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         neigh_parms_data_state_setall(idev->nd_parms);
 838
 839         ifa->addr = *addr;
 840         if (peer_addr)
 841                 ifa->peer_addr = *peer_addr;
 842
 843         spin_lock_init(&ifa->lock);
 844         spin_lock_init(&ifa->state_lock);
 845         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 846                     (unsigned long)ifa);
 847         INIT_HLIST_NODE(&ifa->addr_lst);
 848         ifa->scope = scope;
 849         ifa->prefix_len = pfxlen;
 850         ifa->flags = flags | IFA_F_TENTATIVE;
 851         ifa->valid_lft = valid_lft;
 852         ifa->prefered_lft = prefered_lft;
 853         ifa->cstamp = ifa->tstamp = jiffies;
 854         ifa->tokenized = false;
 855
 856         ifa->rt = rt;
 857
 858         ifa->idev = idev;
 859         in6_dev_hold(idev);
 860         /* For caller */
 861         in6_ifa_hold(ifa);
 862
 863         /* Add to big hash table */
 864         hash = inet6_addr_hash(addr);
 865
 866         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 867         spin_unlock(&addrconf_hash_lock);
 868
 869         write_lock(&idev->lock);
 870         /* Add to inet6_dev unicast addr list. */
 871         ipv6_link_dev_addr(idev, ifa);
 872
 873         if (ifa->flags&IFA_F_TEMPORARY) {
 874                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 875                 in6_ifa_hold(ifa);
 876         }
 877
 878         in6_ifa_hold(ifa);
 879         write_unlock(&idev->lock);
 880 out2:
 881         rcu_read_unlock_bh();
 882
 883         if (likely(err == 0))
 884                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 885         else {
 886                 kfree(ifa);
 887                 ifa = ERR_PTR(err);
 888         }
 889
 890         return ifa;
 891 out:
 892         spin_unlock(&addrconf_hash_lock);
 893         goto out2;
 894 }
 895
 896 /* This function wants to get referenced ifp and releases it before return */
 897
 898 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 899 {
 900         struct inet6_ifaddr *ifa, *ifn;
 901         struct inet6_dev *idev = ifp->idev;
 902         int state;
 903         int deleted = 0, onlink = 0;
 904         unsigned long expires = jiffies;
 905
 906         spin_lock_bh(&ifp->state_lock);
 907         state = ifp->state;
 908         ifp->state = INET6_IFADDR_STATE_DEAD;
 909         spin_unlock_bh(&ifp->state_lock);
 910
 911         if (state == INET6_IFADDR_STATE_DEAD)
 912                 goto out;
 913
 914         spin_lock_bh(&addrconf_hash_lock);
 915         hlist_del_init_rcu(&ifp->addr_lst);
 916         spin_unlock_bh(&addrconf_hash_lock);
 917
 918         write_lock_bh(&idev->lock);
 919
 920         if (ifp->flags&IFA_F_TEMPORARY) {
 921                 list_del(&ifp->tmp_list);
 922                 if (ifp->ifpub) {
 923                         in6_ifa_put(ifp->ifpub);
 924                         ifp->ifpub = NULL;
 925                 }
 926                 __in6_ifa_put(ifp);
 927         }
 928
 929         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 930                 if (ifa == ifp) {
 931                         list_del_init(&ifp->if_list);
 932                         __in6_ifa_put(ifp);
 933
 934                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 935                                 break;
 936                         deleted = 1;
 937                         continue;
 938                 } else if (ifp->flags & IFA_F_PERMANENT) {
 939                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 940                                               ifp->prefix_len)) {
 941                                 if (ifa->flags & IFA_F_PERMANENT) {
 942                                         onlink = 1;
 943                                         if (deleted)
 944                                                 break;
 945                                 } else {
 946                                         unsigned long lifetime;
 947
 948                                         if (!onlink)
 949                                                 onlink = -1;
 950
 951                                         spin_lock(&ifa->lock);
 952
 953                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 954                                         /*
 955                                          * Note: Because this address is
 956                                          * not permanent, lifetime <
 957                                          * LONG_MAX / HZ here.
