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                       unsigned char 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                     unsigned char 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 #if IS_ENABLED(CONFIG_IPV6_SIT)
2000 static void sit_route_add(struct net_device *dev)
2001 {
2002         struct fib6_config cfg = {
2003                 .fc_table = RT6_TABLE_MAIN,
2004                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2005                 .fc_ifindex = dev->ifindex,
2006                 .fc_dst_len = 96,
2007                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2008                 .fc_nlinfo.nl_net = dev_net(dev),
2009         };
2010
2011         /* prefix length - 96 bits "::d.d.d.d" */
2012         ip6_route_add(&cfg);
2013 }
2014 #endif
2015
2016 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2017 {
2018         struct inet6_dev *idev;
2019
2020         ASSERT_RTNL();
2021
2022         idev = ipv6_find_idev(dev);
2023         if (!idev)
2024                 return ERR_PTR(-ENOBUFS);
2025
2026         if (idev->cnf.disable_ipv6)
2027                 return ERR_PTR(-EACCES);
2028
2029         /* Add default multicast route */
2030         if (!(dev->flags & IFF_LOOPBACK))
2031                 addrconf_add_mroute(dev);
2032
2033         return idev;
2034 }
2035
2036 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2037 {
2038         struct prefix_info *pinfo;
2039         __u32 valid_lft;
2040         __u32 prefered_lft;
2041         int addr_type;
2042         struct inet6_dev *in6_dev;
2043         struct net *net = dev_net(dev);
2044
2045         pinfo = (struct prefix_info *) opt;
2046
2047         if (len < sizeof(struct prefix_info)) {
2048                 ADBG("addrconf: prefix option too short\n");
2049                 return;
2050         }
2051
2052         /*
2053          *      Validation checks ([ADDRCONF], page 19)
2054          */
2055
2056         addr_type = ipv6_addr_type(&pinfo->prefix);
2057
2058         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2059                 return;
2060
2061         valid_lft = ntohl(pinfo->valid);
2062         prefered_lft = ntohl(pinfo->prefered);
2063
2064         if (prefered_lft > valid_lft) {
2065                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2066                 return;
2067         }
2068
2069         in6_dev = in6_dev_get(dev);
2070
2071         if (in6_dev == NULL) {
2072                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2073                                     dev->name);
2074                 return;
2075         }
2076
2077         /*
2078          *      Two things going on here:
2079          *      1) Add routes for on-link prefixes
2080          *      2) Configure prefixes with the auto flag set
2081          */
2082
2083         if (pinfo->onlink) {
2084                 struct rt6_info *rt;
2085                 unsigned long rt_expires;
2086
2087                 /* Avoid arithmetic overflow. Really, we could
2088                  * save rt_expires in seconds, likely valid_lft,
2089                  * but it would require division in fib gc, that it
2090                  * not good.
2091                  */
2092                 if (HZ > USER_HZ)
2093                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2094                 else
2095                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2096
2097                 if (addrconf_finite_timeout(rt_expires))
2098                         rt_expires *= HZ;
2099
2100                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2101                                                pinfo->prefix_len,
2102                                                dev,
2103                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2104                                                RTF_GATEWAY | RTF_DEFAULT);
2105
2106                 if (rt) {
2107                         /* Autoconf prefix route */
2108                         if (valid_lft == 0) {
2109                                 ip6_del_rt(rt);
2110                                 rt = NULL;
2111                         } else if (addrconf_finite_timeout(rt_expires)) {
2112                                 /* not infinity */
2113                                 rt6_set_expires(rt, jiffies + rt_expires);
2114                         } else {
2115                                 rt6_clean_expires(rt);
2116                         }
2117                 } else if (valid_lft) {
2118                         clock_t expires = 0;
2119                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2120                         if (addrconf_finite_timeout(rt_expires)) {
2121                                 /* not infinity */
2122                                 flags |= RTF_EXPIRES;
2123                                 expires = jiffies_to_clock_t(rt_expires);
2124                         }
2125                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2126                                               dev, expires, flags);
2127                 }
2128                 ip6_rt_put(rt);
2129         }
2130
2131         /* Try to figure out our local address for this prefix */
2132
2133         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2134                 struct inet6_ifaddr *ifp;
2135                 struct in6_addr addr;
2136                 int create = 0, update_lft = 0;
2137                 bool tokenized = false;
2138
2139                 if (pinfo->prefix_len == 64) {
2140                         memcpy(&addr, &pinfo->prefix, 8);
2141
2142                         if (!ipv6_addr_any(&in6_dev->token)) {
2143                                 read_lock_bh(&in6_dev->lock);
2144                                 memcpy(addr.s6_addr + 8,
2145                                        in6_dev->token.s6_addr + 8, 8);
2146                                 read_unlock_bh(&in6_dev->lock);
2147                                 tokenized = true;
2148                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2149                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2150                                 in6_dev_put(in6_dev);
2151                                 return;
2152                         }
2153                         goto ok;
2154                 }
2155                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2156                                     pinfo->prefix_len);
2157                 in6_dev_put(in6_dev);
2158                 return;
2159
2160 ok:
2161
2162                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2163
2164                 if (ifp == NULL && valid_lft) {
2165                         int max_addresses = in6_dev->cnf.max_addresses;
2166                         u32 addr_flags = 0;
2167
2168 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2169                         if (in6_dev->cnf.optimistic_dad &&
2170                             !net->ipv6.devconf_all->forwarding && sllao)
2171                                 addr_flags = IFA_F_OPTIMISTIC;
2172 #endif
2173
2174                         /* Do not allow to create too much of autoconfigured
2175                          * addresses; this would be too easy way to crash kernel.
2176                          */
2177                         if (!max_addresses ||
2178                             ipv6_count_addresses(in6_dev) < max_addresses)
2179                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2180                                                     pinfo->prefix_len,
2181                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2182                                                     addr_flags, valid_lft,
2183                                                     prefered_lft);
2184
2185                         if (IS_ERR_OR_NULL(ifp)) {
2186                                 in6_dev_put(in6_dev);
2187                                 return;
2188                         }
2189
2190                         update_lft = 0;
2191                         create = 1;
2192                         ifp->cstamp = jiffies;
2193                         ifp->tokenized = tokenized;
2194                         addrconf_dad_start(ifp);
2195                 }
2196
2197                 if (ifp) {
2198                         int flags;
2199                         unsigned long now;
2200                         struct inet6_ifaddr *ift;
2201                         u32 stored_lft;
2202
2203                         /* update lifetime (RFC2462 5.5.3 e) */
2204                         spin_lock(&ifp->lock);
2205                         now = jiffies;
2206                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2207                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2208                         else
2209                                 stored_lft = 0;
2210                         if (!update_lft && !create && stored_lft) {
2211                                 const u32 minimum_lft = min(
2212                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2213                                 valid_lft = max(valid_lft, minimum_lft);
2214
2215                                 /* RFC4862 Section 5.5.3e:
2216                                  * "Note that the preferred lifetime of the
2217                                  *  corresponding address is always reset to
2218                                  *  the Preferred Lifetime in the received
2219                                  *  Prefix Information option, regardless of
2220                                  *  whether the valid lifetime is also reset or
2221                                  *  ignored."
2222                                  *
2223                                  * So we should always update prefered_lft here.
2224                                  */
2225                                 update_lft = 1;
2226                         }
2227
2228                         if (update_lft) {
2229                                 ifp->valid_lft = valid_lft;
2230                                 ifp->prefered_lft = prefered_lft;
2231                                 ifp->tstamp = now;
2232                                 flags = ifp->flags;
2233                                 ifp->flags &= ~IFA_F_DEPRECATED;
2234                                 spin_unlock(&ifp->lock);
2235
2236                                 if (!(flags&IFA_F_TENTATIVE))
2237                                         ipv6_ifa_notify(0, ifp);
2238                         } else
2239                                 spin_unlock(&ifp->lock);
2240
2241                         read_lock_bh(&in6_dev->lock);
2242                         /* update all temporary addresses in the list */
2243                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2244                                             tmp_list) {
2245                                 int age, max_valid, max_prefered;
2246
2247                                 if (ifp != ift->ifpub)
2248                                         continue;
2249
2250                                 /*
2251                                  * RFC 4941 section 3.3:
2252                                  * If a received option will extend the lifetime
2253                                  * of a public address, the lifetimes of
2254                                  * temporary addresses should be extended,
2255                                  * subject to the overall constraint that no
2256                                  * temporary addresses should ever remain
2257                                  * "valid" or "preferred" for a time longer than
2258                                  * (TEMP_VALID_LIFETIME) or
2259                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2260                                  * respectively.
2261                                  */
2262                                 age = (now - ift->cstamp) / HZ;
2263                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2264                                 if (max_valid < 0)
2265                                         max_valid = 0;
2266
2267                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2268                                                in6_dev->cnf.max_desync_factor -
2269                                                age;
2270                                 if (max_prefered < 0)
2271                                         max_prefered = 0;
2272
2273                                 if (valid_lft > max_valid)
2274                                         valid_lft = max_valid;
2275
2276                                 if (prefered_lft > max_prefered)
2277                                         prefered_lft = max_prefered;
2278
2279                                 spin_lock(&ift->lock);
2280                                 flags = ift->flags;
2281                                 ift->valid_lft = valid_lft;
2282                                 ift->prefered_lft = prefered_lft;
2283                                 ift->tstamp = now;
2284                                 if (prefered_lft > 0)
2285                                         ift->flags &= ~IFA_F_DEPRECATED;
2286
2287                                 spin_unlock(&ift->lock);
2288                                 if (!(flags&IFA_F_TENTATIVE))
2289                                         ipv6_ifa_notify(0, ift);
2290                         }
2291
2292                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2293                                 /*
2294                                  * When a new public address is created as
2295                                  * described in [ADDRCONF], also create a new
2296                                  * temporary address. Also create a temporary
2297                                  * address if it's enabled but no temporary
2298                                  * address currently exists.
2299                                  */
2300                                 read_unlock_bh(&in6_dev->lock);
2301                                 ipv6_create_tempaddr(ifp, NULL);
2302                         } else {
2303                                 read_unlock_bh(&in6_dev->lock);
2304                         }
2305
2306                         in6_ifa_put(ifp);
2307                         addrconf_verify(0);
2308                 }
2309         }
2310         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2311         in6_dev_put(in6_dev);
2312 }
2313
2314 /*
2315  *      Set destination address.
2316  *      Special case for SIT interfaces where we create a new "virtual"
2317  *      device.
