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 #include <linux/random.h>
  87 #include <linux/uaccess.h>
  88 #include <asm/unaligned.h>
  89
  90 #include <linux/proc_fs.h>
  91 #include <linux/seq_file.h>
  92 #include <linux/export.h>
  93
  94 /* Set to 3 to get tracing... */
  95 #define ACONF_DEBUG 2
  96
  97 #if ACONF_DEBUG >= 3
  98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
  99 #else
 100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 101 #endif
 102
 103 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 104
 105 static inline u32 cstamp_delta(unsigned long cstamp)
 106 {
 107         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 108 }
 109
 110 #ifdef CONFIG_SYSCTL
 111 static void addrconf_sysctl_register(struct inet6_dev *idev);
 112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 113 #else
 114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 115 {
 116 }
 117
 118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 119 {
 120 }
 121 #endif
 122
 123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 125 static void ipv6_regen_rndid(unsigned long data);
 126
 127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 128 static int ipv6_count_addresses(struct inet6_dev *idev);
 129
 130 /*
 131  *      Configured unicast address hash table
 132  */
 133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 134 static DEFINE_SPINLOCK(addrconf_hash_lock);
 135
 136 static void addrconf_verify(unsigned long);
 137
 138 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 139 static DEFINE_SPINLOCK(addrconf_verify_lock);
 140
 141 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 142 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 143
 144 static void addrconf_type_change(struct net_device *dev,
 145                                  unsigned long event);
 146 static int addrconf_ifdown(struct net_device *dev, int how);
 147
 148 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 149                                                   int plen,
 150                                                   const struct net_device *dev,
 151                                                   u32 flags, u32 noflags);
 152
 153 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_timer(unsigned long data);
 155 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 156 static void addrconf_dad_run(struct inet6_dev *idev);
 157 static void addrconf_rs_timer(unsigned long data);
 158 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 160
 161 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 162                                 struct prefix_info *pinfo);
 163 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 164                                struct net_device *dev);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .mldv1_unsolicited_report_interval = 10 * HZ,
 175         .mldv2_unsolicited_report_interval = HZ,
 176         .dad_transmits          = 1,
 177         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 178         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 179         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199         .suppress_frag_ndisc    = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .force_mld_version      = 0,
 210         .mldv1_unsolicited_report_interval = 10 * HZ,
 211         .mldv2_unsolicited_report_interval = HZ,
 212         .dad_transmits          = 1,
 213         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 214         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 215         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 216         .use_tempaddr           = 0,
 217         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 218         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 219         .regen_max_retry        = REGEN_MAX_RETRY,
 220         .max_desync_factor      = MAX_DESYNC_FACTOR,
 221         .max_addresses          = IPV6_MAX_ADDRESSES,
 222         .accept_ra_defrtr       = 1,
 223         .accept_ra_pinfo        = 1,
 224 #ifdef CONFIG_IPV6_ROUTER_PREF
 225         .accept_ra_rtr_pref     = 1,
 226         .rtr_probe_interval     = 60 * HZ,
 227 #ifdef CONFIG_IPV6_ROUTE_INFO
 228         .accept_ra_rt_info_max_plen = 0,
 229 #endif
 230 #endif
 231         .proxy_ndp              = 0,
 232         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 233         .disable_ipv6           = 0,
 234         .accept_dad             = 1,
 235         .suppress_frag_ndisc    = 1,
 236 };
 237
 238 /* Check if a valid qdisc is available */
 239 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 240 {
 241         return !qdisc_tx_is_noop(dev);
 242 }
 243
 244 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 245 {
 246         if (del_timer(&idev->rs_timer))
 247                 __in6_dev_put(idev);
 248 }
 249
 250 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 251 {
 252         if (del_timer(&ifp->dad_timer))
 253                 __in6_ifa_put(ifp);
 254 }
 255
 256 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 257                                   unsigned long when)
 258 {
 259         if (!timer_pending(&idev->rs_timer))
 260                 in6_dev_hold(idev);
 261         mod_timer(&idev->rs_timer, jiffies + when);
 262 }
 263
 264 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 265                                    unsigned long when)
 266 {
 267         if (!timer_pending(&ifp->dad_timer))
 268                 in6_ifa_hold(ifp);
 269         mod_timer(&ifp->dad_timer, jiffies + when);
 270 }
 271
 272 static int snmp6_alloc_dev(struct inet6_dev *idev)
 273 {
 274         int i;
 275
 276         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 277                           sizeof(struct ipstats_mib),
 278                           __alignof__(struct ipstats_mib)) < 0)
 279                 goto err_ip;
 280
 281         for_each_possible_cpu(i) {
 282                 struct ipstats_mib *addrconf_stats;
 283                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i);
 284                 u64_stats_init(&addrconf_stats->syncp);
 285 #if SNMP_ARRAY_SZ == 2
 286                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i);
 287                 u64_stats_init(&addrconf_stats->syncp);
 288 #endif
 289         }
 290
 291
 292         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 293                                         GFP_KERNEL);
 294         if (!idev->stats.icmpv6dev)
 295                 goto err_icmp;
 296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 297                                            GFP_KERNEL);
 298         if (!idev->stats.icmpv6msgdev)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         kfree(idev->stats.icmpv6dev);
 305 err_icmp:
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 312 {
 313         struct inet6_dev *ndev;
 314
 315         ASSERT_RTNL();
 316
 317         if (dev->mtu < IPV6_MIN_MTU)
 318                 return NULL;
 319
 320         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 321
 322         if (ndev == NULL)
 323                 return NULL;
 324
 325         rwlock_init(&ndev->lock);
 326         ndev->dev = dev;
 327         INIT_LIST_HEAD(&ndev->addr_list);
 328         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 329                     (unsigned long)ndev);
 330         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 331         ndev->cnf.mtu6 = dev->mtu;
 332         ndev->cnf.sysctl = NULL;
 333         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 334         if (ndev->nd_parms == NULL) {
 335                 kfree(ndev);
 336                 return NULL;
 337         }
 338         if (ndev->cnf.forwarding)
 339                 dev_disable_lro(dev);
 340         /* We refer to the device */
 341         dev_hold(dev);
 342
 343         if (snmp6_alloc_dev(ndev) < 0) {
 344                 ADBG(KERN_WARNING
 345                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 346                         __func__, dev->name);
 347                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 348                 dev_put(dev);
 349                 kfree(ndev);
 350                 return NULL;
 351         }
 352
 353         if (snmp6_register_dev(ndev) < 0) {
 354                 ADBG(KERN_WARNING
 355                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 356                         __func__, dev->name);
 357                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 358                 ndev->dead = 1;
 359                 in6_dev_finish_destroy(ndev);
 360                 return NULL;
 361         }
 362
 363         /* One reference from device.  We must do this before
 364          * we invoke __ipv6_regen_rndid().
 365          */
 366         in6_dev_hold(ndev);
 367
 368         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 369                 ndev->cnf.accept_dad = -1;
 370
 371 #if IS_ENABLED(CONFIG_IPV6_SIT)
 372         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 373                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 374                 ndev->cnf.rtr_solicits = 0;
 375         }
 376 #endif
 377
 378         INIT_LIST_HEAD(&ndev->tempaddr_list);
 379         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 380         if ((dev->flags&IFF_LOOPBACK) ||
 381             dev->type == ARPHRD_TUNNEL ||
 382             dev->type == ARPHRD_TUNNEL6 ||
 383             dev->type == ARPHRD_SIT ||
 384             dev->type == ARPHRD_NONE) {
 385                 ndev->cnf.use_tempaddr = -1;
 386         } else {
 387                 in6_dev_hold(ndev);
 388                 ipv6_regen_rndid((unsigned long) ndev);
 389         }
 390
 391         ndev->token = in6addr_any;
 392
 393         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 394                 ndev->if_flags |= IF_READY;
 395
 396         ipv6_mc_init_dev(ndev);
 397         ndev->tstamp = jiffies;
 398         addrconf_sysctl_register(ndev);
 399         /* protected by rtnl_lock */
 400         rcu_assign_pointer(dev->ip6_ptr, ndev);
 401
 402         /* Join interface-local all-node multicast group */
 403         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 404
 405         /* Join all-node multicast group */
 406         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 407
 408         /* Join all-router multicast group if forwarding is set */
 409         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 410                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 411
 412         return ndev;
 413 }
 414
 415 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 416 {
 417         struct inet6_dev *idev;
 418
 419         ASSERT_RTNL();
 420
 421         idev = __in6_dev_get(dev);
 422         if (!idev) {
 423                 idev = ipv6_add_dev(dev);
 424                 if (!idev)
 425                         return NULL;
 426         }
 427
 428         if (dev->flags&IFF_UP)
 429                 ipv6_mc_up(idev);
 430         return idev;
 431 }
 432
 433 static int inet6_netconf_msgsize_devconf(int type)
 434 {
 435         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 436                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 437
 438         /* type -1 is used for ALL */
 439         if (type == -1 || type == NETCONFA_FORWARDING)
 440                 size += nla_total_size(4);
 441 #ifdef CONFIG_IPV6_MROUTE
 442         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 443                 size += nla_total_size(4);
 444 #endif
 445
 446         return size;
 447 }
 448
 449 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 450                                       struct ipv6_devconf *devconf, u32 portid,
 451                                       u32 seq, int event, unsigned int flags,
 452                                       int type)
 453 {
 454         struct nlmsghdr  *nlh;
 455         struct netconfmsg *ncm;
 456
 457         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 458                         flags);
 459         if (nlh == NULL)
 460                 return -EMSGSIZE;
 461
 462         ncm = nlmsg_data(nlh);
 463         ncm->ncm_family = AF_INET6;
 464
 465         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 466                 goto nla_put_failure;
 467
 468         /* type -1 is used for ALL */
 469         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 470             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 471                 goto nla_put_failure;
 472 #ifdef CONFIG_IPV6_MROUTE
 473         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 474             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 475                         devconf->mc_forwarding) < 0)
 476                 goto nla_put_failure;
 477 #endif
 478         return nlmsg_end(skb, nlh);
 479
 480 nla_put_failure:
 481         nlmsg_cancel(skb, nlh);
 482         return -EMSGSIZE;
 483 }
 484
 485 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 486                                   struct ipv6_devconf *devconf)
 487 {
 488         struct sk_buff *skb;
 489         int err = -ENOBUFS;
 490
 491         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 492         if (skb == NULL)
 493                 goto errout;
 494
 495         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 496                                          RTM_NEWNETCONF, 0, type);
 497         if (err < 0) {
 498                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 499                 WARN_ON(err == -EMSGSIZE);
 500                 kfree_skb(skb);
 501                 goto errout;
 502         }
 503         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 504         return;
 505 errout:
 506         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 507 }
 508
 509 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 510         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 511         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 512 };
 513
 514 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 515                                      struct nlmsghdr *nlh)
 516 {
 517         struct net *net = sock_net(in_skb->sk);
 518         struct nlattr *tb[NETCONFA_MAX+1];
 519         struct netconfmsg *ncm;
 520         struct sk_buff *skb;
 521         struct ipv6_devconf *devconf;
 522         struct inet6_dev *in6_dev;
 523         struct net_device *dev;
 524         int ifindex;
 525         int err;
 526
 527         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 528                           devconf_ipv6_policy);
 529         if (err < 0)
 530                 goto errout;
 531
 532         err = EINVAL;
 533         if (!tb[NETCONFA_IFINDEX])
 534                 goto errout;
 535
 536         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 537         switch (ifindex) {
 538         case NETCONFA_IFINDEX_ALL:
 539                 devconf = net->ipv6.devconf_all;
 540                 break;
 541         case NETCONFA_IFINDEX_DEFAULT:
 542                 devconf = net->ipv6.devconf_dflt;
 543                 break;
 544         default:
 545                 dev = __dev_get_by_index(net, ifindex);
 546                 if (dev == NULL)
 547                         goto errout;
 548                 in6_dev = __in6_dev_get(dev);
 549                 if (in6_dev == NULL)
 550                         goto errout;
 551                 devconf = &in6_dev->cnf;
 552                 break;
 553         }
 554
 555         err = -ENOBUFS;
 556         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 557         if (skb == NULL)
 558                 goto errout;
 559
 560         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 561                                          NETLINK_CB(in_skb).portid,
 562                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 563                                          -1);
 564         if (err < 0) {
 565                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 566                 WARN_ON(err == -EMSGSIZE);
 567                 kfree_skb(skb);
 568                 goto errout;
 569         }
 570         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 571 errout:
 572         return err;
 573 }
 574
 575 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 576                                       struct netlink_callback *cb)
 577 {
 578         struct net *net = sock_net(skb->sk);
 579         int h, s_h;
 580         int idx, s_idx;
 581         struct net_device *dev;
 582         struct inet6_dev *idev;
 583         struct hlist_head *head;
 584
 585         s_h = cb->args[0];
 586         s_idx = idx = cb->args[1];
 587
 588         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 589                 idx = 0;
 590                 head = &net->dev_index_head[h];
 591                 rcu_read_lock();
 592                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 593                           net->dev_base_seq;
 594                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 595                         if (idx < s_idx)
 596                                 goto cont;
 597                         idev = __in6_dev_get(dev);
 598                         if (!idev)
 599                                 goto cont;
 600
 601                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 602                                                        &idev->cnf,
 603                                                        NETLINK_CB(cb->skb).portid,
 604                                                        cb->nlh->nlmsg_seq,
 605                                                        RTM_NEWNETCONF,
 606                                                        NLM_F_MULTI,
 607                                                        -1) <= 0) {
 608                                 rcu_read_unlock();
 609                                 goto done;
 610                         }
 611                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 612 cont:
 613                         idx++;
 614                 }
 615                 rcu_read_unlock();
 616         }
 617         if (h == NETDEV_HASHENTRIES) {
 618                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 619                                                net->ipv6.devconf_all,
 620                                                NETLINK_CB(cb->skb).portid,
 621                                                cb->nlh->nlmsg_seq,
 622                                                RTM_NEWNETCONF, NLM_F_MULTI,
 623                                                -1) <= 0)
 624                         goto done;
 625                 else
 626                         h++;
 627         }
 628         if (h == NETDEV_HASHENTRIES + 1) {
 629                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 630                                                net->ipv6.devconf_dflt,
 631                                                NETLINK_CB(cb->skb).portid,
 632                                                cb->nlh->nlmsg_seq,
 633                                                RTM_NEWNETCONF, NLM_F_MULTI,
 634                                                -1) <= 0)
 635                         goto done;
 636                 else
 637                         h++;
 638         }
 639 done:
 640         cb->args[0] = h;
 641         cb->args[1] = idx;
 642
 643         return skb->len;
 644 }
 645
 646 #ifdef CONFIG_SYSCTL
 647 static void dev_forward_change(struct inet6_dev *idev)
 648 {
 649         struct net_device *dev;
 650         struct inet6_ifaddr *ifa;
 651
 652         if (!idev)
 653                 return;
 654         dev = idev->dev;
 655         if (idev->cnf.forwarding)
 656                 dev_disable_lro(dev);
 657         if (dev->flags & IFF_MULTICAST) {
 658                 if (idev->cnf.forwarding) {
 659                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 660                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 661                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 662                 } else {
 663                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 664                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 665                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 666                 }
 667         }
 668
 669         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 670                 if (ifa->flags&IFA_F_TENTATIVE)
 671                         continue;
 672                 if (idev->cnf.forwarding)
 673                         addrconf_join_anycast(ifa);
 674                 else
 675                         addrconf_leave_anycast(ifa);
 676         }
 677         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 678                                      dev->ifindex, &idev->cnf);
 679 }
 680
 681
 682 static void addrconf_forward_change(struct net *net, __s32 newf)
 683 {
 684         struct net_device *dev;
 685         struct inet6_dev *idev;
 686
 687         for_each_netdev(net, dev) {
 688                 idev = __in6_dev_get(dev);
 689                 if (idev) {
 690                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 691                         idev->cnf.forwarding = newf;
 692                         if (changed)
 693                                 dev_forward_change(idev);
 694                 }
 695         }
 696 }
 697
 698 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 699 {
 700         struct net *net;
 701         int old;
 702
 703         if (!rtnl_trylock())
 704                 return restart_syscall();
 705
 706         net = (struct net *)table->extra2;
 707         old = *p;
 708         *p = newf;
 709
 710         if (p == &net->ipv6.devconf_dflt->forwarding) {
 711                 if ((!newf) ^ (!old))
 712                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 713                                                      NETCONFA_IFINDEX_DEFAULT,
 714                                                      net->ipv6.devconf_dflt);
 715                 rtnl_unlock();
 716                 return 0;
 717         }
 718
 719         if (p == &net->ipv6.devconf_all->forwarding) {
 720                 net->ipv6.devconf_dflt->forwarding = newf;
 721                 addrconf_forward_change(net, newf);
 722                 if ((!newf) ^ (!old))
 723                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 724                                                      NETCONFA_IFINDEX_ALL,
 725                                                      net->ipv6.devconf_all);
 726         } else if ((!newf) ^ (!old))
 727                 dev_forward_change((struct inet6_dev *)table->extra1);
 728         rtnl_unlock();
 729
 730         if (newf)
 731                 rt6_purge_dflt_routers(net);
 732         return 1;
 733 }
 734 #endif
 735
 736 /* Nobody refers to this ifaddr, destroy it */
 737 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 738 {
 739         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 740
 741 #ifdef NET_REFCNT_DEBUG
 742         pr_debug("%s\n", __func__);
 743 #endif
 744
 745         in6_dev_put(ifp->idev);
 746
 747         if (del_timer(&ifp->dad_timer))
 748                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 749
 750         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 751                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 752                 return;
 753         }
 754         ip6_rt_put(ifp->rt);
 755
 756         kfree_rcu(ifp, rcu);
 757 }
 758
 759 static void
 760 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 761 {
 762         struct list_head *p;
 763         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 764
 765         /*
 766          * Each device address list is sorted in order of scope -
 767          * global before linklocal.
