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 ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 173
 174 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 175         .forwarding             = 0,
 176         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 177         .mtu6                   = IPV6_MIN_MTU,
 178         .accept_ra              = 1,
 179         .accept_redirects       = 1,
 180         .autoconf               = 1,
 181         .force_mld_version      = 0,
 182         .dad_transmits          = 1,
 183         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 184         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 185         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 186 #ifdef CONFIG_IPV6_PRIVACY
 187         .use_tempaddr           = 0,
 188         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 189         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 190         .regen_max_retry        = REGEN_MAX_RETRY,
 191         .max_desync_factor      = MAX_DESYNC_FACTOR,
 192 #endif
 193         .max_addresses          = IPV6_MAX_ADDRESSES,
 194         .accept_ra_defrtr       = 1,
 195         .accept_ra_pinfo        = 1,
 196 #ifdef CONFIG_IPV6_ROUTER_PREF
 197         .accept_ra_rtr_pref     = 1,
 198         .rtr_probe_interval     = 60 * HZ,
 199 #ifdef CONFIG_IPV6_ROUTE_INFO
 200         .accept_ra_rt_info_max_plen = 0,
 201 #endif
 202 #endif
 203         .proxy_ndp              = 0,
 204         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 205         .disable_ipv6           = 0,
 206         .accept_dad             = 1,
 207 };
 208
 209 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 210         .forwarding             = 0,
 211         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 212         .mtu6                   = IPV6_MIN_MTU,
 213         .accept_ra              = 1,
 214         .accept_redirects       = 1,
 215         .autoconf               = 1,
 216         .dad_transmits          = 1,
 217         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 218         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 219         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 220 #ifdef CONFIG_IPV6_PRIVACY
 221         .use_tempaddr           = 0,
 222         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 223         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 224         .regen_max_retry        = REGEN_MAX_RETRY,
 225         .max_desync_factor      = MAX_DESYNC_FACTOR,
 226 #endif
 227         .max_addresses          = IPV6_MAX_ADDRESSES,
 228         .accept_ra_defrtr       = 1,
 229         .accept_ra_pinfo        = 1,
 230 #ifdef CONFIG_IPV6_ROUTER_PREF
 231         .accept_ra_rtr_pref     = 1,
 232         .rtr_probe_interval     = 60 * HZ,
 233 #ifdef CONFIG_IPV6_ROUTE_INFO
 234         .accept_ra_rt_info_max_plen = 0,
 235 #endif
 236 #endif
 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                 atomic_notifier_call_chain(&inet6addr_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         atomic_notifier_call_chain(&inet6addr_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          */
1117         if (tmp_prefered_lft <= regen_advance) {
1118                 in6_ifa_put(ifp);
1119                 in6_dev_put(idev);
1120                 ret = -1;
1121                 goto out;
1122         }
1123
1124         addr_flags = IFA_F_TEMPORARY;
1125         /* set in addrconf_prefix_rcv() */
1126         if (ifp->flags & IFA_F_OPTIMISTIC)
1127                 addr_flags |= IFA_F_OPTIMISTIC;
1128
1129         ift = !max_addresses ||
1130               ipv6_count_addresses(idev) < max_addresses ?
1131                 ipv6_add_addr(idev, &addr, tmp_plen,
1132                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1133                               addr_flags) : NULL;
1134         if (IS_ERR_OR_NULL(ift)) {
1135                 in6_ifa_put(ifp);
1136                 in6_dev_put(idev);
1137                 pr_info("%s: retry temporary address regeneration\n", __func__);
1138                 tmpaddr = &addr;
1139                 write_lock(&idev->lock);
1140                 goto retry;
1141         }
1142
1143         spin_lock_bh(&ift->lock);
1144         ift->ifpub = ifp;
1145         ift->valid_lft = tmp_valid_lft;
1146         ift->prefered_lft = tmp_prefered_lft;
1147         ift->cstamp = now;
1148         ift->tstamp = tmp_tstamp;
1149         spin_unlock_bh(&ift->lock);
1150
1151         addrconf_dad_start(ift);
1152         in6_ifa_put(ift);
1153         in6_dev_put(idev);
1154 out:
1155         return ret;
1156 }
1157 #endif
1158
1159 /*
1160  *      Choose an appropriate source address (RFC3484)
1161  */
1162 enum {
1163         IPV6_SADDR_RULE_INIT = 0,
1164         IPV6_SADDR_RULE_LOCAL,
1165         IPV6_SADDR_RULE_SCOPE,
1166         IPV6_SADDR_RULE_PREFERRED,
1167 #ifdef CONFIG_IPV6_MIP6
1168         IPV6_SADDR_RULE_HOA,
1169 #endif
1170         IPV6_SADDR_RULE_OIF,
1171         IPV6_SADDR_RULE_LABEL,
1172 #ifdef CONFIG_IPV6_PRIVACY
1173         IPV6_SADDR_RULE_PRIVACY,
1174 #endif
1175         IPV6_SADDR_RULE_ORCHID,
1176         IPV6_SADDR_RULE_PREFIX,
1177         IPV6_SADDR_RULE_MAX
1178 };
1179
1180 struct ipv6_saddr_score {
1181         int                     rule;
1182         int                     addr_type;
1183         struct inet6_ifaddr     *ifa;
1184         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1185         int                     scopedist;
1186         int                     matchlen;
1187 };
1188
1189 struct ipv6_saddr_dst {
1190         const struct in6_addr *addr;
1191         int ifindex;
1192         int scope;
1193         int label;
1194         unsigned int prefs;
1195 };
1196
1197 static inline int ipv6_saddr_preferred(int type)
1198 {
1199         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1200                 return 1;
1201         return 0;
1202 }
1203
1204 static int ipv6_get_saddr_eval(struct net *net,
1205                                struct ipv6_saddr_score *score,
1206                                struct ipv6_saddr_dst *dst,
1207                                int i)
1208 {
1209         int ret;
1210
1211         if (i <= score->rule) {
1212                 switch (i) {
1213                 case IPV6_SADDR_RULE_SCOPE:
1214                         ret = score->scopedist;
1215                         break;
1216                 case IPV6_SADDR_RULE_PREFIX:
1217                         ret = score->matchlen;
1218                         break;
1219                 default:
1220                         ret = !!test_bit(i, score->scorebits);
1221                 }
1222                 goto out;
1223         }
1224
1225         switch (i) {
1226         case IPV6_SADDR_RULE_INIT:
1227                 /* Rule 0: remember if hiscore is not ready yet */
1228                 ret = !!score->ifa;
1229                 break;
1230         case IPV6_SADDR_RULE_LOCAL:
1231                 /* Rule 1: Prefer same address */
1232                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1233                 break;
1234         case IPV6_SADDR_RULE_SCOPE:
1235                 /* Rule 2: Prefer appropriate scope
1236                  *
1237                  *      ret
1238                  *       ^
1239                  *    -1 |  d 15
1240                  *    ---+--+-+---> scope
1241                  *       |
1242                  *       |             d is scope of the destination.
1243                  *  B-d  |  \
1244                  *       |   \      <- smaller scope is better if
1245                  *  B-15 |    \        if scope is enough for destinaion.
1246                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1247                  * d-C-1 | /
1248                  *       |/         <- greater is better
1249                  *   -C  /             if scope is not enough for destination.
1250                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1251                  *
1252                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1253                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1254                  * Assume B = 0 and we get C > 29.
1255                  */
1256                 ret = __ipv6_addr_src_scope(score->addr_type);
1257                 if (ret >= dst->scope)
1258                         ret = -ret;
1259                 else
1260                         ret -= 128;     /* 30 is enough */
1261                 score->scopedist = ret;
1262                 break;
1263         case IPV6_SADDR_RULE_PREFERRED:
1264                 /* Rule 3: Avoid deprecated and optimistic addresses */
1265                 ret = ipv6_saddr_preferred(score->addr_type) ||
1266                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1267                 break;
1268 #ifdef CONFIG_IPV6_MIP6
1269         case IPV6_SADDR_RULE_HOA:
1270             {
1271                 /* Rule 4: Prefer home address */
1272                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1273                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1274                 break;
1275             }
1276 #endif
1277         case IPV6_SADDR_RULE_OIF:
1278                 /* Rule 5: Prefer outgoing interface */
1279                 ret = (!dst->ifindex ||
1280                        dst->ifindex == score->ifa->idev->dev->ifindex);
1281                 break;
1282         case IPV6_SADDR_RULE_LABEL:
1283                 /* Rule 6: Prefer matching label */
1284                 ret = ipv6_addr_label(net,
1285                                       &score->ifa->addr, score->addr_type,
1286                                       score->ifa->idev->dev->ifindex) == dst->label;
1287                 break;
1288 #ifdef CONFIG_IPV6_PRIVACY
1289         case IPV6_SADDR_RULE_PRIVACY:
1290             {
1291                 /* Rule 7: Prefer public address
1292                  * Note: prefer temporary address if use_tempaddr >= 2
1293                  */
1294                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1295                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1296                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1297                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1298                 break;
1299             }
1300 #endif
1301         case IPV6_SADDR_RULE_ORCHID:
1302                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1303                  *          non-ORCHID vs non-ORCHID
1304                  */
1305                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1306                         ipv6_addr_orchid(dst->addr));
1307                 break;
1308         case IPV6_SADDR_RULE_PREFIX:
1309                 /* Rule 8: Use longest matching prefix */
1310                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1311                 if (ret > score->ifa->prefix_len)
1312                         ret = score->ifa->prefix_len;
1313                 score->matchlen = ret;
1314                 break;
1315         default:
1316                 ret = 0;
1317         }
1318
1319         if (ret)
1320                 __set_bit(i, score->scorebits);
1321         score->rule = i;
1322 out:
1323         return ret;
1324 }
1325
1326 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1327                        const struct in6_addr *daddr, unsigned int prefs,
1328                        struct in6_addr *saddr)
1329 {
1330         struct ipv6_saddr_score scores[2],
1331                                 *score = &scores[0], *hiscore = &scores[1];
1332         struct ipv6_saddr_dst dst;
1333         struct net_device *dev;
1334         int dst_type;
1335
1336         dst_type = __ipv6_addr_type(daddr);
1337         dst.addr = daddr;
1338         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1339         dst.scope = __ipv6_addr_src_scope(dst_type);
1340         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1341         dst.prefs = prefs;
1342
1343         hiscore->rule = -1;
1344         hiscore->ifa = NULL;
1345
1346         rcu_read_lock();
1347
1348         for_each_netdev_rcu(net, dev) {
1349                 struct inet6_dev *idev;
1350
1351                 /* Candidate Source Address (section 4)
1352                  *  - multicast and link-local destination address,
1353                  *    the set of candidate source address MUST only
1354                  *    include addresses assigned to interfaces
1355                  *    belonging to the same link as the outgoing
1356                  *    interface.
1357                  * (- For site-local destination addresses, the
1358                  *    set of candidate source addresses MUST only
1359                  *    include addresses assigned to interfaces
1360                  *    belonging to the same site as the outgoing
1361                  *    interface.)
1362                  */
1363                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1364                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1365                     dst.ifindex && dev->ifindex != dst.ifindex)
1366                         continue;
1367
1368                 idev = __in6_dev_get(dev);
1369                 if (!idev)
1370                         continue;
1371
1372                 read_lock_bh(&idev->lock);
1373                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1374                         int i;
1375
1376                         /*
1377                          * - Tentative Address (RFC2462 section 5.4)
1378                          *  - A tentative address is not considered
1379                          *    "assigned to an interface" in the traditional
1380                          *    sense, unless it is also flagged as optimistic.
1381                          * - Candidate Source Address (section 4)
1382                          *  - In any case, anycast addresses, multicast
1383                          *    addresses, and the unspecified address MUST
1384                          *    NOT be included in a candidate set.
1385                          */
1386                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1387                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1388                                 continue;
1389
1390                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1391
1392                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1393                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1394                                 LIMIT_NETDEBUG(KERN_DEBUG
1395                                                "ADDRCONF: unspecified / multicast address "
1396                                                "assigned as unicast address on %s",
1397                                                dev->name);
1398                                 continue;
1399                         }
1400
1401                         score->rule = -1;
1402                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1403
1404                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1405                                 int minihiscore, miniscore;
1406
1407                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1408                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1409
1410                                 if (minihiscore > miniscore) {
1411                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1412                                             score->scopedist > 0) {
1413                                                 /*
1414                                                  * special case:
1415                                                  * each remaining entry
1416                                                  * has too small (not enough)
1417                                                  * scope, because ifa entries
1418                                                  * are sorted by their scope
1419                                                  * values.