 958                                          */
 959                                         if (time_before(expires,
 960                                                         ifa->tstamp + lifetime * HZ))
 961                                                 expires = ifa->tstamp + lifetime * HZ;
 962                                         spin_unlock(&ifa->lock);
 963                                 }
 964                         }
 965                 }
 966         }
 967         write_unlock_bh(&idev->lock);
 968
 969         addrconf_del_dad_timer(ifp);
 970
 971         ipv6_ifa_notify(RTM_DELADDR, ifp);
 972
 973         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
 974
 975         /*
 976          * Purge or update corresponding prefix
 977          *
 978          * 1) we don't purge prefix here if address was not permanent.
 979          *    prefix is managed by its own lifetime.
 980          * 2) if there're no addresses, delete prefix.
 981          * 3) if there're still other permanent address(es),
 982          *    corresponding prefix is still permanent.
 983          * 4) otherwise, update prefix lifetime to the
 984          *    longest valid lifetime among the corresponding
 985          *    addresses on the device.
 986          *    Note: subsequent RA will update lifetime.
 987          *
 988          * --yoshfuji
 989          */
 990         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 991                 struct in6_addr prefix;
 992                 struct rt6_info *rt;
 993
 994                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 995
 996                 rt = addrconf_get_prefix_route(&prefix,
 997                                                ifp->prefix_len,
 998                                                ifp->idev->dev,
 999                                                0, RTF_GATEWAY | RTF_DEFAULT);
1000
1001                 if (rt) {
1002                         if (onlink == 0) {
1003                                 ip6_del_rt(rt);
1004                                 rt = NULL;
1005                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1006                                 rt6_set_expires(rt, expires);
1007                         }
1008                 }
1009                 ip6_rt_put(rt);
1010         }
1011
1012         /* clean up prefsrc entries */
1013         rt6_remove_prefsrc(ifp);
1014 out:
1015         in6_ifa_put(ifp);
1016 }
1017
1018 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1019 {
1020         struct inet6_dev *idev = ifp->idev;
1021         struct in6_addr addr, *tmpaddr;
1022         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1023         unsigned long regen_advance;
1024         int tmp_plen;
1025         int ret = 0;
1026         u32 addr_flags;
1027         unsigned long now = jiffies;
1028
1029         write_lock_bh(&idev->lock);
1030         if (ift) {
1031                 spin_lock_bh(&ift->lock);
1032                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1033                 spin_unlock_bh(&ift->lock);
1034                 tmpaddr = &addr;
1035         } else {
1036                 tmpaddr = NULL;
1037         }
1038 retry:
1039         in6_dev_hold(idev);
1040         if (idev->cnf.use_tempaddr <= 0) {
1041                 write_unlock_bh(&idev->lock);
1042                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1043                 in6_dev_put(idev);
1044                 ret = -1;
1045                 goto out;
1046         }
1047         spin_lock_bh(&ifp->lock);
1048         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1049                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1050                 spin_unlock_bh(&ifp->lock);
1051                 write_unlock_bh(&idev->lock);
1052                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1053                         __func__);
1054                 in6_dev_put(idev);
1055                 ret = -1;
1056                 goto out;
1057         }
1058         in6_ifa_hold(ifp);
1059         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1060         __ipv6_try_regen_rndid(idev, tmpaddr);
1061         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1062         age = (now - ifp->tstamp) / HZ;
1063         tmp_valid_lft = min_t(__u32,
1064                               ifp->valid_lft,
1065                               idev->cnf.temp_valid_lft + age);
1066         tmp_prefered_lft = min_t(__u32,
1067                                  ifp->prefered_lft,
1068                                  idev->cnf.temp_prefered_lft + age -
1069                                  idev->cnf.max_desync_factor);
1070         tmp_plen = ifp->prefix_len;
1071         tmp_tstamp = ifp->tstamp;
1072         spin_unlock_bh(&ifp->lock);
1073
1074         regen_advance = idev->cnf.regen_max_retry *
1075                         idev->cnf.dad_transmits *
1076                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1077         write_unlock_bh(&idev->lock);
1078
1079         /* A temporary address is created only if this calculated Preferred
1080          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1081          * an implementation must not create a temporary address with a zero
1082          * Preferred Lifetime.