 768          */
 769         list_for_each(p, &idev->addr_list) {
 770                 struct inet6_ifaddr *ifa
 771                         = list_entry(p, struct inet6_ifaddr, if_list);
 772                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 773                         break;
 774         }
 775
 776         list_add_tail(&ifp->if_list, p);
 777 }
 778
 779 static u32 inet6_addr_hash(const struct in6_addr *addr)
 780 {
 781         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 782 }
 783
 784 /* On success it returns ifp with increased reference count */
 785
 786 static struct inet6_ifaddr *
 787 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 788               const struct in6_addr *peer_addr, int pfxlen,
 789               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 790 {
 791         struct inet6_ifaddr *ifa = NULL;
 792         struct rt6_info *rt;
 793         unsigned int hash;
 794         int err = 0;
 795         int addr_type = ipv6_addr_type(addr);
 796
 797         if (addr_type == IPV6_ADDR_ANY ||
 798             addr_type & IPV6_ADDR_MULTICAST ||
 799             (!(idev->dev->flags & IFF_LOOPBACK) &&
 800              addr_type & IPV6_ADDR_LOOPBACK))
 801                 return ERR_PTR(-EADDRNOTAVAIL);
 802
 803         rcu_read_lock_bh();
 804         if (idev->dead) {
 805                 err = -ENODEV;                  /*XXX*/
 806                 goto out2;
 807         }
 808
 809         if (idev->cnf.disable_ipv6) {
 810                 err = -EACCES;
 811                 goto out2;
 812         }
 813
 814         spin_lock(&addrconf_hash_lock);
 815
 816         /* Ignore adding duplicate addresses on an interface */
 817         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 818                 ADBG("ipv6_add_addr: already assigned\n");
 819                 err = -EEXIST;
 820                 goto out;
 821         }
 822
 823         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 824
 825         if (ifa == NULL) {
 826                 ADBG("ipv6_add_addr: malloc failed\n");
 827                 err = -ENOBUFS;
 828                 goto out;
 829         }
 830
 831         rt = addrconf_dst_alloc(idev, addr, false);
 832         if (IS_ERR(rt)) {
 833                 err = PTR_ERR(rt);
 834                 goto out;
 835         }
 836
 837         ifa->addr = *addr;
 838         if (peer_addr)
 839                 ifa->peer_addr = *peer_addr;
 840
 841         spin_lock_init(&ifa->lock);
 842         spin_lock_init(&ifa->state_lock);
 843         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 844                     (unsigned long)ifa);
 845         INIT_HLIST_NODE(&ifa->addr_lst);
 846         ifa->scope = scope;
 847         ifa->prefix_len = pfxlen;
 848         ifa->flags = flags | IFA_F_TENTATIVE;
 849         ifa->valid_lft = valid_lft;
 850         ifa->prefered_lft = prefered_lft;
 851         ifa->cstamp = ifa->tstamp = jiffies;
 852         ifa->tokenized = false;
 853
 854         ifa->rt = rt;
 855
 856         ifa->idev = idev;
 857         in6_dev_hold(idev);
 858         /* For caller */
 859         in6_ifa_hold(ifa);
 860
 861         /* Add to big hash table */
 862         hash = inet6_addr_hash(addr);
 863
 864         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 865         spin_unlock(&addrconf_hash_lock);
 866
 867         write_lock(&idev->lock);
 868         /* Add to inet6_dev unicast addr list. */
 869         ipv6_link_dev_addr(idev, ifa);
 870
 871         if (ifa->flags&IFA_F_TEMPORARY) {
 872                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 873                 in6_ifa_hold(ifa);
 874         }
 875
 876         in6_ifa_hold(ifa);
 877         write_unlock(&idev->lock);
 878 out2:
 879         rcu_read_unlock_bh();
 880
 881         if (likely(err == 0))
 882                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 883         else {
 884                 kfree(ifa);
 885                 ifa = ERR_PTR(err);
 886         }
 887
 888         return ifa;
 889 out:
 890         spin_unlock(&addrconf_hash_lock);
 891         goto out2;
 892 }
 893
 894 /* This function wants to get referenced ifp and releases it before return */
 895
 896 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 897 {
 898         struct inet6_ifaddr *ifa, *ifn;
 899         struct inet6_dev *idev = ifp->idev;
 900         int state;
 901         int deleted = 0, onlink = 0;
 902         unsigned long expires = jiffies;
 903
 904         spin_lock_bh(&ifp->state_lock);
 905         state = ifp->state;
 906         ifp->state = INET6_IFADDR_STATE_DEAD;
 907         spin_unlock_bh(&ifp->state_lock);
 908
 909         if (state == INET6_IFADDR_STATE_DEAD)
 910                 goto out;
 911
 912         spin_lock_bh(&addrconf_hash_lock);
 913         hlist_del_init_rcu(&ifp->addr_lst);
 914         spin_unlock_bh(&addrconf_hash_lock);
 915
 916         write_lock_bh(&idev->lock);
 917
 918         if (ifp->flags&IFA_F_TEMPORARY) {
 919                 list_del(&ifp->tmp_list);
 920                 if (ifp->ifpub) {
 921                         in6_ifa_put(ifp->ifpub);
 922                         ifp->ifpub = NULL;
 923                 }
 924                 __in6_ifa_put(ifp);
 925         }
 926
 927         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 928                 if (ifa == ifp) {
 929                         list_del_init(&ifp->if_list);
 930                         __in6_ifa_put(ifp);
 931
 932                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 933                                 break;
 934                         deleted = 1;
 935                         continue;
 936                 } else if (ifp->flags & IFA_F_PERMANENT) {
 937                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 938                                               ifp->prefix_len)) {
 939                                 if (ifa->flags & IFA_F_PERMANENT) {
 940                                         onlink = 1;
 941                                         if (deleted)
 942                                                 break;
 943                                 } else {
 944                                         unsigned long lifetime;
 945
 946                                         if (!onlink)
 947                                                 onlink = -1;
 948
 949                                         spin_lock(&ifa->lock);
 950
 951                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 952                                         /*
 953                                          * Note: Because this address is
 954                                          * not permanent, lifetime <
 955                                          * LONG_MAX / HZ here.
 956                                          */
 957                                         if (time_before(expires,
 958                                                         ifa->tstamp + lifetime * HZ))
 959                                                 expires = ifa->tstamp + lifetime * HZ;
 960                                         spin_unlock(&ifa->lock);
 961                                 }
 962                         }
 963                 }
 964         }
 965         write_unlock_bh(&idev->lock);
 966
 967         addrconf_del_dad_timer(ifp);
 968
 969         ipv6_ifa_notify(RTM_DELADDR, ifp);
 970
 971         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
 972
 973         /*
 974          * Purge or update corresponding prefix
 975          *
 976          * 1) we don't purge prefix here if address was not permanent.
 977          *    prefix is managed by its own lifetime.
 978          * 2) if there're no addresses, delete prefix.
 979          * 3) if there're still other permanent address(es),
 980          *    corresponding prefix is still permanent.
 981          * 4) otherwise, update prefix lifetime to the
 982          *    longest valid lifetime among the corresponding
 983          *    addresses on the device.
 984          *    Note: subsequent RA will update lifetime.
 985          *
 986          * --yoshfuji
 987          */
 988         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 989                 struct in6_addr prefix;
 990                 struct rt6_info *rt;
 991
 992                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 993
 994                 rt = addrconf_get_prefix_route(&prefix,
 995                                                ifp->prefix_len,
 996                                                ifp->idev->dev,
 997                                                0, RTF_GATEWAY | RTF_DEFAULT);
 998
 999                 if (rt) {
1000                         if (onlink == 0) {
1001                                 ip6_del_rt(rt);
1002                                 rt = NULL;
1003                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1004                                 rt6_set_expires(rt, expires);
1005                         }
1006                 }
1007                 ip6_rt_put(rt);
1008         }
1009
1010         /* clean up prefsrc entries */
1011         rt6_remove_prefsrc(ifp);
1012 out:
1013         in6_ifa_put(ifp);
1014 }
1015
1016 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1017 {
1018         struct inet6_dev *idev = ifp->idev;
1019         struct in6_addr addr, *tmpaddr;
1020         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1021         unsigned long regen_advance;
1022         int tmp_plen;
1023         int ret = 0;
1024         u32 addr_flags;
1025         unsigned long now = jiffies;
1026
1027         write_lock(&idev->lock);
1028         if (ift) {
1029                 spin_lock_bh(&ift->lock);
1030                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1031                 spin_unlock_bh(&ift->lock);
1032                 tmpaddr = &addr;
1033         } else {
1034                 tmpaddr = NULL;
1035         }
1036 retry:
1037         in6_dev_hold(idev);
1038         if (idev->cnf.use_tempaddr <= 0) {
1039                 write_unlock(&idev->lock);
1040                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1041                 in6_dev_put(idev);
1042                 ret = -1;
1043                 goto out;
1044         }
1045         spin_lock_bh(&ifp->lock);
1046         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1047                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1048                 spin_unlock_bh(&ifp->lock);
1049                 write_unlock(&idev->lock);
1050                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1051                         __func__);
1052                 in6_dev_put(idev);
1053                 ret = -1;
1054                 goto out;
1055         }
1056         in6_ifa_hold(ifp);
1057         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1058         __ipv6_try_regen_rndid(idev, tmpaddr);
1059         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1060         age = (now - ifp->tstamp) / HZ;
1061         tmp_valid_lft = min_t(__u32,
1062                               ifp->valid_lft,
1063                               idev->cnf.temp_valid_lft + age);
1064         tmp_prefered_lft = min_t(__u32,
1065                                  ifp->prefered_lft,
1066                                  idev->cnf.temp_prefered_lft + age -
1067                                  idev->cnf.max_desync_factor);
1068         tmp_plen = ifp->prefix_len;
1069         tmp_tstamp = ifp->tstamp;
1070         spin_unlock_bh(&ifp->lock);
1071
1072         regen_advance = idev->cnf.regen_max_retry *
1073                         idev->cnf.dad_transmits *
1074                         idev->nd_parms->retrans_time / HZ;
1075         write_unlock(&idev->lock);
1076
1077         /* A temporary address is created only if this calculated Preferred
1078          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1079          * an implementation must not create a temporary address with a zero
1080          * Preferred Lifetime.
1081          */
1082         if (tmp_prefered_lft <= regen_advance) {
1083                 in6_ifa_put(ifp);
1084                 in6_dev_put(idev);
1085                 ret = -1;
1086                 goto out;
1087         }
1088
1089         addr_flags = IFA_F_TEMPORARY;
1090         /* set in addrconf_prefix_rcv() */
1091         if (ifp->flags & IFA_F_OPTIMISTIC)
1092                 addr_flags |= IFA_F_OPTIMISTIC;
1093
1094         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1095                             ipv6_addr_scope(&addr), addr_flags,
1096                             tmp_valid_lft, tmp_prefered_lft);
1097         if (IS_ERR(ift)) {
1098                 in6_ifa_put(ifp);
1099                 in6_dev_put(idev);
1100                 pr_info("%s: retry temporary address regeneration\n", __func__);
1101                 tmpaddr = &addr;
1102                 write_lock(&idev->lock);
1103                 goto retry;
1104         }
1105
1106         spin_lock_bh(&ift->lock);
1107         ift->ifpub = ifp;
1108         ift->cstamp = now;
1109         ift->tstamp = tmp_tstamp;
1110         spin_unlock_bh(&ift->lock);
1111
1112         addrconf_dad_start(ift);
1113         in6_ifa_put(ift);
1114         in6_dev_put(idev);
1115 out:
1116         return ret;
1117 }
1118
1119 /*
1120  *      Choose an appropriate source address (RFC3484)
1121  */
1122 enum {
1123         IPV6_SADDR_RULE_INIT = 0,
1124         IPV6_SADDR_RULE_LOCAL,
1125         IPV6_SADDR_RULE_SCOPE,
1126         IPV6_SADDR_RULE_PREFERRED,
1127 #ifdef CONFIG_IPV6_MIP6
1128         IPV6_SADDR_RULE_HOA,
1129 #endif
1130         IPV6_SADDR_RULE_OIF,
1131         IPV6_SADDR_RULE_LABEL,
1132         IPV6_SADDR_RULE_PRIVACY,
1133         IPV6_SADDR_RULE_ORCHID,
1134         IPV6_SADDR_RULE_PREFIX,
1135         IPV6_SADDR_RULE_MAX
1136 };
1137
1138 struct ipv6_saddr_score {
1139         int                     rule;
1140         int                     addr_type;
1141         struct inet6_ifaddr     *ifa;
1142         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1143         int                     scopedist;
1144         int                     matchlen;
1145 };
1146
1147 struct ipv6_saddr_dst {
1148         const struct in6_addr *addr;
1149         int ifindex;
1150         int scope;
1151         int label;
1152         unsigned int prefs;
1153 };
1154
1155 static inline int ipv6_saddr_preferred(int type)
1156 {
1157         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1158                 return 1;
1159         return 0;
1160 }
1161
1162 static int ipv6_get_saddr_eval(struct net *net,
1163                                struct ipv6_saddr_score *score,
1164                                struct ipv6_saddr_dst *dst,
1165                                int i)
1166 {
1167         int ret;
1168
1169         if (i <= score->rule) {
1170                 switch (i) {
1171                 case IPV6_SADDR_RULE_SCOPE:
1172                         ret = score->scopedist;
1173                         break;
1174                 case IPV6_SADDR_RULE_PREFIX:
1175                         ret = score->matchlen;
1176                         break;
1177                 default:
1178                         ret = !!test_bit(i, score->scorebits);
1179                 }
1180                 goto out;
1181         }
1182
1183         switch (i) {
1184         case IPV6_SADDR_RULE_INIT:
1185                 /* Rule 0: remember if hiscore is not ready yet */
1186                 ret = !!score->ifa;
1187                 break;
1188         case IPV6_SADDR_RULE_LOCAL:
1189                 /* Rule 1: Prefer same address */
1190                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1191                 break;
1192         case IPV6_SADDR_RULE_SCOPE:
1193                 /* Rule 2: Prefer appropriate scope
1194                  *
1195                  *      ret
1196                  *       ^
1197                  *    -1 |  d 15
1198                  *    ---+--+-+---> scope
1199                  *       |
1200                  *       |             d is scope of the destination.
1201                  *  B-d  |  \
1202                  *       |   \      <- smaller scope is better if
1203                  *  B-15 |    \        if scope is enough for destinaion.
1204                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1205                  * d-C-1 | /
1206                  *       |/         <- greater is better
1207                  *   -C  /             if scope is not enough for destination.
1208                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1209                  *
1210                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1211                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1212                  * Assume B = 0 and we get C > 29.
1213                  */
1214                 ret = __ipv6_addr_src_scope(score->addr_type);
1215                 if (ret >= dst->scope)
1216                         ret = -ret;
1217                 else
1218                         ret -= 128;     /* 30 is enough */
1219                 score->scopedist = ret;
1220                 break;
1221         case IPV6_SADDR_RULE_PREFERRED:
1222                 /* Rule 3: Avoid deprecated and optimistic addresses */
1223                 ret = ipv6_saddr_preferred(score->addr_type) ||
1224                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1225                 break;
1226 #ifdef CONFIG_IPV6_MIP6
1227         case IPV6_SADDR_RULE_HOA:
1228             {
1229                 /* Rule 4: Prefer home address */
1230                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1231                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1232                 break;
1233             }
1234 #endif
1235         case IPV6_SADDR_RULE_OIF:
1236                 /* Rule 5: Prefer outgoing interface */
1237                 ret = (!dst->ifindex ||
1238                        dst->ifindex == score->ifa->idev->dev->ifindex);
1239                 break;
1240         case IPV6_SADDR_RULE_LABEL:
1241                 /* Rule 6: Prefer matching label */
1242                 ret = ipv6_addr_label(net,
1243                                       &score->ifa->addr, score->addr_type,
1244                                       score->ifa->idev->dev->ifindex) == dst->label;
1245                 break;
1246         case IPV6_SADDR_RULE_PRIVACY:
1247             {
1248                 /* Rule 7: Prefer public address
1249                  * Note: prefer temporary address if use_tempaddr >= 2
1250                  */
1251                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1252                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1253                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1254                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1255                 break;
1256             }
1257         case IPV6_SADDR_RULE_ORCHID:
1258                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1259                  *          non-ORCHID vs non-ORCHID
1260                  */
1261                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1262                         ipv6_addr_orchid(dst->addr));
1263                 break;
1264         case IPV6_SADDR_RULE_PREFIX:
1265                 /* Rule 8: Use longest matching prefix */
1266                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1267                 if (ret > score->ifa->prefix_len)
1268                         ret = score->ifa->prefix_len;
1269                 score->matchlen = ret;
1270                 break;
1271         default:
1272                 ret = 0;
1273         }
1274
1275         if (ret)
1276                 __set_bit(i, score->scorebits);
1277         score->rule = i;
1278 out:
1279         return ret;
1280 }
1281
1282 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1283                        const struct in6_addr *daddr, unsigned int prefs,
1284                        struct in6_addr *saddr)
1285 {
1286         struct ipv6_saddr_score scores[2],
1287                                 *score = &scores[0], *hiscore = &scores[1];
1288         struct ipv6_saddr_dst dst;
1289         struct net_device *dev;
1290         int dst_type;
1291
1292         dst_type = __ipv6_addr_type(daddr);
1293         dst.addr = daddr;
1294         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1295         dst.scope = __ipv6_addr_src_scope(dst_type);
1296         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1297         dst.prefs = prefs;
1298
1299         hiscore->rule = -1;
1300         hiscore->ifa = NULL;
1301
1302         rcu_read_lock();
1303
1304         for_each_netdev_rcu(net, dev) {
1305                 struct inet6_dev *idev;
1306
1307                 /* Candidate Source Address (section 4)
1308                  *  - multicast and link-local destination address,
1309                  *    the set of candidate source address MUST only
1310                  *    include addresses assigned to interfaces
1311                  *    belonging to the same link as the outgoing
1312                  *    interface.
1313                  * (- For site-local destination addresses, the
1314                  *    set of candidate source addresses MUST only
1315                  *    include addresses assigned to interfaces
1316                  *    belonging to the same site as the outgoing
1317                  *    interface.)
1318                  */
1319                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1320                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1321                     dst.ifindex && dev->ifindex != dst.ifindex)
1322                         continue;
1323
1324                 idev = __in6_dev_get(dev);
1325                 if (!idev)
1326                         continue;
1327
1328                 read_lock_bh(&idev->lock);
1329                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1330                         int i;
1331
1332                         /*
1333                          * - Tentative Address (RFC2462 section 5.4)
1334                          *  - A tentative address is not considered
1335                          *    "assigned to an interface" in the traditional
1336                          *    sense, unless it is also flagged as optimistic.
1337                          * - Candidate Source Address (section 4)
1338                          *  - In any case, anycast addresses, multicast
1339                          *    addresses, and the unspecified address MUST
1340                          *    NOT be included in a candidate set.