1420                                                  */
1421                                                 goto try_nextdev;
1422                                         }
1423                                         break;
1424                                 } else if (minihiscore < miniscore) {
1425                                         if (hiscore->ifa)
1426                                                 in6_ifa_put(hiscore->ifa);
1427
1428                                         in6_ifa_hold(score->ifa);
1429
1430                                         swap(hiscore, score);
1431
1432                                         /* restore our iterator */
1433                                         score->ifa = hiscore->ifa;
1434
1435                                         break;
1436                                 }
1437                         }
1438                 }
1439 try_nextdev:
1440                 read_unlock_bh(&idev->lock);
1441         }
1442         rcu_read_unlock();
1443
1444         if (!hiscore->ifa)
1445                 return -EADDRNOTAVAIL;
1446
1447         *saddr = hiscore->ifa->addr;
1448         in6_ifa_put(hiscore->ifa);
1449         return 0;
1450 }
1451 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1452
1453 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1454                     unsigned char banned_flags)
1455 {
1456         struct inet6_dev *idev;
1457         int err = -EADDRNOTAVAIL;
1458
1459         rcu_read_lock();
1460         idev = __in6_dev_get(dev);
1461         if (idev) {
1462                 struct inet6_ifaddr *ifp;
1463
1464                 read_lock_bh(&idev->lock);
1465                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1466                         if (ifp->scope == IFA_LINK &&
1467                             !(ifp->flags & banned_flags)) {
1468                                 *addr = ifp->addr;
1469                                 err = 0;
1470                                 break;
1471                         }
1472                 }
1473                 read_unlock_bh(&idev->lock);
1474         }
1475         rcu_read_unlock();
1476         return err;
1477 }
1478
1479 static int ipv6_count_addresses(struct inet6_dev *idev)
1480 {
1481         int cnt = 0;
1482         struct inet6_ifaddr *ifp;
1483
1484         read_lock_bh(&idev->lock);
1485         list_for_each_entry(ifp, &idev->addr_list, if_list)
1486                 cnt++;
1487         read_unlock_bh(&idev->lock);
1488         return cnt;
1489 }
1490
1491 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1492                   struct net_device *dev, int strict)
1493 {
1494         struct inet6_ifaddr *ifp;
1495         unsigned int hash = inet6_addr_hash(addr);
1496
1497         rcu_read_lock_bh();
1498         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1499                 if (!net_eq(dev_net(ifp->idev->dev), net))
1500                         continue;
1501                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1502                     !(ifp->flags&IFA_F_TENTATIVE) &&
1503                     (dev == NULL || ifp->idev->dev == dev ||
1504                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1505                         rcu_read_unlock_bh();
1506                         return 1;
1507                 }
1508         }
1509
1510         rcu_read_unlock_bh();
1511         return 0;
1512 }
1513 EXPORT_SYMBOL(ipv6_chk_addr);
1514
1515 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1516                                struct net_device *dev)
1517 {
1518         unsigned int hash = inet6_addr_hash(addr);
1519         struct inet6_ifaddr *ifp;
1520
1521         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1522                 if (!net_eq(dev_net(ifp->idev->dev), net))
1523                         continue;
1524                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1525                         if (dev == NULL || ifp->idev->dev == dev)
1526                                 return true;
1527                 }
1528         }
1529         return false;
1530 }
1531
1532 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1533 {
1534         struct inet6_dev *idev;
1535         struct inet6_ifaddr *ifa;
1536         int     onlink;
1537
1538         onlink = 0;
1539         rcu_read_lock();
1540         idev = __in6_dev_get(dev);
1541         if (idev) {
1542                 read_lock_bh(&idev->lock);
1543                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1544                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1545                                                    ifa->prefix_len);
1546                         if (onlink)
1547                                 break;
1548                 }
1549                 read_unlock_bh(&idev->lock);
1550         }
1551         rcu_read_unlock();
1552         return onlink;
1553 }
1554 EXPORT_SYMBOL(ipv6_chk_prefix);
1555
1556 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1557                                      struct net_device *dev, int strict)
1558 {
1559         struct inet6_ifaddr *ifp, *result = NULL;
1560         unsigned int hash = inet6_addr_hash(addr);
1561
1562         rcu_read_lock_bh();
1563         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1564                 if (!net_eq(dev_net(ifp->idev->dev), net))
1565                         continue;
1566                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1567                         if (dev == NULL || ifp->idev->dev == dev ||
1568                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1569                                 result = ifp;
1570                                 in6_ifa_hold(ifp);
1571                                 break;
1572                         }
1573                 }
1574         }
1575         rcu_read_unlock_bh();
1576
1577         return result;
1578 }
1579
1580 /* Gets referenced address, destroys ifaddr */
1581
1582 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1583 {
1584         if (ifp->flags&IFA_F_PERMANENT) {
1585                 spin_lock_bh(&ifp->lock);
1586                 addrconf_del_timer(ifp);
1587                 ifp->flags |= IFA_F_TENTATIVE;
1588                 if (dad_failed)
1589                         ifp->flags |= IFA_F_DADFAILED;
1590                 spin_unlock_bh(&ifp->lock);
1591                 if (dad_failed)
1592                         ipv6_ifa_notify(0, ifp);
1593                 in6_ifa_put(ifp);
1594 #ifdef CONFIG_IPV6_PRIVACY
1595         } else if (ifp->flags&IFA_F_TEMPORARY) {
1596                 struct inet6_ifaddr *ifpub;
1597                 spin_lock_bh(&ifp->lock);
1598                 ifpub = ifp->ifpub;
1599                 if (ifpub) {
1600                         in6_ifa_hold(ifpub);
1601                         spin_unlock_bh(&ifp->lock);
1602                         ipv6_create_tempaddr(ifpub, ifp);
1603                         in6_ifa_put(ifpub);
1604                 } else {
1605                         spin_unlock_bh(&ifp->lock);
1606                 }
1607                 ipv6_del_addr(ifp);
1608 #endif
1609         } else
1610                 ipv6_del_addr(ifp);
1611 }
1612
1613 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1614 {
1615         int err = -ENOENT;
1616
1617         spin_lock(&ifp->state_lock);
1618         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1619                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1620                 err = 0;
1621         }
1622         spin_unlock(&ifp->state_lock);
1623
1624         return err;
1625 }
1626
1627 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1628 {
1629         struct inet6_dev *idev = ifp->idev;
1630
1631         if (addrconf_dad_end(ifp)) {
1632                 in6_ifa_put(ifp);
1633                 return;
1634         }
1635
1636         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1637                              ifp->idev->dev->name, &ifp->addr);
1638
1639         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1640                 struct in6_addr addr;
1641
1642                 addr.s6_addr32[0] = htonl(0xfe800000);
1643                 addr.s6_addr32[1] = 0;
1644
1645                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1646                     ipv6_addr_equal(&ifp->addr, &addr)) {
1647                         /* DAD failed for link-local based on MAC address */
1648                         idev->cnf.disable_ipv6 = 1;
1649
1650                         pr_info("%s: IPv6 being disabled!\n",
1651                                 ifp->idev->dev->name);
1652                 }
1653         }
1654
1655         addrconf_dad_stop(ifp, 1);
1656 }
1657
1658 /* Join to solicited addr multicast group. */
1659
1660 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1661 {
1662         struct in6_addr maddr;
1663
1664         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1665                 return;
1666
1667         addrconf_addr_solict_mult(addr, &maddr);
1668         ipv6_dev_mc_inc(dev, &maddr);
1669 }
1670
1671 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1672 {
1673         struct in6_addr maddr;
1674
1675         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1676                 return;
1677
1678         addrconf_addr_solict_mult(addr, &maddr);
1679         __ipv6_dev_mc_dec(idev, &maddr);
1680 }
1681
1682 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1683 {
1684         struct in6_addr addr;
1685         if (ifp->prefix_len == 127) /* RFC 6164 */
1686                 return;
1687         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1688         if (ipv6_addr_any(&addr))
1689                 return;
1690         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1691 }
1692
1693 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1694 {
1695         struct in6_addr addr;
1696         if (ifp->prefix_len == 127) /* RFC 6164 */
1697                 return;
1698         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1699         if (ipv6_addr_any(&addr))
1700                 return;
1701         __ipv6_dev_ac_dec(ifp->idev, &addr);
1702 }
1703
1704 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1705 {
1706         if (dev->addr_len != ETH_ALEN)
1707                 return -1;
1708         memcpy(eui, dev->dev_addr, 3);
1709         memcpy(eui + 5, dev->dev_addr + 3, 3);
1710
1711         /*
1712          * The zSeries OSA network cards can be shared among various
1713          * OS instances, but the OSA cards have only one MAC address.
1714          * This leads to duplicate address conflicts in conjunction
1715          * with IPv6 if more than one instance uses the same card.
1716          *
1717          * The driver for these cards can deliver a unique 16-bit
1718          * identifier for each instance sharing the same card.  It is
1719          * placed instead of 0xFFFE in the interface identifier.  The
1720          * "u" bit of the interface identifier is not inverted in this
1721          * case.  Hence the resulting interface identifier has local
1722          * scope according to RFC2373.
1723          */
1724         if (dev->dev_id) {
1725                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1726                 eui[4] = dev->dev_id & 0xFF;
1727         } else {
1728                 eui[3] = 0xFF;
1729                 eui[4] = 0xFE;
1730                 eui[0] ^= 2;
1731         }
1732         return 0;
1733 }
1734
1735 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1736 {
1737         if (dev->addr_len != IEEE802154_ADDR_LEN)
1738                 return -1;
1739         memcpy(eui, dev->dev_addr, 8);
1740         eui[0] ^= 2;
1741         return 0;
1742 }
1743
1744 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1745 {
1746         union fwnet_hwaddr *ha;
1747
1748         if (dev->addr_len != FWNET_ALEN)
1749                 return -1;
1750
1751         ha = (union fwnet_hwaddr *)dev->dev_addr;
1752
1753         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1754         eui[0] ^= 2;
1755         return 0;
1756 }
1757
1758 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1759 {
1760         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1761         if (dev->addr_len != ARCNET_ALEN)
1762                 return -1;
1763         memset(eui, 0, 7);
1764         eui[7] = *(u8 *)dev->dev_addr;
1765         return 0;
1766 }
1767
1768 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1769 {
1770         if (dev->addr_len != INFINIBAND_ALEN)
1771                 return -1;
1772         memcpy(eui, dev->dev_addr + 12, 8);
1773         eui[0] |= 2;
1774         return 0;
1775 }
1776
1777 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1778 {
1779         if (addr == 0)
1780                 return -1;
1781         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1782                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1783                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1784                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1785                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1786                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1787         eui[1] = 0;
1788         eui[2] = 0x5E;
1789         eui[3] = 0xFE;
1790         memcpy(eui + 4, &addr, 4);
1791         return 0;
1792 }
1793
1794 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1795 {
1796         if (dev->priv_flags & IFF_ISATAP)
1797                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1798         return -1;
1799 }
1800
1801 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1802 {
1803         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1804 }
1805
1806 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1807 {
1808         switch (dev->type) {
1809         case ARPHRD_ETHER:
1810         case ARPHRD_FDDI:
1811                 return addrconf_ifid_eui48(eui, dev);
1812         case ARPHRD_ARCNET:
1813                 return addrconf_ifid_arcnet(eui, dev);
1814         case ARPHRD_INFINIBAND:
1815                 return addrconf_ifid_infiniband(eui, dev);
1816         case ARPHRD_SIT:
1817                 return addrconf_ifid_sit(eui, dev);
1818         case ARPHRD_IPGRE:
1819                 return addrconf_ifid_gre(eui, dev);
1820         case ARPHRD_IEEE802154:
1821                 return addrconf_ifid_eui64(eui, dev);
1822         case ARPHRD_IEEE1394:
1823                 return addrconf_ifid_ieee1394(eui, dev);
1824         }
1825         return -1;
1826 }
1827
1828 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1829 {
1830         int err = -1;
1831         struct inet6_ifaddr *ifp;
1832
1833         read_lock_bh(&idev->lock);
1834         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1835                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1836                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1837                         err = 0;
1838                         break;
1839                 }
1840         }
1841         read_unlock_bh(&idev->lock);
1842         return err;
1843 }
1844
1845 #ifdef CONFIG_IPV6_PRIVACY
1846 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1847 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1848 {
1849 regen:
1850         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1851         idev->rndid[0] &= ~0x02;
1852
1853         /*
1854          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1855          * check if generated address is not inappropriate
1856          *
1857          *  - Reserved subnet anycast (RFC 2526)
1858          *      11111101 11....11 1xxxxxxx
1859          *  - ISATAP (RFC4214) 6.1
1860          *      00-00-5E-FE-xx-xx-xx-xx
1861          *  - value 0
1862          *  - XXX: already assigned to an address on the device
1863          */
1864         if (idev->rndid[0] == 0xfd &&
1865             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1866             (idev->rndid[7]&0x80))
1867                 goto regen;
1868         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1869                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1870                         goto regen;
1871                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1872                         goto regen;
1873         }
1874 }
1875
1876 static void ipv6_regen_rndid(unsigned long data)
1877 {
1878         struct inet6_dev *idev = (struct inet6_dev *) data;
1879         unsigned long expires;
1880
1881         rcu_read_lock_bh();
1882         write_lock_bh(&idev->lock);
1883
1884         if (idev->dead)
1885                 goto out;
1886
1887         __ipv6_regen_rndid(idev);
1888
1889         expires = jiffies +
1890                 idev->cnf.temp_prefered_lft * HZ -
1891                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1892                 idev->cnf.max_desync_factor * HZ;
1893         if (time_before(expires, jiffies)) {
1894                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1895                         __func__, idev->dev->name);
1896                 goto out;
1897         }
1898
1899         if (!mod_timer(&idev->regen_timer, expires))
1900                 in6_dev_hold(idev);
1901
1902 out:
1903         write_unlock_bh(&idev->lock);
1904         rcu_read_unlock_bh();
1905         in6_dev_put(idev);
1906 }
1907
1908 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1909 {
1910         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1911                 __ipv6_regen_rndid(idev);
1912 }
1913 #endif
1914
1915 /*
1916  *      Add prefix route.
1917  */
1918
1919 static void
1920 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1921                       unsigned long expires, u32 flags)
1922 {
1923         struct fib6_config cfg = {
1924                 .fc_table = RT6_TABLE_PREFIX,
1925                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1926                 .fc_ifindex = dev->ifindex,
1927                 .fc_expires = expires,
1928                 .fc_dst_len = plen,
1929                 .fc_flags = RTF_UP | flags,
1930                 .fc_nlinfo.nl_net = dev_net(dev),
1931                 .fc_protocol = RTPROT_KERNEL,
1932         };
1933
1934         cfg.fc_dst = *pfx;
1935
1936         /* Prevent useless cloning on PtP SIT.
1937            This thing is done here expecting that the whole
1938            class of non-broadcast devices need not cloning.
1939          */
1940 #if IS_ENABLED(CONFIG_IPV6_SIT)
1941         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1942                 cfg.fc_flags |= RTF_NONEXTHOP;
1943 #endif
1944
1945         ip6_route_add(&cfg);
1946 }
1947
1948
1949 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1950                                                   int plen,
1951                                                   const struct net_device *dev,
1952                                                   u32 flags, u32 noflags)
1953 {
1954         struct fib6_node *fn;
1955         struct rt6_info *rt = NULL;
1956         struct fib6_table *table;
1957
1958         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1959         if (table == NULL)
1960                 return NULL;
1961
1962         read_lock_bh(&table->tb6_lock);
1963         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1964         if (!fn)
1965                 goto out;
1966         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1967                 if (rt->dst.dev->ifindex != dev->ifindex)
1968                         continue;
1969                 if ((rt->rt6i_flags & flags) != flags)
1970                         continue;
1971                 if ((rt->rt6i_flags & noflags) != 0)
1972                         continue;
1973                 dst_hold(&rt->dst);
1974                 break;
1975         }
1976 out:
1977         read_unlock_bh(&table->tb6_lock);
1978         return rt;
1979 }
1980
1981
1982 /* Create "default" multicast route to the interface */
1983
1984 static void addrconf_add_mroute(struct net_device *dev)
1985 {
1986         struct fib6_config cfg = {
1987                 .fc_table = RT6_TABLE_LOCAL,
1988                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1989                 .fc_ifindex = dev->ifindex,
1990                 .fc_dst_len = 8,
1991                 .fc_flags = RTF_UP,
1992                 .fc_nlinfo.nl_net = dev_net(dev),
1993         };
1994
1995         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1996
1997         ip6_route_add(&cfg);
1998 }
1999
2000 #if IS_ENABLED(CONFIG_IPV6_SIT)
2001 static void sit_route_add(struct net_device *dev)
2002 {
2003         struct fib6_config cfg = {
2004                 .fc_table = RT6_TABLE_MAIN,
2005                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2006                 .fc_ifindex = dev->ifindex,
2007                 .fc_dst_len = 96,
2008                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2009                 .fc_nlinfo.nl_net = dev_net(dev),
2010         };
2011
2012         /* prefix length - 96 bits "::d.d.d.d" */
2013         ip6_route_add(&cfg);
2014 }
2015 #endif
2016
2017 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2018 {
2019         struct inet6_dev *idev;
2020
2021         ASSERT_RTNL();
2022
2023         idev = ipv6_find_idev(dev);
2024         if (!idev)
2025                 return ERR_PTR(-ENOBUFS);
2026
2027         if (idev->cnf.disable_ipv6)
2028                 return ERR_PTR(-EACCES);
2029
2030         /* Add default multicast route */
2031         if (!(dev->flags & IFF_LOOPBACK))
2032                 addrconf_add_mroute(dev);
2033
2034         return idev;
2035 }
2036
2037 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2038 {
2039         struct prefix_info *pinfo;
2040         __u32 valid_lft;
2041         __u32 prefered_lft;
2042         int addr_type;
2043         struct inet6_dev *in6_dev;
2044         struct net *net = dev_net(dev);
2045
2046         pinfo = (struct prefix_info *) opt;
2047
2048         if (len < sizeof(struct prefix_info)) {
2049                 ADBG(("addrconf: prefix option too short\n"));
2050                 return;
2051         }
2052
2053         /*
2054          *      Validation checks ([ADDRCONF], page 19)
2055          */
2056
2057         addr_type = ipv6_addr_type(&pinfo->prefix);
2058
2059         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2060                 return;
2061
2062         valid_lft = ntohl(pinfo->valid);
2063         prefered_lft = ntohl(pinfo->prefered);
2064
2065         if (prefered_lft > valid_lft) {
2066                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2067                 return;
2068         }
2069
2070         in6_dev = in6_dev_get(dev);
2071
2072         if (in6_dev == NULL) {
2073                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2074                                     dev->name);
2075                 return;
2076         }
2077
2078         /*
2079          *      Two things going on here:
2080          *      1) Add routes for on-link prefixes
2081          *      2) Configure prefixes with the auto flag set
2082          */
2083
2084         if (pinfo->onlink) {
2085                 struct rt6_info *rt;
2086                 unsigned long rt_expires;
2087
2088                 /* Avoid arithmetic overflow. Really, we could
2089                  * save rt_expires in seconds, likely valid_lft,
2090                  * but it would require division in fib gc, that it
2091                  * not good.