2318  */
2319 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2320 {
2321         struct in6_ifreq ireq;
2322         struct net_device *dev;
2323         int err = -EINVAL;
2324
2325         rtnl_lock();
2326
2327         err = -EFAULT;
2328         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2329                 goto err_exit;
2330
2331         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2332
2333         err = -ENODEV;
2334         if (dev == NULL)
2335                 goto err_exit;
2336
2337 #if IS_ENABLED(CONFIG_IPV6_SIT)
2338         if (dev->type == ARPHRD_SIT) {
2339                 const struct net_device_ops *ops = dev->netdev_ops;
2340                 struct ifreq ifr;
2341                 struct ip_tunnel_parm p;
2342
2343                 err = -EADDRNOTAVAIL;
2344                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2345                         goto err_exit;
2346
2347                 memset(&p, 0, sizeof(p));
2348                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2349                 p.iph.saddr = 0;
2350                 p.iph.version = 4;
2351                 p.iph.ihl = 5;
2352                 p.iph.protocol = IPPROTO_IPV6;
2353                 p.iph.ttl = 64;
2354                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2355
2356                 if (ops->ndo_do_ioctl) {
2357                         mm_segment_t oldfs = get_fs();
2358
2359                         set_fs(KERNEL_DS);
2360                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2361                         set_fs(oldfs);
2362                 } else
2363                         err = -EOPNOTSUPP;
2364
2365                 if (err == 0) {
2366                         err = -ENOBUFS;
2367                         dev = __dev_get_by_name(net, p.name);
2368                         if (!dev)
2369                                 goto err_exit;
2370                         err = dev_open(dev);
2371                 }
2372         }
2373 #endif
2374
2375 err_exit:
2376         rtnl_unlock();
2377         return err;
2378 }
2379
2380 /*
2381  *      Manual configuration of address on an interface
2382  */
2383 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2384                           const struct in6_addr *peer_pfx,
2385                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2386                           __u32 valid_lft)
2387 {
2388         struct inet6_ifaddr *ifp;
2389         struct inet6_dev *idev;
2390         struct net_device *dev;
2391         int scope;
2392         u32 flags;
2393         clock_t expires;
2394         unsigned long timeout;
2395
2396         ASSERT_RTNL();
2397
2398         if (plen > 128)
2399                 return -EINVAL;
2400
2401         /* check the lifetime */
2402         if (!valid_lft || prefered_lft > valid_lft)
2403                 return -EINVAL;
2404
2405         dev = __dev_get_by_index(net, ifindex);
2406         if (!dev)
2407                 return -ENODEV;
2408
2409         idev = addrconf_add_dev(dev);
2410         if (IS_ERR(idev))
2411                 return PTR_ERR(idev);
2412
2413         scope = ipv6_addr_scope(pfx);
2414
2415         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2416         if (addrconf_finite_timeout(timeout)) {
2417                 expires = jiffies_to_clock_t(timeout * HZ);
2418                 valid_lft = timeout;
2419                 flags = RTF_EXPIRES;
2420         } else {
2421                 expires = 0;
2422                 flags = 0;
2423                 ifa_flags |= IFA_F_PERMANENT;
2424         }
2425
2426         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2427         if (addrconf_finite_timeout(timeout)) {
2428                 if (timeout == 0)
2429                         ifa_flags |= IFA_F_DEPRECATED;
2430                 prefered_lft = timeout;
2431         }
2432
2433         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2434                             valid_lft, prefered_lft);
2435
2436         if (!IS_ERR(ifp)) {
2437                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2438                                       expires, flags);
2439                 /*
2440                  * Note that section 3.1 of RFC 4429 indicates
2441                  * that the Optimistic flag should not be set for
2442                  * manually configured addresses
2443                  */
2444                 addrconf_dad_start(ifp);
2445                 in6_ifa_put(ifp);
2446                 addrconf_verify(0);
2447                 return 0;
2448         }
2449
2450         return PTR_ERR(ifp);
2451 }
2452
2453 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2454                           unsigned int plen)
2455 {
2456         struct inet6_ifaddr *ifp;
2457         struct inet6_dev *idev;
2458         struct net_device *dev;
2459
2460         if (plen > 128)
2461                 return -EINVAL;
2462
2463         dev = __dev_get_by_index(net, ifindex);
2464         if (!dev)
2465                 return -ENODEV;
2466
2467         if ((idev = __in6_dev_get(dev)) == NULL)
2468                 return -ENXIO;
2469
2470         read_lock_bh(&idev->lock);
2471         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2472                 if (ifp->prefix_len == plen &&
2473                     ipv6_addr_equal(pfx, &ifp->addr)) {
2474                         in6_ifa_hold(ifp);
2475                         read_unlock_bh(&idev->lock);
2476
2477                         ipv6_del_addr(ifp);
2478                         return 0;
2479                 }
2480         }
2481         read_unlock_bh(&idev->lock);
2482         return -EADDRNOTAVAIL;
2483 }
2484
2485
2486 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2487 {
2488         struct in6_ifreq ireq;
2489         int err;
2490
2491         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2492                 return -EPERM;
2493
2494         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2495                 return -EFAULT;
2496
2497         rtnl_lock();
2498         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2499                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2500                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2501         rtnl_unlock();
2502         return err;
2503 }
2504
2505 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2506 {
2507         struct in6_ifreq ireq;
2508         int err;
2509
2510         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2511                 return -EPERM;
2512
2513         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2514                 return -EFAULT;
2515
2516         rtnl_lock();
2517         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2518                              ireq.ifr6_prefixlen);
2519         rtnl_unlock();
2520         return err;
2521 }
2522
2523 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2524                      int plen, int scope)
2525 {
2526         struct inet6_ifaddr *ifp;
2527
2528         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2529                             scope, IFA_F_PERMANENT, 0, 0);
2530         if (!IS_ERR(ifp)) {
2531                 spin_lock_bh(&ifp->lock);
2532                 ifp->flags &= ~IFA_F_TENTATIVE;
2533                 spin_unlock_bh(&ifp->lock);
2534                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2535                 in6_ifa_put(ifp);
2536         }
2537 }
2538
2539 #if IS_ENABLED(CONFIG_IPV6_SIT)
2540 static void sit_add_v4_addrs(struct inet6_dev *idev)
2541 {
2542         struct in6_addr addr;
2543         struct net_device *dev;
2544         struct net *net = dev_net(idev->dev);
2545         int scope;
2546
2547         ASSERT_RTNL();
2548
2549         memset(&addr, 0, sizeof(struct in6_addr));
2550         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2551
2552         if (idev->dev->flags&IFF_POINTOPOINT) {
2553                 addr.s6_addr32[0] = htonl(0xfe800000);
2554                 scope = IFA_LINK;
2555         } else {
2556                 scope = IPV6_ADDR_COMPATv4;
2557         }
2558
2559         if (addr.s6_addr32[3]) {
2560                 add_addr(idev, &addr, 128, scope);
2561                 return;
2562         }
2563
2564         for_each_netdev(net, dev) {
2565                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2566                 if (in_dev && (dev->flags & IFF_UP)) {
2567                         struct in_ifaddr *ifa;
2568
2569                         int flag = scope;
2570
2571                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2572                                 int plen;
2573
2574                                 addr.s6_addr32[3] = ifa->ifa_local;
2575
2576                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2577                                         continue;
2578                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2579                                         if (idev->dev->flags&IFF_POINTOPOINT)
2580                                                 continue;
2581                                         flag |= IFA_HOST;
2582                                 }
2583                                 if (idev->dev->flags&IFF_POINTOPOINT)
2584                                         plen = 64;
2585                                 else
2586                                         plen = 96;
2587
2588                                 add_addr(idev, &addr, plen, flag);
2589                         }
2590                 }
2591         }
2592 }
2593 #endif
2594
2595 static void init_loopback(struct net_device *dev)
2596 {
2597         struct inet6_dev  *idev;
2598         struct net_device *sp_dev;
2599         struct inet6_ifaddr *sp_ifa;
2600         struct rt6_info *sp_rt;
2601
2602         /* ::1 */
2603
2604         ASSERT_RTNL();
2605
2606         if ((idev = ipv6_find_idev(dev)) == NULL) {
2607                 pr_debug("%s: add_dev failed\n", __func__);
2608                 return;
2609         }
2610
2611         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2612
2613         /* Add routes to other interface's IPv6 addresses */
2614         for_each_netdev(dev_net(dev), sp_dev) {
2615                 if (!strcmp(sp_dev->name, dev->name))
2616                         continue;
2617
2618                 idev = __in6_dev_get(sp_dev);
2619                 if (!idev)
2620                         continue;
2621
2622                 read_lock_bh(&idev->lock);
2623                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2624
2625                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2626                                 continue;
2627
2628                         if (sp_ifa->rt)
2629                                 continue;
2630
2631                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2632
2633                         /* Failure cases are ignored */
2634                         if (!IS_ERR(sp_rt)) {
2635                                 sp_ifa->rt = sp_rt;
2636                                 ip6_ins_rt(sp_rt);
2637                         }
2638                 }
2639                 read_unlock_bh(&idev->lock);
2640         }
2641 }
2642
2643 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2644 {
2645         struct inet6_ifaddr *ifp;
2646         u32 addr_flags = IFA_F_PERMANENT;
2647
2648 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2649         if (idev->cnf.optimistic_dad &&
2650             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2651                 addr_flags |= IFA_F_OPTIMISTIC;
2652 #endif
2653
2654
2655         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2656         if (!IS_ERR(ifp)) {
2657                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2658                 addrconf_dad_start(ifp);
2659                 in6_ifa_put(ifp);
2660         }
2661 }
2662
2663 static void addrconf_dev_config(struct net_device *dev)
2664 {
2665         struct in6_addr addr;
2666         struct inet6_dev *idev;
2667
2668         ASSERT_RTNL();
2669
2670         if ((dev->type != ARPHRD_ETHER) &&
2671             (dev->type != ARPHRD_FDDI) &&
2672             (dev->type != ARPHRD_ARCNET) &&
2673             (dev->type != ARPHRD_INFINIBAND) &&
2674             (dev->type != ARPHRD_IEEE802154) &&
2675             (dev->type != ARPHRD_IEEE1394) &&
2676             (dev->type != ARPHRD_TUNNEL6)) {
2677                 /* Alas, we support only Ethernet autoconfiguration. */
2678                 return;
2679         }
2680
2681         idev = addrconf_add_dev(dev);
2682         if (IS_ERR(idev))
2683                 return;
2684
2685         memset(&addr, 0, sizeof(struct in6_addr));
2686         addr.s6_addr32[0] = htonl(0xFE800000);
2687
2688         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2689                 addrconf_add_linklocal(idev, &addr);
2690 }
2691
2692 #if IS_ENABLED(CONFIG_IPV6_SIT)
2693 static void addrconf_sit_config(struct net_device *dev)
2694 {
2695         struct inet6_dev *idev;
2696
2697         ASSERT_RTNL();
2698
2699         /*
2700          * Configure the tunnel with one of our IPv4
2701          * addresses... we should configure all of
2702          * our v4 addrs in the tunnel
2703          */
2704
2705         if ((idev = ipv6_find_idev(dev)) == NULL) {
2706                 pr_debug("%s: add_dev failed\n", __func__);
2707                 return;
2708         }
2709
2710         if (dev->priv_flags & IFF_ISATAP) {
2711                 struct in6_addr addr;
2712
2713                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2714                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2715                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2716                         addrconf_add_linklocal(idev, &addr);
2717                 return;
2718         }
2719
2720         sit_add_v4_addrs(idev);
2721
2722         if (dev->flags&IFF_POINTOPOINT)
2723                 addrconf_add_mroute(dev);
2724         else
2725                 sit_route_add(dev);
2726 }
2727 #endif
2728
2729 #if IS_ENABLED(CONFIG_NET_IPGRE)
2730 static void addrconf_gre_config(struct net_device *dev)
2731 {
2732         struct inet6_dev *idev;
2733         struct in6_addr addr;
2734
2735         ASSERT_RTNL();
2736
2737         if ((idev = ipv6_find_idev(dev)) == NULL) {
2738                 pr_debug("%s: add_dev failed\n", __func__);
2739                 return;
2740         }
2741
2742         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2743         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2744
2745         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2746                 addrconf_add_linklocal(idev, &addr);
2747 }
2748 #endif
2749
2750 static inline int
2751 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2752 {
2753         struct in6_addr lladdr;
2754
2755         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2756                 addrconf_add_linklocal(idev, &lladdr);
2757                 return 0;
2758         }
2759         return -1;
2760 }
2761
2762 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2763                            void *ptr)
2764 {
2765         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2766         struct inet6_dev *idev = __in6_dev_get(dev);
2767         int run_pending = 0;
2768         int err;
2769
2770         switch (event) {
2771         case NETDEV_REGISTER:
2772                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2773                         idev = ipv6_add_dev(dev);
2774                         if (!idev)
2775                                 return notifier_from_errno(-ENOMEM);
2776                 }
2777                 break;
2778
2779         case NETDEV_UP:
2780         case NETDEV_CHANGE:
2781                 if (dev->flags & IFF_SLAVE)
2782                         break;
2783
2784                 if (event == NETDEV_UP) {
2785                         if (!addrconf_qdisc_ok(dev)) {
2786                                 /* device is not ready yet. */
2787                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2788                                         dev->name);
2789                                 break;
2790                         }
2791
2792                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2793                                 idev = ipv6_add_dev(dev);
2794
2795                         if (idev) {
2796                                 idev->if_flags |= IF_READY;
2797                                 run_pending = 1;
2798                         }
2799                 } else {
2800                         if (!addrconf_qdisc_ok(dev)) {
2801                                 /* device is still not ready. */
2802                                 break;
2803                         }
2804
2805                         if (idev) {
2806                                 if (idev->if_flags & IF_READY)
2807                                         /* device is already configured. */
2808                                         break;
2809                                 idev->if_flags |= IF_READY;
2810                         }
2811
2812                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2813                                 dev->name);
2814
2815                         run_pending = 1;
2816                 }
2817
2818                 switch (dev->type) {
2819 #if IS_ENABLED(CONFIG_IPV6_SIT)
2820                 case ARPHRD_SIT:
2821                         addrconf_sit_config(dev);
2822                         break;
2823 #endif
2824 #if IS_ENABLED(CONFIG_NET_IPGRE)
2825                 case ARPHRD_IPGRE:
2826                         addrconf_gre_config(dev);
2827                         break;
2828 #endif
2829                 case ARPHRD_LOOPBACK:
2830                         init_loopback(dev);
2831                         break;
2832
2833                 default:
2834                         addrconf_dev_config(dev);
2835                         break;
2836                 }
2837
2838                 if (idev) {
2839                         if (run_pending)
2840                                 addrconf_dad_run(idev);
2841
2842                         /*
2843                          * If the MTU changed during the interface down,
2844                          * when the interface up, the changed MTU must be
2845                          * reflected in the idev as well as routers.