1341                          */
1342                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1343                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1344                                 continue;
1345
1346                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1347
1348                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1349                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1350                                 LIMIT_NETDEBUG(KERN_DEBUG
1351                                                "ADDRCONF: unspecified / multicast address "
1352                                                "assigned as unicast address on %s",
1353                                                dev->name);
1354                                 continue;
1355                         }
1356
1357                         score->rule = -1;
1358                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1359
1360                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1361                                 int minihiscore, miniscore;
1362
1363                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1364                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1365
1366                                 if (minihiscore > miniscore) {
1367                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1368                                             score->scopedist > 0) {
1369                                                 /*
1370                                                  * special case:
1371                                                  * each remaining entry
1372                                                  * has too small (not enough)
1373                                                  * scope, because ifa entries
1374                                                  * are sorted by their scope
1375                                                  * values.
1376                                                  */
1377                                                 goto try_nextdev;
1378                                         }
1379                                         break;
1380                                 } else if (minihiscore < miniscore) {
1381                                         if (hiscore->ifa)
1382                                                 in6_ifa_put(hiscore->ifa);
1383
1384                                         in6_ifa_hold(score->ifa);
1385
1386                                         swap(hiscore, score);
1387
1388                                         /* restore our iterator */
1389                                         score->ifa = hiscore->ifa;
1390
1391                                         break;
1392                                 }
1393                         }
1394                 }
1395 try_nextdev:
1396                 read_unlock_bh(&idev->lock);
1397         }
1398         rcu_read_unlock();
1399
1400         if (!hiscore->ifa)
1401                 return -EADDRNOTAVAIL;
1402
1403         *saddr = hiscore->ifa->addr;
1404         in6_ifa_put(hiscore->ifa);
1405         return 0;
1406 }
1407 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1408
1409 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1410                       u32 banned_flags)
1411 {
1412         struct inet6_ifaddr *ifp;
1413         int err = -EADDRNOTAVAIL;
1414
1415         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1416                 if (ifp->scope == IFA_LINK &&
1417                     !(ifp->flags & banned_flags)) {
1418                         *addr = ifp->addr;
1419                         err = 0;
1420                         break;
1421                 }
1422         }
1423         return err;
1424 }
1425
1426 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1427                     u32 banned_flags)
1428 {
1429         struct inet6_dev *idev;
1430         int err = -EADDRNOTAVAIL;
1431
1432         rcu_read_lock();
1433         idev = __in6_dev_get(dev);
1434         if (idev) {
1435                 read_lock_bh(&idev->lock);
1436                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1437                 read_unlock_bh(&idev->lock);
1438         }
1439         rcu_read_unlock();
1440         return err;
1441 }
1442
1443 static int ipv6_count_addresses(struct inet6_dev *idev)
1444 {
1445         int cnt = 0;
1446         struct inet6_ifaddr *ifp;
1447
1448         read_lock_bh(&idev->lock);
1449         list_for_each_entry(ifp, &idev->addr_list, if_list)
1450                 cnt++;
1451         read_unlock_bh(&idev->lock);
1452         return cnt;
1453 }
1454
1455 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1456                   const struct net_device *dev, int strict)
1457 {
1458         struct inet6_ifaddr *ifp;
1459         unsigned int hash = inet6_addr_hash(addr);
1460
1461         rcu_read_lock_bh();
1462         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1463                 if (!net_eq(dev_net(ifp->idev->dev), net))
1464                         continue;
1465                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1466                     !(ifp->flags&IFA_F_TENTATIVE) &&
1467                     (dev == NULL || ifp->idev->dev == dev ||
1468                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1469                         rcu_read_unlock_bh();
1470                         return 1;
1471                 }
1472         }
1473
1474         rcu_read_unlock_bh();
1475         return 0;
1476 }
1477 EXPORT_SYMBOL(ipv6_chk_addr);
1478
1479 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1480                                struct net_device *dev)
1481 {
1482         unsigned int hash = inet6_addr_hash(addr);
1483         struct inet6_ifaddr *ifp;
1484
1485         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1486                 if (!net_eq(dev_net(ifp->idev->dev), net))
1487                         continue;
1488                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1489                         if (dev == NULL || ifp->idev->dev == dev)
1490                                 return true;
1491                 }
1492         }
1493         return false;
1494 }
1495
1496 /* Compares an address/prefix_len with addresses on device @dev.
1497  * If one is found it returns true.
1498  */
1499 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1500         const unsigned int prefix_len, struct net_device *dev)
1501 {
1502         struct inet6_dev *idev;
1503         struct inet6_ifaddr *ifa;
1504         bool ret = false;
1505
1506         rcu_read_lock();
1507         idev = __in6_dev_get(dev);
1508         if (idev) {
1509                 read_lock_bh(&idev->lock);
1510                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1511                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1512                         if (ret)
1513                                 break;
1514                 }
1515                 read_unlock_bh(&idev->lock);
1516         }
1517         rcu_read_unlock();
1518
1519         return ret;
1520 }
1521 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1522
1523 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1524 {
1525         struct inet6_dev *idev;
1526         struct inet6_ifaddr *ifa;
1527         int     onlink;
1528
1529         onlink = 0;
1530         rcu_read_lock();
1531         idev = __in6_dev_get(dev);
1532         if (idev) {
1533                 read_lock_bh(&idev->lock);
1534                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1535                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1536                                                    ifa->prefix_len);
1537                         if (onlink)
1538                                 break;
1539                 }
1540                 read_unlock_bh(&idev->lock);
1541         }
1542         rcu_read_unlock();
1543         return onlink;
1544 }
1545 EXPORT_SYMBOL(ipv6_chk_prefix);
1546
1547 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1548                                      struct net_device *dev, int strict)
1549 {
1550         struct inet6_ifaddr *ifp, *result = NULL;
1551         unsigned int hash = inet6_addr_hash(addr);
1552
1553         rcu_read_lock_bh();
1554         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1555                 if (!net_eq(dev_net(ifp->idev->dev), net))
1556                         continue;
1557                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1558                         if (dev == NULL || ifp->idev->dev == dev ||
1559                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1560                                 result = ifp;
1561                                 in6_ifa_hold(ifp);
1562                                 break;
1563                         }
1564                 }
1565         }
1566         rcu_read_unlock_bh();
1567
1568         return result;
1569 }
1570
1571 /* Gets referenced address, destroys ifaddr */
1572
1573 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1574 {
1575         if (ifp->flags&IFA_F_PERMANENT) {
1576                 spin_lock_bh(&ifp->lock);
1577                 addrconf_del_dad_timer(ifp);
1578                 ifp->flags |= IFA_F_TENTATIVE;
1579                 if (dad_failed)
1580                         ifp->flags |= IFA_F_DADFAILED;
1581                 spin_unlock_bh(&ifp->lock);
1582                 if (dad_failed)
1583                         ipv6_ifa_notify(0, ifp);
1584                 in6_ifa_put(ifp);
1585         } else if (ifp->flags&IFA_F_TEMPORARY) {
1586                 struct inet6_ifaddr *ifpub;
1587                 spin_lock_bh(&ifp->lock);
1588                 ifpub = ifp->ifpub;
1589                 if (ifpub) {
1590                         in6_ifa_hold(ifpub);
1591                         spin_unlock_bh(&ifp->lock);
1592                         ipv6_create_tempaddr(ifpub, ifp);
1593                         in6_ifa_put(ifpub);
1594                 } else {
1595                         spin_unlock_bh(&ifp->lock);
1596                 }
1597                 ipv6_del_addr(ifp);
1598         } else
1599                 ipv6_del_addr(ifp);
1600 }
1601
1602 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1603 {
1604         int err = -ENOENT;
1605
1606         spin_lock(&ifp->state_lock);
1607         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1608                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1609                 err = 0;
1610         }
1611         spin_unlock(&ifp->state_lock);
1612
1613         return err;
1614 }
1615
1616 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1617 {
1618         struct inet6_dev *idev = ifp->idev;
1619
1620         if (addrconf_dad_end(ifp)) {
1621                 in6_ifa_put(ifp);
1622                 return;
1623         }
1624
1625         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1626                              ifp->idev->dev->name, &ifp->addr);
1627
1628         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1629                 struct in6_addr addr;
1630
1631                 addr.s6_addr32[0] = htonl(0xfe800000);
1632                 addr.s6_addr32[1] = 0;
1633
1634                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1635                     ipv6_addr_equal(&ifp->addr, &addr)) {
1636                         /* DAD failed for link-local based on MAC address */
1637                         idev->cnf.disable_ipv6 = 1;
1638
1639                         pr_info("%s: IPv6 being disabled!\n",
1640                                 ifp->idev->dev->name);
1641                 }
1642         }
1643
1644         addrconf_dad_stop(ifp, 1);
1645 }
1646
1647 /* Join to solicited addr multicast group. */
1648
1649 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1650 {
1651         struct in6_addr maddr;
1652
1653         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1654                 return;
1655
1656         addrconf_addr_solict_mult(addr, &maddr);
1657         ipv6_dev_mc_inc(dev, &maddr);
1658 }
1659
1660 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1661 {
1662         struct in6_addr maddr;
1663
1664         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1665                 return;
1666
1667         addrconf_addr_solict_mult(addr, &maddr);
1668         __ipv6_dev_mc_dec(idev, &maddr);
1669 }
1670
1671 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1672 {
1673         struct in6_addr addr;
1674         if (ifp->prefix_len == 127) /* RFC 6164 */
1675                 return;
1676         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1677         if (ipv6_addr_any(&addr))
1678                 return;
1679         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1680 }
1681
1682 static void addrconf_leave_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_dec(ifp->idev, &addr);
1691 }
1692
1693 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1694 {
1695         if (dev->addr_len != ETH_ALEN)
1696                 return -1;
1697         memcpy(eui, dev->dev_addr, 3);
1698         memcpy(eui + 5, dev->dev_addr + 3, 3);
1699
1700         /*
1701          * The zSeries OSA network cards can be shared among various
1702          * OS instances, but the OSA cards have only one MAC address.
1703          * This leads to duplicate address conflicts in conjunction
1704          * with IPv6 if more than one instance uses the same card.
1705          *
1706          * The driver for these cards can deliver a unique 16-bit
1707          * identifier for each instance sharing the same card.  It is
1708          * placed instead of 0xFFFE in the interface identifier.  The
1709          * "u" bit of the interface identifier is not inverted in this
1710          * case.  Hence the resulting interface identifier has local
1711          * scope according to RFC2373.
1712          */
1713         if (dev->dev_id) {
1714                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1715                 eui[4] = dev->dev_id & 0xFF;
1716         } else {
1717                 eui[3] = 0xFF;
1718                 eui[4] = 0xFE;
1719                 eui[0] ^= 2;
1720         }
1721         return 0;
1722 }
1723
1724 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1725 {
1726         if (dev->addr_len != IEEE802154_ADDR_LEN)
1727                 return -1;
1728         memcpy(eui, dev->dev_addr, 8);
1729         eui[0] ^= 2;
1730         return 0;
1731 }
1732
1733 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1734 {
1735         union fwnet_hwaddr *ha;
1736
1737         if (dev->addr_len != FWNET_ALEN)
1738                 return -1;
1739
1740         ha = (union fwnet_hwaddr *)dev->dev_addr;
1741
1742         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1743         eui[0] ^= 2;
1744         return 0;
1745 }
1746
1747 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1748 {
1749         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1750         if (dev->addr_len != ARCNET_ALEN)
1751                 return -1;
1752         memset(eui, 0, 7);
1753         eui[7] = *(u8 *)dev->dev_addr;
1754         return 0;
1755 }
1756
1757 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1758 {
1759         if (dev->addr_len != INFINIBAND_ALEN)
1760                 return -1;
1761         memcpy(eui, dev->dev_addr + 12, 8);
1762         eui[0] |= 2;
1763         return 0;
1764 }
1765
1766 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1767 {
1768         if (addr == 0)
1769                 return -1;
1770         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1771                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1772                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1773                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1774                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1775                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1776         eui[1] = 0;
1777         eui[2] = 0x5E;
1778         eui[3] = 0xFE;
1779         memcpy(eui + 4, &addr, 4);
1780         return 0;
1781 }
1782
1783 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1784 {
1785         if (dev->priv_flags & IFF_ISATAP)
1786                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1787         return -1;
1788 }
1789
1790 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1791 {
1792         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1793 }
1794
1795 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1796 {
1797         memcpy(eui, dev->perm_addr, 3);
1798         memcpy(eui + 5, dev->perm_addr + 3, 3);
1799         eui[3] = 0xFF;
1800         eui[4] = 0xFE;
1801         eui[0] ^= 2;
1802         return 0;
1803 }
1804
1805 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1806 {
1807         switch (dev->type) {
1808         case ARPHRD_ETHER:
1809         case ARPHRD_FDDI:
1810                 return addrconf_ifid_eui48(eui, dev);
1811         case ARPHRD_ARCNET:
1812                 return addrconf_ifid_arcnet(eui, dev);
1813         case ARPHRD_INFINIBAND:
1814                 return addrconf_ifid_infiniband(eui, dev);
1815         case ARPHRD_SIT:
1816                 return addrconf_ifid_sit(eui, dev);
1817         case ARPHRD_IPGRE:
1818                 return addrconf_ifid_gre(eui, dev);
1819         case ARPHRD_IEEE802154:
1820                 return addrconf_ifid_eui64(eui, dev);
1821         case ARPHRD_IEEE1394:
1822                 return addrconf_ifid_ieee1394(eui, dev);
1823         case ARPHRD_TUNNEL6:
1824                 return addrconf_ifid_ip6tnl(eui, dev);
1825         }
1826         return -1;
1827 }
1828
1829 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1830 {
1831         int err = -1;
1832         struct inet6_ifaddr *ifp;
1833
1834         read_lock_bh(&idev->lock);
1835         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1836                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1837                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1838                         err = 0;
1839                         break;
1840                 }
1841         }
1842         read_unlock_bh(&idev->lock);
1843         return err;
1844 }
1845
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
1914 /*
1915  *      Add prefix route.
1916  */
1917
1918 static void
1919 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1920                       unsigned long expires, u32 flags)
1921 {
1922         struct fib6_config cfg = {
1923                 .fc_table = RT6_TABLE_PREFIX,
1924                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1925                 .fc_ifindex = dev->ifindex,
1926                 .fc_expires = expires,
1927                 .fc_dst_len = plen,
1928                 .fc_flags = RTF_UP | flags,
1929                 .fc_nlinfo.nl_net = dev_net(dev),
1930                 .fc_protocol = RTPROT_KERNEL,
1931         };
1932
1933         cfg.fc_dst = *pfx;
1934
1935         /* Prevent useless cloning on PtP SIT.
1936            This thing is done here expecting that the whole
1937            class of non-broadcast devices need not cloning.
1938          */
1939 #if IS_ENABLED(CONFIG_IPV6_SIT)
1940         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1941                 cfg.fc_flags |= RTF_NONEXTHOP;
1942 #endif
1943
1944         ip6_route_add(&cfg);
1945 }
1946
1947
1948 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1949                                                   int plen,
1950                                                   const struct net_device *dev,
1951                                                   u32 flags, u32 noflags)
1952 {
1953         struct fib6_node *fn;
1954         struct rt6_info *rt = NULL;
1955         struct fib6_table *table;
1956
1957         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1958         if (table == NULL)
1959                 return NULL;
1960
1961         read_lock_bh(&table->tb6_lock);
1962         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1963         if (!fn)
1964                 goto out;
1965         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1966                 if (rt->dst.dev->ifindex != dev->ifindex)
1967                         continue;
1968                 if ((rt->rt6i_flags & flags) != flags)
1969                         continue;
1970                 if ((rt->rt6i_flags & noflags) != 0)
1971                         continue;
1972                 dst_hold(&rt->dst);
1973                 break;
1974         }
1975 out:
1976         read_unlock_bh(&table->tb6_lock);
1977         return rt;
1978 }
1979
1980
1981 /* Create "default" multicast route to the interface */
1982
1983 static void addrconf_add_mroute(struct net_device *dev)
1984 {
1985         struct fib6_config cfg = {
1986                 .fc_table = RT6_TABLE_LOCAL,
1987                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1988                 .fc_ifindex = dev->ifindex,
1989                 .fc_dst_len = 8,
1990                 .fc_flags = RTF_UP,
1991                 .fc_nlinfo.nl_net = dev_net(dev),
1992         };
1993
1994         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1995
1996         ip6_route_add(&cfg);
1997 }
1998
1999 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2000 {
2001         struct inet6_dev *idev;
2002
2003         ASSERT_RTNL();
2004
2005         idev = ipv6_find_idev(dev);
2006         if (!idev)
2007                 return ERR_PTR(-ENOBUFS);
2008
2009         if (idev->cnf.disable_ipv6)
2010                 return ERR_PTR(-EACCES);
2011
2012         /* Add default multicast route */
2013         if (!(dev->flags & IFF_LOOPBACK))
2014                 addrconf_add_mroute(dev);
2015
2016         return idev;
2017 }
2018
2019 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2020 {
2021         struct prefix_info *pinfo;
2022         __u32 valid_lft;
2023         __u32 prefered_lft;
2024         int addr_type;
2025         struct inet6_dev *in6_dev;
2026         struct net *net = dev_net(dev);
2027
2028         pinfo = (struct prefix_info *) opt;
2029
2030         if (len < sizeof(struct prefix_info)) {
2031                 ADBG("addrconf: prefix option too short\n");
2032                 return;
2033         }
2034
2035         /*
2036          *      Validation checks ([ADDRCONF], page 19)
2037          */
2038
2039         addr_type = ipv6_addr_type(&pinfo->prefix);
2040
2041         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2042                 return;
2043
2044         valid_lft = ntohl(pinfo->valid);
2045         prefered_lft = ntohl(pinfo->prefered);
2046
2047         if (prefered_lft > valid_lft) {
2048                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2049                 return;
2050         }
2051
2052         in6_dev = in6_dev_get(dev);
2053
2054         if (in6_dev == NULL) {
2055                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2056                                     dev->name);
2057                 return;
2058         }
2059
2060         /*
2061          *      Two things going on here:
2062          *      1) Add routes for on-link prefixes
2063          *      2) Configure prefixes with the auto flag set
2064          */
2065
2066         if (pinfo->onlink) {
2067                 struct rt6_info *rt;
2068                 unsigned long rt_expires;
2069
2070                 /* Avoid arithmetic overflow. Really, we could
2071                  * save rt_expires in seconds, likely valid_lft,
2072                  * but it would require division in fib gc, that it
2073                  * not good.