2092                  */
2093                 if (HZ > USER_HZ)
2094                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2095                 else
2096                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2097
2098                 if (addrconf_finite_timeout(rt_expires))
2099                         rt_expires *= HZ;
2100
2101                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2102                                                pinfo->prefix_len,
2103                                                dev,
2104                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2105                                                RTF_GATEWAY | RTF_DEFAULT);
2106
2107                 if (rt) {
2108                         /* Autoconf prefix route */
2109                         if (valid_lft == 0) {
2110                                 ip6_del_rt(rt);
2111                                 rt = NULL;
2112                         } else if (addrconf_finite_timeout(rt_expires)) {
2113                                 /* not infinity */
2114                                 rt6_set_expires(rt, jiffies + rt_expires);
2115                         } else {
2116                                 rt6_clean_expires(rt);
2117                         }
2118                 } else if (valid_lft) {
2119                         clock_t expires = 0;
2120                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2121                         if (addrconf_finite_timeout(rt_expires)) {
2122                                 /* not infinity */
2123                                 flags |= RTF_EXPIRES;
2124                                 expires = jiffies_to_clock_t(rt_expires);
2125                         }
2126                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2127                                               dev, expires, flags);
2128                 }
2129                 ip6_rt_put(rt);
2130         }
2131
2132         /* Try to figure out our local address for this prefix */
2133
2134         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2135                 struct inet6_ifaddr *ifp;
2136                 struct in6_addr addr;
2137                 int create = 0, update_lft = 0;
2138                 bool tokenized = false;
2139
2140                 if (pinfo->prefix_len == 64) {
2141                         memcpy(&addr, &pinfo->prefix, 8);
2142
2143                         if (!ipv6_addr_any(&in6_dev->token)) {
2144                                 read_lock_bh(&in6_dev->lock);
2145                                 memcpy(addr.s6_addr + 8,
2146                                        in6_dev->token.s6_addr + 8, 8);
2147                                 read_unlock_bh(&in6_dev->lock);
2148                                 tokenized = true;
2149                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2150                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2151                                 in6_dev_put(in6_dev);
2152                                 return;
2153                         }
2154                         goto ok;
2155                 }
2156                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2157                                     pinfo->prefix_len);
2158                 in6_dev_put(in6_dev);
2159                 return;
2160
2161 ok:
2162
2163                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2164
2165                 if (ifp == NULL && valid_lft) {
2166                         int max_addresses = in6_dev->cnf.max_addresses;
2167                         u32 addr_flags = 0;
2168
2169 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2170                         if (in6_dev->cnf.optimistic_dad &&
2171                             !net->ipv6.devconf_all->forwarding && sllao)
2172                                 addr_flags = IFA_F_OPTIMISTIC;
2173 #endif
2174
2175                         /* Do not allow to create too much of autoconfigured
2176                          * addresses; this would be too easy way to crash kernel.
2177                          */
2178                         if (!max_addresses ||
2179                             ipv6_count_addresses(in6_dev) < max_addresses)
2180                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2181                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2182                                                     addr_flags);
2183
2184                         if (IS_ERR_OR_NULL(ifp)) {
2185                                 in6_dev_put(in6_dev);
2186                                 return;
2187                         }
2188
2189                         update_lft = create = 1;
2190                         ifp->cstamp = jiffies;
2191                         ifp->tokenized = tokenized;
2192                         addrconf_dad_start(ifp);
2193                 }
2194
2195                 if (ifp) {
2196                         int flags;
2197                         unsigned long now;
2198 #ifdef CONFIG_IPV6_PRIVACY
2199                         struct inet6_ifaddr *ift;
2200 #endif
2201                         u32 stored_lft;
2202
2203                         /* update lifetime (RFC2462 5.5.3 e) */
2204                         spin_lock(&ifp->lock);
2205                         now = jiffies;
2206                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2207                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2208                         else
2209                                 stored_lft = 0;
2210                         if (!update_lft && stored_lft) {
2211                                 if (valid_lft > MIN_VALID_LIFETIME ||
2212                                     valid_lft > stored_lft)
2213                                         update_lft = 1;
2214                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2215                                         /* valid_lft <= stored_lft is always true */
2216                                         /*
2217                                          * RFC 4862 Section 5.5.3e:
2218                                          * "Note that the preferred lifetime of
2219                                          *  the corresponding address is always
2220                                          *  reset to the Preferred Lifetime in
2221                                          *  the received Prefix Information
2222                                          *  option, regardless of whether the
2223                                          *  valid lifetime is also reset or
2224                                          *  ignored."
2225                                          *
2226                                          *  So if the preferred lifetime in
2227                                          *  this advertisement is different
2228                                          *  than what we have stored, but the
2229                                          *  valid lifetime is invalid, just
2230                                          *  reset prefered_lft.
2231                                          *
2232                                          *  We must set the valid lifetime
2233                                          *  to the stored lifetime since we'll
2234                                          *  be updating the timestamp below,
2235                                          *  else we'll set it back to the
2236                                          *  minimum.
2237                                          */
2238                                         if (prefered_lft != ifp->prefered_lft) {
2239                                                 valid_lft = stored_lft;
2240                                                 update_lft = 1;
2241                                         }
2242                                 } else {
2243                                         valid_lft = MIN_VALID_LIFETIME;
2244                                         if (valid_lft < prefered_lft)
2245                                                 prefered_lft = valid_lft;
2246                                         update_lft = 1;
2247                                 }
2248                         }
2249
2250                         if (update_lft) {
2251                                 ifp->valid_lft = valid_lft;
2252                                 ifp->prefered_lft = prefered_lft;
2253                                 ifp->tstamp = now;
2254                                 flags = ifp->flags;
2255                                 ifp->flags &= ~IFA_F_DEPRECATED;
2256                                 spin_unlock(&ifp->lock);
2257
2258                                 if (!(flags&IFA_F_TENTATIVE))
2259                                         ipv6_ifa_notify(0, ifp);
2260                         } else
2261                                 spin_unlock(&ifp->lock);
2262
2263 #ifdef CONFIG_IPV6_PRIVACY
2264                         read_lock_bh(&in6_dev->lock);
2265                         /* update all temporary addresses in the list */
2266                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2267                                             tmp_list) {
2268                                 int age, max_valid, max_prefered;
2269
2270                                 if (ifp != ift->ifpub)
2271                                         continue;
2272
2273                                 /*
2274                                  * RFC 4941 section 3.3:
2275                                  * If a received option will extend the lifetime
2276                                  * of a public address, the lifetimes of
2277                                  * temporary addresses should be extended,
2278                                  * subject to the overall constraint that no
2279                                  * temporary addresses should ever remain
2280                                  * "valid" or "preferred" for a time longer than
2281                                  * (TEMP_VALID_LIFETIME) or
2282                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2283                                  * respectively.
2284                                  */
2285                                 age = (now - ift->cstamp) / HZ;
2286                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2287                                 if (max_valid < 0)
2288                                         max_valid = 0;
2289
2290                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2291                                                in6_dev->cnf.max_desync_factor -
2292                                                age;
2293                                 if (max_prefered < 0)
2294                                         max_prefered = 0;
2295
2296                                 if (valid_lft > max_valid)
2297                                         valid_lft = max_valid;
2298
2299                                 if (prefered_lft > max_prefered)
2300                                         prefered_lft = max_prefered;
2301
2302                                 spin_lock(&ift->lock);
2303                                 flags = ift->flags;
2304                                 ift->valid_lft = valid_lft;
2305                                 ift->prefered_lft = prefered_lft;
2306                                 ift->tstamp = now;
2307                                 if (prefered_lft > 0)
2308                                         ift->flags &= ~IFA_F_DEPRECATED;
2309
2310                                 spin_unlock(&ift->lock);
2311                                 if (!(flags&IFA_F_TENTATIVE))
2312                                         ipv6_ifa_notify(0, ift);
2313                         }
2314
2315                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2316                                 /*
2317                                  * When a new public address is created as
2318                                  * described in [ADDRCONF], also create a new
2319                                  * temporary address. Also create a temporary
2320                                  * address if it's enabled but no temporary
2321                                  * address currently exists.
2322                                  */
2323                                 read_unlock_bh(&in6_dev->lock);
2324                                 ipv6_create_tempaddr(ifp, NULL);
2325                         } else {
2326                                 read_unlock_bh(&in6_dev->lock);
2327                         }
2328 #endif
2329                         in6_ifa_put(ifp);
2330                         addrconf_verify(0);
2331                 }
2332         }
2333         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2334         in6_dev_put(in6_dev);
2335 }
2336
2337 /*
2338  *      Set destination address.
2339  *      Special case for SIT interfaces where we create a new "virtual"
2340  *      device.
2341  */
2342 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2343 {
2344         struct in6_ifreq ireq;
2345         struct net_device *dev;
2346         int err = -EINVAL;
2347
2348         rtnl_lock();
2349
2350         err = -EFAULT;
2351         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2352                 goto err_exit;
2353
2354         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2355
2356         err = -ENODEV;
2357         if (dev == NULL)
2358                 goto err_exit;
2359
2360 #if IS_ENABLED(CONFIG_IPV6_SIT)
2361         if (dev->type == ARPHRD_SIT) {
2362                 const struct net_device_ops *ops = dev->netdev_ops;
2363                 struct ifreq ifr;
2364                 struct ip_tunnel_parm p;
2365
2366                 err = -EADDRNOTAVAIL;
2367                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2368                         goto err_exit;
2369
2370                 memset(&p, 0, sizeof(p));
2371                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2372                 p.iph.saddr = 0;
2373                 p.iph.version = 4;
2374                 p.iph.ihl = 5;
2375                 p.iph.protocol = IPPROTO_IPV6;
2376                 p.iph.ttl = 64;
2377                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2378
2379                 if (ops->ndo_do_ioctl) {
2380                         mm_segment_t oldfs = get_fs();
2381
2382                         set_fs(KERNEL_DS);
2383                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2384                         set_fs(oldfs);
2385                 } else
2386                         err = -EOPNOTSUPP;
2387
2388                 if (err == 0) {
2389                         err = -ENOBUFS;
2390                         dev = __dev_get_by_name(net, p.name);
2391                         if (!dev)
2392                                 goto err_exit;
2393                         err = dev_open(dev);
2394                 }
2395         }
2396 #endif
2397
2398 err_exit:
2399         rtnl_unlock();
2400         return err;
2401 }
2402
2403 /*
2404  *      Manual configuration of address on an interface
2405  */
2406 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2407                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2408                           __u32 valid_lft)
2409 {
2410         struct inet6_ifaddr *ifp;
2411         struct inet6_dev *idev;
2412         struct net_device *dev;
2413         int scope;
2414         u32 flags;
2415         clock_t expires;
2416         unsigned long timeout;
2417
2418         ASSERT_RTNL();
2419
2420         if (plen > 128)
2421                 return -EINVAL;
2422
2423         /* check the lifetime */
2424         if (!valid_lft || prefered_lft > valid_lft)
2425                 return -EINVAL;
2426
2427         dev = __dev_get_by_index(net, ifindex);
2428         if (!dev)
2429                 return -ENODEV;
2430
2431         idev = addrconf_add_dev(dev);
2432         if (IS_ERR(idev))
2433                 return PTR_ERR(idev);
2434
2435         scope = ipv6_addr_scope(pfx);
2436
2437         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2438         if (addrconf_finite_timeout(timeout)) {
2439                 expires = jiffies_to_clock_t(timeout * HZ);
2440                 valid_lft = timeout;
2441                 flags = RTF_EXPIRES;
2442         } else {
2443                 expires = 0;
2444                 flags = 0;
2445                 ifa_flags |= IFA_F_PERMANENT;
2446         }
2447
2448         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2449         if (addrconf_finite_timeout(timeout)) {
2450                 if (timeout == 0)
2451                         ifa_flags |= IFA_F_DEPRECATED;
2452                 prefered_lft = timeout;
2453         }
2454
2455         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2456
2457         if (!IS_ERR(ifp)) {
2458                 spin_lock_bh(&ifp->lock);
2459                 ifp->valid_lft = valid_lft;
2460                 ifp->prefered_lft = prefered_lft;
2461                 ifp->tstamp = jiffies;
2462                 spin_unlock_bh(&ifp->lock);
2463
2464                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2465                                       expires, flags);
2466                 /*
2467                  * Note that section 3.1 of RFC 4429 indicates
2468                  * that the Optimistic flag should not be set for
2469                  * manually configured addresses
2470                  */
2471                 addrconf_dad_start(ifp);
2472                 in6_ifa_put(ifp);
2473                 addrconf_verify(0);
2474                 return 0;
2475         }
2476
2477         return PTR_ERR(ifp);
2478 }
2479
2480 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2481                           unsigned int plen)
2482 {
2483         struct inet6_ifaddr *ifp;
2484         struct inet6_dev *idev;
2485         struct net_device *dev;
2486
2487         if (plen > 128)
2488                 return -EINVAL;
2489
2490         dev = __dev_get_by_index(net, ifindex);
2491         if (!dev)
2492                 return -ENODEV;
2493
2494         if ((idev = __in6_dev_get(dev)) == NULL)
2495                 return -ENXIO;
2496
2497         read_lock_bh(&idev->lock);
2498         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2499                 if (ifp->prefix_len == plen &&
2500                     ipv6_addr_equal(pfx, &ifp->addr)) {
2501                         in6_ifa_hold(ifp);
2502                         read_unlock_bh(&idev->lock);
2503
2504                         ipv6_del_addr(ifp);
2505
2506                         /* If the last address is deleted administratively,
2507                            disable IPv6 on this interface.