1083          */
1084         if (tmp_prefered_lft <= regen_advance) {
1085                 in6_ifa_put(ifp);
1086                 in6_dev_put(idev);
1087                 ret = -1;
1088                 goto out;
1089         }
1090
1091         addr_flags = IFA_F_TEMPORARY;
1092         /* set in addrconf_prefix_rcv() */
1093         if (ifp->flags & IFA_F_OPTIMISTIC)
1094                 addr_flags |= IFA_F_OPTIMISTIC;
1095
1096         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1097                             ipv6_addr_scope(&addr), addr_flags,
1098                             tmp_valid_lft, tmp_prefered_lft);
1099         if (IS_ERR(ift)) {
1100                 in6_ifa_put(ifp);
1101                 in6_dev_put(idev);
1102                 pr_info("%s: retry temporary address regeneration\n", __func__);
1103                 tmpaddr = &addr;
1104                 write_lock_bh(&idev->lock);
1105                 goto retry;
1106         }
1107
1108         spin_lock_bh(&ift->lock);
1109         ift->ifpub = ifp;
1110         ift->cstamp = now;
1111         ift->tstamp = tmp_tstamp;
1112         spin_unlock_bh(&ift->lock);
1113
1114         addrconf_dad_start(ift);
1115         in6_ifa_put(ift);
1116         in6_dev_put(idev);
1117 out:
1118         return ret;
1119 }
1120
1121 /*
1122  *      Choose an appropriate source address (RFC3484)
1123  */
1124 enum {
1125         IPV6_SADDR_RULE_INIT = 0,
1126         IPV6_SADDR_RULE_LOCAL,
1127         IPV6_SADDR_RULE_SCOPE,
1128         IPV6_SADDR_RULE_PREFERRED,
1129 #ifdef CONFIG_IPV6_MIP6
1130         IPV6_SADDR_RULE_HOA,
1131 #endif
1132         IPV6_SADDR_RULE_OIF,
1133         IPV6_SADDR_RULE_LABEL,
1134         IPV6_SADDR_RULE_PRIVACY,
1135         IPV6_SADDR_RULE_ORCHID,
1136         IPV6_SADDR_RULE_PREFIX,
1137         IPV6_SADDR_RULE_MAX
1138 };
1139
1140 struct ipv6_saddr_score {
1141         int                     rule;
1142         int                     addr_type;
1143         struct inet6_ifaddr     *ifa;
1144         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1145         int                     scopedist;
1146         int                     matchlen;
1147 };
1148
1149 struct ipv6_saddr_dst {
1150         const struct in6_addr *addr;
1151         int ifindex;
1152         int scope;
1153         int label;
1154         unsigned int prefs;
1155 };
1156
1157 static inline int ipv6_saddr_preferred(int type)
1158 {
1159         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1160                 return 1;
1161         return 0;
1162 }
1163
1164 static int ipv6_get_saddr_eval(struct net *net,
1165                                struct ipv6_saddr_score *score,
1166                                struct ipv6_saddr_dst *dst,
1167                                int i)
1168 {
1169         int ret;
1170
1171         if (i <= score->rule) {
1172                 switch (i) {
1173                 case IPV6_SADDR_RULE_SCOPE:
1174                         ret = score->scopedist;
1175                         break;
1176                 case IPV6_SADDR_RULE_PREFIX:
1177                         ret = score->matchlen;
1178                         break;
1179                 default:
1180                         ret = !!test_bit(i, score->scorebits);
1181                 }
1182                 goto out;
1183         }
1184
1185         switch (i) {
1186         case IPV6_SADDR_RULE_INIT:
1187                 /* Rule 0: remember if hiscore is not ready yet */
1188                 ret = !!score->ifa;
1189                 break;
1190         case IPV6_SADDR_RULE_LOCAL:
1191                 /* Rule 1: Prefer same address */
1192                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1193                 break;
1194         case IPV6_SADDR_RULE_SCOPE:
1195                 /* Rule 2: Prefer appropriate scope
1196                  *
1197                  *      ret
1198                  *       ^
1199                  *    -1 |  d 15
1200                  *    ---+--+-+---> scope
1201                  *       |
1202                  *       |             d is scope of the destination.
1203                  *  B-d  |  \
1204                  *       |   \      <- smaller scope is better if
1205                  *  B-15 |    \        if scope is enough for destinaion.