2074                  */
2075                 if (HZ > USER_HZ)
2076                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2077                 else
2078                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2079
2080                 if (addrconf_finite_timeout(rt_expires))
2081                         rt_expires *= HZ;
2082
2083                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2084                                                pinfo->prefix_len,
2085                                                dev,
2086                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2087                                                RTF_GATEWAY | RTF_DEFAULT);
2088
2089                 if (rt) {
2090                         /* Autoconf prefix route */
2091                         if (valid_lft == 0) {
2092                                 ip6_del_rt(rt);
2093                                 rt = NULL;
2094                         } else if (addrconf_finite_timeout(rt_expires)) {
2095                                 /* not infinity */
2096                                 rt6_set_expires(rt, jiffies + rt_expires);
2097                         } else {
2098                                 rt6_clean_expires(rt);
2099                         }
2100                 } else if (valid_lft) {
2101                         clock_t expires = 0;
2102                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2103                         if (addrconf_finite_timeout(rt_expires)) {
2104                                 /* not infinity */
2105                                 flags |= RTF_EXPIRES;
2106                                 expires = jiffies_to_clock_t(rt_expires);
2107                         }
2108                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2109                                               dev, expires, flags);
2110                 }
2111                 ip6_rt_put(rt);
2112         }
2113
2114         /* Try to figure out our local address for this prefix */
2115
2116         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2117                 struct inet6_ifaddr *ifp;
2118                 struct in6_addr addr;
2119                 int create = 0, update_lft = 0;
2120                 bool tokenized = false;
2121
2122                 if (pinfo->prefix_len == 64) {
2123                         memcpy(&addr, &pinfo->prefix, 8);
2124
2125                         if (!ipv6_addr_any(&in6_dev->token)) {
2126                                 read_lock_bh(&in6_dev->lock);
2127                                 memcpy(addr.s6_addr + 8,
2128                                        in6_dev->token.s6_addr + 8, 8);
2129                                 read_unlock_bh(&in6_dev->lock);
2130                                 tokenized = true;
2131                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2132                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2133                                 in6_dev_put(in6_dev);
2134                                 return;
2135                         }
2136                         goto ok;
2137                 }
2138                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2139                                     pinfo->prefix_len);
2140                 in6_dev_put(in6_dev);
2141                 return;
2142
2143 ok:
2144
2145                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2146
2147                 if (ifp == NULL && valid_lft) {
2148                         int max_addresses = in6_dev->cnf.max_addresses;
2149                         u32 addr_flags = 0;
2150
2151 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2152                         if (in6_dev->cnf.optimistic_dad &&
2153                             !net->ipv6.devconf_all->forwarding && sllao)
2154                                 addr_flags = IFA_F_OPTIMISTIC;
2155 #endif
2156
2157                         /* Do not allow to create too much of autoconfigured
2158                          * addresses; this would be too easy way to crash kernel.
2159                          */
2160                         if (!max_addresses ||
2161                             ipv6_count_addresses(in6_dev) < max_addresses)
2162                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2163                                                     pinfo->prefix_len,
2164                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2165                                                     addr_flags, valid_lft,
2166                                                     prefered_lft);
2167
2168                         if (IS_ERR_OR_NULL(ifp)) {
2169                                 in6_dev_put(in6_dev);
2170                                 return;
2171                         }
2172
2173                         update_lft = 0;
2174                         create = 1;
2175                         ifp->cstamp = jiffies;
2176                         ifp->tokenized = tokenized;
2177                         addrconf_dad_start(ifp);
2178                 }
2179
2180                 if (ifp) {
2181                         u32 flags;
2182                         unsigned long now;
2183                         struct inet6_ifaddr *ift;
2184                         u32 stored_lft;
2185
2186                         /* update lifetime (RFC2462 5.5.3 e) */
2187                         spin_lock(&ifp->lock);
2188                         now = jiffies;
2189                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2190                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2191                         else
2192                                 stored_lft = 0;
2193                         if (!update_lft && !create && stored_lft) {
2194                                 const u32 minimum_lft = min(
2195                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2196                                 valid_lft = max(valid_lft, minimum_lft);
2197
2198                                 /* RFC4862 Section 5.5.3e:
2199                                  * "Note that the preferred lifetime of the
2200                                  *  corresponding address is always reset to
2201                                  *  the Preferred Lifetime in the received
2202                                  *  Prefix Information option, regardless of
2203                                  *  whether the valid lifetime is also reset or
2204                                  *  ignored."
2205                                  *
2206                                  * So we should always update prefered_lft here.
2207                                  */
2208                                 update_lft = 1;
2209                         }
2210
2211                         if (update_lft) {
2212                                 ifp->valid_lft = valid_lft;
2213                                 ifp->prefered_lft = prefered_lft;
2214                                 ifp->tstamp = now;
2215                                 flags = ifp->flags;
2216                                 ifp->flags &= ~IFA_F_DEPRECATED;
2217                                 spin_unlock(&ifp->lock);
2218
2219                                 if (!(flags&IFA_F_TENTATIVE))
2220                                         ipv6_ifa_notify(0, ifp);
2221                         } else
2222                                 spin_unlock(&ifp->lock);
2223
2224                         read_lock_bh(&in6_dev->lock);
2225                         /* update all temporary addresses in the list */
2226                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2227                                             tmp_list) {
2228                                 int age, max_valid, max_prefered;
2229
2230                                 if (ifp != ift->ifpub)
2231                                         continue;
2232
2233                                 /*
2234                                  * RFC 4941 section 3.3:
2235                                  * If a received option will extend the lifetime
2236                                  * of a public address, the lifetimes of
2237                                  * temporary addresses should be extended,
2238                                  * subject to the overall constraint that no
2239                                  * temporary addresses should ever remain
2240                                  * "valid" or "preferred" for a time longer than
2241                                  * (TEMP_VALID_LIFETIME) or
2242                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2243                                  * respectively.
2244                                  */
2245                                 age = (now - ift->cstamp) / HZ;
2246                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2247                                 if (max_valid < 0)
2248                                         max_valid = 0;
2249
2250                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2251                                                in6_dev->cnf.max_desync_factor -
2252                                                age;
2253                                 if (max_prefered < 0)
2254                                         max_prefered = 0;
2255
2256                                 if (valid_lft > max_valid)
2257                                         valid_lft = max_valid;
2258
2259                                 if (prefered_lft > max_prefered)
2260                                         prefered_lft = max_prefered;
2261
2262                                 spin_lock(&ift->lock);
2263                                 flags = ift->flags;
2264                                 ift->valid_lft = valid_lft;
2265                                 ift->prefered_lft = prefered_lft;
2266                                 ift->tstamp = now;
2267                                 if (prefered_lft > 0)
2268                                         ift->flags &= ~IFA_F_DEPRECATED;
2269
2270                                 spin_unlock(&ift->lock);
2271                                 if (!(flags&IFA_F_TENTATIVE))
2272                                         ipv6_ifa_notify(0, ift);
2273                         }
2274
2275                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2276                                 /*
2277                                  * When a new public address is created as
2278                                  * described in [ADDRCONF], also create a new
2279                                  * temporary address. Also create a temporary
2280                                  * address if it's enabled but no temporary
2281                                  * address currently exists.
2282                                  */
2283                                 read_unlock_bh(&in6_dev->lock);
2284                                 ipv6_create_tempaddr(ifp, NULL);
2285                         } else {
2286                                 read_unlock_bh(&in6_dev->lock);
2287                         }
2288
2289                         in6_ifa_put(ifp);
2290                         addrconf_verify(0);
2291                 }
2292         }
2293         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2294         in6_dev_put(in6_dev);
2295 }
2296
2297 /*
2298  *      Set destination address.
2299  *      Special case for SIT interfaces where we create a new "virtual"
2300  *      device.
2301  */
2302 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2303 {
2304         struct in6_ifreq ireq;
2305         struct net_device *dev;
2306         int err = -EINVAL;
2307
2308         rtnl_lock();
2309
2310         err = -EFAULT;
2311         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2312                 goto err_exit;
2313
2314         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2315
2316         err = -ENODEV;
2317         if (dev == NULL)
2318                 goto err_exit;
2319
2320 #if IS_ENABLED(CONFIG_IPV6_SIT)
2321         if (dev->type == ARPHRD_SIT) {
2322                 const struct net_device_ops *ops = dev->netdev_ops;
2323                 struct ifreq ifr;
2324                 struct ip_tunnel_parm p;
2325
2326                 err = -EADDRNOTAVAIL;
2327                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2328                         goto err_exit;
2329
2330                 memset(&p, 0, sizeof(p));
2331                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2332                 p.iph.saddr = 0;
2333                 p.iph.version = 4;
2334                 p.iph.ihl = 5;
2335                 p.iph.protocol = IPPROTO_IPV6;
2336                 p.iph.ttl = 64;
2337                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2338
2339                 if (ops->ndo_do_ioctl) {
2340                         mm_segment_t oldfs = get_fs();
2341
2342                         set_fs(KERNEL_DS);
2343                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2344                         set_fs(oldfs);
2345                 } else
2346                         err = -EOPNOTSUPP;
2347
2348                 if (err == 0) {
2349                         err = -ENOBUFS;
2350                         dev = __dev_get_by_name(net, p.name);
2351                         if (!dev)
2352                                 goto err_exit;
2353                         err = dev_open(dev);
2354                 }
2355         }
2356 #endif
2357
2358 err_exit:
2359         rtnl_unlock();
2360         return err;
2361 }
2362
2363 /*
2364  *      Manual configuration of address on an interface
2365  */
2366 static int inet6_addr_add(struct net *net, int ifindex,
2367                           const struct in6_addr *pfx,
2368                           const struct in6_addr *peer_pfx,
2369                           unsigned int plen, __u32 ifa_flags,
2370                           __u32 prefered_lft, __u32 valid_lft)
2371 {
2372         struct inet6_ifaddr *ifp;
2373         struct inet6_dev *idev;
2374         struct net_device *dev;
2375         int scope;
2376         u32 flags;
2377         clock_t expires;
2378         unsigned long timeout;
2379
2380         ASSERT_RTNL();
2381
2382         if (plen > 128)
2383                 return -EINVAL;
2384
2385         /* check the lifetime */
2386         if (!valid_lft || prefered_lft > valid_lft)
2387                 return -EINVAL;
2388
2389         dev = __dev_get_by_index(net, ifindex);
2390         if (!dev)
2391                 return -ENODEV;
2392
2393         idev = addrconf_add_dev(dev);
2394         if (IS_ERR(idev))
2395                 return PTR_ERR(idev);
2396
2397         scope = ipv6_addr_scope(pfx);
2398
2399         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2400         if (addrconf_finite_timeout(timeout)) {
2401                 expires = jiffies_to_clock_t(timeout * HZ);
2402                 valid_lft = timeout;
2403                 flags = RTF_EXPIRES;
2404         } else {
2405                 expires = 0;
2406                 flags = 0;
2407                 ifa_flags |= IFA_F_PERMANENT;
2408         }
2409
2410         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2411         if (addrconf_finite_timeout(timeout)) {
2412                 if (timeout == 0)
2413                         ifa_flags |= IFA_F_DEPRECATED;
2414                 prefered_lft = timeout;
2415         }
2416
2417         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2418                             valid_lft, prefered_lft);
2419
2420         if (!IS_ERR(ifp)) {
2421                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2422                                       expires, flags);
2423                 /*
2424                  * Note that section 3.1 of RFC 4429 indicates
2425                  * that the Optimistic flag should not be set for
2426                  * manually configured addresses
2427                  */
2428                 addrconf_dad_start(ifp);
2429                 in6_ifa_put(ifp);
2430                 addrconf_verify(0);
2431                 return 0;
2432         }
2433
2434         return PTR_ERR(ifp);
2435 }
2436
2437 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2438                           unsigned int plen)
2439 {
2440         struct inet6_ifaddr *ifp;
2441         struct inet6_dev *idev;
2442         struct net_device *dev;
2443
2444         if (plen > 128)
2445                 return -EINVAL;
2446
2447         dev = __dev_get_by_index(net, ifindex);
2448         if (!dev)
2449                 return -ENODEV;
2450
2451         if ((idev = __in6_dev_get(dev)) == NULL)
2452                 return -ENXIO;
2453
2454         read_lock_bh(&idev->lock);
2455         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2456                 if (ifp->prefix_len == plen &&
2457                     ipv6_addr_equal(pfx, &ifp->addr)) {
2458                         in6_ifa_hold(ifp);
2459                         read_unlock_bh(&idev->lock);
2460
2461                         ipv6_del_addr(ifp);
2462                         return 0;
2463                 }
2464         }
2465         read_unlock_bh(&idev->lock);
2466         return -EADDRNOTAVAIL;
2467 }
2468
2469
2470 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2471 {
2472         struct in6_ifreq ireq;
2473         int err;
2474
2475         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2476                 return -EPERM;
2477
2478         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2479                 return -EFAULT;
2480
2481         rtnl_lock();
2482         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2483                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2484                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2485         rtnl_unlock();
2486         return err;
2487 }
2488
2489 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2490 {
2491         struct in6_ifreq ireq;
2492         int err;
2493
2494         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2495                 return -EPERM;
2496
2497         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2498                 return -EFAULT;
2499
2500         rtnl_lock();
2501         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2502                              ireq.ifr6_prefixlen);
2503         rtnl_unlock();
2504         return err;
2505 }
2506
2507 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2508                      int plen, int scope)
2509 {
2510         struct inet6_ifaddr *ifp;
2511
2512         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2513                             scope, IFA_F_PERMANENT, 0, 0);
2514         if (!IS_ERR(ifp)) {
2515                 spin_lock_bh(&ifp->lock);
2516                 ifp->flags &= ~IFA_F_TENTATIVE;
2517                 spin_unlock_bh(&ifp->lock);
2518                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2519                 in6_ifa_put(ifp);
2520         }
2521 }
2522
2523 #if IS_ENABLED(CONFIG_IPV6_SIT)
2524 static void sit_add_v4_addrs(struct inet6_dev *idev)
2525 {
2526         struct in6_addr addr;
2527         struct net_device *dev;
2528         struct net *net = dev_net(idev->dev);
2529         int scope, plen;
2530         u32 pflags = 0;
2531
2532         ASSERT_RTNL();
2533
2534         memset(&addr, 0, sizeof(struct in6_addr));
2535         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2536
2537         if (idev->dev->flags&IFF_POINTOPOINT) {
2538                 addr.s6_addr32[0] = htonl(0xfe800000);
2539                 scope = IFA_LINK;
2540                 plen = 64;
2541         } else {
2542                 scope = IPV6_ADDR_COMPATv4;
2543                 plen = 96;
2544                 pflags |= RTF_NONEXTHOP;
2545         }
2546
2547         if (addr.s6_addr32[3]) {
2548                 add_addr(idev, &addr, plen, scope);
2549                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2550                 return;
2551         }
2552
2553         for_each_netdev(net, dev) {
2554                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2555                 if (in_dev && (dev->flags & IFF_UP)) {
2556                         struct in_ifaddr *ifa;
2557
2558                         int flag = scope;
2559
2560                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2561
2562                                 addr.s6_addr32[3] = ifa->ifa_local;
2563
2564                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2565                                         continue;
2566                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2567                                         if (idev->dev->flags&IFF_POINTOPOINT)
2568                                                 continue;
2569                                         flag |= IFA_HOST;
2570                                 }
2571
2572                                 add_addr(idev, &addr, plen, flag);
2573                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2574                                                       pflags);
2575                         }
2576                 }
2577         }
2578 }
2579 #endif
2580
2581 static void init_loopback(struct net_device *dev)
2582 {
2583         struct inet6_dev  *idev;
2584         struct net_device *sp_dev;
2585         struct inet6_ifaddr *sp_ifa;
2586         struct rt6_info *sp_rt;
2587
2588         /* ::1 */
2589
2590         ASSERT_RTNL();
2591
2592         if ((idev = ipv6_find_idev(dev)) == NULL) {
2593                 pr_debug("%s: add_dev failed\n", __func__);
2594                 return;
2595         }
2596
2597         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2598
2599         /* Add routes to other interface's IPv6 addresses */
2600         for_each_netdev(dev_net(dev), sp_dev) {
2601                 if (!strcmp(sp_dev->name, dev->name))
2602                         continue;
2603
2604                 idev = __in6_dev_get(sp_dev);
2605                 if (!idev)
2606                         continue;
2607
2608                 read_lock_bh(&idev->lock);
2609                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2610
2611                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2612                                 continue;
2613
2614                         if (sp_ifa->rt)
2615                                 continue;
2616
2617                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2618
2619                         /* Failure cases are ignored */
2620                         if (!IS_ERR(sp_rt)) {
2621                                 sp_ifa->rt = sp_rt;
2622                                 ip6_ins_rt(sp_rt);
2623                         }
2624                 }
2625                 read_unlock_bh(&idev->lock);
2626         }
2627 }
2628
2629 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2630 {
2631         struct inet6_ifaddr *ifp;
2632         u32 addr_flags = IFA_F_PERMANENT;
2633
2634 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2635         if (idev->cnf.optimistic_dad &&
2636             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2637                 addr_flags |= IFA_F_OPTIMISTIC;
2638 #endif
2639
2640
2641         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2642         if (!IS_ERR(ifp)) {
2643                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2644                 addrconf_dad_start(ifp);
2645                 in6_ifa_put(ifp);
2646         }
2647 }
2648
2649 static void addrconf_dev_config(struct net_device *dev)
2650 {
2651         struct in6_addr addr;
2652         struct inet6_dev *idev;
2653
2654         ASSERT_RTNL();
2655
2656         if ((dev->type != ARPHRD_ETHER) &&
2657             (dev->type != ARPHRD_FDDI) &&
2658             (dev->type != ARPHRD_ARCNET) &&
2659             (dev->type != ARPHRD_INFINIBAND) &&
2660             (dev->type != ARPHRD_IEEE802154) &&
2661             (dev->type != ARPHRD_IEEE1394) &&
2662             (dev->type != ARPHRD_TUNNEL6)) {
2663                 /* Alas, we support only Ethernet autoconfiguration. */
2664                 return;
2665         }
2666
2667         idev = addrconf_add_dev(dev);
2668         if (IS_ERR(idev))
2669                 return;
2670
2671         memset(&addr, 0, sizeof(struct in6_addr));
2672         addr.s6_addr32[0] = htonl(0xFE800000);
2673
2674         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2675                 addrconf_add_linklocal(idev, &addr);
2676 }
2677
2678 #if IS_ENABLED(CONFIG_IPV6_SIT)
2679 static void addrconf_sit_config(struct net_device *dev)
2680 {
2681         struct inet6_dev *idev;
2682
2683         ASSERT_RTNL();
2684
2685         /*
2686          * Configure the tunnel with one of our IPv4
2687          * addresses... we should configure all of
2688          * our v4 addrs in the tunnel
2689          */
2690
2691         if ((idev = ipv6_find_idev(dev)) == NULL) {
2692                 pr_debug("%s: add_dev failed\n", __func__);
2693                 return;
2694         }
2695
2696         if (dev->priv_flags & IFF_ISATAP) {
2697                 struct in6_addr addr;
2698
2699                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2700                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2701                         addrconf_add_linklocal(idev, &addr);
2702                 return;
2703         }
2704
2705         sit_add_v4_addrs(idev);
2706
2707         if (dev->flags&IFF_POINTOPOINT)
2708                 addrconf_add_mroute(dev);
2709 }
2710 #endif
2711
2712 #if IS_ENABLED(CONFIG_NET_IPGRE)
2713 static void addrconf_gre_config(struct net_device *dev)
2714 {
2715         struct inet6_dev *idev;
2716         struct in6_addr addr;
2717
2718         ASSERT_RTNL();
2719
2720         if ((idev = ipv6_find_idev(dev)) == NULL) {
2721                 pr_debug("%s: add_dev failed\n", __func__);
2722                 return;
2723         }
2724
2725         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2726         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2727                 addrconf_add_linklocal(idev, &addr);
2728 }
2729 #endif
2730
2731 static inline int
2732 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2733 {
2734         struct in6_addr lladdr;
2735
2736         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2737                 addrconf_add_linklocal(idev, &lladdr);
2738                 return 0;
2739         }
2740         return -1;
2741 }
2742
2743 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2744                            void *ptr)
2745 {
2746         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2747         struct inet6_dev *idev = __in6_dev_get(dev);
2748         int run_pending = 0;
2749         int err;
2750
2751         switch (event) {
2752         case NETDEV_REGISTER:
2753                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2754                         idev = ipv6_add_dev(dev);
2755                         if (!idev)
2756                                 return notifier_from_errno(-ENOMEM);
2757                 }
2758                 break;
2759
2760         case NETDEV_UP:
2761         case NETDEV_CHANGE:
2762                 if (dev->flags & IFF_SLAVE)
2763                         break;
2764
2765                 if (event == NETDEV_UP) {
2766                         if (!addrconf_qdisc_ok(dev)) {
2767                                 /* device is not ready yet. */
2768                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2769                                         dev->name);
2770                                 break;
2771                         }
2772
2773                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2774                                 idev = ipv6_add_dev(dev);
2775
2776                         if (idev) {
2777                                 idev->if_flags |= IF_READY;
2778                                 run_pending = 1;
2779                         }
2780                 } else {
2781                         if (!addrconf_qdisc_ok(dev)) {
2782                                 /* device is still not ready. */
2783                                 break;
2784                         }
2785
2786                         if (idev) {
2787                                 if (idev->if_flags & IF_READY)
2788                                         /* device is already configured. */
2789                                         break;
2790                                 idev->if_flags |= IF_READY;
2791                         }
2792
2793                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2794                                 dev->name);
2795
2796                         run_pending = 1;
2797                 }
2798
2799                 switch (dev->type) {
2800 #if IS_ENABLED(CONFIG_IPV6_SIT)
2801                 case ARPHRD_SIT:
2802                         addrconf_sit_config(dev);
2803                         break;
2804 #endif
2805 #if IS_ENABLED(CONFIG_NET_IPGRE)
2806                 case ARPHRD_IPGRE:
2807                         addrconf_gre_config(dev);
2808                         break;
2809 #endif
2810                 case ARPHRD_LOOPBACK:
2811                         init_loopback(dev);
2812                         break;
2813
2814                 default:
2815                         addrconf_dev_config(dev);
2816                         break;
2817                 }
2818
2819                 if (idev) {
2820                         if (run_pending)
2821                                 addrconf_dad_run(idev);
2822
2823                         /*
2824                          * If the MTU changed during the interface down,
2825                          * when the interface up, the changed MTU must be
2826                          * reflected in the idev as well as routers.