2508                          */
2509                         if (list_empty(&idev->addr_list))
2510                                 addrconf_ifdown(idev->dev, 1);
2511                         return 0;
2512                 }
2513         }
2514         read_unlock_bh(&idev->lock);
2515         return -EADDRNOTAVAIL;
2516 }
2517
2518
2519 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2520 {
2521         struct in6_ifreq ireq;
2522         int err;
2523
2524         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2525                 return -EPERM;
2526
2527         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2528                 return -EFAULT;
2529
2530         rtnl_lock();
2531         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2532                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2533                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2534         rtnl_unlock();
2535         return err;
2536 }
2537
2538 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2539 {
2540         struct in6_ifreq ireq;
2541         int err;
2542
2543         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2544                 return -EPERM;
2545
2546         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2547                 return -EFAULT;
2548
2549         rtnl_lock();
2550         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2551                              ireq.ifr6_prefixlen);
2552         rtnl_unlock();
2553         return err;
2554 }
2555
2556 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2557                      int plen, int scope)
2558 {
2559         struct inet6_ifaddr *ifp;
2560
2561         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2562         if (!IS_ERR(ifp)) {
2563                 spin_lock_bh(&ifp->lock);
2564                 ifp->flags &= ~IFA_F_TENTATIVE;
2565                 spin_unlock_bh(&ifp->lock);
2566                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2567                 in6_ifa_put(ifp);
2568         }
2569 }
2570
2571 #if IS_ENABLED(CONFIG_IPV6_SIT)
2572 static void sit_add_v4_addrs(struct inet6_dev *idev)
2573 {
2574         struct in6_addr addr;
2575         struct net_device *dev;
2576         struct net *net = dev_net(idev->dev);
2577         int scope;
2578
2579         ASSERT_RTNL();
2580
2581         memset(&addr, 0, sizeof(struct in6_addr));
2582         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2583
2584         if (idev->dev->flags&IFF_POINTOPOINT) {
2585                 addr.s6_addr32[0] = htonl(0xfe800000);
2586                 scope = IFA_LINK;
2587         } else {
2588                 scope = IPV6_ADDR_COMPATv4;
2589         }
2590
2591         if (addr.s6_addr32[3]) {
2592                 add_addr(idev, &addr, 128, scope);
2593                 return;
2594         }
2595
2596         for_each_netdev(net, dev) {
2597                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2598                 if (in_dev && (dev->flags & IFF_UP)) {
2599                         struct in_ifaddr *ifa;
2600
2601                         int flag = scope;
2602
2603                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2604                                 int plen;
2605
2606                                 addr.s6_addr32[3] = ifa->ifa_local;
2607
2608                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2609                                         continue;
2610                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2611                                         if (idev->dev->flags&IFF_POINTOPOINT)
2612                                                 continue;
2613                                         flag |= IFA_HOST;
2614                                 }
2615                                 if (idev->dev->flags&IFF_POINTOPOINT)
2616                                         plen = 64;
2617                                 else
2618                                         plen = 96;
2619
2620                                 add_addr(idev, &addr, plen, flag);
2621                         }
2622                 }
2623         }
2624 }
2625 #endif
2626
2627 static void init_loopback(struct net_device *dev)
2628 {
2629         struct inet6_dev  *idev;
2630         struct net_device *sp_dev;
2631         struct inet6_ifaddr *sp_ifa;
2632         struct rt6_info *sp_rt;
2633
2634         /* ::1 */
2635
2636         ASSERT_RTNL();
2637
2638         if ((idev = ipv6_find_idev(dev)) == NULL) {
2639                 pr_debug("%s: add_dev failed\n", __func__);
2640                 return;
2641         }
2642
2643         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2644
2645         /* Add routes to other interface's IPv6 addresses */
2646         for_each_netdev(dev_net(dev), sp_dev) {
2647                 if (!strcmp(sp_dev->name, dev->name))
2648                         continue;
2649
2650                 idev = __in6_dev_get(sp_dev);
2651                 if (!idev)
2652                         continue;
2653
2654                 read_lock_bh(&idev->lock);
2655                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2656
2657                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2658                                 continue;
2659
2660                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2661
2662                         /* Failure cases are ignored */
2663                         if (!IS_ERR(sp_rt))
2664                                 ip6_ins_rt(sp_rt);
2665                 }
2666                 read_unlock_bh(&idev->lock);
2667         }
2668 }
2669
2670 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2671 {
2672         struct inet6_ifaddr *ifp;
2673         u32 addr_flags = IFA_F_PERMANENT;
2674
2675 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2676         if (idev->cnf.optimistic_dad &&
2677             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2678                 addr_flags |= IFA_F_OPTIMISTIC;
2679 #endif
2680
2681
2682         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2683         if (!IS_ERR(ifp)) {
2684                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2685                 addrconf_dad_start(ifp);
2686                 in6_ifa_put(ifp);
2687         }
2688 }
2689
2690 static void addrconf_dev_config(struct net_device *dev)
2691 {
2692         struct in6_addr addr;
2693         struct inet6_dev *idev;
2694
2695         ASSERT_RTNL();
2696
2697         if ((dev->type != ARPHRD_ETHER) &&
2698             (dev->type != ARPHRD_FDDI) &&
2699             (dev->type != ARPHRD_ARCNET) &&
2700             (dev->type != ARPHRD_INFINIBAND) &&
2701             (dev->type != ARPHRD_IEEE802154) &&
2702             (dev->type != ARPHRD_IEEE1394)) {
2703                 /* Alas, we support only Ethernet autoconfiguration. */
2704                 return;
2705         }
2706
2707         idev = addrconf_add_dev(dev);
2708         if (IS_ERR(idev))
2709                 return;
2710
2711         memset(&addr, 0, sizeof(struct in6_addr));
2712         addr.s6_addr32[0] = htonl(0xFE800000);
2713
2714         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2715                 addrconf_add_linklocal(idev, &addr);
2716 }
2717
2718 #if IS_ENABLED(CONFIG_IPV6_SIT)
2719 static void addrconf_sit_config(struct net_device *dev)
2720 {
2721         struct inet6_dev *idev;
2722
2723         ASSERT_RTNL();
2724
2725         /*
2726          * Configure the tunnel with one of our IPv4
2727          * addresses... we should configure all of
2728          * our v4 addrs in the tunnel
2729          */
2730
2731         if ((idev = ipv6_find_idev(dev)) == NULL) {
2732                 pr_debug("%s: add_dev failed\n", __func__);
2733                 return;
2734         }
2735
2736         if (dev->priv_flags & IFF_ISATAP) {
2737                 struct in6_addr addr;
2738
2739                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2740                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2741                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2742                         addrconf_add_linklocal(idev, &addr);
2743                 return;
2744         }
2745
2746         sit_add_v4_addrs(idev);
2747
2748         if (dev->flags&IFF_POINTOPOINT)
2749                 addrconf_add_mroute(dev);
2750         else
2751                 sit_route_add(dev);
2752 }
2753 #endif
2754
2755 #if IS_ENABLED(CONFIG_NET_IPGRE)
2756 static void addrconf_gre_config(struct net_device *dev)
2757 {
2758         struct inet6_dev *idev;
2759         struct in6_addr addr;
2760
2761         pr_info("%s(%s)\n", __func__, dev->name);
2762
2763         ASSERT_RTNL();
2764
2765         if ((idev = ipv6_find_idev(dev)) == NULL) {
2766                 pr_debug("%s: add_dev failed\n", __func__);
2767                 return;
2768         }
2769
2770         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2771         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2772
2773         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2774                 addrconf_add_linklocal(idev, &addr);
2775 }
2776 #endif
2777
2778 static inline int
2779 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2780 {
2781         struct in6_addr lladdr;
2782
2783         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2784                 addrconf_add_linklocal(idev, &lladdr);
2785                 return 0;
2786         }
2787         return -1;
2788 }
2789
2790 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2791 {
2792         struct net_device *link_dev;
2793         struct net *net = dev_net(idev->dev);
2794
2795         /* first try to inherit the link-local address from the link device */
2796         if (idev->dev->iflink &&
2797             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2798                 if (!ipv6_inherit_linklocal(idev, link_dev))
2799                         return;
2800         }
2801         /* then try to inherit it from any device */
2802         for_each_netdev(net, link_dev) {
2803                 if (!ipv6_inherit_linklocal(idev, link_dev))
2804                         return;
2805         }
2806         pr_debug("init ip6-ip6: add_linklocal failed\n");
2807 }
2808
2809 /*
2810  * Autoconfigure tunnel with a link-local address so routing protocols,
2811  * DHCPv6, MLD etc. can be run over the virtual link
2812  */
2813
2814 static void addrconf_ip6_tnl_config(struct net_device *dev)
2815 {
2816         struct inet6_dev *idev;
2817
2818         ASSERT_RTNL();
2819
2820         idev = addrconf_add_dev(dev);
2821         if (IS_ERR(idev)) {
2822                 pr_debug("init ip6-ip6: add_dev failed\n");
2823                 return;
2824         }
2825         ip6_tnl_add_linklocal(idev);
2826 }
2827
2828 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2829                            void *data)
2830 {
2831         struct net_device *dev = (struct net_device *) data;
2832         struct inet6_dev *idev = __in6_dev_get(dev);
2833         int run_pending = 0;
2834         int err;
2835
2836         switch (event) {
2837         case NETDEV_REGISTER:
2838                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2839                         idev = ipv6_add_dev(dev);
2840                         if (!idev)
2841                                 return notifier_from_errno(-ENOMEM);
2842                 }
2843                 break;
2844
2845         case NETDEV_UP:
2846         case NETDEV_CHANGE:
2847                 if (dev->flags & IFF_SLAVE)
2848                         break;
2849
2850                 if (event == NETDEV_UP) {
2851                         if (!addrconf_qdisc_ok(dev)) {
2852                                 /* device is not ready yet. */
2853                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2854                                         dev->name);
2855                                 break;
2856                         }
2857
2858                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2859                                 idev = ipv6_add_dev(dev);
2860
2861                         if (idev) {
2862                                 idev->if_flags |= IF_READY;
2863                                 run_pending = 1;
2864                         }
2865                 } else {
2866                         if (!addrconf_qdisc_ok(dev)) {
2867                                 /* device is still not ready. */
2868                                 break;
2869                         }
2870
2871                         if (idev) {
2872                                 if (idev->if_flags & IF_READY)
2873                                         /* device is already configured. */
2874                                         break;
2875                                 idev->if_flags |= IF_READY;
2876                         }
2877
2878                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2879                                 dev->name);
2880
2881                         run_pending = 1;
2882                 }
2883
2884                 switch (dev->type) {
2885 #if IS_ENABLED(CONFIG_IPV6_SIT)
2886                 case ARPHRD_SIT:
2887                         addrconf_sit_config(dev);
2888                         break;
2889 #endif
2890 #if IS_ENABLED(CONFIG_NET_IPGRE)
2891                 case ARPHRD_IPGRE:
2892                         addrconf_gre_config(dev);
2893                         break;
2894 #endif
2895                 case ARPHRD_TUNNEL6:
2896                         addrconf_ip6_tnl_config(dev);
2897                         break;
2898                 case ARPHRD_LOOPBACK:
2899                         init_loopback(dev);
2900                         break;
2901
2902                 default:
2903                         addrconf_dev_config(dev);
2904                         break;
2905                 }
2906
2907                 if (idev) {
2908                         if (run_pending)
2909                                 addrconf_dad_run(idev);
2910
2911                         /*
2912                          * If the MTU changed during the interface down,
2913                          * when the interface up, the changed MTU must be
2914                          * reflected in the idev as well as routers.
2915                          */
2916                         if (idev->cnf.mtu6 != dev->mtu &&
2917                             dev->mtu >= IPV6_MIN_MTU) {
2918                                 rt6_mtu_change(dev, dev->mtu);
2919                                 idev->cnf.mtu6 = dev->mtu;
2920                         }
2921                         idev->tstamp = jiffies;
2922                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2923
2924                         /*
2925                          * If the changed mtu during down is lower than
2926                          * IPV6_MIN_MTU stop IPv6 on this interface.
2927                          */
2928                         if (dev->mtu < IPV6_MIN_MTU)
2929                                 addrconf_ifdown(dev, 1);
2930                 }
2931                 break;
2932
2933         case NETDEV_CHANGEMTU:
2934                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2935                         rt6_mtu_change(dev, dev->mtu);
2936                         idev->cnf.mtu6 = dev->mtu;
2937                         break;
2938                 }
2939
2940                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2941                         idev = ipv6_add_dev(dev);
2942                         if (idev)
2943                                 break;
2944                 }
2945
2946                 /*
2947                  * MTU falled under IPV6_MIN_MTU.
2948                  * Stop IPv6 on this interface.
2949                  */
2950
2951         case NETDEV_DOWN:
2952         case NETDEV_UNREGISTER:
2953                 /*
2954                  *      Remove all addresses from this interface.
2955                  */
2956                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2957                 break;
2958
2959         case NETDEV_CHANGENAME:
2960                 if (idev) {
2961                         snmp6_unregister_dev(idev);
2962                         addrconf_sysctl_unregister(idev);
2963                         addrconf_sysctl_register(idev);
2964                         err = snmp6_register_dev(idev);
2965                         if (err)
2966                                 return notifier_from_errno(err);
2967                 }
2968                 break;
2969
2970         case NETDEV_PRE_TYPE_CHANGE:
2971         case NETDEV_POST_TYPE_CHANGE:
2972                 addrconf_type_change(dev, event);
2973                 break;
2974         }
2975
2976         return NOTIFY_OK;
2977 }
2978
2979 /*
2980  *      addrconf module should be notified of a device going up
2981  */
2982 static struct notifier_block ipv6_dev_notf = {
2983         .notifier_call = addrconf_notify,
2984 };
2985
2986 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2987 {
2988         struct inet6_dev *idev;
2989         ASSERT_RTNL();
2990
2991         idev = __in6_dev_get(dev);
2992
2993         if (event == NETDEV_POST_TYPE_CHANGE)
2994                 ipv6_mc_remap(idev);
2995         else if (event == NETDEV_PRE_TYPE_CHANGE)
2996                 ipv6_mc_unmap(idev);
2997 }
2998
2999 static int addrconf_ifdown(struct net_device *dev, int how)
3000 {
3001         struct net *net = dev_net(dev);
3002         struct inet6_dev *idev;
3003         struct inet6_ifaddr *ifa;
3004         int state, i;
3005
3006         ASSERT_RTNL();
3007
3008         rt6_ifdown(net, dev);
3009         neigh_ifdown(&nd_tbl, dev);
3010
3011         idev = __in6_dev_get(dev);
3012         if (idev == NULL)
3013                 return -ENODEV;
3014
3015         /*
3016          * Step 1: remove reference to ipv6 device from parent device.
3017          *         Do not dev_put!