2846                          */
2847                         if (idev->cnf.mtu6 != dev->mtu &&
2848                             dev->mtu >= IPV6_MIN_MTU) {
2849                                 rt6_mtu_change(dev, dev->mtu);
2850                                 idev->cnf.mtu6 = dev->mtu;
2851                         }
2852                         idev->tstamp = jiffies;
2853                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2854
2855                         /*
2856                          * If the changed mtu during down is lower than
2857                          * IPV6_MIN_MTU stop IPv6 on this interface.
2858                          */
2859                         if (dev->mtu < IPV6_MIN_MTU)
2860                                 addrconf_ifdown(dev, 1);
2861                 }
2862                 break;
2863
2864         case NETDEV_CHANGEMTU:
2865                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2866                         rt6_mtu_change(dev, dev->mtu);
2867                         idev->cnf.mtu6 = dev->mtu;
2868                         break;
2869                 }
2870
2871                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2872                         idev = ipv6_add_dev(dev);
2873                         if (idev)
2874                                 break;
2875                 }
2876
2877                 /*
2878                  * MTU falled under IPV6_MIN_MTU.
2879                  * Stop IPv6 on this interface.
2880                  */
2881
2882         case NETDEV_DOWN:
2883         case NETDEV_UNREGISTER:
2884                 /*
2885                  *      Remove all addresses from this interface.
2886                  */
2887                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2888                 break;
2889
2890         case NETDEV_CHANGENAME:
2891                 if (idev) {
2892                         snmp6_unregister_dev(idev);
2893                         addrconf_sysctl_unregister(idev);
2894                         addrconf_sysctl_register(idev);
2895                         err = snmp6_register_dev(idev);
2896                         if (err)
2897                                 return notifier_from_errno(err);
2898                 }
2899                 break;
2900
2901         case NETDEV_PRE_TYPE_CHANGE:
2902         case NETDEV_POST_TYPE_CHANGE:
2903                 addrconf_type_change(dev, event);
2904                 break;
2905         }
2906
2907         return NOTIFY_OK;
2908 }
2909
2910 /*
2911  *      addrconf module should be notified of a device going up
2912  */
2913 static struct notifier_block ipv6_dev_notf = {
2914         .notifier_call = addrconf_notify,
2915 };
2916
2917 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2918 {
2919         struct inet6_dev *idev;
2920         ASSERT_RTNL();
2921
2922         idev = __in6_dev_get(dev);
2923
2924         if (event == NETDEV_POST_TYPE_CHANGE)
2925                 ipv6_mc_remap(idev);
2926         else if (event == NETDEV_PRE_TYPE_CHANGE)
2927                 ipv6_mc_unmap(idev);
2928 }
2929
2930 static int addrconf_ifdown(struct net_device *dev, int how)
2931 {
2932         struct net *net = dev_net(dev);
2933         struct inet6_dev *idev;
2934         struct inet6_ifaddr *ifa;
2935         int state, i;
2936
2937         ASSERT_RTNL();
2938
2939         rt6_ifdown(net, dev);
2940         neigh_ifdown(&nd_tbl, dev);
2941
2942         idev = __in6_dev_get(dev);
2943         if (idev == NULL)
2944                 return -ENODEV;
2945
2946         /*
2947          * Step 1: remove reference to ipv6 device from parent device.
2948          *         Do not dev_put!
2949          */
2950         if (how) {
2951                 idev->dead = 1;
2952
2953                 /* protected by rtnl_lock */
2954                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2955
2956                 /* Step 1.5: remove snmp6 entry */
2957                 snmp6_unregister_dev(idev);
2958
2959         }
2960
2961         /* Step 2: clear hash table */
2962         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2963                 struct hlist_head *h = &inet6_addr_lst[i];
2964
2965                 spin_lock_bh(&addrconf_hash_lock);
2966         restart:
2967                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2968                         if (ifa->idev == idev) {
2969                                 hlist_del_init_rcu(&ifa->addr_lst);
2970                                 addrconf_del_dad_timer(ifa);
2971                                 goto restart;
2972                         }
2973                 }
2974                 spin_unlock_bh(&addrconf_hash_lock);
2975         }
2976
2977         write_lock_bh(&idev->lock);
2978
2979         addrconf_del_rs_timer(idev);
2980
2981         /* Step 2: clear flags for stateless addrconf */
2982         if (!how)
2983                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2984
2985         if (how && del_timer(&idev->regen_timer))
2986                 in6_dev_put(idev);
2987
2988         /* Step 3: clear tempaddr list */
2989         while (!list_empty(&idev->tempaddr_list)) {
2990                 ifa = list_first_entry(&idev->tempaddr_list,
2991                                        struct inet6_ifaddr, tmp_list);
2992                 list_del(&ifa->tmp_list);
2993                 write_unlock_bh(&idev->lock);
2994                 spin_lock_bh(&ifa->lock);
2995
2996                 if (ifa->ifpub) {
2997                         in6_ifa_put(ifa->ifpub);
2998                         ifa->ifpub = NULL;
2999                 }
3000                 spin_unlock_bh(&ifa->lock);
3001                 in6_ifa_put(ifa);
3002                 write_lock_bh(&idev->lock);
3003         }
3004
3005         while (!list_empty(&idev->addr_list)) {
3006                 ifa = list_first_entry(&idev->addr_list,
3007                                        struct inet6_ifaddr, if_list);
3008                 addrconf_del_dad_timer(ifa);
3009
3010                 list_del(&ifa->if_list);
3011
3012                 write_unlock_bh(&idev->lock);
3013
3014                 spin_lock_bh(&ifa->state_lock);
3015                 state = ifa->state;
3016                 ifa->state = INET6_IFADDR_STATE_DEAD;
3017                 spin_unlock_bh(&ifa->state_lock);
3018
3019                 if (state != INET6_IFADDR_STATE_DEAD) {
3020                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3021                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3022                 }
3023                 in6_ifa_put(ifa);
3024
3025                 write_lock_bh(&idev->lock);
3026         }
3027
3028         write_unlock_bh(&idev->lock);
3029
3030         /* Step 5: Discard multicast list */
3031         if (how)
3032                 ipv6_mc_destroy_dev(idev);
3033         else
3034                 ipv6_mc_down(idev);
3035
3036         idev->tstamp = jiffies;
3037
3038         /* Last: Shot the device (if unregistered) */
3039         if (how) {
3040                 addrconf_sysctl_unregister(idev);
3041                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3042                 neigh_ifdown(&nd_tbl, dev);
3043                 in6_dev_put(idev);
3044         }
3045         return 0;
3046 }
3047
3048 static void addrconf_rs_timer(unsigned long data)
3049 {
3050         struct inet6_dev *idev = (struct inet6_dev *)data;
3051         struct net_device *dev = idev->dev;
3052         struct in6_addr lladdr;
3053
3054         write_lock(&idev->lock);
3055         if (idev->dead || !(idev->if_flags & IF_READY))
3056                 goto out;
3057
3058         if (!ipv6_accept_ra(idev))
3059                 goto out;
3060
3061         /* Announcement received after solicitation was sent */
3062         if (idev->if_flags & IF_RA_RCVD)
3063                 goto out;
3064
3065         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3066                 write_unlock(&idev->lock);
3067                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3068                         ndisc_send_rs(dev, &lladdr,
3069                                       &in6addr_linklocal_allrouters);
3070                 else
3071                         goto put;
3072
3073                 write_lock(&idev->lock);
3074                 /* The wait after the last probe can be shorter */
3075                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3076                                              idev->cnf.rtr_solicits) ?
3077                                       idev->cnf.rtr_solicit_delay :
3078                                       idev->cnf.rtr_solicit_interval);
3079         } else {
3080                 /*
3081                  * Note: we do not support deprecated "all on-link"
3082                  * assumption any longer.
3083                  */
3084                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3085         }
3086
3087 out:
3088         write_unlock(&idev->lock);
3089 put:
3090         in6_dev_put(idev);
3091 }
3092
3093 /*
3094  *      Duplicate Address Detection
3095  */
3096 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3097 {
3098         unsigned long rand_num;
3099         struct inet6_dev *idev = ifp->idev;
3100
3101         if (ifp->flags & IFA_F_OPTIMISTIC)
3102                 rand_num = 0;
3103         else
3104                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3105
3106         ifp->dad_probes = idev->cnf.dad_transmits;
3107         addrconf_mod_dad_timer(ifp, rand_num);
3108 }
3109
3110 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3111 {
3112         struct inet6_dev *idev = ifp->idev;
3113         struct net_device *dev = idev->dev;
3114
3115         addrconf_join_solict(dev, &ifp->addr);
3116
3117         net_srandom(ifp->addr.s6_addr32[3]);
3118
3119         read_lock_bh(&idev->lock);
3120         spin_lock(&ifp->lock);
3121         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3122                 goto out;
3123
3124         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3125             idev->cnf.accept_dad < 1 ||
3126             !(ifp->flags&IFA_F_TENTATIVE) ||
3127             ifp->flags & IFA_F_NODAD) {
3128                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3129                 spin_unlock(&ifp->lock);
3130                 read_unlock_bh(&idev->lock);
3131
3132                 addrconf_dad_completed(ifp);
3133                 return;
3134         }
3135
3136         if (!(idev->if_flags & IF_READY)) {
3137                 spin_unlock(&ifp->lock);
3138                 read_unlock_bh(&idev->lock);
3139                 /*
3140                  * If the device is not ready:
3141                  * - keep it tentative if it is a permanent address.
3142                  * - otherwise, kill it.
3143                  */
3144                 in6_ifa_hold(ifp);
3145                 addrconf_dad_stop(ifp, 0);
3146                 return;
3147         }
3148
3149         /*
3150          * Optimistic nodes can start receiving
3151          * Frames right away
3152          */
3153         if (ifp->flags & IFA_F_OPTIMISTIC)
3154                 ip6_ins_rt(ifp->rt);
3155
3156         addrconf_dad_kick(ifp);
3157 out:
3158         spin_unlock(&ifp->lock);
3159         read_unlock_bh(&idev->lock);
3160 }
3161
3162 static void addrconf_dad_timer(unsigned long data)
3163 {
3164         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3165         struct inet6_dev *idev = ifp->idev;
3166         struct in6_addr mcaddr;
3167
3168         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3169                 goto out;
3170
3171         write_lock(&idev->lock);
3172         if (idev->dead || !(idev->if_flags & IF_READY)) {
3173                 write_unlock(&idev->lock);
3174                 goto out;
3175         }
3176
3177         spin_lock(&ifp->lock);
3178         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3179                 spin_unlock(&ifp->lock);
3180                 write_unlock(&idev->lock);
3181                 goto out;
3182         }
3183
3184         if (ifp->dad_probes == 0) {
3185                 /*
3186                  * DAD was successful
3187                  */
3188
3189                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3190                 spin_unlock(&ifp->lock);
3191                 write_unlock(&idev->lock);
3192
3193                 addrconf_dad_completed(ifp);
3194
3195                 goto out;
3196         }
3197
3198         ifp->dad_probes--;
3199         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3200         spin_unlock(&ifp->lock);
3201         write_unlock(&idev->lock);
3202
3203         /* send a neighbour solicitation for our addr */
3204         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3205         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3206 out:
3207         in6_ifa_put(ifp);
3208 }
3209
3210 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3211 {
3212         struct net_device *dev = ifp->idev->dev;
3213         struct in6_addr lladdr;
3214         bool send_rs, send_mld;
3215
3216         addrconf_del_dad_timer(ifp);
3217
3218         /*
3219          *      Configure the address for reception. Now it is valid.