2827                          */
2828                         if (idev->cnf.mtu6 != dev->mtu &&
2829                             dev->mtu >= IPV6_MIN_MTU) {
2830                                 rt6_mtu_change(dev, dev->mtu);
2831                                 idev->cnf.mtu6 = dev->mtu;
2832                         }
2833                         idev->tstamp = jiffies;
2834                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2835
2836                         /*
2837                          * If the changed mtu during down is lower than
2838                          * IPV6_MIN_MTU stop IPv6 on this interface.
2839                          */
2840                         if (dev->mtu < IPV6_MIN_MTU)
2841                                 addrconf_ifdown(dev, 1);
2842                 }
2843                 break;
2844
2845         case NETDEV_CHANGEMTU:
2846                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2847                         rt6_mtu_change(dev, dev->mtu);
2848                         idev->cnf.mtu6 = dev->mtu;
2849                         break;
2850                 }
2851
2852                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2853                         idev = ipv6_add_dev(dev);
2854                         if (idev)
2855                                 break;
2856                 }
2857
2858                 /*
2859                  * MTU falled under IPV6_MIN_MTU.
2860                  * Stop IPv6 on this interface.
2861                  */
2862
2863         case NETDEV_DOWN:
2864         case NETDEV_UNREGISTER:
2865                 /*
2866                  *      Remove all addresses from this interface.
2867                  */
2868                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2869                 break;
2870
2871         case NETDEV_CHANGENAME:
2872                 if (idev) {
2873                         snmp6_unregister_dev(idev);
2874                         addrconf_sysctl_unregister(idev);
2875                         addrconf_sysctl_register(idev);
2876                         err = snmp6_register_dev(idev);
2877                         if (err)
2878                                 return notifier_from_errno(err);
2879                 }
2880                 break;
2881
2882         case NETDEV_PRE_TYPE_CHANGE:
2883         case NETDEV_POST_TYPE_CHANGE:
2884                 addrconf_type_change(dev, event);
2885                 break;
2886         }
2887
2888         return NOTIFY_OK;
2889 }
2890
2891 /*
2892  *      addrconf module should be notified of a device going up
2893  */
2894 static struct notifier_block ipv6_dev_notf = {
2895         .notifier_call = addrconf_notify,
2896 };
2897
2898 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2899 {
2900         struct inet6_dev *idev;
2901         ASSERT_RTNL();
2902
2903         idev = __in6_dev_get(dev);
2904
2905         if (event == NETDEV_POST_TYPE_CHANGE)
2906                 ipv6_mc_remap(idev);
2907         else if (event == NETDEV_PRE_TYPE_CHANGE)
2908                 ipv6_mc_unmap(idev);
2909 }
2910
2911 static int addrconf_ifdown(struct net_device *dev, int how)
2912 {
2913         struct net *net = dev_net(dev);
2914         struct inet6_dev *idev;
2915         struct inet6_ifaddr *ifa;
2916         int state, i;
2917
2918         ASSERT_RTNL();
2919
2920         rt6_ifdown(net, dev);
2921         neigh_ifdown(&nd_tbl, dev);
2922
2923         idev = __in6_dev_get(dev);
2924         if (idev == NULL)
2925                 return -ENODEV;
2926
2927         /*
2928          * Step 1: remove reference to ipv6 device from parent device.
2929          *         Do not dev_put!
2930          */
2931         if (how) {
2932                 idev->dead = 1;
2933
2934                 /* protected by rtnl_lock */
2935                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2936
2937                 /* Step 1.5: remove snmp6 entry */
2938                 snmp6_unregister_dev(idev);
2939
2940         }
2941
2942         /* Step 2: clear hash table */
2943         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2944                 struct hlist_head *h = &inet6_addr_lst[i];
2945
2946                 spin_lock_bh(&addrconf_hash_lock);
2947         restart:
2948                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2949                         if (ifa->idev == idev) {
2950                                 hlist_del_init_rcu(&ifa->addr_lst);
2951                                 addrconf_del_dad_timer(ifa);
2952                                 goto restart;
2953                         }
2954                 }
2955                 spin_unlock_bh(&addrconf_hash_lock);
2956         }
2957
2958         write_lock_bh(&idev->lock);
2959
2960         addrconf_del_rs_timer(idev);
2961
2962         /* Step 2: clear flags for stateless addrconf */
2963         if (!how)
2964                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2965
2966         if (how && del_timer(&idev->regen_timer))
2967                 in6_dev_put(idev);
2968
2969         /* Step 3: clear tempaddr list */
2970         while (!list_empty(&idev->tempaddr_list)) {
2971                 ifa = list_first_entry(&idev->tempaddr_list,
2972                                        struct inet6_ifaddr, tmp_list);
2973                 list_del(&ifa->tmp_list);
2974                 write_unlock_bh(&idev->lock);
2975                 spin_lock_bh(&ifa->lock);
2976
2977                 if (ifa->ifpub) {
2978                         in6_ifa_put(ifa->ifpub);
2979                         ifa->ifpub = NULL;
2980                 }
2981                 spin_unlock_bh(&ifa->lock);
2982                 in6_ifa_put(ifa);
2983                 write_lock_bh(&idev->lock);
2984         }
2985
2986         while (!list_empty(&idev->addr_list)) {
2987                 ifa = list_first_entry(&idev->addr_list,
2988                                        struct inet6_ifaddr, if_list);
2989                 addrconf_del_dad_timer(ifa);
2990
2991                 list_del(&ifa->if_list);
2992
2993                 write_unlock_bh(&idev->lock);
2994
2995                 spin_lock_bh(&ifa->state_lock);
2996                 state = ifa->state;
2997                 ifa->state = INET6_IFADDR_STATE_DEAD;
2998                 spin_unlock_bh(&ifa->state_lock);
2999
3000                 if (state != INET6_IFADDR_STATE_DEAD) {
3001                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3002                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3003                 }
3004                 in6_ifa_put(ifa);
3005
3006                 write_lock_bh(&idev->lock);
3007         }
3008
3009         write_unlock_bh(&idev->lock);
3010
3011         /* Step 5: Discard multicast list */
3012         if (how)
3013                 ipv6_mc_destroy_dev(idev);
3014         else
3015                 ipv6_mc_down(idev);
3016
3017         idev->tstamp = jiffies;
3018
3019         /* Last: Shot the device (if unregistered) */
3020         if (how) {
3021                 addrconf_sysctl_unregister(idev);
3022                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3023                 neigh_ifdown(&nd_tbl, dev);
3024                 in6_dev_put(idev);
3025         }
3026         return 0;
3027 }
3028
3029 static void addrconf_rs_timer(unsigned long data)
3030 {
3031         struct inet6_dev *idev = (struct inet6_dev *)data;
3032         struct net_device *dev = idev->dev;
3033         struct in6_addr lladdr;
3034
3035         write_lock(&idev->lock);
3036         if (idev->dead || !(idev->if_flags & IF_READY))
3037                 goto out;
3038
3039         if (!ipv6_accept_ra(idev))
3040                 goto out;
3041
3042         /* Announcement received after solicitation was sent */
3043         if (idev->if_flags & IF_RA_RCVD)
3044                 goto out;
3045
3046         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3047                 write_unlock(&idev->lock);
3048                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3049                         ndisc_send_rs(dev, &lladdr,
3050                                       &in6addr_linklocal_allrouters);
3051                 else
3052                         goto put;
3053
3054                 write_lock(&idev->lock);
3055                 /* The wait after the last probe can be shorter */
3056                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3057                                              idev->cnf.rtr_solicits) ?
3058                                       idev->cnf.rtr_solicit_delay :
3059                                       idev->cnf.rtr_solicit_interval);
3060         } else {
3061                 /*
3062                  * Note: we do not support deprecated "all on-link"
3063                  * assumption any longer.
3064                  */
3065                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3066         }
3067
3068 out:
3069         write_unlock(&idev->lock);
3070 put:
3071         in6_dev_put(idev);
3072 }
3073
3074 /*
3075  *      Duplicate Address Detection
3076  */
3077 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3078 {
3079         unsigned long rand_num;
3080         struct inet6_dev *idev = ifp->idev;
3081
3082         if (ifp->flags & IFA_F_OPTIMISTIC)
3083                 rand_num = 0;
3084         else
3085                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3086
3087         ifp->dad_probes = idev->cnf.dad_transmits;
3088         addrconf_mod_dad_timer(ifp, rand_num);
3089 }
3090
3091 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3092 {
3093         struct inet6_dev *idev = ifp->idev;
3094         struct net_device *dev = idev->dev;
3095
3096         addrconf_join_solict(dev, &ifp->addr);
3097
3098         net_srandom(ifp->addr.s6_addr32[3]);
3099
3100         read_lock_bh(&idev->lock);
3101         spin_lock(&ifp->lock);
3102         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3103                 goto out;
3104
3105         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3106             idev->cnf.accept_dad < 1 ||
3107             !(ifp->flags&IFA_F_TENTATIVE) ||
3108             ifp->flags & IFA_F_NODAD) {
3109                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3110                 spin_unlock(&ifp->lock);
3111                 read_unlock_bh(&idev->lock);
3112
3113                 addrconf_dad_completed(ifp);
3114                 return;
3115         }
3116
3117         if (!(idev->if_flags & IF_READY)) {
3118                 spin_unlock(&ifp->lock);
3119                 read_unlock_bh(&idev->lock);
3120                 /*
3121                  * If the device is not ready:
3122                  * - keep it tentative if it is a permanent address.
3123                  * - otherwise, kill it.
3124                  */
3125                 in6_ifa_hold(ifp);
3126                 addrconf_dad_stop(ifp, 0);
3127                 return;
3128         }
3129
3130         /*
3131          * Optimistic nodes can start receiving
3132          * Frames right away
3133          */
3134         if (ifp->flags & IFA_F_OPTIMISTIC)
3135                 ip6_ins_rt(ifp->rt);
3136
3137         addrconf_dad_kick(ifp);
3138 out:
3139         spin_unlock(&ifp->lock);
3140         read_unlock_bh(&idev->lock);
3141 }
3142
3143 static void addrconf_dad_timer(unsigned long data)
3144 {
3145         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3146         struct inet6_dev *idev = ifp->idev;
3147         struct in6_addr mcaddr;
3148
3149         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3150                 goto out;
3151
3152         write_lock(&idev->lock);
3153         if (idev->dead || !(idev->if_flags & IF_READY)) {
3154                 write_unlock(&idev->lock);
3155                 goto out;
3156         }
3157
3158         spin_lock(&ifp->lock);
3159         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3160                 spin_unlock(&ifp->lock);
3161                 write_unlock(&idev->lock);
3162                 goto out;
3163         }
3164
3165         if (ifp->dad_probes == 0) {
3166                 /*
3167                  * DAD was successful
3168                  */
3169
3170                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3171                 spin_unlock(&ifp->lock);
3172                 write_unlock(&idev->lock);
3173
3174                 addrconf_dad_completed(ifp);
3175
3176                 goto out;
3177         }
3178
3179         ifp->dad_probes--;
3180         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3181         spin_unlock(&ifp->lock);
3182         write_unlock(&idev->lock);
3183
3184         /* send a neighbour solicitation for our addr */
3185         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3186         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3187 out:
3188         in6_ifa_put(ifp);
3189 }
3190
3191 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3192 {
3193         struct net_device *dev = ifp->idev->dev;
3194         struct in6_addr lladdr;
3195         bool send_rs, send_mld;
3196
3197         addrconf_del_dad_timer(ifp);
3198
3199         /*
3200          *      Configure the address for reception. Now it is valid.
3201          */
3202
3203         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3204
3205         /* If added prefix is link local and we are prepared to process
3206            router advertisements, start sending router solicitations.
3207          */
3208
3209         read_lock_bh(&ifp->idev->lock);
3210         spin_lock(&ifp->lock);
3211         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3212                    ifp->idev->valid_ll_addr_cnt == 1;
3213         send_rs = send_mld &&
3214                   ipv6_accept_ra(ifp->idev) &&
3215                   ifp->idev->cnf.rtr_solicits > 0 &&
3216                   (dev->flags&IFF_LOOPBACK) == 0;
3217         spin_unlock(&ifp->lock);
3218         read_unlock_bh(&ifp->idev->lock);
3219
3220         /* While dad is in progress mld report's source address is in6_addrany.
3221          * Resend with proper ll now.