1206                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1207                  * d-C-1 | /
1208                  *       |/         <- greater is better
1209                  *   -C  /             if scope is not enough for destination.
1210                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1211                  *
1212                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1213                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1214                  * Assume B = 0 and we get C > 29.
1215                  */
1216                 ret = __ipv6_addr_src_scope(score->addr_type);
1217                 if (ret >= dst->scope)
1218                         ret = -ret;
1219                 else
1220                         ret -= 128;     /* 30 is enough */
1221                 score->scopedist = ret;
1222                 break;
1223         case IPV6_SADDR_RULE_PREFERRED:
1224                 /* Rule 3: Avoid deprecated and optimistic addresses */
1225                 ret = ipv6_saddr_preferred(score->addr_type) ||
1226                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1227                 break;
1228 #ifdef CONFIG_IPV6_MIP6
1229         case IPV6_SADDR_RULE_HOA:
1230             {
1231                 /* Rule 4: Prefer home address */
1232                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1233                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1234                 break;
1235             }
1236 #endif
1237         case IPV6_SADDR_RULE_OIF:
1238                 /* Rule 5: Prefer outgoing interface */
1239                 ret = (!dst->ifindex ||
1240                        dst->ifindex == score->ifa->idev->dev->ifindex);
1241                 break;
1242         case IPV6_SADDR_RULE_LABEL:
1243                 /* Rule 6: Prefer matching label */
1244                 ret = ipv6_addr_label(net,
1245                                       &score->ifa->addr, score->addr_type,
1246                                       score->ifa->idev->dev->ifindex) == dst->label;
1247                 break;
1248         case IPV6_SADDR_RULE_PRIVACY:
1249             {
1250                 /* Rule 7: Prefer public address
1251                  * Note: prefer temporary address if use_tempaddr >= 2
1252                  */
1253                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1254                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1255                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1256                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1257                 break;
1258             }
1259         case IPV6_SADDR_RULE_ORCHID:
1260                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1261                  *          non-ORCHID vs non-ORCHID
1262                  */
1263                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1264                         ipv6_addr_orchid(dst->addr));
1265                 break;
1266         case IPV6_SADDR_RULE_PREFIX:
1267                 /* Rule 8: Use longest matching prefix */
1268                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1269                 if (ret > score->ifa->prefix_len)
1270                         ret = score->ifa->prefix_len;
1271                 score->matchlen = ret;
1272                 break;
1273         default:
1274                 ret = 0;
1275         }
1276
1277         if (ret)
1278                 __set_bit(i, score->scorebits);
1279         score->rule = i;
1280 out:
1281         return ret;
1282 }
1283
1284 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1285                        const struct in6_addr *daddr, unsigned int prefs,
1286                        struct in6_addr *saddr)
1287 {
1288         struct ipv6_saddr_score scores[2],
1289                                 *score = &scores[0], *hiscore = &scores[1];
1290         struct ipv6_saddr_dst dst;
1291         struct net_device *dev;
1292         int dst_type;
1293
1294         dst_type = __ipv6_addr_type(daddr);
1295         dst.addr = daddr;
1296         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1297         dst.scope = __ipv6_addr_src_scope(dst_type);
1298         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1299         dst.prefs = prefs;
1300
1301         hiscore->rule = -1;
1302         hiscore->ifa = NULL;
1303
1304         rcu_read_lock();
1305
1306         for_each_netdev_rcu(net, dev) {
1307                 struct inet6_dev *idev;
1308
1309                 /* Candidate Source Address (section 4)
1310                  *  - multicast and link-local destination address,
1311                  *    the set of candidate source address MUST only
1312                  *    include addresses assigned to interfaces
1313                  *    belonging to the same link as the outgoing
1314                  *    interface.
1315                  * (- For site-local destination addresses, the
1316                  *    set of candidate source addresses MUST only
1317                  *    include addresses assigned to interfaces
1318                  *    belonging to the same site as the outgoing
1319                  *    interface.)
1320                  */
1321                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1322                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1323                     dst.ifindex && dev->ifindex != dst.ifindex)
1324                         continue;
1325
1326                 idev = __in6_dev_get(dev);
1327                 if (!idev)
1328                         continue;
1329
1330                 read_lock_bh(&idev->lock);
1331                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1332                         int i;
1333
1334                         /*
1335                          * - Tentative Address (RFC2462 section 5.4)
1336                          *  - A tentative address is not considered
1337                          *    "assigned to an interface" in the traditional
1338                          *    sense, unless it is also flagged as optimistic.
