla == NULL)
4320                 goto nla_put_failure;
4321         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4322
4323         /* XXX - MC not implemented */
4324
4325         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4326         if (nla == NULL)
4327                 goto nla_put_failure;
4328         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4329
4330         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4331         if (nla == NULL)
4332                 goto nla_put_failure;
4333         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4334
4335         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4336         if (nla == NULL)
4337                 goto nla_put_failure;
4338         read_lock_bh(&idev->lock);
4339         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4340         read_unlock_bh(&idev->lock);
4341
4342         return 0;
4343
4344 nla_put_failure:
4345         return -EMSGSIZE;
4346 }
4347
4348 static size_t inet6_get_link_af_size(const struct net_device *dev)
4349 {
4350         if (!__in6_dev_get(dev))
4351                 return 0;
4352
4353         return inet6_ifla6_size();
4354 }
4355
4356 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4357 {
4358         struct inet6_dev *idev = __in6_dev_get(dev);
4359
4360         if (!idev)
4361                 return -ENODATA;
4362
4363         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4364                 return -EMSGSIZE;
4365
4366         return 0;
4367 }
4368
4369 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4370 {
4371         struct inet6_ifaddr *ifp;
4372         struct net_device *dev = idev->dev;
4373         bool update_rs = false;
4374         struct in6_addr ll_addr;
4375
4376         if (token == NULL)
4377                 return -EINVAL;
4378         if (ipv6_addr_any(token))
4379                 return -EINVAL;
4380         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4381                 return -EINVAL;
4382         if (!ipv6_accept_ra(idev))
4383                 return -EINVAL;
4384         if (idev->cnf.rtr_solicits <= 0)
4385                 return -EINVAL;
4386
4387         write_lock_bh(&idev->lock);
4388
4389         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4390         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4391
4392         write_unlock_bh(&idev->lock);
4393
4394         if (!idev->dead && (idev->if_flags & IF_READY) &&
4395             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4396                              IFA_F_OPTIMISTIC)) {
4397
4398                 /* If we're not ready, then normal ifup will take care
4399                  * of this. Otherwise, we need to request our rs here.
4400                  */
4401                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4402                 update_rs = true;
4403         }
4404
4405         write_lock_bh(&idev->lock);
4406
4407         if (update_rs)
4408                 idev->if_flags |= IF_RS_SENT;
4409
4410         /* Well, that's kinda nasty ... */
4411         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4412                 spin_lock(&ifp->lock);
4413                 if (ifp->tokenized) {
4414                         ifp->valid_lft = 0;
4415                         ifp->prefered_lft = 0;
4416                 }
4417                 spin_unlock(&ifp->lock);
4418         }
4419
4420         write_unlock_bh(&idev->lock);
4421         return 0;
4422 }
4423
4424 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4425 {
4426         int err = -EINVAL;
4427         struct inet6_dev *idev = __in6_dev_get(dev);
4428         struct nlattr *tb[IFLA_INET6_MAX + 1];
4429
4430         if (!idev)
4431                 return -EAFNOSUPPORT;
4432
4433         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4434                 BUG();
4435
4436         if (tb[IFLA_INET6_TOKEN])
4437                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4438
4439         return err;
4440 }
4441
4442 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4443                              u32 portid, u32 seq, int event, unsigned int flags)
4444 {
4445         struct net_device *dev = idev->dev;
4446         struct ifinfomsg *hdr;
4447         struct nlmsghdr *nlh;
4448         void *protoinfo;
4449
4450         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4451         if (nlh == NULL)
4452                 return -EMSGSIZE;
4453
4454         hdr = nlmsg_data(nlh);
4455         hdr->ifi_family = AF_INET6;
4456         hdr->__ifi_pad = 0;
4457         hdr->ifi_type = dev->type;
4458         hdr->ifi_index = dev->ifindex;
4459         hdr->ifi_flags = dev_get_flags(dev);
4460         hdr->ifi_change = 0;
4461
4462         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4463             (dev->addr_len &&
4464              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4465             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4466             (dev->ifindex != dev->iflink &&