3018          */
3019         if (how) {
3020                 idev->dead = 1;
3021
3022                 /* protected by rtnl_lock */
3023                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3024
3025                 /* Step 1.5: remove snmp6 entry */
3026                 snmp6_unregister_dev(idev);
3027
3028         }
3029
3030         /* Step 2: clear hash table */
3031         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3032                 struct hlist_head *h = &inet6_addr_lst[i];
3033
3034                 spin_lock_bh(&addrconf_hash_lock);
3035         restart:
3036                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3037                         if (ifa->idev == idev) {
3038                                 hlist_del_init_rcu(&ifa->addr_lst);
3039                                 addrconf_del_timer(ifa);
3040                                 goto restart;
3041                         }
3042                 }
3043                 spin_unlock_bh(&addrconf_hash_lock);
3044         }
3045
3046         write_lock_bh(&idev->lock);
3047
3048         /* Step 2: clear flags for stateless addrconf */
3049         if (!how)
3050                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3051
3052 #ifdef CONFIG_IPV6_PRIVACY
3053         if (how && del_timer(&idev->regen_timer))
3054                 in6_dev_put(idev);
3055
3056         /* Step 3: clear tempaddr list */
3057         while (!list_empty(&idev->tempaddr_list)) {
3058                 ifa = list_first_entry(&idev->tempaddr_list,
3059                                        struct inet6_ifaddr, tmp_list);
3060                 list_del(&ifa->tmp_list);
3061                 write_unlock_bh(&idev->lock);
3062                 spin_lock_bh(&ifa->lock);
3063
3064                 if (ifa->ifpub) {
3065                         in6_ifa_put(ifa->ifpub);
3066                         ifa->ifpub = NULL;
3067                 }
3068                 spin_unlock_bh(&ifa->lock);
3069                 in6_ifa_put(ifa);
3070                 write_lock_bh(&idev->lock);
3071         }
3072 #endif
3073
3074         while (!list_empty(&idev->addr_list)) {
3075                 ifa = list_first_entry(&idev->addr_list,
3076                                        struct inet6_ifaddr, if_list);
3077                 addrconf_del_timer(ifa);
3078
3079                 list_del(&ifa->if_list);
3080
3081                 write_unlock_bh(&idev->lock);
3082
3083                 spin_lock_bh(&ifa->state_lock);
3084                 state = ifa->state;
3085                 ifa->state = INET6_IFADDR_STATE_DEAD;
3086                 spin_unlock_bh(&ifa->state_lock);
3087
3088                 if (state != INET6_IFADDR_STATE_DEAD) {
3089                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3090                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
3091                 }
3092                 in6_ifa_put(ifa);
3093
3094                 write_lock_bh(&idev->lock);
3095         }
3096
3097         write_unlock_bh(&idev->lock);
3098
3099         /* Step 5: Discard multicast list */
3100         if (how)
3101                 ipv6_mc_destroy_dev(idev);
3102         else
3103                 ipv6_mc_down(idev);
3104
3105         idev->tstamp = jiffies;
3106
3107         /* Last: Shot the device (if unregistered) */
3108         if (how) {
3109                 addrconf_sysctl_unregister(idev);
3110                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3111                 neigh_ifdown(&nd_tbl, dev);
3112                 in6_dev_put(idev);
3113         }
3114         return 0;
3115 }
3116
3117 static void addrconf_rs_timer(unsigned long data)
3118 {
3119         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3120         struct inet6_dev *idev = ifp->idev;
3121
3122         read_lock(&idev->lock);
3123         if (idev->dead || !(idev->if_flags & IF_READY))
3124                 goto out;
3125
3126         if (!ipv6_accept_ra(idev))
3127                 goto out;
3128
3129         /* Announcement received after solicitation was sent */
3130         if (idev->if_flags & IF_RA_RCVD)
3131                 goto out;
3132
3133         spin_lock(&ifp->lock);
3134         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3135                 /* The wait after the last probe can be shorter */
3136                 addrconf_mod_timer(ifp, AC_RS,
3137                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3138                                    idev->cnf.rtr_solicit_delay :
3139                                    idev->cnf.rtr_solicit_interval);
3140                 spin_unlock(&ifp->lock);
3141
3142                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3143         } else {
3144                 spin_unlock(&ifp->lock);
3145                 /*
3146                  * Note: we do not support deprecated "all on-link"
3147                  * assumption any longer.
3148                  */
3149                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3150         }
3151
3152 out:
3153         read_unlock(&idev->lock);
3154         in6_ifa_put(ifp);
3155 }
3156
3157 /*
3158  *      Duplicate Address Detection
3159  */
3160 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3161 {
3162         unsigned long rand_num;
3163         struct inet6_dev *idev = ifp->idev;
3164
3165         if (ifp->flags & IFA_F_OPTIMISTIC)
3166                 rand_num = 0;
3167         else
3168                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3169
3170         ifp->probes = idev->cnf.dad_transmits;
3171         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3172 }
3173
3174 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3175 {
3176         struct inet6_dev *idev = ifp->idev;
3177         struct net_device *dev = idev->dev;
3178
3179         addrconf_join_solict(dev, &ifp->addr);
3180
3181         net_srandom(ifp->addr.s6_addr32[3]);
3182
3183         read_lock_bh(&idev->lock);
3184         spin_lock(&ifp->lock);
3185         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3186                 goto out;
3187
3188         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3189             idev->cnf.accept_dad < 1 ||
3190             !(ifp->flags&IFA_F_TENTATIVE) ||
3191             ifp->flags & IFA_F_NODAD) {
3192                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3193                 spin_unlock(&ifp->lock);
3194                 read_unlock_bh(&idev->lock);
3195
3196                 addrconf_dad_completed(ifp);
3197                 return;
3198         }
3199
3200         if (!(idev->if_flags & IF_READY)) {
3201                 spin_unlock(&ifp->lock);
3202                 read_unlock_bh(&idev->lock);
3203                 /*
3204                  * If the device is not ready:
3205                  * - keep it tentative if it is a permanent address.
3206                  * - otherwise, kill it.
3207                  */
3208                 in6_ifa_hold(ifp);
3209                 addrconf_dad_stop(ifp, 0);
3210                 return;
3211         }
3212
3213         /*
3214          * Optimistic nodes can start receiving
3215          * Frames right away
3216          */
3217         if (ifp->flags & IFA_F_OPTIMISTIC)
3218                 ip6_ins_rt(ifp->rt);
3219
3220         addrconf_dad_kick(ifp);
3221 out:
3222         spin_unlock(&ifp->lock);
3223         read_unlock_bh(&idev->lock);
3224 }
3225
3226 static void addrconf_dad_timer(unsigned long data)
3227 {
3228         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3229         struct inet6_dev *idev = ifp->idev;
3230         struct in6_addr mcaddr;
3231
3232         if (!ifp->probes && addrconf_dad_end(ifp))
3233                 goto out;
3234
3235         read_lock(&idev->lock);
3236         if (idev->dead || !(idev->if_flags & IF_READY)) {
3237                 read_unlock(&idev->lock);
3238                 goto out;
3239         }
3240
3241         spin_lock(&ifp->lock);
3242         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3243                 spin_unlock(&ifp->lock);
3244                 read_unlock(&idev->lock);
3245                 goto out;
3246         }
3247
3248         if (ifp->probes == 0) {
3249                 /*
3250                  * DAD was successful
3251                  */
3252
3253                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3254                 spin_unlock(&ifp->lock);
3255                 read_unlock(&idev->lock);
3256
3257                 addrconf_dad_completed(ifp);
3258
3259                 goto out;
3260         }
3261
3262         ifp->probes--;
3263         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3264         spin_unlock(&ifp->lock);
3265         read_unlock(&idev->lock);
3266
3267         /* send a neighbour solicitation for our addr */
3268         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3269         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3270 out:
3271         in6_ifa_put(ifp);
3272 }
3273
3274 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3275 {
3276         struct net_device *dev = ifp->idev->dev;
3277
3278         /*
3279          *      Configure the address for reception. Now it is valid.
3280          */
3281
3282         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3283
3284         /* If added prefix is link local and we are prepared to process
3285            router advertisements, start sending router solicitations.
3286          */
3287
3288         if (ipv6_accept_ra(ifp->idev) &&
3289             ifp->idev->cnf.rtr_solicits > 0 &&
3290             (dev->flags&IFF_LOOPBACK) == 0 &&
3291             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3292                 /*
3293                  *      If a host as already performed a random delay
3294                  *      [...] as part of DAD [...] there is no need
3295                  *      to delay again before sending the first RS
3296                  */
3297                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3298
3299                 spin_lock_bh(&ifp->lock);
3300                 ifp->probes = 1;
3301                 ifp->idev->if_flags |= IF_RS_SENT;
3302                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3303                 spin_unlock_bh(&ifp->lock);
3304         }
3305 }
3306
3307 static void addrconf_dad_run(struct inet6_dev *idev)
3308 {
3309         struct inet6_ifaddr *ifp;
3310
3311         read_lock_bh(&idev->lock);
3312         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3313                 spin_lock(&ifp->lock);
3314                 if (ifp->flags & IFA_F_TENTATIVE &&
3315                     ifp->state == INET6_IFADDR_STATE_DAD)
3316                         addrconf_dad_kick(ifp);
3317                 spin_unlock(&ifp->lock);
3318         }
3319         read_unlock_bh(&idev->lock);
3320 }
3321
3322 #ifdef CONFIG_PROC_FS
3323 struct if6_iter_state {
3324         struct seq_net_private p;
3325         int bucket;
3326         int offset;
3327 };
3328
3329 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3330 {
3331         struct inet6_ifaddr *ifa = NULL;
3332         struct if6_iter_state *state = seq->private;
3333         struct net *net = seq_file_net(seq);
3334         int p = 0;
3335
3336         /* initial bucket if pos is 0 */
3337         if (pos == 0) {
3338                 state->bucket = 0;
3339                 state->offset = 0;
3340         }
3341
3342         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3343                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3344                                          addr_lst) {
3345                         if (!net_eq(dev_net(ifa->idev->dev), net))
3346                                 continue;
3347                         /* sync with offset */
3348                         if (p < state->offset) {
3349                                 p++;
3350                                 continue;
3351                         }
3352                         state->offset++;
3353                         return ifa;
3354                 }
3355
3356                 /* prepare for next bucket */
3357                 state->offset = 0;
3358                 p = 0;
3359         }
3360         return NULL;
3361 }
3362
3363 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3364                                          struct inet6_ifaddr *ifa)
3365 {
3366         struct if6_iter_state *state = seq->private;
3367         struct net *net = seq_file_net(seq);
3368
3369         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3370                 if (!net_eq(dev_net(ifa->idev->dev), net))
3371                         continue;
3372                 state->offset++;
3373                 return ifa;
3374         }
3375
3376         while (++state->bucket < IN6_ADDR_HSIZE) {
3377                 state->offset = 0;
3378                 hlist_for_each_entry_rcu_bh(ifa,
3379                                      &inet6_addr_lst[state->bucket], addr_lst) {
3380                         if (!net_eq(dev_net(ifa->idev->dev), net))
3381                                 continue;
3382                         state->offset++;
3383                         return ifa;
3384                 }
3385         }
3386
3387         return NULL;
3388 }
3389
3390 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3391         __acquires(rcu_bh)
3392 {
3393         rcu_read_lock_bh();
3394         return if6_get_first(seq, *pos);
3395 }
3396
3397 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3398 {
3399         struct inet6_ifaddr *ifa;
3400
3401         ifa = if6_get_next(seq, v);
3402         ++*pos;
3403         return ifa;
3404 }
3405
3406 static void if6_seq_stop(struct seq_file *seq, void *v)
3407         __releases(rcu_bh)
3408 {
3409         rcu_read_unlock_bh();
3410 }
3411
3412 static int if6_seq_show(struct seq_file *seq, void *v)
3413 {
3414         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3415         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3416                    &ifp->addr,
3417                    ifp->idev->dev->ifindex,
3418                    ifp->prefix_len,
3419                    ifp->scope,
3420                    ifp->flags,
3421                    ifp->idev->dev->name);
3422         return 0;
3423 }
3424
3425 static const struct seq_operations if6_seq_ops = {
3426         .start  = if6_seq_start,
3427         .next   = if6_seq_next,
3428         .show   = if6_seq_show,
3429         .stop   = if6_seq_stop,
3430 };
3431
3432 static int if6_seq_open(struct inode *inode, struct file *file)
3433 {
3434         return seq_open_net(inode, file, &if6_seq_ops,
3435                             sizeof(struct if6_iter_state));
3436 }
3437
3438 static const struct file_operations if6_fops = {
3439         .owner          = THIS_MODULE,
3440         .open           = if6_seq_open,
3441         .read           = seq_read,
3442         .llseek         = seq_lseek,
3443         .release        = seq_release_net,
3444 };
3445
3446 static int __net_init if6_proc_net_init(struct net *net)
3447 {
3448         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3449                 return -ENOMEM;
3450         return 0;
3451 }
3452
3453 static void __net_exit if6_proc_net_exit(struct net *net)
3454 {
3455         remove_proc_entry("if_inet6", net->proc_net);
3456 }
3457
3458 static struct pernet_operations if6_proc_net_ops = {
3459        .init = if6_proc_net_init,
3460        .exit = if6_proc_net_exit,
3461 };
3462
3463 int __init if6_proc_init(void)
3464 {
3465         return register_pernet_subsys(&if6_proc_net_ops);
3466 }
3467
3468 void if6_proc_exit(void)
3469 {
3470         unregister_pernet_subsys(&if6_proc_net_ops);
3471 }
3472 #endif  /* CONFIG_PROC_FS */
3473
3474 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3475 /* Check if address is a home address configured on any interface. */
3476 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3477 {
3478         int ret = 0;
3479         struct inet6_ifaddr *ifp = NULL;
3480         unsigned int hash = inet6_addr_hash(addr);
3481
3482         rcu_read_lock_bh();
3483         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3484                 if (!net_eq(dev_net(ifp->idev->dev), net))
3485                         continue;
3486                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3487                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3488                         ret = 1;
3489                         break;
3490                 }
3491         }
3492         rcu_read_unlock_bh();
3493         return ret;
3494 }
3495 #endif
3496
3497 /*
3498  *      Periodic address status verification
3499  */
3500
3501 static void addrconf_verify(unsigned long foo)
3502 {
3503         unsigned long now, next, next_sec, next_sched;
3504         struct inet6_ifaddr *ifp;
3505         int i;
3506
3507         rcu_read_lock_bh();
3508         spin_lock(&addrconf_verify_lock);
3509         now = jiffies;
3510         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3511
3512         del_timer(&addr_chk_timer);
3513
3514         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3515 restart:
3516                 hlist_for_each_entry_rcu_bh(ifp,
3517                                          &inet6_addr_lst[i], addr_lst) {
3518                         unsigned long age;
3519
3520                         if (ifp->flags & IFA_F_PERMANENT)
3521                                 continue;
3522
3523                         spin_lock(&ifp->lock);
3524                         /* We try to batch several events at once. */
3525                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3526
3527                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3528                             age >= ifp->valid_lft) {
3529                                 spin_unlock(&ifp->lock);
3530                                 in6_ifa_hold(ifp);
3531                                 ipv6_del_addr(ifp);
3532                                 goto restart;
3533                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3534                                 spin_unlock(&ifp->lock);
3535                                 continue;
3536                         } else if (age >= ifp->prefered_lft) {
3537                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3538                                 int deprecate = 0;
3539
3540                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3541                                         deprecate = 1;
3542                                         ifp->flags |= IFA_F_DEPRECATED;
3543                                 }
3544
3545                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3546                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3547
3548                                 spin_unlock(&ifp->lock);
3549
3550                                 if (deprecate) {
3551                                         in6_ifa_hold(ifp);
3552
3553                                         ipv6_ifa_notify(0, ifp);
3554                                         in6_ifa_put(ifp);
3555                                         goto restart;
3556                                 }
3557 #ifdef CONFIG_IPV6_PRIVACY
3558                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3559                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3560                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3561                                         ifp->idev->cnf.dad_transmits *
3562                                         ifp->idev->nd_parms->retrans_time / HZ;
3563
3564                                 if (age >= ifp->prefered_lft - regen_advance) {
3565                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3566                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3567                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3568                                         if (!ifp->regen_count && ifpub) {
3569                                                 ifp->regen_count++;
3570                                                 in6_ifa_hold(ifp);
3571                                                 in6_ifa_hold(ifpub);
3572                                                 spin_unlock(&ifp->lock);
3573
3574                                                 spin_lock(&ifpub->lock);
3575                                                 ifpub->regen_count = 0;
3576                                                 spin_unlock(&ifpub->lock);
3577                                                 ipv6_create_tempaddr(ifpub, ifp);
3578                                                 in6_ifa_put(ifpub);
3579                                                 in6_ifa_put(ifp);
3580                                                 goto restart;
3581                                         }
3582                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3583                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3584                                 spin_unlock(&ifp->lock);
3585 #endif
3586                         } else {
3587                                 /* ifp->prefered_lft <= ifp->valid_lft */
3588                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3589                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3590                                 spin_unlock(&ifp->lock);
3591                         }
3592                 }
3593         }
3594
3595         next_sec = round_jiffies_up(next);
3596         next_sched = next;
3597
3598         /* If rounded timeout is accurate enough, accept it. */
3599         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3600                 next_sched = next_sec;
3601
3602         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3603         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3604                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3605
3606         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3607               now, next, next_sec, next_sched));
3608
3609         addr_chk_timer.expires = next_sched;
3610         add_timer(&addr_chk_timer);
3611         spin_unlock(&addrconf_verify_lock);
3612         rcu_read_unlock_bh();
3613 }
3614
3615 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3616 {
3617         struct in6_addr *pfx = NULL;
3618
3619         if (addr)
3620                 pfx = nla_data(addr);
3621
3622         if (local) {
3623                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3624                         pfx = NULL;
3625                 else
3626                         pfx = nla_data(local);
3627         }
3628
3629         return pfx;
3630 }
3631
3632 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3633         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3634         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3635         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3636 };
3637
3638 static int
3639 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3640 {
3641         struct net *net = sock_net(skb->sk);
3642         struct ifaddrmsg *ifm;
3643         struct nlattr *tb[IFA_MAX+1];
3644         struct in6_addr *pfx;
3645         int err;
3646
3647         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3648         if (err < 0)
3649                 return err;
3650
3651         ifm = nlmsg_data(nlh);
3652         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3653         if (pfx == NULL)
3654                 return -EINVAL;
3655
3656         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3657 }
3658
3659 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3660                              u32 prefered_lft, u32 valid_lft)
3661 {
3662         u32 flags;
3663         clock_t expires;
3664         unsigned long timeout;
3665
3666         if (!valid_lft || (prefered_lft > valid_lft))
3667                 return -EINVAL;
3668
3669         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3670         if (addrconf_finite_timeout(timeout)) {
3671                 expires = jiffies_to_clock_t(timeout * HZ);
3672                 valid_lft = timeout;
3673                 flags = RTF_EXPIRES;
3674         } else {
3675                 expires = 0;
3676                 flags = 0;
3677                 ifa_flags |= IFA_F_PERMANENT;
3678         }
3679
3680         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3681         if (addrconf_finite_timeout(timeout)) {
3682                 if (timeout == 0)
3683                         ifa_flags |= IFA_F_DEPRECATED;
3684                 prefered_lft = timeout;
3685         }
3686
3687         spin_lock_bh(&ifp->lock);
3688         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3689         ifp->tstamp = jiffies;
3690         ifp->valid_lft = valid_lft;
3691         ifp->prefered_lft = prefered_lft;
3692
3693         spin_unlock_bh(&ifp->lock);
3694         if (!(ifp->flags&IFA_F_TENTATIVE))
3695                 ipv6_ifa_notify(0, ifp);
3696
3697         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3698                               expires, flags);
3699         addrconf_verify(0);
3700
3701         return 0;
3702 }
3703
3704 static int
3705 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3706 {
3707         struct net *net = sock_net(skb->sk);
3708         struct ifaddrmsg *ifm;
3709         struct nlattr *tb[IFA_MAX+1];
3710         struct in6_addr *pfx;
3711         struct inet6_ifaddr *ifa;
3712         struct net_device *dev;
3713         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3714         u8 ifa_flags;
3715         int err;
3716
3717         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3718         if (err < 0)
3719                 return err;
3720
3721         ifm = nlmsg_data(nlh);
3722         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3723         if (pfx == NULL)
3724                 return -EINVAL;
3725
3726         if (tb[IFA_CACHEINFO]) {
3727                 struct ifa_cacheinfo *ci;
3728
3729                 ci = nla_data(tb[IFA_CACHEINFO]);
3730                 valid_lft = ci->ifa_valid;
3731                 preferred_lft = ci->ifa_prefered;
3732         } else {
3733                 preferred_lft = INFINITY_LIFE_TIME;
3734                 valid_lft = INFINITY_LIFE_TIME;
3735         }
3736
3737         dev =  __dev_get_by_index(net, ifm->ifa_index);
3738         if (dev == NULL)
3739                 return -ENODEV;
3740
3741         /* We ignore other flags so far. */
3742         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3743
3744         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3745         if (ifa == NULL) {
3746                 /*
3747                  * It would be best to check for !NLM_F_CREATE here but
3748                  * userspace alreay relies on not having to provide this.