3220          */
3221
3222         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3223
3224         /* If added prefix is link local and we are prepared to process
3225            router advertisements, start sending router solicitations.
3226          */
3227
3228         read_lock_bh(&ifp->idev->lock);
3229         spin_lock(&ifp->lock);
3230         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3231                    ifp->idev->valid_ll_addr_cnt == 1;
3232         send_rs = send_mld &&
3233                   ipv6_accept_ra(ifp->idev) &&
3234                   ifp->idev->cnf.rtr_solicits > 0 &&
3235                   (dev->flags&IFF_LOOPBACK) == 0;
3236         spin_unlock(&ifp->lock);
3237         read_unlock_bh(&ifp->idev->lock);
3238
3239         /* While dad is in progress mld report's source address is in6_addrany.
3240          * Resend with proper ll now.
3241          */
3242         if (send_mld)
3243                 ipv6_mc_dad_complete(ifp->idev);
3244
3245         if (send_rs) {
3246                 /*
3247                  *      If a host as already performed a random delay
3248                  *      [...] as part of DAD [...] there is no need
3249                  *      to delay again before sending the first RS
3250                  */
3251                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3252                         return;
3253                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3254
3255                 write_lock_bh(&ifp->idev->lock);
3256                 spin_lock(&ifp->lock);
3257                 ifp->idev->rs_probes = 1;
3258                 ifp->idev->if_flags |= IF_RS_SENT;
3259                 addrconf_mod_rs_timer(ifp->idev,
3260                                       ifp->idev->cnf.rtr_solicit_interval);
3261                 spin_unlock(&ifp->lock);
3262                 write_unlock_bh(&ifp->idev->lock);
3263         }
3264 }
3265
3266 static void addrconf_dad_run(struct inet6_dev *idev)
3267 {
3268         struct inet6_ifaddr *ifp;
3269
3270         read_lock_bh(&idev->lock);
3271         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3272                 spin_lock(&ifp->lock);
3273                 if (ifp->flags & IFA_F_TENTATIVE &&
3274                     ifp->state == INET6_IFADDR_STATE_DAD)
3275                         addrconf_dad_kick(ifp);
3276                 spin_unlock(&ifp->lock);
3277         }
3278         read_unlock_bh(&idev->lock);
3279 }
3280
3281 #ifdef CONFIG_PROC_FS
3282 struct if6_iter_state {
3283         struct seq_net_private p;
3284         int bucket;
3285         int offset;
3286 };
3287
3288 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3289 {
3290         struct inet6_ifaddr *ifa = NULL;
3291         struct if6_iter_state *state = seq->private;
3292         struct net *net = seq_file_net(seq);
3293         int p = 0;
3294
3295         /* initial bucket if pos is 0 */
3296         if (pos == 0) {
3297                 state->bucket = 0;
3298                 state->offset = 0;
3299         }
3300
3301         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3302                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3303                                          addr_lst) {
3304                         if (!net_eq(dev_net(ifa->idev->dev), net))
3305                                 continue;
3306                         /* sync with offset */
3307                         if (p < state->offset) {
3308                                 p++;
3309                                 continue;
3310                         }
3311                         state->offset++;
3312                         return ifa;
3313                 }
3314
3315                 /* prepare for next bucket */
3316                 state->offset = 0;
3317                 p = 0;
3318         }
3319         return NULL;
3320 }
3321
3322 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3323                                          struct inet6_ifaddr *ifa)
3324 {
3325         struct if6_iter_state *state = seq->private;
3326         struct net *net = seq_file_net(seq);
3327
3328         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3329                 if (!net_eq(dev_net(ifa->idev->dev), net))
3330                         continue;
3331                 state->offset++;
3332                 return ifa;
3333         }
3334
3335         while (++state->bucket < IN6_ADDR_HSIZE) {
3336                 state->offset = 0;
3337                 hlist_for_each_entry_rcu_bh(ifa,
3338                                      &inet6_addr_lst[state->bucket], addr_lst) {
3339                         if (!net_eq(dev_net(ifa->idev->dev), net))
3340                                 continue;
3341                         state->offset++;
3342                         return ifa;
3343                 }
3344         }
3345
3346         return NULL;
3347 }
3348
3349 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3350         __acquires(rcu_bh)
3351 {
3352         rcu_read_lock_bh();
3353         return if6_get_first(seq, *pos);
3354 }
3355
3356 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3357 {
3358         struct inet6_ifaddr *ifa;
3359
3360         ifa = if6_get_next(seq, v);
3361         ++*pos;
3362         return ifa;
3363 }
3364
3365 static void if6_seq_stop(struct seq_file *seq, void *v)
3366         __releases(rcu_bh)
3367 {
3368         rcu_read_unlock_bh();
3369 }
3370
3371 static int if6_seq_show(struct seq_file *seq, void *v)
3372 {
3373         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3374         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3375                    &ifp->addr,
3376                    ifp->idev->dev->ifindex,
3377                    ifp->prefix_len,
3378                    ifp->scope,
3379                    ifp->flags,
3380                    ifp->idev->dev->name);
3381         return 0;
3382 }
3383
3384 static const struct seq_operations if6_seq_ops = {
3385         .start  = if6_seq_start,
3386         .next   = if6_seq_next,
3387         .show   = if6_seq_show,
3388         .stop   = if6_seq_stop,
3389 };
3390
3391 static int if6_seq_open(struct inode *inode, struct file *file)
3392 {
3393         return seq_open_net(inode, file, &if6_seq_ops,
3394                             sizeof(struct if6_iter_state));
3395 }
3396
3397 static const struct file_operations if6_fops = {
3398         .owner          = THIS_MODULE,
3399         .open           = if6_seq_open,
3400         .read           = seq_read,
3401         .llseek         = seq_lseek,
3402         .release        = seq_release_net,
3403 };
3404
3405 static int __net_init if6_proc_net_init(struct net *net)
3406 {
3407         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3408                 return -ENOMEM;
3409         return 0;
3410 }
3411
3412 static void __net_exit if6_proc_net_exit(struct net *net)
3413 {
3414         remove_proc_entry("if_inet6", net->proc_net);
3415 }
3416
3417 static struct pernet_operations if6_proc_net_ops = {
3418        .init = if6_proc_net_init,
3419        .exit = if6_proc_net_exit,
3420 };
3421
3422 int __init if6_proc_init(void)
3423 {
3424         return register_pernet_subsys(&if6_proc_net_ops);
3425 }
3426
3427 void if6_proc_exit(void)
3428 {
3429         unregister_pernet_subsys(&if6_proc_net_ops);
3430 }
3431 #endif  /* CONFIG_PROC_FS */
3432
3433 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3434 /* Check if address is a home address configured on any interface. */
3435 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3436 {
3437         int ret = 0;
3438         struct inet6_ifaddr *ifp = NULL;
3439         unsigned int hash = inet6_addr_hash(addr);
3440
3441         rcu_read_lock_bh();
3442         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3443                 if (!net_eq(dev_net(ifp->idev->dev), net))
3444                         continue;
3445                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3446                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3447                         ret = 1;
3448                         break;
3449                 }
3450         }
3451         rcu_read_unlock_bh();
3452         return ret;
3453 }
3454 #endif
3455
3456 /*
3457  *      Periodic address status verification
3458  */
3459
3460 static void addrconf_verify(unsigned long foo)
3461 {
3462         unsigned long now, next, next_sec, next_sched;
3463         struct inet6_ifaddr *ifp;
3464         int i;
3465
3466         rcu_read_lock_bh();
3467         spin_lock(&addrconf_verify_lock);
3468         now = jiffies;
3469         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3470
3471         del_timer(&addr_chk_timer);
3472
3473         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3474 restart:
3475                 hlist_for_each_entry_rcu_bh(ifp,
3476                                          &inet6_addr_lst[i], addr_lst) {
3477                         unsigned long age;
3478
3479                         if (ifp->flags & IFA_F_PERMANENT)
3480                                 continue;
3481
3482                         spin_lock(&ifp->lock);
3483                         /* We try to batch several events at once. */
3484                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3485
3486                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3487                             age >= ifp->valid_lft) {
3488                                 spin_unlock(&ifp->lock);
3489                                 in6_ifa_hold(ifp);
3490                                 ipv6_del_addr(ifp);
3491                                 goto restart;
3492                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3493                                 spin_unlock(&ifp->lock);
3494                                 continue;
3495                         } else if (age >= ifp->prefered_lft) {
3496                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3497                                 int deprecate = 0;
3498
3499                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3500                                         deprecate = 1;
3501                                         ifp->flags |= IFA_F_DEPRECATED;
3502                                 }
3503
3504                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3505                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3506
3507                                 spin_unlock(&ifp->lock);
3508
3509                                 if (deprecate) {
3510                                         in6_ifa_hold(ifp);
3511
3512                                         ipv6_ifa_notify(0, ifp);
3513                                         in6_ifa_put(ifp);
3514                                         goto restart;
3515                                 }
3516                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3517                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3518                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3519                                         ifp->idev->cnf.dad_transmits *
3520                                         ifp->idev->nd_parms->retrans_time / HZ;
3521
3522                                 if (age >= ifp->prefered_lft - regen_advance) {
3523                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3524                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3525                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3526                                         if (!ifp->regen_count && ifpub) {
3527                                                 ifp->regen_count++;
3528                                                 in6_ifa_hold(ifp);
3529                                                 in6_ifa_hold(ifpub);
3530                                                 spin_unlock(&ifp->lock);
3531
3532                                                 spin_lock(&ifpub->lock);
3533                                                 ifpub->regen_count = 0;
3534                                                 spin_unlock(&ifpub->lock);
3535                                                 ipv6_create_tempaddr(ifpub, ifp);
3536                                                 in6_ifa_put(ifpub);
3537                                                 in6_ifa_put(ifp);
3538                                                 goto restart;
3539                                         }
3540                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3541                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3542                                 spin_unlock(&ifp->lock);
3543                         } else {
3544                                 /* ifp->prefered_lft <= ifp->valid_lft */
3545                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3546                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3547                                 spin_unlock(&ifp->lock);
3548                         }
3549                 }
3550         }
3551
3552         next_sec = round_jiffies_up(next);
3553         next_sched = next;
3554
3555         /* If rounded timeout is accurate enough, accept it. */
3556         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3557                 next_sched = next_sec;
3558
3559         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3560         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3561                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3562
3563         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3564               now, next, next_sec, next_sched);
3565
3566         addr_chk_timer.expires = next_sched;
3567         add_timer(&addr_chk_timer);
3568         spin_unlock(&addrconf_verify_lock);
3569         rcu_read_unlock_bh();
3570 }
3571
3572 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3573                                      struct in6_addr **peer_pfx)
3574 {
3575         struct in6_addr *pfx = NULL;
3576
3577         *peer_pfx = NULL;
3578
3579         if (addr)
3580                 pfx = nla_data(addr);
3581
3582         if (local) {
3583                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3584                         *peer_pfx = pfx;
3585                 pfx = nla_data(local);
3586         }
3587
3588         return pfx;
3589 }
3590
3591 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3592         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3593         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3594         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3595 };
3596
3597 static int
3598 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3599 {
3600         struct net *net = sock_net(skb->sk);
3601         struct ifaddrmsg *ifm;
3602         struct nlattr *tb[IFA_MAX+1];
3603         struct in6_addr *pfx, *peer_pfx;
3604         int err;
3605
3606         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3607         if (err < 0)
3608                 return err;
3609
3610         ifm = nlmsg_data(nlh);
3611         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3612         if (pfx == NULL)
3613                 return -EINVAL;
3614
3615         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3616 }
3617
3618 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3619                              u32 prefered_lft, u32 valid_lft)
3620 {
3621         u32 flags;
3622         clock_t expires;
3623         unsigned long timeout;
3624
3625         if (!valid_lft || (prefered_lft > valid_lft))
3626                 return -EINVAL;
3627
3628         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3629         if (addrconf_finite_timeout(timeout)) {
3630                 expires = jiffies_to_clock_t(timeout * HZ);
3631                 valid_lft = timeout;
3632                 flags = RTF_EXPIRES;
3633         } else {
3634                 expires = 0;
3635                 flags = 0;
3636                 ifa_flags |= IFA_F_PERMANENT;
3637         }
3638
3639         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3640         if (addrconf_finite_timeout(timeout)) {
3641                 if (timeout == 0)
3642                         ifa_flags |= IFA_F_DEPRECATED;
3643                 prefered_lft = timeout;
3644         }
3645
3646         spin_lock_bh(&ifp->lock);
3647         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3648         ifp->tstamp = jiffies;
3649         ifp->valid_lft = valid_lft;
3650         ifp->prefered_lft = prefered_lft;
3651
3652         spin_unlock_bh(&ifp->lock);
3653         if (!(ifp->flags&IFA_F_TENTATIVE))
3654                 ipv6_ifa_notify(0, ifp);
3655
3656         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3657                               expires, flags);
3658         addrconf_verify(0);
3659
3660         return 0;
3661 }
3662
3663 static int
3664 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3665 {
3666         struct net *net = sock_net(skb->sk);
3667         struct ifaddrmsg *ifm;
3668         struct nlattr *tb[IFA_MAX+1];
3669         struct in6_addr *pfx, *peer_pfx;
3670         struct inet6_ifaddr *ifa;
3671         struct net_device *dev;
3672         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3673         u8 ifa_flags;
3674         int err;
3675
3676         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3677         if (err < 0)
3678                 return err;
3679
3680         ifm = nlmsg_data(nlh);
3681         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3682         if (pfx == NULL)
3683                 return -EINVAL;
3684
3685         if (tb[IFA_CACHEINFO]) {
3686                 struct ifa_cacheinfo *ci;
3687
3688                 ci = nla_data(tb[IFA_CACHEINFO]);
3689                 valid_lft = ci->ifa_valid;
3690                 preferred_lft = ci->ifa_prefered;
3691         } else {
3692                 preferred_lft = INFINITY_LIFE_TIME;
3693                 valid_lft = INFINITY_LIFE_TIME;
3694         }
3695
3696         dev =  __dev_get_by_index(net, ifm->ifa_index);
3697         if (dev == NULL)
3698                 return -ENODEV;
3699
3700         /* We ignore other flags so far. */
3701         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3702
3703         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3704         if (ifa == NULL) {
3705                 /*
3706                  * It would be best to check for !NLM_F_CREATE here but
3707                  * userspace alreay relies on not having to provide this.