4467              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4468                 goto nla_put_failure;
4469         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4470         if (protoinfo == NULL)
4471                 goto nla_put_failure;
4472
4473         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4474                 goto nla_put_failure;
4475
4476         nla_nest_end(skb, protoinfo);
4477         return nlmsg_end(skb, nlh);
4478
4479 nla_put_failure:
4480         nlmsg_cancel(skb, nlh);
4481         return -EMSGSIZE;
4482 }
4483
4484 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4485 {
4486         struct net *net = sock_net(skb->sk);
4487         int h, s_h;
4488         int idx = 0, s_idx;
4489         struct net_device *dev;
4490         struct inet6_dev *idev;
4491         struct hlist_head *head;
4492
4493         s_h = cb->args[0];
4494         s_idx = cb->args[1];
4495
4496         rcu_read_lock();
4497         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4498                 idx = 0;
4499                 head = &net->dev_index_head[h];
4500                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4501                         if (idx < s_idx)
4502                                 goto cont;
4503                         idev = __in6_dev_get(dev);
4504                         if (!idev)
4505                                 goto cont;
4506                         if (inet6_fill_ifinfo(skb, idev,
4507                                               NETLINK_CB(cb->skb).portid,
4508                                               cb->nlh->nlmsg_seq,
4509                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4510                                 goto out;
4511 cont:
4512                         idx++;
4513                 }
4514         }
4515 out:
4516         rcu_read_unlock();
4517         cb->args[1] = idx;
4518         cb->args[0] = h;
4519
4520         return skb->len;
4521 }
4522
4523 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4524 {
4525         struct sk_buff *skb;
4526         struct net *net = dev_net(idev->dev);
4527         int err = -ENOBUFS;
4528
4529         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4530         if (skb == NULL)
4531                 goto errout;
4532
4533         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4534         if (err < 0) {
4535                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4536                 WARN_ON(err == -EMSGSIZE);
4537                 kfree_skb(skb);
4538                 goto errout;
4539         }
4540         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4541         return;
4542 errout:
4543         if (err < 0)
4544                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4545 }
4546
4547 static inline size_t inet6_prefix_nlmsg_size(void)
4548 {
4549         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4550                + nla_total_size(sizeof(struct in6_addr))
4551                + nla_total_size(sizeof(struct prefix_cacheinfo));
4552 }
4553
4554 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4555                              struct prefix_info *pinfo, u32 portid, u32 seq,
4556                              int event, unsigned int flags)
4557 {
4558         struct prefixmsg *pmsg;
4559         struct nlmsghdr *nlh;
4560         struct prefix_cacheinfo ci;
4561
4562         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4563         if (nlh == NULL)
4564                 return -EMSGSIZE;
4565
4566         pmsg = nlmsg_data(nlh);
4567         pmsg->prefix_family = AF_INET6;
4568         pmsg->prefix_pad1 = 0;
4569         pmsg->prefix_pad2 = 0;
4570         pmsg->prefix_ifindex = idev->dev->ifindex;
4571         pmsg->prefix_len = pinfo->prefix_len;
4572         pmsg->prefix_type = pinfo->type;
4573         pmsg->prefix_pad3 = 0;
4574         pmsg->prefix_flags = 0;
4575         if (pinfo->onlink)
4576                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4577         if (pinfo->autoconf)
4578                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4579
4580         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4581                 goto nla_put_failure;
4582         ci.preferred_time = ntohl(pinfo->prefered);
4583         ci.valid_time = ntohl(pinfo->valid);
4584         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4585                 goto nla_put_failure;
4586         return nlmsg_end(skb, nlh);
4587
4588 nla_put_failure:
4589         nlmsg_cancel(skb, nlh);
4590         return -EMSGSIZE;
4591 }
4592
4593 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4594                          struct prefix_info *pinfo)
4595 {
4596         struct sk_buff *skb;
4597         struct net *net = dev_net(idev->dev);
4598         int err = -ENOBUFS;
4599
4600         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4601         if (skb == NULL)
4602                 goto errout;
4603