3749                  */
3750                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3751                                       ifm->ifa_prefixlen, ifa_flags,
3752                                       preferred_lft, valid_lft);
3753         }
3754
3755         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3756             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3757                 err = -EEXIST;
3758         else
3759                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3760
3761         in6_ifa_put(ifa);
3762
3763         return err;
3764 }
3765
3766 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3767                           u8 scope, int ifindex)
3768 {
3769         struct ifaddrmsg *ifm;
3770
3771         ifm = nlmsg_data(nlh);
3772         ifm->ifa_family = AF_INET6;
3773         ifm->ifa_prefixlen = prefixlen;
3774         ifm->ifa_flags = flags;
3775         ifm->ifa_scope = scope;
3776         ifm->ifa_index = ifindex;
3777 }
3778
3779 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3780                          unsigned long tstamp, u32 preferred, u32 valid)
3781 {
3782         struct ifa_cacheinfo ci;
3783
3784         ci.cstamp = cstamp_delta(cstamp);
3785         ci.tstamp = cstamp_delta(tstamp);
3786         ci.ifa_prefered = preferred;
3787         ci.ifa_valid = valid;
3788
3789         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3790 }
3791
3792 static inline int rt_scope(int ifa_scope)
3793 {
3794         if (ifa_scope & IFA_HOST)
3795                 return RT_SCOPE_HOST;
3796         else if (ifa_scope & IFA_LINK)
3797                 return RT_SCOPE_LINK;
3798         else if (ifa_scope & IFA_SITE)
3799                 return RT_SCOPE_SITE;
3800         else
3801                 return RT_SCOPE_UNIVERSE;
3802 }
3803
3804 static inline int inet6_ifaddr_msgsize(void)
3805 {
3806         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3807                + nla_total_size(16) /* IFA_ADDRESS */
3808                + nla_total_size(sizeof(struct ifa_cacheinfo));
3809 }
3810
3811 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3812                              u32 portid, u32 seq, int event, unsigned int flags)
3813 {
3814         struct nlmsghdr  *nlh;
3815         u32 preferred, valid;
3816
3817         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3818         if (nlh == NULL)
3819                 return -EMSGSIZE;
3820
3821         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3822                       ifa->idev->dev->ifindex);
3823
3824         if (!(ifa->flags&IFA_F_PERMANENT)) {
3825                 preferred = ifa->prefered_lft;
3826                 valid = ifa->valid_lft;
3827                 if (preferred != INFINITY_LIFE_TIME) {
3828                         long tval = (jiffies - ifa->tstamp)/HZ;
3829                         if (preferred > tval)
3830                                 preferred -= tval;
3831                         else
3832                                 preferred = 0;
3833                         if (valid != INFINITY_LIFE_TIME) {
3834                                 if (valid > tval)
3835                                         valid -= tval;
3836                                 else
3837                                         valid = 0;
3838                         }
3839                 }
3840         } else {
3841                 preferred = INFINITY_LIFE_TIME;
3842                 valid = INFINITY_LIFE_TIME;
3843         }
3844
3845         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3846             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3847                 nlmsg_cancel(skb, nlh);
3848                 return -EMSGSIZE;
3849         }
3850
3851         return nlmsg_end(skb, nlh);
3852 }
3853
3854 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3855                                 u32 portid, u32 seq, int event, u16 flags)
3856 {
3857         struct nlmsghdr  *nlh;
3858         u8 scope = RT_SCOPE_UNIVERSE;
3859         int ifindex = ifmca->idev->dev->ifindex;
3860
3861         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3862                 scope = RT_SCOPE_SITE;
3863
3864         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3865         if (nlh == NULL)
3866                 return -EMSGSIZE;
3867
3868         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3869         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3870             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3871                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3872                 nlmsg_cancel(skb, nlh);
3873                 return -EMSGSIZE;
3874         }
3875
3876         return nlmsg_end(skb, nlh);
3877 }
3878
3879 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3880                                 u32 portid, u32 seq, int event, unsigned int flags)
3881 {
3882         struct nlmsghdr  *nlh;
3883         u8 scope = RT_SCOPE_UNIVERSE;
3884         int ifindex = ifaca->aca_idev->dev->ifindex;
3885
3886         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3887                 scope = RT_SCOPE_SITE;
3888
3889         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3890         if (nlh == NULL)
3891                 return -EMSGSIZE;
3892
3893         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3894         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3895             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3896                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3897                 nlmsg_cancel(skb, nlh);
3898                 return -EMSGSIZE;
3899         }
3900
3901         return nlmsg_end(skb, nlh);
3902 }
3903
3904 enum addr_type_t {
3905         UNICAST_ADDR,
3906         MULTICAST_ADDR,
3907         ANYCAST_ADDR,
3908 };
3909
3910 /* called with rcu_read_lock() */
3911 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3912                           struct netlink_callback *cb, enum addr_type_t type,
3913                           int s_ip_idx, int *p_ip_idx)
3914 {
3915         struct ifmcaddr6 *ifmca;
3916         struct ifacaddr6 *ifaca;
3917         int err = 1;
3918         int ip_idx = *p_ip_idx;
3919
3920         read_lock_bh(&idev->lock);
3921         switch (type) {
3922         case UNICAST_ADDR: {
3923                 struct inet6_ifaddr *ifa;
3924
3925                 /* unicast address incl. temp addr */
3926                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3927                         if (++ip_idx < s_ip_idx)
3928                                 continue;
3929                         err = inet6_fill_ifaddr(skb, ifa,
3930                                                 NETLINK_CB(cb->skb).portid,
3931                                                 cb->nlh->nlmsg_seq,
3932                                                 RTM_NEWADDR,
3933                                                 NLM_F_MULTI);
3934                         if (err <= 0)
3935                                 break;
3936                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3937                 }
3938                 break;
3939         }
3940         case MULTICAST_ADDR:
3941                 /* multicast address */
3942                 for (ifmca = idev->mc_list; ifmca;
3943                      ifmca = ifmca->next, ip_idx++) {
3944                         if (ip_idx < s_ip_idx)
3945                                 continue;
3946                         err = inet6_fill_ifmcaddr(skb, ifmca,
3947                                                   NETLINK_CB(cb->skb).portid,
3948                                                   cb->nlh->nlmsg_seq,
3949                                                   RTM_GETMULTICAST,
3950                                                   NLM_F_MULTI);
3951                         if (err <= 0)
3952                                 break;
3953                 }
3954                 break;
3955         case ANYCAST_ADDR:
3956                 /* anycast address */
3957                 for (ifaca = idev->ac_list; ifaca;
3958                      ifaca = ifaca->aca_next, ip_idx++) {
3959                         if (ip_idx < s_ip_idx)
3960                                 continue;
3961                         err = inet6_fill_ifacaddr(skb, ifaca,
3962                                                   NETLINK_CB(cb->skb).portid,
3963                                                   cb->nlh->nlmsg_seq,
3964                                                   RTM_GETANYCAST,
3965                                                   NLM_F_MULTI);
3966                         if (err <= 0)
3967                                 break;
3968                 }
3969                 break;
3970         default:
3971                 break;
3972         }
3973         read_unlock_bh(&idev->lock);
3974         *p_ip_idx = ip_idx;
3975         return err;
3976 }
3977
3978 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3979                            enum addr_type_t type)
3980 {
3981         struct net *net = sock_net(skb->sk);
3982         int h, s_h;
3983         int idx, ip_idx;
3984         int s_idx, s_ip_idx;
3985         struct net_device *dev;
3986         struct inet6_dev *idev;
3987         struct hlist_head *head;
3988
3989         s_h = cb->args[0];
3990         s_idx = idx = cb->args[1];
3991         s_ip_idx = ip_idx = cb->args[2];
3992
3993         rcu_read_lock();
3994         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3995         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3996                 idx = 0;
3997                 head = &net->dev_index_head[h];
3998                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3999                         if (idx < s_idx)
4000                                 goto cont;
4001                         if (h > s_h || idx > s_idx)
4002                                 s_ip_idx = 0;
4003                         ip_idx = 0;
4004                         idev = __in6_dev_get(dev);
4005                         if (!idev)
4006                                 goto cont;
4007
4008                         if (in6_dump_addrs(idev, skb, cb, type,
4009                                            s_ip_idx, &ip_idx) <= 0)
4010                                 goto done;
4011 cont:
4012                         idx++;
4013                 }
4014         }
4015 done:
4016         rcu_read_unlock();
4017         cb->args[0] = h;
4018         cb->args[1] = idx;
4019         cb->args[2] = ip_idx;
4020
4021         return skb->len;
4022 }
4023
4024 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4025 {
4026         enum addr_type_t type = UNICAST_ADDR;
4027
4028         return inet6_dump_addr(skb, cb, type);
4029 }
4030
4031 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4032 {
4033         enum addr_type_t type = MULTICAST_ADDR;
4034
4035         return inet6_dump_addr(skb, cb, type);
4036 }
4037
4038
4039 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4040 {
4041         enum addr_type_t type = ANYCAST_ADDR;
4042
4043         return inet6_dump_addr(skb, cb, type);
4044 }
4045
4046 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4047 {
4048         struct net *net = sock_net(in_skb->sk);
4049         struct ifaddrmsg *ifm;
4050         struct nlattr *tb[IFA_MAX+1];
4051         struct in6_addr *addr = NULL;
4052         struct net_device *dev = NULL;
4053         struct inet6_ifaddr *ifa;
4054         struct sk_buff *skb;
4055         int err;
4056
4057         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4058         if (err < 0)
4059                 goto errout;
4060
4061         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4062         if (addr == NULL) {
4063                 err = -EINVAL;
4064                 goto errout;
4065         }
4066
4067         ifm = nlmsg_data(nlh);
4068         if (ifm->ifa_index)
4069                 dev = __dev_get_by_index(net, ifm->ifa_index);
4070
4071         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4072         if (!ifa) {
4073                 err = -EADDRNOTAVAIL;
4074                 goto errout;
4075         }
4076
4077         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4078         if (!skb) {
4079                 err = -ENOBUFS;
4080                 goto errout_ifa;
4081         }
4082
4083         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4084                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4085         if (err < 0) {
4086                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4087                 WARN_ON(err == -EMSGSIZE);
4088                 kfree_skb(skb);
4089                 goto errout_ifa;
4090         }
4091         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4092 errout_ifa:
4093         in6_ifa_put(ifa);
4094 errout:
4095         return err;
4096 }
4097
4098 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4099 {
4100         struct sk_buff *skb;
4101         struct net *net = dev_net(ifa->idev->dev);
4102         int err = -ENOBUFS;
4103
4104         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4105         if (skb == NULL)
4106                 goto errout;
4107
4108         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4109         if (err < 0) {
4110                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4111                 WARN_ON(err == -EMSGSIZE);
4112                 kfree_skb(skb);
4113                 goto errout;
4114         }
4115         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4116         return;
4117 errout:
4118         if (err < 0)
4119                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4120 }
4121
4122 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4123                                 __s32 *array, int bytes)
4124 {
4125         BUG_ON(bytes < (DEVCONF_MAX * 4));
4126
4127         memset(array, 0, bytes);
4128         array[DEVCONF_FORWARDING] = cnf->forwarding;
4129         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4130         array[DEVCONF_MTU6] = cnf->mtu6;
4131         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4132         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4133         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4134         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4135         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4136         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4137                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4138         array[DEVCONF_RTR_SOLICIT_DELAY] =
4139                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4140         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4141 #ifdef CONFIG_IPV6_PRIVACY
4142         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4143         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4144         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4145         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4146         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4147 #endif
4148         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4149         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4150         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4151 #ifdef CONFIG_IPV6_ROUTER_PREF
4152         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4153         array[DEVCONF_RTR_PROBE_INTERVAL] =
4154                 jiffies_to_msecs(cnf->rtr_probe_interval);
4155 #ifdef CONFIG_IPV6_ROUTE_INFO
4156         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4157 #endif
4158 #endif
4159         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4160         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4161 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4162         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4163 #endif
4164 #ifdef CONFIG_IPV6_MROUTE
4165         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4166 #endif
4167         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4168         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4169         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4170         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4171 }
4172
4173 static inline size_t inet6_ifla6_size(void)
4174 {
4175         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4176              + nla_total_size(sizeof(struct ifla_cacheinfo))
4177              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4178              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4179              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4180              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4181 }
4182
4183 static inline size_t inet6_if_nlmsg_size(void)
4184 {
4185         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4186                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4187                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4188                + nla_total_size(4) /* IFLA_MTU */
4189                + nla_total_size(4) /* IFLA_LINK */
4190                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4191 }
4192
4193 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4194                                       int items, int bytes)
4195 {
4196         int i;
4197         int pad = bytes - sizeof(u64) * items;
4198         BUG_ON(pad < 0);
4199
4200         /* Use put_unaligned() because stats may not be aligned for u64. */
4201         put_unaligned(items, &stats[0]);
4202         for (i = 1; i < items; i++)
4203                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4204
4205         memset(&stats[items], 0, pad);
4206 }
4207
4208 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4209                                       int items, int bytes, size_t syncpoff)
4210 {
4211         int i;
4212         int pad = bytes - sizeof(u64) * items;
4213         BUG_ON(pad < 0);
4214
4215         /* Use put_unaligned() because stats may not be aligned for u64. */
4216         put_unaligned(items, &stats[0]);
4217         for (i = 1; i < items; i++)
4218                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4219
4220         memset(&stats[items], 0, pad);
4221 }
4222
4223 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4224                              int bytes)
4225 {
4226         switch (attrtype) {
4227         case IFLA_INET6_STATS:
4228                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4229                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4230                 break;
4231         case IFLA_INET6_ICMP6STATS:
4232                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4233                 break;
4234         }
4235 }
4236
4237 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4238 {
4239         struct nlattr *nla;
4240         struct ifla_cacheinfo ci;
4241
4242         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4243                 goto nla_put_failure;
4244         ci.max_reasm_len = IPV6_MAXPLEN;
4245         ci.tstamp = cstamp_delta(idev->tstamp);
4246         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4247         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4248         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4249                 goto nla_put_failure;
4250         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4251         if (nla == NULL)
4252                 goto nla_put_failure;
4253         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4254
4255         /* XXX - MC not implemented */
4256
4257         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4258         if (nla == NULL)
4259                 goto nla_put_failure;
4260         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4261
4262         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4263         if (nla == NULL)
4264                 goto nla_put_failure;
4265         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4266
4267         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4268         if (nla == NULL)
4269                 goto nla_put_failure;
4270         read_lock_bh(&idev->lock);
4271         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4272         read_unlock_bh(&idev->lock);
4273
4274         return 0;
4275
4276 nla_put_failure:
4277         return -EMSGSIZE;
4278 }
4279
4280 static size_t inet6_get_link_af_size(const struct net_device *dev)
4281 {
4282         if (!__in6_dev_get(dev))
4283                 return 0;
4284
4285         return inet6_ifla6_size();
4286 }
4287
4288 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4289 {
4290         struct inet6_dev *idev = __in6_dev_get(dev);
4291
4292         if (!idev)
4293                 return -ENODATA;
4294
4295         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4296                 return -EMSGSIZE;
4297
4298         return 0;
4299 }
4300
4301 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4302 {
4303         struct inet6_ifaddr *ifp;
4304         struct net_device *dev = idev->dev;
4305         bool update_rs = false;
4306
4307         if (token == NULL)
4308                 return -EINVAL;
4309         if (ipv6_addr_any(token))
4310                 return -EINVAL;
4311         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4312                 return -EINVAL;
4313         if (!ipv6_accept_ra(idev))
4314                 return -EINVAL;
4315         if (idev->cnf.rtr_solicits <= 0)
4316                 return -EINVAL;
4317
4318         write_lock_bh(&idev->lock);
4319
4320         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4321         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4322
4323         write_unlock_bh(&idev->lock);
4324
4325         if (!idev->dead && (idev->if_flags & IF_READY)) {
4326                 struct in6_addr ll_addr;
4327
4328                 ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4329                                 IFA_F_OPTIMISTIC);
4330
4331                 /* If we're not ready, then normal ifup will take care
4332                  * of this. Otherwise, we need to request our rs here.