1339                          * - Candidate Source Address (section 4)
1340                          *  - In any case, anycast addresses, multicast
1341                          *    addresses, and the unspecified address MUST
1342                          *    NOT be included in a candidate set.
1343                          */
1344                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1345                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1346                                 continue;
1347
1348                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1349
1350                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1351                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1352                                 LIMIT_NETDEBUG(KERN_DEBUG
1353                                                "ADDRCONF: unspecified / multicast address "
1354                                                "assigned as unicast address on %s",
1355                                                dev->name);
1356                                 continue;
1357                         }
1358
1359                         score->rule = -1;
1360                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1361
1362                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1363                                 int minihiscore, miniscore;
1364
1365                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1366                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1367
1368                                 if (minihiscore > miniscore) {
1369                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1370                                             score->scopedist > 0) {
1371                                                 /*
1372                                                  * special case:
1373                                                  * each remaining entry
1374                                                  * has too small (not enough)
1375                                                  * scope, because ifa entries
1376                                                  * are sorted by their scope
1377                                                  * values.
1378                                                  */
1379                                                 goto try_nextdev;
1380                                         }
1381                                         break;
1382                                 } else if (minihiscore < miniscore) {
1383                                         if (hiscore->ifa)
1384                                                 in6_ifa_put(hiscore->ifa);
1385
1386                                         in6_ifa_hold(score->ifa);
1387
1388                                         swap(hiscore, score);
1389
1390                                         /* restore our iterator */
1391                                         score->ifa = hiscore->ifa;
1392
1393                                         break;
1394                                 }
1395                         }
1396                 }
1397 try_nextdev:
1398                 read_unlock_bh(&idev->lock);
1399         }
1400         rcu_read_unlock();
1401
1402         if (!hiscore->ifa)
1403                 return -EADDRNOTAVAIL;
1404
1405         *saddr = hiscore->ifa->addr;
1406         in6_ifa_put(hiscore->ifa);
1407         return 0;
1408 }
1409 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1410
1411 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1412                       u32 banned_flags)
1413 {
1414         struct inet6_ifaddr *ifp;
1415         int err = -EADDRNOTAVAIL;
1416
1417         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1418                 if (ifp->scope == IFA_LINK &&
1419                     !(ifp->flags & banned_flags)) {
1420                         *addr = ifp->addr;
1421                         err = 0;
1422                         break;
1423                 }
1424         }
1425         return err;
1426 }
1427
1428 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1429                     u32 banned_flags)
1430 {
1431         struct inet6_dev *idev;
1432         int err = -EADDRNOTAVAIL;
1433
1434         rcu_read_lock();
1435         idev = __in6_dev_get(dev);
1436         if (idev) {
1437                 read_lock_bh(&idev->lock);
1438                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1439                 read_unlock_bh(&idev->lock);
1440         }
1441         rcu_read_unlock();
1442         return err;
1443 }
1444
1445 static int ipv6_count_addresses(struct inet6_dev *idev)
1446 {
1447         int cnt = 0;
1448         struct inet6_ifaddr *ifp;
1449
1450         read_lock_bh(&idev->lock);
1451         list_for_each_entry(ifp, &idev->addr_list, if_list)
1452                 cnt++;
1453         read_unlock_bh(&idev->lock);
1454         return cnt;
1455 }
1456
1457 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1458                   const struct net_device *dev, int strict)
1459 {
1460         struct inet6_ifaddr *ifp;
1461         unsigned int hash = inet6_addr_hash(addr);
1462
1463         rcu_read_lock_bh();
1464         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1465                 if (!net_eq(dev_net(ifp->idev->dev), net))
1466                         continue;
1467                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1468                     !(ifp->flags&IFA_F_TENTATIVE) &&
1469                     (dev == NULL || ifp->idev->dev == dev ||
1470                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1471                         rcu_read_unlock_bh();
1472                         return 1;
1473                 }
1474         }
1475
1476         rcu_read_unlock_bh();
1477         return 0;
1478 }
1479 EXPORT_SYMBOL(ipv6_chk_addr);
1480
1481 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1482                                struct net_device *dev)
1483 {
1484         unsigned int hash = inet6_addr_hash(addr);
1485         struct inet6_ifaddr *ifp;
1486
1487         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1488                 if (!net_eq(dev_net(ifp->idev->dev), net))
1489                         continue;
1490                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1491                         if (dev == NULL || ifp->idev->dev == dev)
1492                                 return true;
1493                 }
1494         }
1495         return false;
1496 }
1497
1498 /* Compares an address/prefix_len with addresses on device @dev.