3222          */
3223         if (send_mld)
3224                 ipv6_mc_dad_complete(ifp->idev);
3225
3226         if (send_rs) {
3227                 /*
3228                  *      If a host as already performed a random delay
3229                  *      [...] as part of DAD [...] there is no need
3230                  *      to delay again before sending the first RS
3231                  */
3232                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3233                         return;
3234                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3235
3236                 write_lock_bh(&ifp->idev->lock);
3237                 spin_lock(&ifp->lock);
3238                 ifp->idev->rs_probes = 1;
3239                 ifp->idev->if_flags |= IF_RS_SENT;
3240                 addrconf_mod_rs_timer(ifp->idev,
3241                                       ifp->idev->cnf.rtr_solicit_interval);
3242                 spin_unlock(&ifp->lock);
3243                 write_unlock_bh(&ifp->idev->lock);
3244         }
3245 }
3246
3247 static void addrconf_dad_run(struct inet6_dev *idev)
3248 {
3249         struct inet6_ifaddr *ifp;
3250
3251         read_lock_bh(&idev->lock);
3252         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3253                 spin_lock(&ifp->lock);
3254                 if (ifp->flags & IFA_F_TENTATIVE &&
3255                     ifp->state == INET6_IFADDR_STATE_DAD)
3256                         addrconf_dad_kick(ifp);
3257                 spin_unlock(&ifp->lock);
3258         }
3259         read_unlock_bh(&idev->lock);
3260 }
3261
3262 #ifdef CONFIG_PROC_FS
3263 struct if6_iter_state {
3264         struct seq_net_private p;
3265         int bucket;
3266         int offset;
3267 };
3268
3269 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3270 {
3271         struct inet6_ifaddr *ifa = NULL;
3272         struct if6_iter_state *state = seq->private;
3273         struct net *net = seq_file_net(seq);
3274         int p = 0;
3275
3276         /* initial bucket if pos is 0 */
3277         if (pos == 0) {
3278                 state->bucket = 0;
3279                 state->offset = 0;
3280         }
3281
3282         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3283                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3284                                          addr_lst) {
3285                         if (!net_eq(dev_net(ifa->idev->dev), net))
3286                                 continue;
3287                         /* sync with offset */
3288                         if (p < state->offset) {
3289                                 p++;
3290                                 continue;
3291                         }
3292                         state->offset++;
3293                         return ifa;
3294                 }
3295
3296                 /* prepare for next bucket */
3297                 state->offset = 0;
3298                 p = 0;
3299         }
3300         return NULL;
3301 }
3302
3303 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3304                                          struct inet6_ifaddr *ifa)
3305 {
3306         struct if6_iter_state *state = seq->private;
3307         struct net *net = seq_file_net(seq);
3308
3309         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3310                 if (!net_eq(dev_net(ifa->idev->dev), net))
3311                         continue;
3312                 state->offset++;
3313                 return ifa;
3314         }
3315
3316         while (++state->bucket < IN6_ADDR_HSIZE) {
3317                 state->offset = 0;
3318                 hlist_for_each_entry_rcu_bh(ifa,
3319                                      &inet6_addr_lst[state->bucket], addr_lst) {
3320                         if (!net_eq(dev_net(ifa->idev->dev), net))
3321                                 continue;
3322                         state->offset++;
3323                         return ifa;
3324                 }
3325         }
3326
3327         return NULL;
3328 }
3329
3330 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3331         __acquires(rcu_bh)
3332 {
3333         rcu_read_lock_bh();
3334         return if6_get_first(seq, *pos);
3335 }
3336
3337 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3338 {
3339         struct inet6_ifaddr *ifa;
3340
3341         ifa = if6_get_next(seq, v);
3342         ++*pos;
3343         return ifa;
3344 }
3345
3346 static void if6_seq_stop(struct seq_file *seq, void *v)
3347         __releases(rcu_bh)
3348 {
3349         rcu_read_unlock_bh();
3350 }
3351
3352 static int if6_seq_show(struct seq_file *seq, void *v)
3353 {
3354         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3355         seq_printf(seq, "%pi6 %02x %02x %02x %03x %8s\n",
3356                    &ifp->addr,
3357                    ifp->idev->dev->ifindex,
3358                    ifp->prefix_len,
3359                    ifp->scope,
3360                    ifp->flags,
3361                    ifp->idev->dev->name);
3362         return 0;
3363 }
3364
3365 static const struct seq_operations if6_seq_ops = {
3366         .start  = if6_seq_start,
3367         .next   = if6_seq_next,
3368         .show   = if6_seq_show,
3369         .stop   = if6_seq_stop,
3370 };
3371
3372 static int if6_seq_open(struct inode *inode, struct file *file)
3373 {
3374         return seq_open_net(inode, file, &if6_seq_ops,
3375                             sizeof(struct if6_iter_state));
3376 }
3377
3378 static const struct file_operations if6_fops = {
3379         .owner          = THIS_MODULE,
3380         .open           = if6_seq_open,
3381         .read           = seq_read,
3382         .llseek         = seq_lseek,
3383         .release        = seq_release_net,
3384 };
3385
3386 static int __net_init if6_proc_net_init(struct net *net)
3387 {
3388         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3389                 return -ENOMEM;
3390         return 0;
3391 }
3392
3393 static void __net_exit if6_proc_net_exit(struct net *net)
3394 {
3395         remove_proc_entry("if_inet6", net->proc_net);
3396 }
3397
3398 static struct pernet_operations if6_proc_net_ops = {
3399        .init = if6_proc_net_init,
3400        .exit = if6_proc_net_exit,
3401 };
3402
3403 int __init if6_proc_init(void)
3404 {
3405         return register_pernet_subsys(&if6_proc_net_ops);
3406 }
3407
3408 void if6_proc_exit(void)
3409 {
3410         unregister_pernet_subsys(&if6_proc_net_ops);
3411 }
3412 #endif  /* CONFIG_PROC_FS */
3413
3414 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3415 /* Check if address is a home address configured on any interface. */
3416 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3417 {
3418         int ret = 0;
3419         struct inet6_ifaddr *ifp = NULL;
3420         unsigned int hash = inet6_addr_hash(addr);
3421
3422         rcu_read_lock_bh();
3423         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3424                 if (!net_eq(dev_net(ifp->idev->dev), net))
3425                         continue;
3426                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3427                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3428                         ret = 1;
3429                         break;
3430                 }
3431         }
3432         rcu_read_unlock_bh();
3433         return ret;
3434 }
3435 #endif
3436
3437 /*
3438  *      Periodic address status verification
3439  */
3440
3441 static void addrconf_verify(unsigned long foo)
3442 {
3443         unsigned long now, next, next_sec, next_sched;
3444         struct inet6_ifaddr *ifp;
3445         int i;
3446
3447         rcu_read_lock_bh();
3448         spin_lock(&addrconf_verify_lock);
3449         now = jiffies;
3450         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3451
3452         del_timer(&addr_chk_timer);
3453
3454         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3455 restart:
3456                 hlist_for_each_entry_rcu_bh(ifp,
3457                                          &inet6_addr_lst[i], addr_lst) {
3458                         unsigned long age;
3459
3460                         if (ifp->flags & IFA_F_PERMANENT)
3461                                 continue;
3462
3463                         spin_lock(&ifp->lock);
3464                         /* We try to batch several events at once. */
3465                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3466
3467                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3468                             age >= ifp->valid_lft) {
3469                                 spin_unlock(&ifp->lock);
3470                                 in6_ifa_hold(ifp);
3471                                 ipv6_del_addr(ifp);
3472                                 goto restart;
3473                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3474                                 spin_unlock(&ifp->lock);
3475                                 continue;
3476                         } else if (age >= ifp->prefered_lft) {
3477                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3478                                 int deprecate = 0;
3479
3480                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3481                                         deprecate = 1;
3482                                         ifp->flags |= IFA_F_DEPRECATED;
3483                                 }
3484
3485                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3486                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3487
3488                                 spin_unlock(&ifp->lock);
3489
3490                                 if (deprecate) {
3491                                         in6_ifa_hold(ifp);
3492
3493                                         ipv6_ifa_notify(0, ifp);
3494                                         in6_ifa_put(ifp);
3495                                         goto restart;
3496                                 }
3497                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3498                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3499                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3500                                         ifp->idev->cnf.dad_transmits *
3501                                         ifp->idev->nd_parms->retrans_time / HZ;
3502
3503                                 if (age >= ifp->prefered_lft - regen_advance) {
3504                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3505                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3506                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3507                                         if (!ifp->regen_count && ifpub) {
3508                                                 ifp->regen_count++;
3509                                                 in6_ifa_hold(ifp);
3510                                                 in6_ifa_hold(ifpub);
3511                                                 spin_unlock(&ifp->lock);
3512
3513                                                 spin_lock(&ifpub->lock);
3514                                                 ifpub->regen_count = 0;
3515                                                 spin_unlock(&ifpub->lock);
3516                                                 ipv6_create_tempaddr(ifpub, ifp);
3517                                                 in6_ifa_put(ifpub);
3518                                                 in6_ifa_put(ifp);
3519                                                 goto restart;
3520                                         }
3521                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3522                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3523                                 spin_unlock(&ifp->lock);
3524                         } else {
3525                                 /* ifp->prefered_lft <= ifp->valid_lft */
3526                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3527                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3528                                 spin_unlock(&ifp->lock);
3529                         }
3530                 }
3531         }
3532
3533         next_sec = round_jiffies_up(next);
3534         next_sched = next;
3535
3536         /* If rounded timeout is accurate enough, accept it. */
3537         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3538                 next_sched = next_sec;
3539
3540         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3541         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3542                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3543
3544         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3545               now, next, next_sec, next_sched);
3546
3547         addr_chk_timer.expires = next_sched;
3548         add_timer(&addr_chk_timer);
3549         spin_unlock(&addrconf_verify_lock);
3550         rcu_read_unlock_bh();
3551 }
3552
3553 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3554                                      struct in6_addr **peer_pfx)
3555 {
3556         struct in6_addr *pfx = NULL;
3557
3558         *peer_pfx = NULL;
3559
3560         if (addr)
3561                 pfx = nla_data(addr);
3562
3563         if (local) {
3564                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3565                         *peer_pfx = pfx;
3566                 pfx = nla_data(local);
3567         }
3568
3569         return pfx;
3570 }
3571
3572 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3573         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3574         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3575         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3576         [IFA_FLAGS]             = { .len = sizeof(u32) },
3577 };
3578
3579 static int
3580 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3581 {
3582         struct net *net = sock_net(skb->sk);
3583         struct ifaddrmsg *ifm;
3584         struct nlattr *tb[IFA_MAX+1];
3585         struct in6_addr *pfx, *peer_pfx;
3586         int err;
3587
3588         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3589         if (err < 0)
3590                 return err;
3591
3592         ifm = nlmsg_data(nlh);
3593         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3594         if (pfx == NULL)
3595                 return -EINVAL;
3596
3597         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3598 }
3599
3600 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3601                              u32 prefered_lft, u32 valid_lft)
3602 {
3603         u32 flags;
3604         clock_t expires;
3605         unsigned long timeout;
3606
3607         if (!valid_lft || (prefered_lft > valid_lft))
3608                 return -EINVAL;
3609
3610         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3611         if (addrconf_finite_timeout(timeout)) {
3612                 expires = jiffies_to_clock_t(timeout * HZ);
3613                 valid_lft = timeout;
3614                 flags = RTF_EXPIRES;
3615         } else {
3616                 expires = 0;
3617                 flags = 0;
3618                 ifa_flags |= IFA_F_PERMANENT;
3619         }
3620
3621         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3622         if (addrconf_finite_timeout(timeout)) {
3623                 if (timeout == 0)
3624                         ifa_flags |= IFA_F_DEPRECATED;
3625                 prefered_lft = timeout;
3626         }
3627
3628         spin_lock_bh(&ifp->lock);
3629         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3630         ifp->tstamp = jiffies;
3631         ifp->valid_lft = valid_lft;
3632         ifp->prefered_lft = prefered_lft;
3633
3634         spin_unlock_bh(&ifp->lock);
3635         if (!(ifp->flags&IFA_F_TENTATIVE))
3636                 ipv6_ifa_notify(0, ifp);
3637
3638         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3639                               expires, flags);
3640         addrconf_verify(0);
3641
3642         return 0;
3643 }
3644
3645 static int
3646 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3647 {
3648         struct net *net = sock_net(skb->sk);
3649         struct ifaddrmsg *ifm;
3650         struct nlattr *tb[IFA_MAX+1];
3651         struct in6_addr *pfx, *peer_pfx;
3652         struct inet6_ifaddr *ifa;
3653         struct net_device *dev;
3654         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3655         u32 ifa_flags;
3656         int err;
3657
3658         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3659         if (err < 0)
3660                 return err;
3661
3662         ifm = nlmsg_data(nlh);
3663         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3664         if (pfx == NULL)
3665                 return -EINVAL;
3666
3667         if (tb[IFA_CACHEINFO]) {
3668                 struct ifa_cacheinfo *ci;
3669
3670                 ci = nla_data(tb[IFA_CACHEINFO]);
3671                 valid_lft = ci->ifa_valid;
3672                 preferred_lft = ci->ifa_prefered;
3673         } else {
3674                 preferred_lft = INFINITY_LIFE_TIME;
3675                 valid_lft = INFINITY_LIFE_TIME;
3676         }
3677
3678         dev =  __dev_get_by_index(net, ifm->ifa_index);
3679         if (dev == NULL)
3680                 return -ENODEV;
3681
3682         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3683
3684         /* We ignore other flags so far. */
3685         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS;
3686
3687         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3688         if (ifa == NULL) {
3689                 /*
3690                  * It would be best to check for !NLM_F_CREATE here but
3691                  * userspace alreay relies on not having to provide this.
3692                  */
3693                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3694                                       ifm->ifa_prefixlen, ifa_flags,
3695                                       preferred_lft, valid_lft);
3696         }
3697
3698         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3699             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3700                 err = -EEXIST;
3701         else
3702                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3703
3704         in6_ifa_put(ifa);
3705
3706         return err;
3707 }
3708
3709 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3710                           u8 scope, int ifindex)
3711 {
3712         struct ifaddrmsg *ifm;
3713
3714         ifm = nlmsg_data(nlh);
3715         ifm->ifa_family = AF_INET6;
3716         ifm->ifa_prefixlen = prefixlen;
3717         ifm->ifa_flags = flags;
3718         ifm->ifa_scope = scope;
3719         ifm->ifa_index = ifindex;
3720 }
3721
3722 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3723                          unsigned long tstamp, u32 preferred, u32 valid)
3724 {
3725         struct ifa_cacheinfo ci;
3726
3727         ci.cstamp = cstamp_delta(cstamp);
3728         ci.tstamp = cstamp_delta(tstamp);
3729         ci.ifa_prefered = preferred;
3730         ci.ifa_valid = valid;
3731
3732         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3733 }
3734
3735 static inline int rt_scope(int ifa_scope)
3736 {
3737         if (ifa_scope & IFA_HOST)
3738                 return RT_SCOPE_HOST;
3739         else if (ifa_scope & IFA_LINK)
3740                 return RT_SCOPE_LINK;
3741         else if (ifa_scope & IFA_SITE)
3742                 return RT_SCOPE_SITE;
3743         else
3744                 return RT_SCOPE_UNIVERSE;
3745 }
3746
3747 static inline int inet6_ifaddr_msgsize(void)
3748 {
3749         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3750                + nla_total_size(16) /* IFA_LOCAL */
3751                + nla_total_size(16) /* IFA_ADDRESS */
3752                + nla_total_size(sizeof(struct ifa_cacheinfo))
3753                + nla_total_size(4)  /* IFA_FLAGS */;
3754 }
3755
3756 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3757                              u32 portid, u32 seq, int event, unsigned int flags)
3758 {
3759         struct nlmsghdr  *nlh;
3760         u32 preferred, valid;
3761
3762         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3763         if (nlh == NULL)
3764                 return -EMSGSIZE;
3765
3766         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3767                       ifa->idev->dev->ifindex);
3768
3769         if (!(ifa->flags&IFA_F_PERMANENT)) {
3770                 preferred = ifa->prefered_lft;
3771                 valid = ifa->valid_lft;
3772                 if (preferred != INFINITY_LIFE_TIME) {
3773                         long tval = (jiffies - ifa->tstamp)/HZ;
3774                         if (preferred > tval)
3775                                 preferred -= tval;
3776                         else
3777                                 preferred = 0;
3778                         if (valid != INFINITY_LIFE_TIME) {
3779                                 if (valid > tval)
3780                                         valid -= tval;
3781                                 else
3782                                         valid = 0;
3783                         }
3784                 }
3785         } else {
3786                 preferred = INFINITY_LIFE_TIME;
3787                 valid = INFINITY_LIFE_TIME;
3788         }
3789
3790         if (!ipv6_addr_any(&ifa->peer_addr)) {
3791                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3792                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3793                         goto error;
3794         } else
3795                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3796                         goto error;
3797
3798         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3799                 goto error;
3800
3801         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
3802                 goto error;
3803
3804         return nlmsg_end(skb, nlh);
3805
3806 error:
3807         nlmsg_cancel(skb, nlh);
3808         return -EMSGSIZE;
3809 }
3810
3811 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3812                                 u32 portid, u32 seq, int event, u16 flags)
3813 {
3814         struct nlmsghdr  *nlh;
3815         u8 scope = RT_SCOPE_UNIVERSE;
3816         int ifindex = ifmca->idev->dev->ifindex;
3817
3818         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3819                 scope = RT_SCOPE_SITE;
3820
3821         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3822         if (nlh == NULL)
3823                 return -EMSGSIZE;
3824
3825         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3826         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3827             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3828                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3829                 nlmsg_cancel(skb, nlh);
3830                 return -EMSGSIZE;
3831         }
3832
3833         return nlmsg_end(skb, nlh);
3834 }
3835
3836 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3837                                 u32 portid, u32 seq, int event, unsigned int flags)
3838 {
3839         struct nlmsghdr  *nlh;
3840         u8 scope = RT_SCOPE_UNIVERSE;
3841         int ifindex = ifaca->aca_idev->dev->ifindex;
3842
3843         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3844                 scope = RT_SCOPE_SITE;
3845
3846         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3847         if (nlh == NULL)
3848                 return -EMSGSIZE;
3849
3850         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3851         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3852             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3853                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3854                 nlmsg_cancel(skb, nlh);
3855                 return -EMSGSIZE;
3856         }
3857
3858         return nlmsg_end(skb, nlh);
3859 }
3860
3861 enum addr_type_t {
3862         UNICAST_ADDR,
3863         MULTICAST_ADDR,
3864         ANYCAST_ADDR,
3865 };
3866
3867 /* called with rcu_read_lock() */
3868 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3869                           struct netlink_callback *cb, enum addr_type_t type,
3870                           int s_ip_idx, int *p_ip_idx)
3871 {
3872         struct ifmcaddr6 *ifmca;
3873         struct ifacaddr6 *ifaca;
3874         int err = 1;
3875         int ip_idx = *p_ip_idx;
3876
3877         read_lock_bh(&idev->lock);
3878         switch (type) {
3879         case UNICAST_ADDR: {
3880                 struct inet6_ifaddr *ifa;
3881
3882                 /* unicast address incl. temp addr */
3883                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3884                         if (++ip_idx < s_ip_idx)
3885                                 continue;
3886                         err = inet6_fill_ifaddr(skb, ifa,
3887                                                 NETLINK_CB(cb->skb).portid,
3888                                                 cb->nlh->nlmsg_seq,
3889                                                 RTM_NEWADDR,
3890                                                 NLM_F_MULTI);
3891                         if (err <= 0)
3892                                 break;
3893                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3894                 }
3895                 break;
3896         }
3897         case MULTICAST_ADDR:
3898                 /* multicast address */
3899                 for (ifmca = idev->mc_list; ifmca;
3900                      ifmca = ifmca->next, ip_idx++) {
3901                         if (ip_idx < s_ip_idx)
3902                                 continue;
3903                         err = inet6_fill_ifmcaddr(skb, ifmca,
3904                                                   NETLINK_CB(cb->skb).portid,
3905                                                   cb->nlh->nlmsg_seq,
3906                                                   RTM_GETMULTICAST,