3708                  */
3709                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3710                                       ifm->ifa_prefixlen, ifa_flags,
3711                                       preferred_lft, valid_lft);
3712         }
3713
3714         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3715             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3716                 err = -EEXIST;
3717         else
3718                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3719
3720         in6_ifa_put(ifa);
3721
3722         return err;
3723 }
3724
3725 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3726                           u8 scope, int ifindex)
3727 {
3728         struct ifaddrmsg *ifm;
3729
3730         ifm = nlmsg_data(nlh);
3731         ifm->ifa_family = AF_INET6;
3732         ifm->ifa_prefixlen = prefixlen;
3733         ifm->ifa_flags = flags;
3734         ifm->ifa_scope = scope;
3735         ifm->ifa_index = ifindex;
3736 }
3737
3738 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3739                          unsigned long tstamp, u32 preferred, u32 valid)
3740 {
3741         struct ifa_cacheinfo ci;
3742
3743         ci.cstamp = cstamp_delta(cstamp);
3744         ci.tstamp = cstamp_delta(tstamp);
3745         ci.ifa_prefered = preferred;
3746         ci.ifa_valid = valid;
3747
3748         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3749 }
3750
3751 static inline int rt_scope(int ifa_scope)
3752 {
3753         if (ifa_scope & IFA_HOST)
3754                 return RT_SCOPE_HOST;
3755         else if (ifa_scope & IFA_LINK)
3756                 return RT_SCOPE_LINK;
3757         else if (ifa_scope & IFA_SITE)
3758                 return RT_SCOPE_SITE;
3759         else
3760                 return RT_SCOPE_UNIVERSE;
3761 }
3762
3763 static inline int inet6_ifaddr_msgsize(void)
3764 {
3765         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3766                + nla_total_size(16) /* IFA_LOCAL */
3767                + nla_total_size(16) /* IFA_ADDRESS */
3768                + nla_total_size(sizeof(struct ifa_cacheinfo));
3769 }
3770
3771 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3772                              u32 portid, u32 seq, int event, unsigned int flags)
3773 {
3774         struct nlmsghdr  *nlh;
3775         u32 preferred, valid;
3776
3777         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3778         if (nlh == NULL)
3779                 return -EMSGSIZE;
3780
3781         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3782                       ifa->idev->dev->ifindex);
3783
3784         if (!(ifa->flags&IFA_F_PERMANENT)) {
3785                 preferred = ifa->prefered_lft;
3786                 valid = ifa->valid_lft;
3787                 if (preferred != INFINITY_LIFE_TIME) {
3788                         long tval = (jiffies - ifa->tstamp)/HZ;
3789                         if (preferred > tval)
3790                                 preferred -= tval;
3791                         else
3792                                 preferred = 0;
3793                         if (valid != INFINITY_LIFE_TIME) {
3794                                 if (valid > tval)
3795                                         valid -= tval;
3796                                 else
3797                                         valid = 0;
3798                         }
3799                 }
3800         } else {
3801                 preferred = INFINITY_LIFE_TIME;
3802                 valid = INFINITY_LIFE_TIME;
3803         }
3804
3805         if (!ipv6_addr_any(&ifa->peer_addr)) {
3806                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3807                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3808                         goto error;
3809         } else
3810                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3811                         goto error;
3812
3813         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3814                 goto error;
3815
3816         return nlmsg_end(skb, nlh);
3817
3818 error:
3819         nlmsg_cancel(skb, nlh);
3820         return -EMSGSIZE;
3821 }
3822
3823 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3824                                 u32 portid, u32 seq, int event, u16 flags)
3825 {
3826         struct nlmsghdr  *nlh;
3827         u8 scope = RT_SCOPE_UNIVERSE;
3828         int ifindex = ifmca->idev->dev->ifindex;
3829
3830         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3831                 scope = RT_SCOPE_SITE;
3832
3833         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3834         if (nlh == NULL)
3835                 return -EMSGSIZE;
3836
3837         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3838         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3839             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3840                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3841                 nlmsg_cancel(skb, nlh);
3842                 return -EMSGSIZE;
3843         }
3844
3845         return nlmsg_end(skb, nlh);
3846 }
3847
3848 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3849                                 u32 portid, u32 seq, int event, unsigned int flags)
3850 {
3851         struct nlmsghdr  *nlh;
3852         u8 scope = RT_SCOPE_UNIVERSE;
3853         int ifindex = ifaca->aca_idev->dev->ifindex;
3854
3855         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3856                 scope = RT_SCOPE_SITE;
3857
3858         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3859         if (nlh == NULL)
3860                 return -EMSGSIZE;
3861
3862         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3863         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3864             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3865                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3866                 nlmsg_cancel(skb, nlh);
3867                 return -EMSGSIZE;
3868         }
3869
3870         return nlmsg_end(skb, nlh);
3871 }
3872
3873 enum addr_type_t {
3874         UNICAST_ADDR,
3875         MULTICAST_ADDR,
3876         ANYCAST_ADDR,
3877 };
3878
3879 /* called with rcu_read_lock() */
3880 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3881                           struct netlink_callback *cb, enum addr_type_t type,
3882                           int s_ip_idx, int *p_ip_idx)
3883 {
3884         struct ifmcaddr6 *ifmca;
3885         struct ifacaddr6 *ifaca;
3886         int err = 1;
3887         int ip_idx = *p_ip_idx;
3888
3889         read_lock_bh(&idev->lock);
3890         switch (type) {
3891         case UNICAST_ADDR: {
3892                 struct inet6_ifaddr *ifa;
3893
3894                 /* unicast address incl. temp addr */
3895                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3896                         if (++ip_idx < s_ip_idx)
3897                                 continue;
3898                         err = inet6_fill_ifaddr(skb, ifa,
3899                                                 NETLINK_CB(cb->skb).portid,
3900                                                 cb->nlh->nlmsg_seq,
3901                                                 RTM_NEWADDR,
3902                                                 NLM_F_MULTI);
3903                         if (err <= 0)
3904                                 break;
3905                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3906                 }
3907                 break;
3908         }
3909         case MULTICAST_ADDR:
3910                 /* multicast address */
3911                 for (ifmca = idev->mc_list; ifmca;
3912                      ifmca = ifmca->next, ip_idx++) {
3913                         if (ip_idx < s_ip_idx)
3914                                 continue;
3915                         err = inet6_fill_ifmcaddr(skb, ifmca,
3916                                                   NETLINK_CB(cb->skb).portid,
3917                                                   cb->nlh->nlmsg_seq,
3918                                                   RTM_GETMULTICAST,
3919                                                   NLM_F_MULTI);
3920                         if (err <= 0)
3921                                 break;
3922                 }
3923                 break;
3924         case ANYCAST_ADDR:
3925                 /* anycast address */
3926                 for (ifaca = idev->ac_list; ifaca;
3927                      ifaca = ifaca->aca_next, ip_idx++) {
3928                         if (ip_idx < s_ip_idx)
3929                                 continue;
3930                         err = inet6_fill_ifacaddr(skb, ifaca,
3931                                                   NETLINK_CB(cb->skb).portid,
3932                                                   cb->nlh->nlmsg_seq,
3933                                                   RTM_GETANYCAST,
3934                                                   NLM_F_MULTI);
3935                         if (err <= 0)
3936                                 break;
3937                 }
3938                 break;
3939         default:
3940                 break;
3941         }
3942         read_unlock_bh(&idev->lock);
3943         *p_ip_idx = ip_idx;
3944         return err;
3945 }
3946
3947 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3948                            enum addr_type_t type)
3949 {
3950         struct net *net = sock_net(skb->sk);
3951         int h, s_h;
3952         int idx, ip_idx;
3953         int s_idx, s_ip_idx;
3954         struct net_device *dev;
3955         struct inet6_dev *idev;
3956         struct hlist_head *head;
3957
3958         s_h = cb->args[0];
3959         s_idx = idx = cb->args[1];
3960         s_ip_idx = ip_idx = cb->args[2];
3961
3962         rcu_read_lock();
3963         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3964         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3965                 idx = 0;
3966                 head = &net->dev_index_head[h];
3967                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3968                         if (idx < s_idx)
3969                                 goto cont;
3970                         if (h > s_h || idx > s_idx)
3971                                 s_ip_idx = 0;
3972                         ip_idx = 0;
3973                         idev = __in6_dev_get(dev);
3974                         if (!idev)
3975                                 goto cont;
3976
3977                         if (in6_dump_addrs(idev, skb, cb, type,
3978                                            s_ip_idx, &ip_idx) <= 0)
3979                                 goto done;
3980 cont:
3981                         idx++;
3982                 }
3983         }
3984 done:
3985         rcu_read_unlock();
3986         cb->args[0] = h;
3987         cb->args[1] = idx;
3988         cb->args[2] = ip_idx;
3989
3990         return skb->len;
3991 }
3992
3993 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3994 {
3995         enum addr_type_t type = UNICAST_ADDR;
3996
3997         return inet6_dump_addr(skb, cb, type);
3998 }
3999
4000 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4001 {
4002         enum addr_type_t type = MULTICAST_ADDR;
4003
4004         return inet6_dump_addr(skb, cb, type);
4005 }
4006
4007
4008 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4009 {
4010         enum addr_type_t type = ANYCAST_ADDR;
4011
4012         return inet6_dump_addr(skb, cb, type);
4013 }
4014
4015 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4016 {
4017         struct net *net = sock_net(in_skb->sk);
4018         struct ifaddrmsg *ifm;
4019         struct nlattr *tb[IFA_MAX+1];
4020         struct in6_addr *addr = NULL, *peer;
4021         struct net_device *dev = NULL;
4022         struct inet6_ifaddr *ifa;
4023         struct sk_buff *skb;
4024         int err;
4025
4026         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4027         if (err < 0)
4028                 goto errout;
4029
4030         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4031         if (addr == NULL) {
4032                 err = -EINVAL;
4033                 goto errout;
4034         }
4035
4036         ifm = nlmsg_data(nlh);
4037         if (ifm->ifa_index)
4038                 dev = __dev_get_by_index(net, ifm->ifa_index);
4039
4040         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4041         if (!ifa) {
4042                 err = -EADDRNOTAVAIL;
4043                 goto errout;
4044         }
4045
4046         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4047         if (!skb) {
4048                 err = -ENOBUFS;
4049                 goto errout_ifa;
4050         }
4051
4052         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4053                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4054         if (err < 0) {
4055                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4056                 WARN_ON(err == -EMSGSIZE);
4057                 kfree_skb(skb);
4058                 goto errout_ifa;
4059         }
4060         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4061 errout_ifa:
4062         in6_ifa_put(ifa);
4063 errout:
4064         return err;
4065 }
4066
4067 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4068 {
4069         struct sk_buff *skb;
4070         struct net *net = dev_net(ifa->idev->dev);
4071         int err = -ENOBUFS;
4072
4073         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4074         if (skb == NULL)
4075                 goto errout;
4076
4077         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4078         if (err < 0) {
4079                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4080                 WARN_ON(err == -EMSGSIZE);
4081                 kfree_skb(skb);
4082                 goto errout;
4083         }
4084         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4085         return;
4086 errout:
4087         if (err < 0)
4088                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4089 }
4090
4091 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4092                                 __s32 *array, int bytes)
4093 {
4094         BUG_ON(bytes < (DEVCONF_MAX * 4));
4095
4096         memset(array, 0, bytes);
4097         array[DEVCONF_FORWARDING] = cnf->forwarding;
4098         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4099         array[DEVCONF_MTU6] = cnf->mtu6;