1499  * If one is found it returns true.
1500  */
1501 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1502         const unsigned int prefix_len, struct net_device *dev)
1503 {
1504         struct inet6_dev *idev;
1505         struct inet6_ifaddr *ifa;
1506         bool ret = false;
1507
1508         rcu_read_lock();
1509         idev = __in6_dev_get(dev);
1510         if (idev) {
1511                 read_lock_bh(&idev->lock);
1512                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1513                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1514                         if (ret)
1515                                 break;
1516                 }
1517                 read_unlock_bh(&idev->lock);
1518         }
1519         rcu_read_unlock();
1520
1521         return ret;
1522 }
1523 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1524
1525 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1526 {
1527         struct inet6_dev *idev;
1528         struct inet6_ifaddr *ifa;
1529         int     onlink;
1530
1531         onlink = 0;
1532         rcu_read_lock();
1533         idev = __in6_dev_get(dev);
1534         if (idev) {
1535                 read_lock_bh(&idev->lock);
1536                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1537                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1538                                                    ifa->prefix_len);
1539                         if (onlink)
1540                                 break;
1541                 }
1542                 read_unlock_bh(&idev->lock);
1543         }
1544         rcu_read_unlock();
1545         return onlink;
1546 }
1547 EXPORT_SYMBOL(ipv6_chk_prefix);
1548
1549 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1550                                      struct net_device *dev, int strict)
1551 {
1552         struct inet6_ifaddr *ifp, *result = NULL;
1553         unsigned int hash = inet6_addr_hash(addr);
1554
1555         rcu_read_lock_bh();
1556         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1557                 if (!net_eq(dev_net(ifp->idev->dev), net))
1558                         continue;
1559                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1560                         if (dev == NULL || ifp->idev->dev == dev ||
1561                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1562                                 result = ifp;
1563                                 in6_ifa_hold(ifp);
1564                                 break;
1565                         }
1566                 }
1567         }
1568         rcu_read_unlock_bh();
1569
1570         return result;
1571 }
1572
1573 /* Gets referenced address, destroys ifaddr */
1574
1575 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1576 {
1577         if (ifp->flags&IFA_F_PERMANENT) {
1578                 spin_lock_bh(&ifp->lock);
1579                 addrconf_del_dad_timer(ifp);
1580                 ifp->flags |= IFA_F_TENTATIVE;
1581                 if (dad_failed)
1582                         ifp->flags |= IFA_F_DADFAILED;
1583                 spin_unlock_bh(&ifp->lock);
1584                 if (dad_failed)
1585                         ipv6_ifa_notify(0, ifp);
1586                 in6_ifa_put(ifp);
1587         } else if (ifp->flags&IFA_F_TEMPORARY) {
1588                 struct inet6_ifaddr *ifpub;
1589                 spin_lock_bh(&ifp->lock);
1590                 ifpub = ifp->ifpub;
1591                 if (ifpub) {
1592                         in6_ifa_hold(ifpub);
1593                         spin_unlock_bh(&ifp->lock);
1594                         ipv6_create_tempaddr(ifpub, ifp);
1595                         in6_ifa_put(ifpub);
1596                 } else {
1597                         spin_unlock_bh(&ifp->lock);
1598                 }
1599                 ipv6_del_addr(ifp);
1600         } else
1601                 ipv6_del_addr(ifp);
1602 }
1603
1604 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1605 {
1606         int err = -ENOENT;
1607
1608         spin_lock(&ifp->state_lock);
1609         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1610                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1611                 err = 0;
1612         }
1613         spin_unlock(&ifp->state_lock);
1614
1615         return err;
1616 }
1617
1618 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1619 {
1620         struct inet6_dev *idev = ifp->idev;
1621
1622         if (addrconf_dad_end(ifp)) {
1623                 in6_ifa_put(ifp);
1624                 return;
1625         }
1626