3907                                                   NLM_F_MULTI);
3908                         if (err <= 0)
3909                                 break;
3910                 }
3911                 break;
3912         case ANYCAST_ADDR:
3913                 /* anycast address */
3914                 for (ifaca = idev->ac_list; ifaca;
3915                      ifaca = ifaca->aca_next, ip_idx++) {
3916                         if (ip_idx < s_ip_idx)
3917                                 continue;
3918                         err = inet6_fill_ifacaddr(skb, ifaca,
3919                                                   NETLINK_CB(cb->skb).portid,
3920                                                   cb->nlh->nlmsg_seq,
3921                                                   RTM_GETANYCAST,
3922                                                   NLM_F_MULTI);
3923                         if (err <= 0)
3924                                 break;
3925                 }
3926                 break;
3927         default:
3928                 break;
3929         }
3930         read_unlock_bh(&idev->lock);
3931         *p_ip_idx = ip_idx;
3932         return err;
3933 }
3934
3935 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3936                            enum addr_type_t type)
3937 {
3938         struct net *net = sock_net(skb->sk);
3939         int h, s_h;
3940         int idx, ip_idx;
3941         int s_idx, s_ip_idx;
3942         struct net_device *dev;
3943         struct inet6_dev *idev;
3944         struct hlist_head *head;
3945
3946         s_h = cb->args[0];
3947         s_idx = idx = cb->args[1];
3948         s_ip_idx = ip_idx = cb->args[2];
3949
3950         rcu_read_lock();
3951         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3952         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3953                 idx = 0;
3954                 head = &net->dev_index_head[h];
3955                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3956                         if (idx < s_idx)
3957                                 goto cont;
3958                         if (h > s_h || idx > s_idx)
3959                                 s_ip_idx = 0;
3960                         ip_idx = 0;
3961                         idev = __in6_dev_get(dev);
3962                         if (!idev)
3963                                 goto cont;
3964
3965                         if (in6_dump_addrs(idev, skb, cb, type,
3966                                            s_ip_idx, &ip_idx) <= 0)
3967                                 goto done;
3968 cont:
3969                         idx++;
3970                 }
3971         }
3972 done:
3973         rcu_read_unlock();
3974         cb->args[0] = h;
3975         cb->args[1] = idx;
3976         cb->args[2] = ip_idx;
3977
3978         return skb->len;
3979 }
3980
3981 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3982 {
3983         enum addr_type_t type = UNICAST_ADDR;
3984
3985         return inet6_dump_addr(skb, cb, type);
3986 }
3987
3988 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3989 {
3990         enum addr_type_t type = MULTICAST_ADDR;
3991
3992         return inet6_dump_addr(skb, cb, type);
3993 }
3994
3995
3996 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3997 {
3998         enum addr_type_t type = ANYCAST_ADDR;
3999
4000         return inet6_dump_addr(skb, cb, type);
4001 }
4002
4003 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4004 {
4005         struct net *net = sock_net(in_skb->sk);
4006         struct ifaddrmsg *ifm;
4007         struct nlattr *tb[IFA_MAX+1];
4008         struct in6_addr *addr = NULL, *peer;
4009         struct net_device *dev = NULL;
4010         struct inet6_ifaddr *ifa;
4011         struct sk_buff *skb;
4012         int err;
4013
4014         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4015         if (err < 0)
4016                 goto errout;
4017
4018         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4019         if (addr == NULL) {
4020                 err = -EINVAL;
4021                 goto errout;
4022         }
4023
4024         ifm = nlmsg_data(nlh);
4025         if (ifm->ifa_index)
4026                 dev = __dev_get_by_index(net, ifm->ifa_index);
4027
4028         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4029         if (!ifa) {
4030                 err = -EADDRNOTAVAIL;
4031                 goto errout;
4032         }
4033
4034         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4035         if (!skb) {
4036                 err = -ENOBUFS;
4037                 goto errout_ifa;
4038         }
4039
4040         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4041                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4042         if (err < 0) {
4043                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4044                 WARN_ON(err == -EMSGSIZE);
4045                 kfree_skb(skb);
4046                 goto errout_ifa;
4047         }
4048         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4049 errout_ifa:
4050         in6_ifa_put(ifa);
4051 errout:
4052         return err;
4053 }
4054
4055 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4056 {
4057         struct sk_buff *skb;
4058         struct net *net = dev_net(ifa->idev->dev);
4059         int err = -ENOBUFS;
4060
4061         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4062         if (skb == NULL)
4063                 goto errout;
4064
4065         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4066         if (err < 0) {
4067                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4068                 WARN_ON(err == -EMSGSIZE);
4069                 kfree_skb(skb);
4070                 goto errout;
4071         }
4072         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4073         return;
4074 errout:
4075         if (err < 0)
4076                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4077 }
4078
4079 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4080                                 __s32 *array, int bytes)
4081 {
4082         BUG_ON(bytes < (DEVCONF_MAX * 4));
4083
4084         memset(array, 0, bytes);
4085         array[DEVCONF_FORWARDING] = cnf->forwarding;
4086         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4087         array[DEVCONF_MTU6] = cnf->mtu6;
4088         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4089         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4090         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4091         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4092         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4093         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4094                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4095         array[DEVCONF_RTR_SOLICIT_DELAY] =
4096                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4097         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4098         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4099                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4100         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4101                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4102         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4103         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4104         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4105         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4106         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4107         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4108         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4109         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4110 #ifdef CONFIG_IPV6_ROUTER_PREF
4111         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4112         array[DEVCONF_RTR_PROBE_INTERVAL] =
4113                 jiffies_to_msecs(cnf->rtr_probe_interval);
4114 #ifdef CONFIG_IPV6_ROUTE_INFO
4115         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4116 #endif
4117 #endif
4118         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4119         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4120 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4121         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4122 #endif
4123 #ifdef CONFIG_IPV6_MROUTE
4124         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4125 #endif
4126         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4127         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4128         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4129         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4130         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4131 }
4132
4133 static inline size_t inet6_ifla6_size(void)
4134 {
4135         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4136              + nla_total_size(sizeof(struct ifla_cacheinfo))
4137              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4138              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4139              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4140              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4141 }
4142
4143 static inline size_t inet6_if_nlmsg_size(void)
4144 {
4145         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4146                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4147                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4148                + nla_total_size(4) /* IFLA_MTU */
4149                + nla_total_size(4) /* IFLA_LINK */
4150                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4151 }
4152
4153 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4154                                       int items, int bytes)
4155 {
4156         int i;
4157         int pad = bytes - sizeof(u64) * items;
4158         BUG_ON(pad < 0);
4159
4160         /* Use put_unaligned() because stats may not be aligned for u64. */
4161         put_unaligned(items, &stats[0]);
4162         for (i = 1; i < items; i++)
4163                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4164
4165         memset(&stats[items], 0, pad);
4166 }
4167
4168 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4169                                       int items, int bytes, size_t syncpoff)
4170 {
4171         int i;
4172         int pad = bytes - sizeof(u64) * items;
4173         BUG_ON(pad < 0);
4174
4175         /* Use put_unaligned() because stats may not be aligned for u64. */
4176         put_unaligned(items, &stats[0]);
4177         for (i = 1; i < items; i++)
4178                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4179
4180         memset(&stats[items], 0, pad);
4181 }
4182
4183 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4184                              int bytes)
4185 {
4186         switch (attrtype) {
4187         case IFLA_INET6_STATS:
4188                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4189                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4190                 break;
4191         case IFLA_INET6_ICMP6STATS:
4192                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4193                 break;
4194         }
4195 }
4196
4197 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4198 {
4199         struct nlattr *nla;
4200         struct ifla_cacheinfo ci;
4201
4202         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4203                 goto nla_put_failure;
4204         ci.max_reasm_len = IPV6_MAXPLEN;
4205         ci.tstamp = cstamp_delta(idev->tstamp);
4206         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4207         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4208         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4209                 goto nla_put_failure;
4210         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4211         if (nla == NULL)
4212                 goto nla_put_failure;
4213         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4214
4215         /* XXX - MC not implemented */
4216
4217         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4218         if (nla == NULL)
4219                 goto nla_put_failure;
4220         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4221
4222         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4223         if (nla == NULL)
4224                 goto nla_put_failure;
4225         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4226
4227         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4228         if (nla == NULL)
4229                 goto nla_put_failure;
4230         read_lock_bh(&idev->lock);
4231         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4232         read_unlock_bh(&idev->lock);
4233
4234         return 0;
4235
4236 nla_put_failure:
4237         return -EMSGSIZE;
4238 }
4239
4240 static size_t inet6_get_link_af_size(const struct net_device *dev)
4241 {
4242         if (!__in6_dev_get(dev))
4243                 return 0;
4244
4245         return inet6_ifla6_size();
4246 }
4247
4248 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4249 {
4250         struct inet6_dev *idev = __in6_dev_get(dev);
4251
4252         if (!idev)
4253                 return -ENODATA;
4254
4255         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4256                 return -EMSGSIZE;
4257
4258         return 0;
4259 }
4260
4261 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4262 {
4263         struct inet6_ifaddr *ifp;
4264         struct net_device *dev = idev->dev;
4265         bool update_rs = false;
4266         struct in6_addr ll_addr;
4267
4268         if (token == NULL)
4269                 return -EINVAL;
4270         if (ipv6_addr_any(token))
4271                 return -EINVAL;
4272         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4273                 return -EINVAL;
4274         if (!ipv6_accept_ra(idev))
4275                 return -EINVAL;
4276         if (idev->cnf.rtr_solicits <= 0)
4277                 return -EINVAL;
4278
4279         write_lock_bh(&idev->lock);
4280
4281         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4282         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4283
4284         write_unlock_bh(&idev->lock);
4285
4286         if (!idev->dead && (idev->if_flags & IF_READY) &&
4287             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4288                              IFA_F_OPTIMISTIC)) {
4289
4290                 /* If we're not ready, then normal ifup will take care
4291                  * of this. Otherwise, we need to request our rs here.
4292                  */
4293                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4294                 update_rs = true;
4295         }
4296
4297         write_lock_bh(&idev->lock);
4298
4299         if (update_rs) {
4300                 idev->if_flags |= IF_RS_SENT;
4301                 idev->rs_probes = 1;
4302                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4303         }
4304
4305         /* Well, that's kinda nasty ... */
4306         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4307                 spin_lock(&ifp->lock);
4308                 if (ifp->tokenized) {
4309                         ifp->valid_lft = 0;
4310                         ifp->prefered_lft = 0;
4311                 }
4312                 spin_unlock(&ifp->lock);
4313         }
4314
4315         write_unlock_bh(&idev->lock);
4316         addrconf_verify(0);
4317         return 0;
4318 }
4319
4320 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4321 {
4322         int err = -EINVAL;
4323         struct inet6_dev *idev = __in6_dev_get(dev);
4324         struct nlattr *tb[IFLA_INET6_MAX + 1];
4325
4326         if (!idev)
4327                 return -EAFNOSUPPORT;
4328
4329         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4330                 BUG();
4331
4332         if (tb[IFLA_INET6_TOKEN])
4333                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4334
4335         return err;
4336 }
4337
4338 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4339                              u32 portid, u32 seq, int event, unsigned int flags)
4340 {
4341         struct net_device *dev = idev->dev;
4342         struct ifinfomsg *hdr;
4343         struct nlmsghdr *nlh;
4344         void *protoinfo;
4345
4346         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4347         if (nlh == NULL)
4348                 return -EMSGSIZE;
4349
4350         hdr = nlmsg_data(nlh);
4351         hdr->ifi_family = AF_INET6;
4352         hdr->__ifi_pad = 0;
4353         hdr->ifi_type = dev->type;
4354         hdr->ifi_index = dev->ifindex;
4355         hdr->ifi_flags = dev_get_flags(dev);
4356         hdr->ifi_change = 0;
4357
4358         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4359             (dev->addr_len &&
4360              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4361             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4362             (dev->ifindex != dev->iflink &&
4363              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4364                 goto nla_put_failure;
4365         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4366         if (protoinfo == NULL)
4367                 goto nla_put_failure;
4368
4369         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4370                 goto nla_put_failure;
4371
4372         nla_nest_end(skb, protoinfo);
4373         return nlmsg_end(skb, nlh);
4374
4375 nla_put_failure:
4376         nlmsg_cancel(skb, nlh);
4377         return -EMSGSIZE;
4378 }
4379
4380 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4381 {
4382         struct net *net = sock_net(skb->sk);
4383         int h, s_h;
4384         int idx = 0, s_idx;
4385         struct net_device *dev;
4386         struct inet6_dev *idev;
4387         struct hlist_head *head;
4388
4389         s_h = cb->args[0];
4390         s_idx = cb->args[1];
4391
4392         rcu_read_lock();
4393         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4394                 idx = 0;
4395                 head = &net->dev_index_head[h];
4396                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4397                         if (idx < s_idx)
4398                                 goto cont;
4399                         idev = __in6_dev_get(dev);
4400                         if (!idev)
4401                                 goto cont;
4402                         if (inet6_fill_ifinfo(skb, idev,
4403                                               NETLINK_CB(cb->skb).portid,
4404                                               cb->nlh->nlmsg_seq,
4405                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4406                                 goto out;
4407 cont:
4408                         idx++;
4409                 }
4410         }
4411 out:
4412         rcu_read_unlock();
4413         cb->args[1] = idx;
4414         cb->args[0] = h;
4415
4416         return skb->len;
4417 }
4418
4419 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4420 {
4421         struct sk_buff *skb;
4422         struct net *net = dev_net(idev->dev);
4423         int err = -ENOBUFS;
4424
4425         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4426         if (skb == NULL)
4427                 goto errout;
4428
4429         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4430         if (err < 0) {
4431                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4432                 WARN_ON(err == -EMSGSIZE);
4433                 kfree_skb(skb);
4434                 goto errout;
4435         }
4436         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4437         return;
4438 errout:
4439         if (err < 0)
4440                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4441 }
4442
4443 static inline size_t inet6_prefix_nlmsg_size(void)
4444 {
4445         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4446                + nla_total_size(sizeof(struct in6_addr))
4447                + nla_total_size(sizeof(struct prefix_cacheinfo));
4448 }
4449
4450 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4451                              struct prefix_info *pinfo, u32 portid, u32 seq,
4452                              int event, unsigned int flags)
4453 {
4454         struct prefixmsg *pmsg;
4455         struct nlmsghdr *nlh;
4456         struct prefix_cacheinfo ci;
4457
4458         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4459         if (nlh == NULL)
4460                 return -EMSGSIZE;
4461
4462         pmsg = nlmsg_data(nlh);
4463         pmsg->prefix_family = AF_INET6;
4464         pmsg->prefix_pad1 = 0;
4465         pmsg->prefix_pad2 = 0;
4466         pmsg->prefix_ifindex = idev->dev->ifindex;
4467         pmsg->prefix_len = pinfo->prefix_len;
4468         pmsg->prefix_type = pinfo->type;
4469         pmsg->prefix_pad3 = 0;
4470         pmsg->prefix_flags = 0;
4471         if (pinfo->onlink)
4472                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4473         if (pinfo->autoconf)
4474                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4475
4476         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4477                 goto nla_put_failure;
4478         ci.preferred_time = ntohl(pinfo->prefered);
4479         ci.valid_time = ntohl(pinfo->valid);
4480         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4481                 goto nla_put_failure;
4482         return nlmsg_end(skb, nlh);
4483
4484 nla_put_failure:
4485         nlmsg_cancel(skb, nlh);
4486         return -EMSGSIZE;
4487 }
4488
4489 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4490                          struct prefix_info *pinfo)
4491 {
4492         struct sk_buff *skb;
4493         struct net *net = dev_net(idev->dev);
4494         int err = -ENOBUFS;
4495
4496         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4497         if (skb == NULL)
4498                 goto errout;
4499
4500         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4501         if (err < 0) {
4502                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4503                 WARN_ON(err == -EMSGSIZE);
4504                 kfree_skb(skb);
4505                 goto errout;
4506         }
4507         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4508         return;
4509 errout:
4510         if (err < 0)
4511                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4512 }
4513
4514 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4515 {
4516         write_lock_bh(&ifp->idev->lock);
4517         spin_lock(&ifp->lock);
4518         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4519                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4520             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4521                 ifp->idev->valid_ll_addr_cnt += count;
4522         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4523         spin_unlock(&ifp->lock);
4524         write_unlock_bh(&ifp->idev->lock);
4525 }
4526
4527 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4528 {
4529         struct net *net = dev_net(ifp->idev->dev);
4530
4531         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4532
4533         switch (event) {
4534         case RTM_NEWADDR:
4535                 update_valid_ll_addr_cnt(ifp, 1);
4536
4537                 /*
4538                  * If the address was optimistic
4539                  * we inserted the route at the start of
4540                  * our DAD process, so we don't need
4541                  * to do it again
4542                  */
4543                 if (!(ifp->rt->rt6i_node))
4544                         ip6_ins_rt(ifp->rt);
4545                 if (ifp->idev->cnf.forwarding)
4546                         addrconf_join_anycast(ifp);
4547                 if (!ipv6_addr_any(&ifp->peer_addr))
4548                         addrconf_prefix_route(&ifp->peer_addr, 128,
4549                                               ifp->idev->dev, 0, 0);
4550                 break;
4551         case RTM_DELADDR:
4552                 update_valid_ll_addr_cnt(ifp, -1);
4553
4554                 if (ifp->idev->cnf.forwarding)
4555                         addrconf_leave_anycast(ifp);
4556                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4557                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4558                         struct rt6_info *rt;
4559                         struct net_device *dev = ifp->idev->dev;
4560
4561                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4562                                         dev->ifindex, 1);
4563                         if (rt) {
4564                                 dst_hold(&rt->dst);
4565                                 if (ip6_del_rt(rt))
4566                                         dst_free(&rt->dst);
4567                         }
4568                 }
4569                 dst_hold(&ifp->rt->dst);
4570
4571                 if (ip6_del_rt(ifp->rt))
4572                         dst_free(&ifp->rt->dst);
4573                 break;
4574         }
4575         atomic_inc(&net->ipv6.dev_addr_genid);
4576         rt_genid_bump_ipv6(net);
4577 }
4578
4579 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4580 {
4581         rcu_read_lock_bh();
4582         if (likely(ifp->idev->dead == 0))
4583                 __ipv6_ifa_notify(event, ifp);
4584         rcu_read_unlock_bh();
4585 }
4586
4587 #ifdef CONFIG_SYSCTL
4588
4589 static
4590 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4591                            void __user *buffer, size_t *lenp, loff_t *ppos)
4592 {
4593         int *valp = ctl->data;
4594         int val = *valp;
4595         loff_t pos = *ppos;
4596         struct ctl_table lctl;
4597         int ret;
4598
4599         /*
4600          * ctl->data points to idev->cnf.forwarding, we should
4601          * not modify it until we get the rtnl lock.