4100         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4101         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4102         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4103         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4104         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4105         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4106                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4107         array[DEVCONF_RTR_SOLICIT_DELAY] =
4108                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4109         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4110         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4111                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4112         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4113                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4114         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4115         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4116         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4117         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4118         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4119         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4120         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4121         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4122 #ifdef CONFIG_IPV6_ROUTER_PREF
4123         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4124         array[DEVCONF_RTR_PROBE_INTERVAL] =
4125                 jiffies_to_msecs(cnf->rtr_probe_interval);
4126 #ifdef CONFIG_IPV6_ROUTE_INFO
4127         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4128 #endif
4129 #endif
4130         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4131         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4132 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4133         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4134 #endif
4135 #ifdef CONFIG_IPV6_MROUTE
4136         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4137 #endif
4138         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4139         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4140         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4141         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4142         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4143 }
4144
4145 static inline size_t inet6_ifla6_size(void)
4146 {
4147         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4148              + nla_total_size(sizeof(struct ifla_cacheinfo))
4149              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4150              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4151              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4152              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4153 }
4154
4155 static inline size_t inet6_if_nlmsg_size(void)
4156 {
4157         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4158                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4159                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4160                + nla_total_size(4) /* IFLA_MTU */
4161                + nla_total_size(4) /* IFLA_LINK */
4162                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4163 }
4164
4165 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4166                                       int items, int bytes)
4167 {
4168         int i;
4169         int pad = bytes - sizeof(u64) * items;
4170         BUG_ON(pad < 0);
4171
4172         /* Use put_unaligned() because stats may not be aligned for u64. */
4173         put_unaligned(items, &stats[0]);
4174         for (i = 1; i < items; i++)
4175                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4176
4177         memset(&stats[items], 0, pad);
4178 }
4179
4180 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4181                                       int items, int bytes, size_t syncpoff)
4182 {
4183         int i;
4184         int pad = bytes - sizeof(u64) * items;
4185         BUG_ON(pad < 0);
4186
4187         /* Use put_unaligned() because stats may not be aligned for u64. */
4188         put_unaligned(items, &stats[0]);
4189         for (i = 1; i < items; i++)
4190                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4191
4192         memset(&stats[items], 0, pad);
4193 }
4194
4195 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4196                              int bytes)
4197 {
4198         switch (attrtype) {
4199         case IFLA_INET6_STATS:
4200                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4201                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4202                 break;
4203         case IFLA_INET6_ICMP6STATS:
4204                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4205                 break;
4206         }
4207 }
4208
4209 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4210 {
4211         struct nlattr *nla;
4212         struct ifla_cacheinfo ci;
4213
4214         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4215                 goto nla_put_failure;
4216         ci.max_reasm_len = IPV6_MAXPLEN;
4217         ci.tstamp = cstamp_delta(idev->tstamp);
4218         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4219         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4220         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4221                 goto nla_put_failure;
4222         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4223         if (nla == NULL)
4224                 goto nla_put_failure;
4225         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4226
4227         /* XXX - MC not implemented */
4228
4229         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4230         if (nla == NULL)
4231                 goto nla_put_failure;
4232         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4233
4234         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4235         if (nla == NULL)
4236                 goto nla_put_failure;
4237         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4238
4239         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4240         if (nla == NULL)
4241                 goto nla_put_failure;
4242         read_lock_bh(&idev->lock);
4243         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4244         read_unlock_bh(&idev->lock);
4245
4246         return 0;
4247
4248 nla_put_failure:
4249         return -EMSGSIZE;
4250 }
4251
4252 static size_t inet6_get_link_af_size(const struct net_device *dev)
4253 {
4254         if (!__in6_dev_get(dev))
4255                 return 0;
4256
4257         return inet6_ifla6_size();
4258 }
4259
4260 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4261 {
4262         struct inet6_dev *idev = __in6_dev_get(dev);
4263
4264         if (!idev)
4265                 return -ENODATA;
4266
4267         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4268                 return -EMSGSIZE;
4269
4270         return 0;
4271 }
4272
4273 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4274 {
4275         struct inet6_ifaddr *ifp;
4276         struct net_device *dev = idev->dev;
4277         bool update_rs = false;
4278         struct in6_addr ll_addr;
4279
4280         if (token == NULL)
4281                 return -EINVAL;
4282         if (ipv6_addr_any(token))
4283                 return -EINVAL;
4284         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4285                 return -EINVAL;
4286         if (!ipv6_accept_ra(idev))
4287                 return -EINVAL;
4288         if (idev->cnf.rtr_solicits <= 0)
4289                 return -EINVAL;
4290
4291         write_lock_bh(&idev->lock);
4292
4293         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4294         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4295
4296         write_unlock_bh(&idev->lock);
4297
4298         if (!idev->dead && (idev->if_flags & IF_READY) &&
4299             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4300                              IFA_F_OPTIMISTIC)) {
4301
4302                 /* If we're not ready, then normal ifup will take care
4303                  * of this. Otherwise, we need to request our rs here.
4304                  */
4305                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4306                 update_rs = true;
4307         }
4308
4309         write_lock_bh(&idev->lock);
4310
4311         if (update_rs) {
4312                 idev->if_flags |= IF_RS_SENT;
4313                 idev->rs_probes = 1;
4314                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4315         }
4316
4317         /* Well, that's kinda nasty ... */
4318         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4319                 spin_lock(&ifp->lock);
4320                 if (ifp->tokenized) {
4321                         ifp->valid_lft = 0;
4322                         ifp->prefered_lft = 0;
4323                 }
4324                 spin_unlock(&ifp->lock);
4325         }
4326
4327         write_unlock_bh(&idev->lock);
4328         addrconf_verify(0);
4329         return 0;
4330 }
4331
4332 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4333 {
4334         int err = -EINVAL;
4335         struct inet6_dev *idev = __in6_dev_get(dev);
4336         struct nlattr *tb[IFLA_INET6_MAX + 1];
4337
4338         if (!idev)
4339                 return -EAFNOSUPPORT;
4340
4341         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4342                 BUG();
4343
4344         if (tb[IFLA_INET6_TOKEN])
4345                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4346
4347         return err;
4348 }
4349
4350 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4351                              u32 portid, u32 seq, int event, unsigned int flags)
4352 {
4353         struct net_device *dev = idev->dev;
4354         struct ifinfomsg *hdr;
4355         struct nlmsghdr *nlh;
4356         void *protoinfo;
4357
4358         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4359         if (nlh == NULL)
4360                 return -EMSGSIZE;
4361
4362         hdr = nlmsg_data(nlh);
4363         hdr->ifi_family = AF_INET6;
4364         hdr->__ifi_pad = 0;
4365         hdr->ifi_type = dev->type;
4366         hdr->ifi_index = dev->ifindex;
4367         hdr->ifi_flags = dev_get_flags(dev);
4368         hdr->ifi_change = 0;
4369
4370         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4371             (dev->addr_len &&
4372              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4373             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4374             (dev->ifindex != dev->iflink &&
4375              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4376                 goto nla_put_failure;
4377         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4378         if (protoinfo == NULL)
4379                 goto nla_put_failure;
4380
4381         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4382                 goto nla_put_failure;
4383
4384         nla_nest_end(skb, protoinfo);
4385         return nlmsg_end(skb, nlh);
4386
4387 nla_put_failure:
4388         nlmsg_cancel(skb, nlh);
4389         return -EMSGSIZE;
4390 }
4391
4392 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4393 {
4394         struct net *net = sock_net(skb->sk);
4395         int h, s_h;
4396         int idx = 0, s_idx;
4397         struct net_device *dev;
4398         struct inet6_dev *idev;
4399         struct hlist_head *head;
4400
4401         s_h = cb->args[0];
4402         s_idx = cb->args[1];
4403
4404         rcu_read_lock();
4405         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4406                 idx = 0;
4407                 head = &net->dev_index_head[h];
4408                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4409                         if (idx < s_idx)
4410                                 goto cont;
4411                         idev = __in6_dev_get(dev);
4412                         if (!idev)
4413                                 goto cont;
4414                         if (inet6_fill_ifinfo(skb, idev,
4415                                               NETLINK_CB(cb->skb).portid,
4416                                               cb->nlh->nlmsg_seq,
4417                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4418                                 goto out;
4419 cont:
4420                         idx++;
4421                 }
4422         }
4423 out:
4424         rcu_read_unlock();
4425         cb->args[1] = idx;
4426         cb->args[0] = h;
4427
4428         return skb->len;
4429 }
4430
4431 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4432 {
4433         struct sk_buff *skb;
4434         struct net *net = dev_net(idev->dev);
4435         int err = -ENOBUFS;
4436
4437         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4438         if (skb == NULL)
4439                 goto errout;
4440
4441         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4442         if (err < 0) {
4443                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4444                 WARN_ON(err == -EMSGSIZE);
4445                 kfree_skb(skb);
4446                 goto errout;
4447         }
4448         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4449         return;
4450 errout:
4451         if (err < 0)
4452                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4453 }
4454
4455 static inline size_t inet6_prefix_nlmsg_size(void)
4456 {
4457         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4458                + nla_total_size(sizeof(struct in6_addr))
4459                + nla_total_size(sizeof(struct prefix_cacheinfo));
4460 }
4461
4462 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4463                              struct prefix_info *pinfo, u32 portid, u32 seq,
4464                              int event, unsigned int flags)
4465 {
4466         struct prefixmsg *pmsg;
4467         struct nlmsghdr *nlh;
4468         struct prefix_cacheinfo ci;
4469
4470         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4471         if (nlh == NULL)
4472                 return -EMSGSIZE;
4473
4474         pmsg = nlmsg_data(nlh);
4475         pmsg->prefix_family = AF_INET6;
4476         pmsg->prefix_pad1 = 0;
4477         pmsg->prefix_pad2 = 0;
4478         pmsg->prefix_ifindex = idev->dev->ifindex;
4479         pmsg->prefix_len = pinfo->prefix_len;
4480         pmsg->prefix_type = pinfo->type;