4333                  */
4334                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4335                 update_rs = true;
4336         }
4337
4338         write_lock_bh(&idev->lock);
4339
4340         if (update_rs)
4341                 idev->if_flags |= IF_RS_SENT;
4342
4343         /* Well, that's kinda nasty ... */
4344         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4345                 spin_lock(&ifp->lock);
4346                 if (ifp->tokenized) {
4347                         ifp->valid_lft = 0;
4348                         ifp->prefered_lft = 0;
4349                 }
4350                 spin_unlock(&ifp->lock);
4351         }
4352
4353         write_unlock_bh(&idev->lock);
4354         return 0;
4355 }
4356
4357 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4358 {
4359         int err = -EINVAL;
4360         struct inet6_dev *idev = __in6_dev_get(dev);
4361         struct nlattr *tb[IFLA_INET6_MAX + 1];
4362
4363         if (!idev)
4364                 return -EAFNOSUPPORT;
4365
4366         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4367                 BUG();
4368
4369         if (tb[IFLA_INET6_TOKEN])
4370                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4371
4372         return err;
4373 }
4374
4375 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4376                              u32 portid, u32 seq, int event, unsigned int flags)
4377 {
4378         struct net_device *dev = idev->dev;
4379         struct ifinfomsg *hdr;
4380         struct nlmsghdr *nlh;
4381         void *protoinfo;
4382
4383         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4384         if (nlh == NULL)
4385                 return -EMSGSIZE;
4386
4387         hdr = nlmsg_data(nlh);
4388         hdr->ifi_family = AF_INET6;
4389         hdr->__ifi_pad = 0;
4390         hdr->ifi_type = dev->type;
4391         hdr->ifi_index = dev->ifindex;
4392         hdr->ifi_flags = dev_get_flags(dev);
4393         hdr->ifi_change = 0;
4394
4395         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4396             (dev->addr_len &&
4397              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4398             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4399             (dev->ifindex != dev->iflink &&
4400              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4401                 goto nla_put_failure;
4402         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4403         if (protoinfo == NULL)
4404                 goto nla_put_failure;
4405
4406         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4407                 goto nla_put_failure;
4408
4409         nla_nest_end(skb, protoinfo);
4410         return nlmsg_end(skb, nlh);
4411
4412 nla_put_failure:
4413         nlmsg_cancel(skb, nlh);
4414         return -EMSGSIZE;
4415 }
4416
4417 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4418 {
4419         struct net *net = sock_net(skb->sk);
4420         int h, s_h;
4421         int idx = 0, s_idx;
4422         struct net_device *dev;
4423         struct inet6_dev *idev;
4424         struct hlist_head *head;
4425
4426         s_h = cb->args[0];
4427         s_idx = cb->args[1];
4428
4429         rcu_read_lock();
4430         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4431                 idx = 0;
4432                 head = &net->dev_index_head[h];
4433                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4434                         if (idx < s_idx)
4435                                 goto cont;
4436                         idev = __in6_dev_get(dev);
4437                         if (!idev)
4438                                 goto cont;
4439                         if (inet6_fill_ifinfo(skb, idev,
4440                                               NETLINK_CB(cb->skb).portid,
4441                                               cb->nlh->nlmsg_seq,
4442                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4443                                 goto out;
4444 cont:
4445                         idx++;
4446                 }
4447         }
4448 out:
4449         rcu_read_unlock();
4450         cb->args[1] = idx;
4451         cb->args[0] = h;
4452
4453         return skb->len;
4454 }
4455
4456 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4457 {
4458         struct sk_buff *skb;
4459         struct net *net = dev_net(idev->dev);
4460         int err = -ENOBUFS;
4461
4462         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4463         if (skb == NULL)
4464                 goto errout;
4465
4466         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4467         if (err < 0) {
4468                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4469                 WARN_ON(err == -EMSGSIZE);
4470                 kfree_skb(skb);
4471                 goto errout;
4472         }
4473         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4474         return;
4475 errout:
4476         if (err < 0)
4477                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4478 }
4479
4480 static inline size_t inet6_prefix_nlmsg_size(void)
4481 {
4482         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4483                + nla_total_size(sizeof(struct in6_addr))
4484                + nla_total_size(sizeof(struct prefix_cacheinfo));
4485 }
4486
4487 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4488                              struct prefix_info *pinfo, u32 portid, u32 seq,
4489                              int event, unsigned int flags)
4490 {
4491         struct prefixmsg *pmsg;
4492         struct nlmsghdr *nlh;
4493         struct prefix_cacheinfo ci;
4494
4495         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4496         if (nlh == NULL)
4497                 return -EMSGSIZE;
4498
4499         pmsg = nlmsg_data(nlh);
4500         pmsg->prefix_family = AF_INET6;
4501         pmsg->prefix_pad1 = 0;
4502         pmsg->prefix_pad2 = 0;
4503         pmsg->prefix_ifindex = idev->dev->ifindex;
4504         pmsg->prefix_len = pinfo->prefix_len;
4505         pmsg->prefix_type = pinfo->type;
4506         pmsg->prefix_pad3 = 0;
4507         pmsg->prefix_flags = 0;
4508         if (pinfo->onlink)
4509                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4510         if (pinfo->autoconf)
4511                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4512
4513         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4514                 goto nla_put_failure;
4515         ci.preferred_time = ntohl(pinfo->prefered);
4516         ci.valid_time = ntohl(pinfo->valid);
4517         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4518                 goto nla_put_failure;
4519         return nlmsg_end(skb, nlh);
4520
4521 nla_put_failure:
4522         nlmsg_cancel(skb, nlh);
4523         return -EMSGSIZE;
4524 }
4525
4526 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4527                          struct prefix_info *pinfo)
4528 {
4529         struct sk_buff *skb;
4530         struct net *net = dev_net(idev->dev);
4531         int err = -ENOBUFS;
4532
4533         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4534         if (skb == NULL)
4535                 goto errout;
4536
4537         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4538         if (err < 0) {
4539                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4540                 WARN_ON(err == -EMSGSIZE);
4541                 kfree_skb(skb);
4542                 goto errout;
4543         }
4544         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4545         return;
4546 errout:
4547         if (err < 0)
4548                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4549 }
4550
4551 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4552 {
4553         struct net *net = dev_net(ifp->idev->dev);
4554
4555         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4556
4557         switch (event) {
4558         case RTM_NEWADDR:
4559                 /*
4560                  * If the address was optimistic
4561                  * we inserted the route at the start of
4562                  * our DAD process, so we don't need
4563                  * to do it again
4564                  */
4565                 if (!(ifp->rt->rt6i_node))
4566                         ip6_ins_rt(ifp->rt);
4567                 if (ifp->idev->cnf.forwarding)
4568                         addrconf_join_anycast(ifp);
4569                 break;
4570         case RTM_DELADDR:
4571                 if (ifp->idev->cnf.forwarding)
4572                         addrconf_leave_anycast(ifp);
4573                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4574                 dst_hold(&ifp->rt->dst);
4575
4576                 if (ip6_del_rt(ifp->rt))
4577                         dst_free(&ifp->rt->dst);
4578                 break;
4579         }
4580         atomic_inc(&net->ipv6.dev_addr_genid);
4581 }
4582
4583 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4584 {
4585         rcu_read_lock_bh();
4586         if (likely(ifp->idev->dead == 0))
4587                 __ipv6_ifa_notify(event, ifp);
4588         rcu_read_unlock_bh();
4589 }
4590
4591 #ifdef CONFIG_SYSCTL
4592
4593 static
4594 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4595                            void __user *buffer, size_t *lenp, loff_t *ppos)
4596 {
4597         int *valp = ctl->data;
4598         int val = *valp;
4599         loff_t pos = *ppos;
4600         ctl_table lctl;
4601         int ret;
4602
4603         /*
4604          * ctl->data points to idev->cnf.forwarding, we should
4605          * not modify it until we get the rtnl lock.
4606          */
4607         lctl = *ctl;
4608         lctl.data = &val;
4609
4610         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4611
4612         if (write)
4613                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4614         if (ret)
4615                 *ppos = pos;
4616         return ret;
4617 }
4618
4619 static void dev_disable_change(struct inet6_dev *idev)
4620 {
4621         if (!idev || !idev->dev)
4622                 return;
4623
4624         if (idev->cnf.disable_ipv6)
4625                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4626         else
4627                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4628 }
4629
4630 static void addrconf_disable_change(struct net *net, __s32 newf)
4631 {
4632         struct net_device *dev;
4633         struct inet6_dev *idev;
4634
4635         rcu_read_lock();
4636         for_each_netdev_rcu(net, dev) {
4637                 idev = __in6_dev_get(dev);
4638                 if (idev) {
4639                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4640                         idev->cnf.disable_ipv6 = newf;
4641                         if (changed)
4642                                 dev_disable_change(idev);
4643                 }
4644         }
4645         rcu_read_unlock();
4646 }
4647
4648 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4649 {
4650         struct net *net;
4651         int old;
4652
4653         if (!rtnl_trylock())
4654                 return restart_syscall();
4655
4656         net = (struct net *)table->extra2;
4657         old = *p;
4658         *p = newf;
4659
4660         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4661                 rtnl_unlock();
4662                 return 0;
4663         }
4664
4665         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4666                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4667                 addrconf_disable_change(net, newf);
4668         } else if ((!newf) ^ (!old))
4669                 dev_disable_change((struct inet6_dev *)table->extra1);
4670
4671         rtnl_unlock();
4672         return 0;
4673 }
4674
4675 static
4676 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4677                             void __user *buffer, size_t *lenp, loff_t *ppos)
4678 {
4679         int *valp = ctl->data;
4680         int val = *valp;
4681         loff_t pos = *ppos;
4682         ctl_table lctl;
4683         int ret;
4684
4685         /*
4686          * ctl->data points to idev->cnf.disable_ipv6, we should
4687          * not modify it until we get the rtnl lock.