4604         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4605         if (err < 0) {
4606                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4607                 WARN_ON(err == -EMSGSIZE);
4608                 kfree_skb(skb);
4609                 goto errout;
4610         }
4611         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4612         return;
4613 errout:
4614         if (err < 0)
4615                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4616 }
4617
4618 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4619 {
4620         struct net *net = dev_net(ifp->idev->dev);
4621
4622         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4623
4624         switch (event) {
4625         case RTM_NEWADDR:
4626                 /*
4627                  * If the address was optimistic
4628                  * we inserted the route at the start of
4629                  * our DAD process, so we don't need
4630                  * to do it again
4631                  */
4632                 if (!(ifp->rt->rt6i_node))
4633                         ip6_ins_rt(ifp->rt);
4634                 if (ifp->idev->cnf.forwarding)
4635                         addrconf_join_anycast(ifp);
4636                 break;
4637         case RTM_DELADDR:
4638                 if (ifp->idev->cnf.forwarding)
4639                         addrconf_leave_anycast(ifp);
4640                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4641                 dst_hold(&ifp->rt->dst);
4642
4643                 if (ip6_del_rt(ifp->rt))
4644                         dst_free(&ifp->rt->dst);
4645                 break;
4646         }
4647         atomic_inc(&net->ipv6.dev_addr_genid);
4648 }
4649
4650 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4651 {
4652         rcu_read_lock_bh();
4653         if (likely(ifp->idev->dead == 0))
4654                 __ipv6_ifa_notify(event, ifp);
4655         rcu_read_unlock_bh();
4656 }
4657
4658 #ifdef CONFIG_SYSCTL
4659
4660 static
4661 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4662                            void __user *buffer, size_t *lenp, loff_t *ppos)
4663 {
4664         int *valp = ctl->data;
4665         int val = *valp;
4666         loff_t pos = *ppos;
4667         ctl_table lctl;
4668         int ret;
4669
4670         /*
4671          * ctl->data points to idev->cnf.forwarding, we should
4672          * not modify it until we get the rtnl lock.
4673          */
4674         lctl = *ctl;
4675         lctl.data = &val;
4676
4677         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4678
4679         if (write)
4680                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4681         if (ret)
4682                 *ppos = pos;
4683         return ret;
4684 }
4685
4686 static void dev_disable_change(struct inet6_dev *idev)
4687 {
4688         if (!idev || !idev->dev)
4689                 return;
4690
4691         if (idev->cnf.disable_ipv6)
4692                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4693         else
4694                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4695 }
4696
4697 static void addrconf_disable_change(struct net *net, __s32 newf)
4698 {
4699         struct net_device *dev;
4700         struct inet6_dev *idev;
4701
4702         rcu_read_lock();
4703         for_each_netdev_rcu(net, dev) {
4704                 idev = __in6_dev_get(dev);
4705                 if (idev) {
4706                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4707                         idev->cnf.disable_ipv6 = newf;
4708                         if (changed)
4709                                 dev_disable_change(idev);
4710                 }
4711         }
4712         rcu_read_unlock();
4713 }
4714
4715 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4716 {
4717         struct net *net;
4718         int old;
4719
4720         if (!rtnl_trylock())
4721                 return restart_syscall();
4722
4723         net = (struct net *)table->extra2;
4724         old = *p;
4725         *p = newf;
4726
4727         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4728                 rtnl_unlock();
4729                 return 0;
4730         }
4731
4732         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4733                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4734                 addrconf_disable_change(net, newf);
4735         } else if ((!newf) ^ (!old))
4736                 dev_disable_change((struct inet6_dev *)table->extra1);
4737
4738         rtnl_unlock();
4739         return 0;
4740 }
4741
4742 static
4743 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4744                             void __user *buffer, size_t *lenp, loff_t *ppos)
4745 {
4746         int *valp = ctl->data;
4747         int val = *valp;
4748         loff_t pos = *ppos;
4749         ctl_table lctl;
4750         int ret;
4751
4752         /*
4753          * ctl->data points to idev->cnf.disable_ipv6, we should
4754          * not modify it until we get the rtnl lock.