4481         pmsg->prefix_pad3 = 0;
4482         pmsg->prefix_flags = 0;
4483         if (pinfo->onlink)
4484                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4485         if (pinfo->autoconf)
4486                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4487
4488         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4489                 goto nla_put_failure;
4490         ci.preferred_time = ntohl(pinfo->prefered);
4491         ci.valid_time = ntohl(pinfo->valid);
4492         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4493                 goto nla_put_failure;
4494         return nlmsg_end(skb, nlh);
4495
4496 nla_put_failure:
4497         nlmsg_cancel(skb, nlh);
4498         return -EMSGSIZE;
4499 }
4500
4501 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4502                          struct prefix_info *pinfo)
4503 {
4504         struct sk_buff *skb;
4505         struct net *net = dev_net(idev->dev);
4506         int err = -ENOBUFS;
4507
4508         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4509         if (skb == NULL)
4510                 goto errout;
4511
4512         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4513         if (err < 0) {
4514                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4515                 WARN_ON(err == -EMSGSIZE);
4516                 kfree_skb(skb);
4517                 goto errout;
4518         }
4519         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4520         return;
4521 errout:
4522         if (err < 0)
4523                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4524 }
4525
4526 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4527 {
4528         write_lock_bh(&ifp->idev->lock);
4529         spin_lock(&ifp->lock);
4530         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4531                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4532             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4533                 ifp->idev->valid_ll_addr_cnt += count;
4534         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4535         spin_unlock(&ifp->lock);
4536         write_unlock_bh(&ifp->idev->lock);
4537 }
4538
4539 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4540 {
4541         struct net *net = dev_net(ifp->idev->dev);
4542
4543         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4544
4545         switch (event) {
4546         case RTM_NEWADDR:
4547                 update_valid_ll_addr_cnt(ifp, 1);
4548
4549                 /*
4550                  * If the address was optimistic
4551                  * we inserted the route at the start of
4552                  * our DAD process, so we don't need
4553                  * to do it again
4554                  */
4555                 if (!(ifp->rt->rt6i_node))
4556                         ip6_ins_rt(ifp->rt);
4557                 if (ifp->idev->cnf.forwarding)
4558                         addrconf_join_anycast(ifp);
4559                 if (!ipv6_addr_any(&ifp->peer_addr))
4560                         addrconf_prefix_route(&ifp->peer_addr, 128,
4561                                               ifp->idev->dev, 0, 0);
4562                 break;
4563         case RTM_DELADDR:
4564                 update_valid_ll_addr_cnt(ifp, -1);
4565
4566                 if (ifp->idev->cnf.forwarding)
4567                         addrconf_leave_anycast(ifp);
4568                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4569                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4570                         struct rt6_info *rt;
4571                         struct net_device *dev = ifp->idev->dev;
4572
4573                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4574                                         dev->ifindex, 1);
4575                         if (rt) {
4576                                 dst_hold(&rt->dst);
4577                                 if (ip6_del_rt(rt))
4578                                         dst_free(&rt->dst);
4579                         }
4580                 }
4581                 dst_hold(&ifp->rt->dst);
4582
4583                 if (ip6_del_rt(ifp->rt))
4584                         dst_free(&ifp->rt->dst);
4585                 break;
4586         }
4587         atomic_inc(&net->ipv6.dev_addr_genid);
4588         rt_genid_bump_ipv6(net);
4589 }
4590
4591 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4592 {
4593         rcu_read_lock_bh();
4594         if (likely(ifp->idev->dead == 0))
4595                 __ipv6_ifa_notify(event, ifp);
4596         rcu_read_unlock_bh();
4597 }
4598
4599 #ifdef CONFIG_SYSCTL
4600
4601 static
4602 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4603                            void __user *buffer, size_t *lenp, loff_t *ppos)
4604 {
4605         int *valp = ctl->data;
4606         int val = *valp;
4607         loff_t pos = *ppos;
4608         struct ctl_table lctl;
4609         int ret;
4610
4611         /*
4612          * ctl->data points to idev->cnf.forwarding, we should
4613          * not modify it until we get the rtnl lock.
4614          */
4615         lctl = *ctl;
4616         lctl.data = &val;
4617
4618         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4619
4620         if (write)
4621                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4622         if (ret)
4623                 *ppos = pos;
4624         return ret;
4625 }
4626
4627 static void dev_disable_change(struct inet6_dev *idev)
4628 {
4629         struct netdev_notifier_info info;
4630
4631         if (!idev || !idev->dev)
4632                 return;
4633
4634         netdev_notifier_info_init(&info, idev->dev);
4635         if (idev->cnf.disable_ipv6)
4636                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4637         else
4638                 addrconf_notify(NULL, NETDEV_UP, &info);
4639 }
4640
4641 static void addrconf_disable_change(struct net *net, __s32 newf)
4642 {
4643         struct net_device *dev;
4644         struct inet6_dev *idev;
4645
4646         rcu_read_lock();
4647         for_each_netdev_rcu(net, dev) {
4648                 idev = __in6_dev_get(dev);
4649                 if (idev) {
4650                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4651                         idev->cnf.disable_ipv6 = newf;
4652                         if (changed)
4653                                 dev_disable_change(idev);
4654                 }
4655         }
4656         rcu_read_unlock();
4657 }
4658
4659 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4660 {
4661         struct net *net;
4662         int old;
4663
4664         if (!rtnl_trylock())
4665                 return restart_syscall();
4666
4667         net = (struct net *)table->extra2;
4668         old = *p;
4669         *p = newf;
4670
4671         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4672                 rtnl_unlock();
4673                 return 0;
4674         }
4675
4676         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4677                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4678                 addrconf_disable_change(net, newf);
4679         } else if ((!newf) ^ (!old))
4680                 dev_disable_change((struct inet6_dev *)table->extra1);
4681
4682         rtnl_unlock();
4683         return 0;
4684 }
4685
4686 static
4687 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4688                             void __user *buffer, size_t *lenp, loff_t *ppos)
4689 {
4690         int *valp = ctl->data;
4691         int val = *valp;
4692         loff_t pos = *ppos;
4693         struct ctl_table lctl;
4694         int ret;
4695
4696         /*
4697          * ctl->data points to idev->cnf.disable_ipv6, we should
4698          * not modify it until we get the rtnl lock.
4699          */
4700         lctl = *ctl;
4701         lctl.data = &val;
4702
4703         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4704
4705         if (write)
4706                 ret = addrconf_disable_ipv6(ctl, valp, val);
4707         if (ret)
4708                 *ppos = pos;
4709         return ret;
4710 }
4711
4712 static struct addrconf_sysctl_table
4713 {
4714         struct ctl_table_header *sysctl_header;
4715         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4716 } addrconf_sysctl __read_mostly = {
4717         .sysctl_header = NULL,
4718         .addrconf_vars = {
4719                 {
4720                         .procname       = "forwarding",
4721                         .data           = &ipv6_devconf.forwarding,
4722                         .maxlen         = sizeof(int),
4723                         .mode           = 0644,
4724                         .proc_handler   = addrconf_sysctl_forward,
4725                 },
4726                 {
4727                         .procname       = "hop_limit",
4728                         .data           = &ipv6_devconf.hop_limit,
4729                         .maxlen         = sizeof(int),
4730                         .mode           = 0644,
4731                         .proc_handler   = proc_dointvec,
4732                 },
4733                 {
4734                         .procname       = "mtu",
4735                         .data           = &ipv6_devconf.mtu6,
4736                         .maxlen         = sizeof(int),
4737                         .mode           = 0644,
4738                         .proc_handler   = proc_dointvec,
4739                 },
4740                 {
4741                         .procname       = "accept_ra",
4742                         .data           = &ipv6_devconf.accept_ra,
4743                         .maxlen         = sizeof(int),
4744                         .mode           = 0644,
4745                         .proc_handler   = proc_dointvec,
4746                 },
4747                 {
4748                         .procname       = "accept_redirects",
4749                         .data           = &ipv6_devconf.accept_redirects,
4750                         .maxlen         = sizeof(int),
4751                         .mode           = 0644,
4752                         .proc_handler   = proc_dointvec,
4753                 },
4754                 {
4755                         .procname       = "autoconf",
4756                         .data           = &ipv6_devconf.autoconf,
4757                         .maxlen         = sizeof(int),
4758                         .mode           = 0644,
4759                         .proc_handler   = proc_dointvec,
4760                 },
4761                 {
4762                         .procname       = "dad_transmits",
4763                         .data           = &ipv6_devconf.dad_transmits,
4764                         .maxlen         = sizeof(int),
4765                         .mode           = 0644,
4766                         .proc_handler   = proc_dointvec,
4767                 },
4768                 {
4769                         .procname       = "router_solicitations",
4770                         .data           = &ipv6_devconf.rtr_solicits,
4771                         .maxlen         = sizeof(int),
4772                         .mode           = 0644,
4773                         .proc_handler   = proc_dointvec,
4774                 },
4775                 {
4776                         .procname       = "router_solicitation_interval",
4777                         .data           = &ipv6_devconf.rtr_solicit_interval,
4778                         .maxlen         = sizeof(int),
4779                         .mode           = 0644,
4780                         .proc_handler   = proc_dointvec_jiffies,
4781                 },
4782                 {
4783                         .procname       = "router_solicitation_delay",
4784                         .data           = &ipv6_devconf.rtr_solicit_delay,
4785                         .maxlen         = sizeof(int),
4786                         .mode           = 0644,
4787                         .proc_handler   = proc_dointvec_jiffies,
4788                 },
4789                 {
4790                         .procname       = "force_mld_version",
4791                         .data           = &ipv6_devconf.force_mld_version,
4792                         .maxlen         = sizeof(int),
4793                         .mode           = 0644,
4794                         .proc_handler   = proc_dointvec,
4795                 },
4796                 {
4797                         .procname       = "mldv1_unsolicited_report_interval",
4798                         .data           =
4799                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4800                         .maxlen         = sizeof(int),
4801                         .mode           = 0644,
4802                         .proc_handler   = proc_dointvec_ms_jiffies,
4803                 },
4804                 {
4805                         .procname       = "mldv2_unsolicited_report_interval",
4806                         .data           =
4807                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4808                         .maxlen         = sizeof(int),
4809                         .mode           = 0644,
4810                         .proc_handler   = proc_dointvec_ms_jiffies,
4811                 },
4812                 {
4813                         .procname       = "use_tempaddr",
4814                         .data           = &ipv6_devconf.use_tempaddr,
4815                         .maxlen         = sizeof(int),
4816                         .mode           = 0644,
4817                         .proc_handler   = proc_dointvec,
4818                 },
4819                 {
4820                         .procname       = "temp_valid_lft",
4821                         .data           = &ipv6_devconf.temp_valid_lft,
4822                         .maxlen         = sizeof(int),
4823                         .mode           = 0644,
4824                         .proc_handler   = proc_dointvec,
4825                 },
4826                 {
4827                         .procname       = "temp_prefered_lft",
4828                         .data           = &ipv6_devconf.temp_prefered_lft,
4829                         .maxlen         = sizeof(int),
4830                         .mode           = 0644,
4831                         .proc_handler   = proc_dointvec,
4832                 },
4833                 {
4834                         .procname       = "regen_max_retry",
4835                         .data           = &ipv6_devconf.regen_max_retry,
4836                         .maxlen         = sizeof(int),
4837                         .mode           = 0644,