4602          */
4603         lctl = *ctl;
4604         lctl.data = &val;
4605
4606         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4607
4608         if (write)
4609                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4610         if (ret)
4611                 *ppos = pos;
4612         return ret;
4613 }
4614
4615 static void dev_disable_change(struct inet6_dev *idev)
4616 {
4617         struct netdev_notifier_info info;
4618
4619         if (!idev || !idev->dev)
4620                 return;
4621
4622         netdev_notifier_info_init(&info, idev->dev);
4623         if (idev->cnf.disable_ipv6)
4624                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4625         else
4626                 addrconf_notify(NULL, NETDEV_UP, &info);
4627 }
4628
4629 static void addrconf_disable_change(struct net *net, __s32 newf)
4630 {
4631         struct net_device *dev;
4632         struct inet6_dev *idev;
4633
4634         rcu_read_lock();
4635         for_each_netdev_rcu(net, dev) {
4636                 idev = __in6_dev_get(dev);
4637                 if (idev) {
4638                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4639                         idev->cnf.disable_ipv6 = newf;
4640                         if (changed)
4641                                 dev_disable_change(idev);
4642                 }
4643         }
4644         rcu_read_unlock();
4645 }
4646
4647 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4648 {
4649         struct net *net;
4650         int old;
4651
4652         if (!rtnl_trylock())
4653                 return restart_syscall();
4654
4655         net = (struct net *)table->extra2;
4656         old = *p;
4657         *p = newf;
4658
4659         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4660                 rtnl_unlock();
4661                 return 0;
4662         }
4663
4664         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4665                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4666                 addrconf_disable_change(net, newf);
4667         } else if ((!newf) ^ (!old))
4668                 dev_disable_change((struct inet6_dev *)table->extra1);
4669
4670         rtnl_unlock();
4671         return 0;
4672 }
4673
4674 static
4675 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4676                             void __user *buffer, size_t *lenp, loff_t *ppos)
4677 {
4678         int *valp = ctl->data;
4679         int val = *valp;
4680         loff_t pos = *ppos;
4681         struct ctl_table lctl;
4682         int ret;
4683
4684         /*
4685          * ctl->data points to idev->cnf.disable_ipv6, we should
4686          * not modify it until we get the rtnl lock.
4687          */
4688         lctl = *ctl;
4689         lctl.data = &val;
4690
4691         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4692
4693         if (write)
4694                 ret = addrconf_disable_ipv6(ctl, valp, val);
4695         if (ret)
4696                 *ppos = pos;
4697         return ret;
4698 }
4699
4700 static struct addrconf_sysctl_table
4701 {
4702         struct ctl_table_header *sysctl_header;
4703         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4704 } addrconf_sysctl __read_mostly = {
4705         .sysctl_header = NULL,
4706         .addrconf_vars = {
4707                 {
4708                         .procname       = "forwarding",
4709                         .data           = &ipv6_devconf.forwarding,
4710                         .maxlen         = sizeof(int),
4711                         .mode           = 0644,
4712                         .proc_handler   = addrconf_sysctl_forward,
4713                 },
4714                 {
4715                         .procname       = "hop_limit",
4716                         .data           = &ipv6_devconf.hop_limit,
4717                         .maxlen         = sizeof(int),
4718                         .mode           = 0644,
4719                         .proc_handler   = proc_dointvec,
4720                 },
4721                 {
4722                         .procname       = "mtu",
4723                         .data           = &ipv6_devconf.mtu6,
4724                         .maxlen         = sizeof(int),
4725                         .mode           = 0644,
4726                         .proc_handler   = proc_dointvec,
4727                 },
4728                 {
4729                         .procname       = "accept_ra",
4730                         .data           = &ipv6_devconf.accept_ra,
4731                         .maxlen         = sizeof(int),
4732                         .mode           = 0644,
4733                         .proc_handler   = proc_dointvec,
4734                 },
4735                 {
4736                         .procname       = "accept_redirects",
4737                         .data           = &ipv6_devconf.accept_redirects,
4738                         .maxlen         = sizeof(int),
4739                         .mode           = 0644,
4740                         .proc_handler   = proc_dointvec,
4741                 },
4742                 {
4743                         .procname       = "autoconf",
4744                         .data           = &ipv6_devconf.autoconf,
4745                         .maxlen         = sizeof(int),
4746                         .mode           = 0644,
4747                         .proc_handler   = proc_dointvec,
4748                 },
4749                 {
4750                         .procname       = "dad_transmits",
4751                         .data           = &ipv6_devconf.dad_transmits,
4752                         .maxlen         = sizeof(int),
4753                         .mode           = 0644,
4754                         .proc_handler   = proc_dointvec,
4755                 },
4756                 {
4757                         .procname       = "router_solicitations",
4758                         .data           = &ipv6_devconf.rtr_solicits,
4759                         .maxlen         = sizeof(int),
4760                         .mode           = 0644,
4761                         .proc_handler   = proc_dointvec,
4762                 },
4763                 {
4764                         .procname       = "router_solicitation_interval",
4765                         .data           = &ipv6_devconf.rtr_solicit_interval,
4766                         .maxlen         = sizeof(int),
4767                         .mode           = 0644,
4768                         .proc_handler   = proc_dointvec_jiffies,
4769                 },
4770                 {
4771                         .procname       = "router_solicitation_delay",
4772                         .data           = &ipv6_devconf.rtr_solicit_delay,
4773                         .maxlen         = sizeof(int),
4774                         .mode           = 0644,
4775                         .proc_handler   = proc_dointvec_jiffies,
4776                 },
4777                 {
4778                         .procname       = "force_mld_version",
4779                         .data           = &ipv6_devconf.force_mld_version,
4780                         .maxlen         = sizeof(int),
4781                         .mode           = 0644,
4782                         .proc_handler   = proc_dointvec,
4783                 },
4784                 {
4785                         .procname       = "mldv1_unsolicited_report_interval",
4786                         .data           =
4787                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4788                         .maxlen         = sizeof(int),
4789                         .mode           = 0644,
4790                         .proc_handler   = proc_dointvec_ms_jiffies,
4791                 },
4792                 {
4793                         .procname       = "mldv2_unsolicited_report_interval",
4794                         .data           =
4795                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4796                         .maxlen         = sizeof(int),
4797                         .mode           = 0644,
4798                         .proc_handler   = proc_dointvec_ms_jiffies,
4799                 },
4800                 {
4801                         .procname       = "use_tempaddr",
4802                         .data           = &ipv6_devconf.use_tempaddr,
4803                         .maxlen         = sizeof(int),
4804                         .mode           = 0644,
4805                         .proc_handler   = proc_dointvec,
4806                 },
4807                 {
4808                         .procname       = "temp_valid_lft",
4809                         .data           = &ipv6_devconf.temp_valid_lft,
4810                         .maxlen         = sizeof(int),
4811                         .mode           = 0644,
4812                         .proc_handler   = proc_dointvec,
4813                 },
4814                 {
4815                         .procname       = "temp_prefered_lft",
4816                         .data           = &ipv6_devconf.temp_prefered_lft,
4817                         .maxlen         = sizeof(int),
4818                         .mode           = 0644,
4819                         .proc_handler   = proc_dointvec,
4820                 },
4821                 {
4822                         .procname       = "regen_max_retry",
4823                         .data           = &ipv6_devconf.regen_max_retry,
4824                         .maxlen         = sizeof(int),
4825                         .mode           = 0644,
4826                         .proc_handler   = proc_dointvec,
4827                 },
4828                 {
4829                         .procname       = "max_desync_factor",
4830                         .data           = &ipv6_devconf.max_desync_factor,
4831                         .maxlen         = sizeof(int),
4832                         .mode           = 0644,
4833                         .proc_handler   = proc_dointvec,
4834                 },
4835                 {
4836                         .procname       = "max_addresses",
4837                         .data           = &ipv6_devconf.max_addresses,
4838                         .maxlen         = sizeof(int),
4839                         .mode           = 0644,
4840                         .proc_handler   = proc_dointvec,
4841                 },
4842                 {
4843                         .procname       = "accept_ra_defrtr",
4844                         .data           = &ipv6_devconf.accept_ra_defrtr,
4845                         .maxlen         = sizeof(int),
4846                         .mode           = 0644,
4847                         .proc_handler   = proc_dointvec,
4848                 },
4849                 {
4850                         .procname       = "accept_ra_pinfo",
4851                         .data           = &ipv6_devconf.accept_ra_pinfo,
4852                         .maxlen         = sizeof(int),
4853                         .mode           = 0644,
4854                         .proc_handler   = proc_dointvec,
4855                 },
4856 #ifdef CONFIG_IPV6_ROUTER_PREF
4857                 {
4858                         .procname       = "accept_ra_rtr_pref",
4859                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4860                         .maxlen         = sizeof(int),
4861                         .mode           = 0644,
4862                         .proc_handler   = proc_dointvec,
4863                 },
4864                 {
4865                         .procname       = "router_probe_interval",
4866                         .data           = &ipv6_devconf.rtr_probe_interval,
4867                         .maxlen         = sizeof(int),
4868                         .mode           = 0644,
4869                         .proc_handler   = proc_dointvec_jiffies,
4870                 },
4871 #ifdef CONFIG_IPV6_ROUTE_INFO
4872                 {
4873                         .procname       = "accept_ra_rt_info_max_plen",
4874                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4875                         .maxlen         = sizeof(int),
4876                         .mode           = 0644,
4877                         .proc_handler   = proc_dointvec,
4878                 },
4879 #endif
4880 #endif
4881                 {
4882                         .procname       = "proxy_ndp",
4883                         .data           = &ipv6_devconf.proxy_ndp,
4884                         .maxlen         = sizeof(int),
4885                         .mode           = 0644,
4886                         .proc_handler   = proc_dointvec,
4887                 },
4888                 {
4889                         .procname       = "accept_source_route",
4890                         .data           = &ipv6_devconf.accept_source_route,
4891                         .maxlen         = sizeof(int),
4892                         .mode           = 0644,
4893                         .proc_handler   = proc_dointvec,
4894                 },
4895 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4896                 {
4897                         .procname       = "optimistic_dad",
4898                         .data           = &ipv6_devconf.optimistic_dad,
4899                         .maxlen         = sizeof(int),
4900                         .mode           = 0644,
4901                         .proc_handler   = proc_dointvec,
4902
4903                 },
4904 #endif
4905 #ifdef CONFIG_IPV6_MROUTE
4906                 {
4907                         .procname       = "mc_forwarding",
4908                         .data           = &ipv6_devconf.mc_forwarding,
4909                         .maxlen         = sizeof(int),
4910                         .mode           = 0444,
4911                         .proc_handler   = proc_dointvec,
4912                 },
4913 #endif
4914                 {
4915                         .procname       = "disable_ipv6",
4916                         .data           = &ipv6_devconf.disable_ipv6,
4917                         .maxlen         = sizeof(int),
4918                         .mode           = 0644,
4919                         .proc_handler   = addrconf_sysctl_disable,
4920                 },
4921                 {
4922                         .procname       = "accept_dad",
4923                         .data           = &ipv6_devconf.accept_dad,
4924                         .maxlen         = sizeof(int),
4925                         .mode           = 0644,
4926                         .proc_handler   = proc_dointvec,
4927                 },
4928                 {
4929                         .procname       = "force_tllao",
4930                         .data           = &ipv6_devconf.force_tllao,
4931                         .maxlen         = sizeof(int),
4932                         .mode           = 0644,
4933                         .proc_handler   = proc_dointvec
4934                 },
4935                 {
4936                         .procname       = "ndisc_notify",
4937                         .data           = &ipv6_devconf.ndisc_notify,
4938                         .maxlen         = sizeof(int),
4939                         .mode           = 0644,
4940                         .proc_handler   = proc_dointvec
4941                 },
4942                 {
4943                         .procname       = "suppress_frag_ndisc",
4944                         .data           = &ipv6_devconf.suppress_frag_ndisc,
4945                         .maxlen         = sizeof(int),
4946                         .mode           = 0644,
4947                         .proc_handler   = proc_dointvec
4948                 },
4949                 {
4950                         /* sentinel */
4951                 }
4952         },
4953 };
4954
4955 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4956                 struct inet6_dev *idev, struct ipv6_devconf *p)
4957 {
4958         int i;
4959         struct addrconf_sysctl_table *t;
4960         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4961
4962         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4963         if (t == NULL)
4964                 goto out;
4965
4966         for (i = 0; t->addrconf_vars[i].data; i++) {
4967                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4968                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4969                 t->addrconf_vars[i].extra2 = net;
4970         }
4971
4972         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4973
4974         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4975         if (t->sysctl_header == NULL)
4976                 goto free;
4977
4978         p->sysctl = t;
4979         return 0;
4980
4981 free:
4982         kfree(t);
4983 out:
4984         return -ENOBUFS;
4985 }
4986
4987 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4988 {
4989         struct addrconf_sysctl_table *t;
4990
4991         if (p->sysctl == NULL)
4992                 return;
4993
4994         t = p->sysctl;
4995         p->sysctl = NULL;
4996         unregister_net_sysctl_table(t->sysctl_header);
4997         kfree(t);
4998 }
4999
5000 static void addrconf_sysctl_register(struct inet6_dev *idev)
5001 {
5002         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5003                               &ndisc_ifinfo_sysctl_change);
5004         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5005                                         idev, &idev->cnf);
5006 }
5007
5008 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5009 {
5010         __addrconf_sysctl_unregister(&idev->cnf);
5011         neigh_sysctl_unregister(idev->nd_parms);
5012 }
5013
5014
5015 #endif
5016
5017 static int __net_init addrconf_init_net(struct net *net)
5018 {
5019         int err = -ENOMEM;
5020         struct ipv6_devconf *all, *dflt;
5021
5022         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5023         if (all == NULL)
5024                 goto err_alloc_all;
5025
5026         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5027         if (dflt == NULL)
5028                 goto err_alloc_dflt;
5029
5030         /* these will be inherited by all namespaces */
5031         dflt->autoconf = ipv6_defaults.autoconf;
5032         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5033
5034         net->ipv6.devconf_all = all;
5035         net->ipv6.devconf_dflt = dflt;
5036
5037 #ifdef CONFIG_SYSCTL
5038         err = __addrconf_sysctl_register(net, "all", NULL, all);
5039         if (err < 0)
5040                 goto err_reg_all;
5041
5042         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5043         if (err < 0)
5044                 goto err_reg_dflt;
5045 #endif
5046         return 0;
5047
5048 #ifdef CONFIG_SYSCTL
5049 err_reg_dflt:
5050         __addrconf_sysctl_unregister(all);
5051 err_reg_all:
5052         kfree(dflt);
5053 #endif
5054 err_alloc_dflt:
5055         kfree(all);
5056 err_alloc_all:
5057         return err;
5058 }
5059
5060 static void __net_exit addrconf_exit_net(struct net *net)
5061 {
5062 #ifdef CONFIG_SYSCTL
5063         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5064         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5065 #endif
5066         if (!net_eq(net, &init_net)) {
5067                 kfree(net->ipv6.devconf_dflt);
5068                 kfree(net->ipv6.devconf_all);
5069         }
5070 }
5071
5072 static struct pernet_operations addrconf_ops = {
5073         .init = addrconf_init_net,
5074         .exit = addrconf_exit_net,
5075 };
5076
5077 static struct rtnl_af_ops inet6_ops = {
5078         .family           = AF_INET6,
5079         .fill_link_af     = inet6_fill_link_af,
5080         .get_link_af_size = inet6_get_link_af_size,
5081         .set_link_af      = inet6_set_link_af,
5082 };
5083
5084 /*
5085  *      Init / cleanup code
5086  */
5087
5088 int __init addrconf_init(void)
5089 {
5090         int i, err;
5091
5092         err = ipv6_addr_label_init();
5093         if (err < 0) {
5094                 pr_crit("%s: cannot initialize default policy table: %d\n",
5095                         __func__, err);
5096                 goto out;
5097         }
5098
5099         err = register_pernet_subsys(&addrconf_ops);
5100         if (err < 0)
5101                 goto out_addrlabel;
5102
5103         /* The addrconf netdev notifier requires that loopback_dev
5104          * has it's ipv6 private information allocated and setup
5105          * before it can bring up and give link-local addresses
5106          * to other devices which are up.
5107          *
5108          * Unfortunately, loopback_dev is not necessarily the first
5109          * entry in the global dev_base list of net devices.  In fact,
5110          * it is likely to be the very last entry on that list.
5111          * So this causes the notifier registry below to try and
5112          * give link-local addresses to all devices besides loopback_dev
5113          * first, then loopback_dev, which cases all the non-loopback_dev
5114          * devices to fail to get a link-local address.
5115          *
5116          * So, as a temporary fix, allocate the ipv6 structure for
5117          * loopback_dev first by hand.
5118          * Longer term, all of the dependencies ipv6 has upon the loopback
5119          * device and it being up should be removed.
5120          */
5121         rtnl_lock();
5122         if (!ipv6_add_dev(init_net.loopback_dev))
5123                 err = -ENOMEM;
5124         rtnl_unlock();
5125         if (err)
5126                 goto errlo;
5127
5128         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5129                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5130
5131         register_netdevice_notifier(&ipv6_dev_notf);
5132
5133         addrconf_verify(0);
5134
5135         err = rtnl_af_register(&inet6_ops);
5136         if (err < 0)
5137                 goto errout_af;
5138
5139         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5140                               NULL);
5141         if (err < 0)
5142                 goto errout;
5143
5144         /* Only the first call to __rtnl_register can fail */
5145         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5146         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5147         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5148                         inet6_dump_ifaddr, NULL);
5149         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5150                         inet6_dump_ifmcaddr, NULL);
5151         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5152                         inet6_dump_ifacaddr, NULL);
5153         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5154                         inet6_netconf_dump_devconf, NULL);
5155
5156         ipv6_addr_label_rtnl_register();
5157
5158         return 0;
5159 errout:
5160         rtnl_af_unregister(&inet6_ops);
5161 errout_af:
5162         unregister_netdevice_notifier(&ipv6_dev_notf);
5163 errlo:
5164         unregister_pernet_subsys(&addrconf_ops);
5165 out_addrlabel:
5166         ipv6_addr_label_cleanup();
5167 out:
5168         return err;
5169 }
5170
5171 void addrconf_cleanup(void)
5172 {
5173         struct net_device *dev;
5174         int i;
5175
5176         unregister_netdevice_notifier(&ipv6_dev_notf);
5177         unregister_pernet_subsys(&addrconf_ops);
5178         ipv6_addr_label_cleanup();
5179
5180         rtnl_lock();
5181
5182         __rtnl_af_unregister(&inet6_ops);
5183
5184         /* clean dev list */
5185         for_each_netdev(&init_net, dev) {
5186                 if (__in6_dev_get(dev) == NULL)
5187                         continue;
5188                 addrconf_ifdown(dev, 1);
5189         }
5190         addrconf_ifdown(init_net.loopback_dev, 2);
5191
5192         /*
5193          *      Check hash table.
5194          */
5195         spin_lock_bh(&addrconf_hash_lock);
5196         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5197                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5198         spin_unlock_bh(&addrconf_hash_lock);
5199
5200         del_timer(&addr_chk_timer);
5201         rtnl_unlock();
5202 }