4755          */
4756         lctl = *ctl;
4757         lctl.data = &val;
4758
4759         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4760
4761         if (write)
4762                 ret = addrconf_disable_ipv6(ctl, valp, val);
4763         if (ret)
4764                 *ppos = pos;
4765         return ret;
4766 }
4767
4768 static struct addrconf_sysctl_table
4769 {
4770         struct ctl_table_header *sysctl_header;
4771         ctl_table addrconf_vars[DEVCONF_MAX+1];
4772 } addrconf_sysctl __read_mostly = {
4773         .sysctl_header = NULL,
4774         .addrconf_vars = {
4775                 {
4776                         .procname       = "forwarding",
4777                         .data           = &ipv6_devconf.forwarding,
4778                         .maxlen         = sizeof(int),
4779                         .mode           = 0644,
4780                         .proc_handler   = addrconf_sysctl_forward,
4781                 },
4782                 {
4783                         .procname       = "hop_limit",
4784                         .data           = &ipv6_devconf.hop_limit,
4785                         .maxlen         = sizeof(int),
4786                         .mode           = 0644,
4787                         .proc_handler   = proc_dointvec,
4788                 },
4789                 {
4790                         .procname       = "mtu",
4791                         .data           = &ipv6_devconf.mtu6,
4792                         .maxlen         = sizeof(int),
4793                         .mode           = 0644,
4794                         .proc_handler   = proc_dointvec,
4795                 },
4796                 {
4797                         .procname       = "accept_ra",
4798                         .data           = &ipv6_devconf.accept_ra,
4799                         .maxlen         = sizeof(int),
4800                         .mode           = 0644,
4801                         .proc_handler   = proc_dointvec,
4802                 },
4803                 {
4804                         .procname       = "accept_redirects",
4805                         .data           = &ipv6_devconf.accept_redirects,
4806                         .maxlen         = sizeof(int),
4807                         .mode           = 0644,
4808                         .proc_handler   = proc_dointvec,
4809                 },
4810                 {
4811                         .procname       = "autoconf",
4812                         .data           = &ipv6_devconf.autoconf,
4813                         .maxlen         = sizeof(int),
4814                         .mode           = 0644,
4815                         .proc_handler   = proc_dointvec,
4816                 },
4817                 {
4818                         .procname       = "dad_transmits",
4819                         .data           = &ipv6_devconf.dad_transmits,
4820                         .maxlen         = sizeof(int),
4821                         .mode           = 0644,
4822                         .proc_handler   = proc_dointvec,
4823                 },
4824                 {
4825                         .procname       = "router_solicitations",
4826                         .data           = &ipv6_devconf.rtr_solicits,
4827                         .maxlen         = sizeof(int),
4828                         .mode           = 0644,
4829                         .proc_handler   = proc_dointvec,
4830                 },
4831                 {
4832                         .procname       = "router_solicitation_interval",
4833                         .data           = &ipv6_devconf.rtr_solicit_interval,
4834                         .maxlen         = sizeof(int),
4835                         .mode           = 0644,
4836                         .proc_handler   = proc_dointvec_jiffies,
4837                 },
4838                 {
4839                         .procname       = "router_solicitation_delay",
4840                         .data           = &ipv6_devconf.rtr_solicit_delay,
4841                         .maxlen         = sizeof(int),
4842                         .mode           = 0644,
4843                         .proc_handler   = proc_dointvec_jiffies,
4844                 },
4845                 {
4846                         .procname       = "force_mld_version",
4847                         .data           = &ipv6_devconf.force_mld_version,
4848                         .maxlen         = sizeof(int),
4849                         .mode           = 0644,
4850                         .proc_handler   = proc_dointvec,
4851                 },
4852 #ifdef CONFIG_IPV6_PRIVACY
4853                 {
4854                         .procname       = "use_tempaddr",
4855                         .data           = &ipv6_devconf.use_tempaddr,
4856                         .maxlen         = sizeof(int),
4857                         .mode           = 0644,
4858                         .proc_handler   = proc_dointvec,
4859                 },
4860                 {
4861                         .procname       = "temp_valid_lft",
4862                         .data           = &ipv6_devconf.temp_valid_lft,
4863                         .maxlen         = sizeof(int),
4864                         .mode           = 0644,
4865                         .proc_handler   = proc_dointvec,
4866                 },
4867                 {
4868                         .procname       = "temp_prefered_lft",
4869                         .data           = &ipv6_devconf.temp_prefered_lft,
4870                         .maxlen         = sizeof(int),
4871                         .mode           = 0644,
4872                         .proc_handler   = proc_dointvec,