4838                         .proc_handler   = proc_dointvec,
4839                 },
4840                 {
4841                         .procname       = "max_desync_factor",
4842                         .data           = &ipv6_devconf.max_desync_factor,
4843                         .maxlen         = sizeof(int),
4844                         .mode           = 0644,
4845                         .proc_handler   = proc_dointvec,
4846                 },
4847                 {
4848                         .procname       = "max_addresses",
4849                         .data           = &ipv6_devconf.max_addresses,
4850                         .maxlen         = sizeof(int),
4851                         .mode           = 0644,
4852                         .proc_handler   = proc_dointvec,
4853                 },
4854                 {
4855                         .procname       = "accept_ra_defrtr",
4856                         .data           = &ipv6_devconf.accept_ra_defrtr,
4857                         .maxlen         = sizeof(int),
4858                         .mode           = 0644,
4859                         .proc_handler   = proc_dointvec,
4860                 },
4861                 {
4862                         .procname       = "accept_ra_pinfo",
4863                         .data           = &ipv6_devconf.accept_ra_pinfo,
4864                         .maxlen         = sizeof(int),
4865                         .mode           = 0644,
4866                         .proc_handler   = proc_dointvec,
4867                 },
4868 #ifdef CONFIG_IPV6_ROUTER_PREF
4869                 {
4870                         .procname       = "accept_ra_rtr_pref",
4871                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4872                         .maxlen         = sizeof(int),
4873                         .mode           = 0644,
4874                         .proc_handler   = proc_dointvec,
4875                 },
4876                 {
4877                         .procname       = "router_probe_interval",
4878                         .data           = &ipv6_devconf.rtr_probe_interval,
4879                         .maxlen         = sizeof(int),
4880                         .mode           = 0644,
4881                         .proc_handler   = proc_dointvec_jiffies,
4882                 },
4883 #ifdef CONFIG_IPV6_ROUTE_INFO
4884                 {
4885                         .procname       = "accept_ra_rt_info_max_plen",
4886                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4887                         .maxlen         = sizeof(int),
4888                         .mode           = 0644,
4889                         .proc_handler   = proc_dointvec,
4890                 },
4891 #endif
4892 #endif
4893                 {
4894                         .procname       = "proxy_ndp",
4895                         .data           = &ipv6_devconf.proxy_ndp,
4896                         .maxlen         = sizeof(int),
4897                         .mode           = 0644,
4898                         .proc_handler   = proc_dointvec,
4899                 },
4900                 {
4901                         .procname       = "accept_source_route",
4902                         .data           = &ipv6_devconf.accept_source_route,
4903                         .maxlen         = sizeof(int),
4904                         .mode           = 0644,
4905                         .proc_handler   = proc_dointvec,
4906                 },
4907 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4908                 {
4909                         .procname       = "optimistic_dad",
4910                         .data           = &ipv6_devconf.optimistic_dad,
4911                         .maxlen         = sizeof(int),
4912                         .mode           = 0644,
4913                         .proc_handler   = proc_dointvec,
4914
4915                 },
4916 #endif
4917 #ifdef CONFIG_IPV6_MROUTE
4918                 {
4919                         .procname       = "mc_forwarding",
4920                         .data           = &ipv6_devconf.mc_forwarding,
4921                         .maxlen         = sizeof(int),
4922                         .mode           = 0444,
4923                         .proc_handler   = proc_dointvec,
4924                 },
4925 #endif
4926                 {
4927                         .procname       = "disable_ipv6",
4928                         .data           = &ipv6_devconf.disable_ipv6,
4929                         .maxlen         = sizeof(int),
4930                         .mode           = 0644,
4931                         .proc_handler   = addrconf_sysctl_disable,
4932                 },
4933                 {
4934                         .procname       = "accept_dad",
4935                         .data           = &ipv6_devconf.accept_dad,
4936                         .maxlen         = sizeof(int),
4937                         .mode           = 0644,
4938                         .proc_handler   = proc_dointvec,
4939                 },
4940                 {
4941                         .procname       = "force_tllao",
4942                         .data           = &ipv6_devconf.force_tllao,
4943                         .maxlen         = sizeof(int),
4944                         .mode           = 0644,
4945                         .proc_handler   = proc_dointvec
4946                 },
4947                 {
4948                         .procname       = "ndisc_notify",
4949                         .data           = &ipv6_devconf.ndisc_notify,
4950                         .maxlen         = sizeof(int),
4951                         .mode           = 0644,
4952                         .proc_handler   = proc_dointvec
4953                 },
4954                 {
4955                         .procname       = "suppress_frag_ndisc",
4956                         .data           = &ipv6_devconf.suppress_frag_ndisc,
4957                         .maxlen         = sizeof(int),
4958                         .mode           = 0644,
4959                         .proc_handler   = proc_dointvec
4960                 },
4961                 {
4962                         /* sentinel */
4963                 }
4964         },
4965 };
4966
4967 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4968                 struct inet6_dev *idev, struct ipv6_devconf *p)
4969 {
4970         int i;
4971         struct addrconf_sysctl_table *t;
4972         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4973
4974         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4975         if (t == NULL)
4976                 goto out;
4977
4978         for (i = 0; t->addrconf_vars[i].data; i++) {
4979                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4980                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4981                 t->addrconf_vars[i].extra2 = net;
4982         }
4983
4984         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4985
4986         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4987         if (t->sysctl_header == NULL)
4988                 goto free;
4989
4990         p->sysctl = t;
4991         return 0;
4992
4993 free:
4994         kfree(t);
4995 out:
4996         return -ENOBUFS;
4997 }
4998
4999 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5000 {
5001         struct addrconf_sysctl_table *t;
5002
5003         if (p->sysctl == NULL)
5004                 return;
5005
5006         t = p->sysctl;
5007         p->sysctl = NULL;
5008         unregister_net_sysctl_table(t->sysctl_header);
5009         kfree(t);
5010 }
5011
5012 static void addrconf_sysctl_register(struct inet6_dev *idev)
5013 {
5014         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5015                               &ndisc_ifinfo_sysctl_change);
5016         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5017                                         idev, &idev->cnf);
5018 }
5019
5020 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5021 {
5022         __addrconf_sysctl_unregister(&idev->cnf);
5023         neigh_sysctl_unregister(idev->nd_parms);
5024 }
5025
5026
5027 #endif
5028
5029 static int __net_init addrconf_init_net(struct net *net)
5030 {
5031         int err = -ENOMEM;
5032         struct ipv6_devconf *all, *dflt;
5033
5034         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5035         if (all == NULL)
5036                 goto err_alloc_all;
5037
5038         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5039         if (dflt == NULL)
5040                 goto err_alloc_dflt;
5041
5042         /* these will be inherited by all namespaces */
5043         dflt->autoconf = ipv6_defaults.autoconf;
5044         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5045
5046         net->ipv6.devconf_all = all;
5047         net->ipv6.devconf_dflt = dflt;
5048
5049 #ifdef CONFIG_SYSCTL
5050         err = __addrconf_sysctl_register(net, "all", NULL, all);
5051         if (err < 0)
5052                 goto err_reg_all;
5053
5054         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5055         if (err < 0)
5056                 goto err_reg_dflt;
5057 #endif
5058         return 0;
5059
5060 #ifdef CONFIG_SYSCTL
5061 err_reg_dflt:
5062         __addrconf_sysctl_unregister(all);
5063 err_reg_all:
5064         kfree(dflt);
5065 #endif
5066 err_alloc_dflt:
5067         kfree(all);
5068 err_alloc_all:
5069         return err;
5070 }
5071
5072 static void __net_exit addrconf_exit_net(struct net *net)
5073 {
5074 #ifdef CONFIG_SYSCTL
5075         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5076         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5077 #endif
5078         if (!net_eq(net, &init_net)) {
5079                 kfree(net->ipv6.devconf_dflt);
5080                 kfree(net->ipv6.devconf_all);
5081         }
5082 }
5083
5084 static struct pernet_operations addrconf_ops = {
5085         .init = addrconf_init_net,
5086         .exit = addrconf_exit_net,
5087 };
5088
5089 static struct rtnl_af_ops inet6_ops = {
5090         .family           = AF_INET6,
5091         .fill_link_af     = inet6_fill_link_af,
5092         .get_link_af_size = inet6_get_link_af_size,
5093         .set_link_af      = inet6_set_link_af,
5094 };
5095
5096 /*
5097  *      Init / cleanup code
5098  */
5099
5100 int __init addrconf_init(void)
5101 {
5102         int i, err;
5103
5104         err = ipv6_addr_label_init();
5105         if (err < 0) {
5106                 pr_crit("%s: cannot initialize default policy table: %d\n",
5107                         __func__, err);
5108                 goto out;
5109         }
5110
5111         err = register_pernet_subsys(&addrconf_ops);
5112         if (err < 0)
5113                 goto out_addrlabel;
5114
5115         /* The addrconf netdev notifier requires that loopback_dev
5116          * has it's ipv6 private information allocated and setup
5117          * before it can bring up and give link-local addresses
5118          * to other devices which are up.
5119          *
5120          * Unfortunately, loopback_dev is not necessarily the first
5121          * entry in the global dev_base list of net devices.  In fact,
5122          * it is likely to be the very last entry on that list.
5123          * So this causes the notifier registry below to try and
5124          * give link-local addresses to all devices besides loopback_dev
5125          * first, then loopback_dev, which cases all the non-loopback_dev
5126          * devices to fail to get a link-local address.
5127          *
5128          * So, as a temporary fix, allocate the ipv6 structure for
5129          * loopback_dev first by hand.
5130          * Longer term, all of the dependencies ipv6 has upon the loopback
5131          * device and it being up should be removed.
5132          */
5133         rtnl_lock();
5134         if (!ipv6_add_dev(init_net.loopback_dev))
5135                 err = -ENOMEM;
5136         rtnl_unlock();
5137         if (err)
5138                 goto errlo;
5139
5140         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5141                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5142
5143         register_netdevice_notifier(&ipv6_dev_notf);
5144
5145         addrconf_verify(0);
5146
5147         err = rtnl_af_register(&inet6_ops);
5148         if (err < 0)
5149                 goto errout_af;
5150
5151         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5152                               NULL);
5153         if (err < 0)
5154                 goto errout;
5155
5156         /* Only the first call to __rtnl_register can fail */
5157         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5158         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5159         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5160                         inet6_dump_ifaddr, NULL);
5161         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5162                         inet6_dump_ifmcaddr, NULL);
5163         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5164                         inet6_dump_ifacaddr, NULL);
5165         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5166                         inet6_netconf_dump_devconf, NULL);
5167
5168         ipv6_addr_label_rtnl_register();
5169
5170         return 0;
5171 errout:
5172         rtnl_af_unregister(&inet6_ops);
5173 errout_af:
5174         unregister_netdevice_notifier(&ipv6_dev_notf);
5175 errlo:
5176         unregister_pernet_subsys(&addrconf_ops);
5177 out_addrlabel:
5178         ipv6_addr_label_cleanup();
5179 out:
5180         return err;
5181 }
5182
5183 void addrconf_cleanup(void)
5184 {
5185         struct net_device *dev;
5186         int i;
5187
5188         unregister_netdevice_notifier(&ipv6_dev_notf);
5189         unregister_pernet_subsys(&addrconf_ops);
5190         ipv6_addr_label_cleanup();
5191
5192         rtnl_lock();
5193
5194         __rtnl_af_unregister(&inet6_ops);
5195
5196         /* clean dev list */
5197         for_each_netdev(&init_net, dev) {
5198                 if (__in6_dev_get(dev) == NULL)
5199                         continue;
5200                 addrconf_ifdown(dev, 1);
5201         }
5202         addrconf_ifdown(init_net.loopback_dev, 2);
5203
5204         /*
5205          *      Check hash table.
5206          */
5207         spin_lock_bh(&addrconf_hash_lock);
5208         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5209                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5210         spin_unlock_bh(&addrconf_hash_lock);
5211
5212         del_timer(&addr_chk_timer);
5213         rtnl_unlock();
5214 }