4688          */
4689         lctl = *ctl;
4690         lctl.data = &val;
4691
4692         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4693
4694         if (write)
4695                 ret = addrconf_disable_ipv6(ctl, valp, val);
4696         if (ret)
4697                 *ppos = pos;
4698         return ret;
4699 }
4700
4701 static struct addrconf_sysctl_table
4702 {
4703         struct ctl_table_header *sysctl_header;
4704         ctl_table addrconf_vars[DEVCONF_MAX+1];
4705 } addrconf_sysctl __read_mostly = {
4706         .sysctl_header = NULL,
4707         .addrconf_vars = {
4708                 {
4709                         .procname       = "forwarding",
4710                         .data           = &ipv6_devconf.forwarding,
4711                         .maxlen         = sizeof(int),
4712                         .mode           = 0644,
4713                         .proc_handler   = addrconf_sysctl_forward,
4714                 },
4715                 {
4716                         .procname       = "hop_limit",
4717                         .data           = &ipv6_devconf.hop_limit,
4718                         .maxlen         = sizeof(int),
4719                         .mode           = 0644,
4720                         .proc_handler   = proc_dointvec,
4721                 },
4722                 {
4723                         .procname       = "mtu",
4724                         .data           = &ipv6_devconf.mtu6,
4725                         .maxlen         = sizeof(int),
4726                         .mode           = 0644,
4727                         .proc_handler   = proc_dointvec,
4728                 },
4729                 {
4730                         .procname       = "accept_ra",
4731                         .data           = &ipv6_devconf.accept_ra,
4732                         .maxlen         = sizeof(int),
4733                         .mode           = 0644,
4734                         .proc_handler   = proc_dointvec,
4735                 },
4736                 {
4737                         .procname       = "accept_redirects",
4738                         .data           = &ipv6_devconf.accept_redirects,
4739                         .maxlen         = sizeof(int),
4740                         .mode           = 0644,
4741                         .proc_handler   = proc_dointvec,
4742                 },
4743                 {
4744                         .procname       = "autoconf",
4745                         .data           = &ipv6_devconf.autoconf,
4746                         .maxlen         = sizeof(int),
4747                         .mode           = 0644,
4748                         .proc_handler   = proc_dointvec,
4749                 },
4750                 {
4751                         .procname       = "dad_transmits",
4752                         .data           = &ipv6_devconf.dad_transmits,
4753                         .maxlen         = sizeof(int),
4754                         .mode           = 0644,
4755                         .proc_handler   = proc_dointvec,
4756                 },
4757                 {
4758                         .procname       = "router_solicitations",
4759                         .data           = &ipv6_devconf.rtr_solicits,
4760                         .maxlen         = sizeof(int),
4761                         .mode           = 0644,
4762                         .proc_handler   = proc_dointvec,
4763                 },
4764                 {
4765                         .procname       = "router_solicitation_interval",
4766                         .data           = &ipv6_devconf.rtr_solicit_interval,
4767                         .maxlen         = sizeof(int),
4768                         .mode           = 0644,
4769                         .proc_handler   = proc_dointvec_jiffies,
4770                 },
4771                 {
4772                         .procname       = "router_solicitation_delay",
4773                         .data           = &ipv6_devconf.rtr_solicit_delay,
4774                         .maxlen         = sizeof(int),
4775                         .mode           = 0644,
4776                         .proc_handler   = proc_dointvec_jiffies,
4777                 },
4778                 {
4779                         .procname       = "force_mld_version",
4780                         .data           = &ipv6_devconf.force_mld_version,
4781                         .maxlen         = sizeof(int),
4782                         .mode           = 0644,
4783                         .proc_handler   = proc_dointvec,
4784                 },
4785 #ifdef CONFIG_IPV6_PRIVACY
4786                 {
4787                         .procname       = "use_tempaddr",
4788                         .data           = &ipv6_devconf.use_tempaddr,
4789                         .maxlen         = sizeof(int),
4790                         .mode           = 0644,
4791                         .proc_handler   = proc_dointvec,
4792                 },
4793                 {
4794                         .procname       = "temp_valid_lft",
4795                         .data           = &ipv6_devconf.temp_valid_lft,
4796                         .maxlen         = sizeof(int),
4797                         .mode           = 0644,
4798                         .proc_handler   = proc_dointvec,
4799                 },
4800                 {
4801                         .procname       = "temp_prefered_lft",
4802                         .data           = &ipv6_devconf.temp_prefered_lft,
4803                         .maxlen         = sizeof(int),
4804                         .mode           = 0644,
4805                         .proc_handler   = proc_dointvec,
4806                 },
4807                 {
4808                         .procname       = "regen_max_retry",
4809                         .data           = &ipv6_devconf.regen_max_retry,
4810                         .maxlen         = sizeof(int),
4811                         .mode           = 0644,
4812                         .proc_handler   = proc_dointvec,
4813                 },
4814                 {
4815                         .procname       = "max_desync_factor",
4816                         .data           = &ipv6_devconf.max_desync_factor,
4817                         .maxlen         = sizeof(int),
4818                         .mode           = 0644,
4819                         .proc_handler   = proc_dointvec,
4820                 },
4821 #endif
4822                 {
4823                         .procname       = "max_addresses",
4824                         .data           = &ipv6_devconf.max_addresses,
4825                         .maxlen         = sizeof(int),
4826                         .mode           = 0644,
4827                         .proc_handler   = proc_dointvec,
4828                 },
4829                 {
4830                         .procname       = "accept_ra_defrtr",
4831                         .data           = &ipv6_devconf.accept_ra_defrtr,
4832                         .maxlen         = sizeof(int),
4833                         .mode           = 0644,
4834                         .proc_handler   = proc_dointvec,
4835                 },
4836                 {
4837                         .procname       = "accept_ra_pinfo",
4838                         .data           = &ipv6_devconf.accept_ra_pinfo,
4839                         .maxlen         = sizeof(int),
4840                         .mode           = 0644,
4841                         .proc_handler   = proc_dointvec,
4842                 },
4843 #ifdef CONFIG_IPV6_ROUTER_PREF
4844                 {
4845                         .procname       = "accept_ra_rtr_pref",
4846                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4847                         .maxlen         = sizeof(int),
4848                         .mode           = 0644,
4849                         .proc_handler   = proc_dointvec,
4850                 },
4851                 {
4852                         .procname       = "router_probe_interval",
4853                         .data           = &ipv6_devconf.rtr_probe_interval,
4854                         .maxlen         = sizeof(int),
4855                         .mode           = 0644,
4856                         .proc_handler   = proc_dointvec_jiffies,
4857                 },
4858 #ifdef CONFIG_IPV6_ROUTE_INFO
4859                 {
4860                         .procname       = "accept_ra_rt_info_max_plen",
4861                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4862                         .maxlen         = sizeof(int),
4863                         .mode           = 0644,
4864                         .proc_handler   = proc_dointvec,
4865                 },
4866 #endif
4867 #endif
4868                 {
4869                         .procname       = "proxy_ndp",
4870                         .data           = &ipv6_devconf.proxy_ndp,
4871                         .maxlen         = sizeof(int),
4872                         .mode           = 0644,
4873                         .proc_handler   = proc_dointvec,
4874                 },
4875                 {
4876                         .procname       = "accept_source_route",
4877                         .data           = &ipv6_devconf.accept_source_route,
4878                         .maxlen         = sizeof(int),
4879                         .mode           = 0644,
4880                         .proc_handler   = proc_dointvec,
4881                 },
4882 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4883                 {
4884                         .procname       = "optimistic_dad",
4885                         .data           = &ipv6_devconf.optimistic_dad,
4886                         .maxlen         = sizeof(int),
4887                         .mode           = 0644,
4888                         .proc_handler   = proc_dointvec,
4889
4890                 },
4891 #endif
4892 #ifdef CONFIG_IPV6_MROUTE
4893                 {
4894                         .procname       = "mc_forwarding",
4895                         .data           = &ipv6_devconf.mc_forwarding,
4896                         .maxlen         = sizeof(int),
4897                         .mode           = 0444,
4898                         .proc_handler   = proc_dointvec,
4899                 },
4900 #endif
4901                 {
4902                         .procname       = "disable_ipv6",
4903                         .data           = &ipv6_devconf.disable_ipv6,
4904                         .maxlen         = sizeof(int),
4905                         .mode           = 0644,
4906                         .proc_handler   = addrconf_sysctl_disable,
4907                 },
4908                 {
4909                         .procname       = "accept_dad",
4910                         .data           = &ipv6_devconf.accept_dad,
4911                         .maxlen         = sizeof(int),
4912                         .mode           = 0644,
4913                         .proc_handler   = proc_dointvec,
4914                 },
4915                 {
4916                         .procname       = "force_tllao",
4917                         .data           = &ipv6_devconf.force_tllao,
4918                         .maxlen         = sizeof(int),
4919                         .mode           = 0644,
4920                         .proc_handler   = proc_dointvec
4921                 },
4922                 {
4923                         .procname       = "ndisc_notify",
4924                         .data           = &ipv6_devconf.ndisc_notify,
4925                         .maxlen         = sizeof(int),
4926                         .mode           = 0644,
4927                         .proc_handler   = proc_dointvec
4928                 },
4929                 {
4930                         /* sentinel */
4931                 }
4932         },
4933 };
4934
4935 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4936                 struct inet6_dev *idev, struct ipv6_devconf *p)
4937 {
4938         int i;
4939         struct addrconf_sysctl_table *t;
4940         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4941
4942         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4943         if (t == NULL)
4944                 goto out;
4945
4946         for (i = 0; t->addrconf_vars[i].data; i++) {
4947                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4948                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4949                 t->addrconf_vars[i].extra2 = net;
4950         }
4951
4952         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4953
4954         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4955         if (t->sysctl_header == NULL)
4956                 goto free;
4957
4958         p->sysctl = t;
4959         return 0;
4960
4961 free:
4962         kfree(t);
4963 out:
4964         return -ENOBUFS;
4965 }
4966
4967 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4968 {
4969         struct addrconf_sysctl_table *t;
4970
4971         if (p->sysctl == NULL)
4972                 return;
4973
4974         t = p->sysctl;
4975         p->sysctl = NULL;
4976         unregister_net_sysctl_table(t->sysctl_header);
4977         kfree(t);
4978 }
4979
4980 static void addrconf_sysctl_register(struct inet6_dev *idev)
4981 {
4982         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4983                               &ndisc_ifinfo_sysctl_change);
4984         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4985                                         idev, &idev->cnf);
4986 }
4987
4988 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4989 {
4990         __addrconf_sysctl_unregister(&idev->cnf);
4991         neigh_sysctl_unregister(idev->nd_parms);
4992 }
4993
4994
4995 #endif
4996
4997 static int __net_init addrconf_init_net(struct net *net)
4998 {
4999         int err = -ENOMEM;
5000         struct ipv6_devconf *all, *dflt;
5001
5002         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5003         if (all == NULL)
5004                 goto err_alloc_all;
5005
5006         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5007         if (dflt == NULL)
5008                 goto err_alloc_dflt;
5009
5010         /* these will be inherited by all namespaces */
5011         dflt->autoconf = ipv6_defaults.autoconf;
5012         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5013
5014         net->ipv6.devconf_all = all;
5015         net->ipv6.devconf_dflt = dflt;
5016
5017 #ifdef CONFIG_SYSCTL
5018         err = __addrconf_sysctl_register(net, "all", NULL, all);
5019         if (err < 0)
5020                 goto err_reg_all;
5021
5022         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5023         if (err < 0)
5024                 goto err_reg_dflt;
5025 #endif
5026         return 0;
5027
5028 #ifdef CONFIG_SYSCTL
5029 err_reg_dflt:
5030         __addrconf_sysctl_unregister(all);
5031 err_reg_all:
5032         kfree(dflt);
5033 #endif
5034 err_alloc_dflt:
5035         kfree(all);
5036 err_alloc_all:
5037         return err;
5038 }
5039
5040 static void __net_exit addrconf_exit_net(struct net *net)
5041 {
5042 #ifdef CONFIG_SYSCTL
5043         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5044         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5045 #endif
5046         if (!net_eq(net, &init_net)) {
5047                 kfree(net->ipv6.devconf_dflt);
5048                 kfree(net->ipv6.devconf_all);
5049         }
5050 }
5051
5052 static struct pernet_operations addrconf_ops = {
5053         .init = addrconf_init_net,
5054         .exit = addrconf_exit_net,
5055 };
5056
5057 /*
5058  *      Device notifier
5059  */
5060
5061 int register_inet6addr_notifier(struct notifier_block *nb)
5062 {
5063         return atomic_notifier_chain_register(&inet6addr_chain, nb);
5064 }
5065 EXPORT_SYMBOL(register_inet6addr_notifier);
5066
5067 int unregister_inet6addr_notifier(struct notifier_block *nb)
5068 {
5069         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
5070 }
5071 EXPORT_SYMBOL(unregister_inet6addr_notifier);
5072
5073 static struct rtnl_af_ops inet6_ops = {
5074         .family           = AF_INET6,
5075         .fill_link_af     = inet6_fill_link_af,
5076         .get_link_af_size = inet6_get_link_af_size,
5077         .set_link_af      = inet6_set_link_af,
5078 };
5079
5080 /*
5081  *      Init / cleanup code
5082  */
5083
5084 int __init addrconf_init(void)
5085 {
5086         int i, err;
5087
5088         err = ipv6_addr_label_init();
5089         if (err < 0) {
5090                 pr_crit("%s: cannot initialize default policy table: %d\n",
5091                         __func__, err);
5092                 goto out;
5093         }
5094
5095         err = register_pernet_subsys(&addrconf_ops);
5096         if (err < 0)
5097                 goto out_addrlabel;
5098
5099         /* The addrconf netdev notifier requires that loopback_dev
5100          * has it's ipv6 private information allocated and setup
5101          * before it can bring up and give link-local addresses
5102          * to other devices which are up.
5103          *
5104          * Unfortunately, loopback_dev is not necessarily the first
5105          * entry in the global dev_base list of net devices.  In fact,
5106          * it is likely to be the very last entry on that list.
5107          * So this causes the notifier registry below to try and
5108          * give link-local addresses to all devices besides loopback_dev
5109          * first, then loopback_dev, which cases all the non-loopback_dev
5110          * devices to fail to get a link-local address.
5111          *
5112          * So, as a temporary fix, allocate the ipv6 structure for
5113          * loopback_dev first by hand.
5114          * Longer term, all of the dependencies ipv6 has upon the loopback
5115          * device and it being up should be removed.
5116          */
5117         rtnl_lock();
5118         if (!ipv6_add_dev(init_net.loopback_dev))
5119                 err = -ENOMEM;
5120         rtnl_unlock();
5121         if (err)
5122                 goto errlo;
5123
5124         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5125                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5126
5127         register_netdevice_notifier(&ipv6_dev_notf);
5128
5129         addrconf_verify(0);
5130
5131         err = rtnl_af_register(&inet6_ops);
5132         if (err < 0)
5133                 goto errout_af;
5134
5135         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5136                               NULL);
5137         if (err < 0)
5138                 goto errout;
5139
5140         /* Only the first call to __rtnl_register can fail */
5141         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5142         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5143         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5144                         inet6_dump_ifaddr, NULL);
5145         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5146                         inet6_dump_ifmcaddr, NULL);
5147         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5148                         inet6_dump_ifacaddr, NULL);
5149         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5150                         inet6_netconf_dump_devconf, NULL);
5151
5152         ipv6_addr_label_rtnl_register();
5153
5154         return 0;
5155 errout:
5156         rtnl_af_unregister(&inet6_ops);
5157 errout_af:
5158         unregister_netdevice_notifier(&ipv6_dev_notf);
5159 errlo:
5160         unregister_pernet_subsys(&addrconf_ops);
5161 out_addrlabel:
5162         ipv6_addr_label_cleanup();
5163 out:
5164         return err;
5165 }
5166
5167 void addrconf_cleanup(void)
5168 {
5169         struct net_device *dev;
5170         int i;
5171
5172         unregister_netdevice_notifier(&ipv6_dev_notf);
5173         unregister_pernet_subsys(&addrconf_ops);
5174         ipv6_addr_label_cleanup();
5175
5176         rtnl_lock();
5177
5178         __rtnl_af_unregister(&inet6_ops);
5179
5180         /* clean dev list */
5181         for_each_netdev(&init_net, dev) {
5182                 if (__in6_dev_get(dev) == NULL)
5183                         continue;
5184                 addrconf_ifdown(dev, 1);
5185         }
5186         addrconf_ifdown(init_net.loopback_dev, 2);
5187
5188         /*
5189          *      Check hash table.
5190          */
5191         spin_lock_bh(&addrconf_hash_lock);
5192         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5193                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5194         spin_unlock_bh(&addrconf_hash_lock);
5195
5196         del_timer(&addr_chk_timer);
5197         rtnl_unlock();
5198 }