4873                 },
4874                 {
4875                         .procname       = "regen_max_retry",
4876                         .data           = &ipv6_devconf.regen_max_retry,
4877                         .maxlen         = sizeof(int),
4878                         .mode           = 0644,
4879                         .proc_handler   = proc_dointvec,
4880                 },
4881                 {
4882                         .procname       = "max_desync_factor",
4883                         .data           = &ipv6_devconf.max_desync_factor,
4884                         .maxlen         = sizeof(int),
4885                         .mode           = 0644,
4886                         .proc_handler   = proc_dointvec,
4887                 },
4888 #endif
4889                 {
4890                         .procname       = "max_addresses",
4891                         .data           = &ipv6_devconf.max_addresses,
4892                         .maxlen         = sizeof(int),
4893                         .mode           = 0644,
4894                         .proc_handler   = proc_dointvec,
4895                 },
4896                 {
4897                         .procname       = "accept_ra_defrtr",
4898                         .data           = &ipv6_devconf.accept_ra_defrtr,
4899                         .maxlen         = sizeof(int),
4900                         .mode           = 0644,
4901                         .proc_handler   = proc_dointvec,
4902                 },
4903                 {
4904                         .procname       = "accept_ra_pinfo",
4905                         .data           = &ipv6_devconf.accept_ra_pinfo,
4906                         .maxlen         = sizeof(int),
4907                         .mode           = 0644,
4908                         .proc_handler   = proc_dointvec,
4909                 },
4910 #ifdef CONFIG_IPV6_ROUTER_PREF
4911                 {
4912                         .procname       = "accept_ra_rtr_pref",
4913                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4914                         .maxlen         = sizeof(int),
4915                         .mode           = 0644,
4916                         .proc_handler   = proc_dointvec,
4917                 },
4918                 {
4919                         .procname       = "router_probe_interval",
4920                         .data           = &ipv6_devconf.rtr_probe_interval,
4921                         .maxlen         = sizeof(int),
4922                         .mode           = 0644,
4923                         .proc_handler   = proc_dointvec_jiffies,
4924                 },
4925 #ifdef CONFIG_IPV6_ROUTE_INFO
4926                 {
4927                         .procname       = "accept_ra_rt_info_max_plen",
4928                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4929                         .maxlen         = sizeof(int),
4930                         .mode           = 0644,
4931                         .proc_handler   = proc_dointvec,
4932                 },
4933 #endif
4934 #endif
4935                 {
4936                         .procname       = "accept_ra_rt_table",
4937                         .data           = &ipv6_devconf.accept_ra_rt_table,
4938                         .maxlen         = sizeof(int),
4939                         .mode           = 0644,
4940                         .proc_handler   = proc_dointvec,
4941                 },
4942                 {
4943                         .procname       = "proxy_ndp",
4944                         .data           = &ipv6_devconf.proxy_ndp,
4945                         .maxlen         = sizeof(int),
4946                         .mode           = 0644,
4947                         .proc_handler   = proc_dointvec,
4948                 },
4949                 {
4950                         .procname       = "accept_source_route",
4951                         .data           = &ipv6_devconf.accept_source_route,
4952                         .maxlen         = sizeof(int),
4953                         .mode           = 0644,
4954                         .proc_handler   = proc_dointvec,
4955                 },
4956 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4957                 {
4958                         .procname       = "optimistic_dad",
4959                         .data           = &ipv6_devconf.optimistic_dad,
4960                         .maxlen         = sizeof(int),
4961                         .mode           = 0644,
4962                         .proc_handler   = proc_dointvec,
4963
4964                 },
4965 #endif
4966 #ifdef CONFIG_IPV6_MROUTE
4967                 {
4968                         .procname       = "mc_forwarding",
4969                         .data           = &ipv6_devconf.mc_forwarding,
4970                         .maxlen         = sizeof(int),
4971                         .mode           = 0444,
4972                         .proc_handler   = proc_dointvec,
4973                 },
4974 #endif
4975                 {
4976                         .procname       = "disable_ipv6",
4977                         .data           = &ipv6_devconf.disable_ipv6,
4978                         .maxlen         = sizeof(int),
4979                         .mode           = 0644,
4980                         .proc_handler   = addrconf_sysctl_disable,
4981                 },
4982                 {
4983                         .procname       = "accept_dad",
4984                         .data           = &ipv6_devconf.accept_dad,
4985                         .maxlen         = sizeof(int),
4986                         .mode           = 0644,
4987                         .proc_handler   = proc_dointvec,
4988                 },
4989                 {
4990                         .procname       = "force_tllao",
4991                         .data           = &ipv6_devconf.force_tllao,
4992                         .maxlen         = sizeof(int),
4993                         .mode           = 0644,
4994                         .proc_handler   = proc_dointvec
4995                 },
4996                 {
4997                         .procname       = "ndisc_notify",
4998                         .data           = &ipv6_devconf.ndisc_notify,
4999                         .maxlen         = sizeof(int),
5000                         .mode           = 0644,
5001                         .proc_handler   = proc_dointvec
5002                 },
5003                 {
5004                         /* sentinel */
5005                 }
5006         },
5007 };
5008
5009 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5010                 struct inet6_dev *idev, struct ipv6_devconf *p)
5011 {
5012         int i;
5013         struct addrconf_sysctl_table *t;
5014         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5015
5016         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5017         if (t == NULL)
5018                 goto out;
5019
5020         for (i = 0; t->addrconf_vars[i].data; i++) {
5021                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5022                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5023                 t->addrconf_vars[i].extra2 = net;
5024         }
5025
5026         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5027
5028         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5029         if (t->sysctl_header == NULL)
5030                 goto free;
5031
5032         p->sysctl = t;
5033         return 0;
5034
5035 free:
5036         kfree(t);
5037 out:
5038         return -ENOBUFS;
5039 }
5040
5041 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5042 {
5043         struct addrconf_sysctl_table *t;
5044
5045         if (p->sysctl == NULL)
5046                 return;
5047
5048         t = p->sysctl;
5049         p->sysctl = NULL;
5050         unregister_net_sysctl_table(t->sysctl_header);
5051         kfree(t);
5052 }
5053
5054 static void addrconf_sysctl_register(struct inet6_dev *idev)
5055 {
5056         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5057                               &ndisc_ifinfo_sysctl_change);
5058         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5059                                         idev, &idev->cnf);
5060 }
5061
5062 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5063 {
5064         __addrconf_sysctl_unregister(&idev->cnf);
5065         neigh_sysctl_unregister(idev->nd_parms);
5066 }
5067
5068
5069 #endif
5070
5071 static int __net_init addrconf_init_net(struct net *net)
5072 {
5073         int err = -ENOMEM;
5074         struct ipv6_devconf *all, *dflt;
5075
5076         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5077         if (all == NULL)
5078                 goto err_alloc_all;
5079
5080         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5081         if (dflt == NULL)
5082                 goto err_alloc_dflt;
5083
5084         /* these will be inherited by all namespaces */
5085         dflt->autoconf = ipv6_defaults.autoconf;
5086         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5087
5088         net->ipv6.devconf_all = all;
5089         net->ipv6.devconf_dflt = dflt;
5090
5091 #ifdef CONFIG_SYSCTL
5092         err = __addrconf_sysctl_register(net, "all", NULL, all);
5093         if (err < 0)
5094                 goto err_reg_all;
5095
5096         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5097         if (err < 0)
5098                 goto err_reg_dflt;
5099 #endif
5100         return 0;
5101
5102 #ifdef CONFIG_SYSCTL
5103 err_reg_dflt:
5104         __addrconf_sysctl_unregister(all);
5105 err_reg_all:
5106         kfree(dflt);
5107 #endif
5108 err_alloc_dflt:
5109         kfree(all);
5110 err_alloc_all:
5111         return err;
5112 }
5113
5114 static void __net_exit addrconf_exit_net(struct net *net)
5115 {
5116 #ifdef CONFIG_SYSCTL
5117         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5118         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5119 #endif
5120         if (!net_eq(net, &init_net)) {
5121                 kfree(net->ipv6.devconf_dflt);
5122                 kfree(net->ipv6.devconf_all);
5123         }
5124 }
5125
5126 static struct pernet_operations addrconf_ops = {
5127         .init = addrconf_init_net,
5128         .exit = addrconf_exit_net,
5129 };
5130
5131 static struct rtnl_af_ops inet6_ops = {
5132         .family           = AF_INET6,
5133         .fill_link_af     = inet6_fill_link_af,
5134         .get_link_af_size = inet6_get_link_af_size,
5135         .set_link_af      = inet6_set_link_af,
5136 };
5137
5138 /*
5139  *      Init / cleanup code
5140  */
5141
5142 int __init addrconf_init(void)
5143 {
5144         int i, err;
5145
5146         err = ipv6_addr_label_init();
5147         if (err < 0) {
5148                 pr_crit("%s: cannot initialize default policy table: %d\n",
5149                         __func__, err);
5150                 goto out;
5151         }
5152
5153         err = register_pernet_subsys(&addrconf_ops);
5154         if (err < 0)
5155                 goto out_addrlabel;
5156
5157         /* The addrconf netdev notifier requires that loopback_dev
5158          * has it's ipv6 private information allocated and setup
5159          * before it can bring up and give link-local addresses
5160          * to other devices which are up.
5161          *
5162          * Unfortunately, loopback_dev is not necessarily the first
5163          * entry in the global dev_base list of net devices.  In fact,
5164          * it is likely to be the very last entry on that list.
5165          * So this causes the notifier registry below to try and
5166          * give link-local addresses to all devices besides loopback_dev
5167          * first, then loopback_dev, which cases all the non-loopback_dev
5168          * devices to fail to get a link-local address.
5169          *
5170          * So, as a temporary fix, allocate the ipv6 structure for
5171          * loopback_dev first by hand.
5172          * Longer term, all of the dependencies ipv6 has upon the loopback
5173          * device and it being up should be removed.
5174          */
5175         rtnl_lock();
5176         if (!ipv6_add_dev(init_net.loopback_dev))
5177                 err = -ENOMEM;
5178         rtnl_unlock();
5179         if (err)
5180                 goto errlo;
5181
5182         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5183                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5184
5185         register_netdevice_notifier(&ipv6_dev_notf);
5186
5187         addrconf_verify(0);
5188
5189         err = rtnl_af_register(&inet6_ops);
5190         if (err < 0)
5191                 goto errout_af;
5192
5193         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5194                               NULL);
5195         if (err < 0)
5196                 goto errout;
5197
5198         /* Only the first call to __rtnl_register can fail */
5199         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5200         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5201         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5202                         inet6_dump_ifaddr, NULL);
5203         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5204                         inet6_dump_ifmcaddr, NULL);
5205         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5206                         inet6_dump_ifacaddr, NULL);
5207         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5208                         inet6_netconf_dump_devconf, NULL);
5209
5210         ipv6_addr_label_rtnl_register();
5211
5212         return 0;
5213 errout:
5214         rtnl_af_unregister(&inet6_ops);
5215 errout_af:
5216         unregister_netdevice_notifier(&ipv6_dev_notf);
5217 errlo:
5218         unregister_pernet_subsys(&addrconf_ops);
5219 out_addrlabel:
5220         ipv6_addr_label_cleanup();
5221 out:
5222         return err;
5223 }
5224
5225 void addrconf_cleanup(void)
5226 {
5227         struct net_device *dev;
5228         int i;
5229
5230         unregister_netdevice_notifier(&ipv6_dev_notf);
5231         unregister_pernet_subsys(&addrconf_ops);
5232         ipv6_addr_label_cleanup();
5233
5234         rtnl_lock();
5235
5236         __rtnl_af_unregister(&inet6_ops);
5237
5238         /* clean dev list */
5239         for_each_netdev(&init_net, dev) {
5240                 if (__in6_dev_get(dev) == NULL)
5241                         continue;
5242                 addrconf_ifdown(dev, 1);
5243         }
5244         addrconf_ifdown(init_net.loopback_dev, 2);
5245
5246         /*
5247          *      Check hash table.
5248          */
5249         spin_lock_bh(&addrconf_hash_lock);
5250         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5251                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5252         spin_unlock_bh(&addrconf_hash_lock);
5253
5254         del_timer(&addr_chk_timer);
5255         rtnl_unlock();
5256 }