1627         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1628                              ifp->idev->dev->name, &ifp->addr);
1629
1630         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1631                 struct in6_addr addr;
1632
1633                 addr.s6_addr32[0] = htonl(0xfe800000);
1634                 addr.s6_addr32[1] = 0;
1635
1636                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1637                     ipv6_addr_equal(&ifp->addr, &addr)) {
1638                         /* DAD failed for link-local based on MAC address */
1639                         idev->cnf.disable_ipv6 = 1;
1640
1641                         pr_info("%s: IPv6 being disabled!\n",
1642                                 ifp->idev->dev->name);
1643                 }
1644         }
1645
1646         addrconf_dad_stop(ifp, 1);
1647 }
1648
1649 /* Join to solicited addr multicast group. */
1650
1651 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1652 {
1653         struct in6_addr maddr;
1654
1655         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1656                 return;
1657
1658         addrconf_addr_solict_mult(addr, &maddr);
1659         ipv6_dev_mc_inc(dev, &maddr);
1660 }
1661
1662 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1663 {
1664         struct in6_addr maddr;
1665
1666         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1667                 return;
1668
1669         addrconf_addr_solict_mult(addr, &maddr);
1670         __ipv6_dev_mc_dec(idev, &maddr);
1671 }
1672
1673 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1674 {
1675         struct in6_addr addr;
1676         if (ifp->prefix_len == 127) /* RFC 6164 */
1677                 return;
1678         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1679         if (ipv6_addr_any(&addr))
1680                 return;
1681         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1682 }
1683
1684 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1685 {
1686         struct in6_addr addr;
1687         if (ifp->prefix_len == 127) /* RFC 6164 */
1688                 return;
1689         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1690         if (ipv6_addr_any(&addr))
1691                 return;
1692         __ipv6_dev_ac_dec(ifp->idev, &addr);
1693 }
1694
1695 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1696 {
1697         if (dev->addr_len != ETH_ALEN)
1698                 return -1;
1699         memcpy(eui, dev->dev_addr, 3);
1700         memcpy(eui + 5, dev->dev_addr + 3, 3);
1701
1702         /*
1703          * The zSeries OSA network cards can be shared among various
1704          * OS instances, but the OSA cards have only one MAC address.
1705          * This leads to duplicate address conflicts in conjunction
1706          * with IPv6 if more than one instance uses the same card.
1707          *
1708          * The driver for these cards can deliver a unique 16-bit
1709          * identifier for each instance sharing the same card.  It is
1710          * placed instead of 0xFFFE in the interface identifier.  The
1711          * "u" bit of the interface identifier is not inverted in this
1712          * case.  Hence the resulting interface identifier has local
1713          * scope according to RFC2373.
1714          */
1715         if (dev->dev_id) {
1716                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1717                 eui[4] = dev->dev_id & 0xFF;
1718         } else {
1719                 eui[3] = 0xFF;
1720                 eui[4] = 0xFE;
1721                 eui[0] ^= 2;
1722         }
1723         return 0;
1724 }
1725
1726 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1727 {
1728         if (dev->addr_len != IEEE802154_ADDR_LEN)
1729                 return -1;
1730         memcpy(eui, dev->dev_addr, 8);
1731         eui[0] ^= 2;
1732         return 0;
1733 }
1734
1735 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1736 {
1737         union fwnet_hwaddr *ha;
1738
1739         if (dev->addr_len != FWNET_ALEN)
1740                 return -1;
1741
1742         ha = (union fwnet_hwaddr *)dev->dev_addr;
1743
1744         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1745         eui[0] ^= 2;
1746         return 0;
1747 }
1748
1749 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1750 {
1751         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1752         if (dev->addr_len != ARCNET_ALEN)
1753                 return -1;
1754         memset(eui, 0, 7);
1755         eui[7] = *(u8 *)dev->dev_addr;
1756         return 0;
1757 }
1758
1759 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1760 {
1761         if (dev->addr_len != INFINIBAND_ALEN)
1762                 return -1;
1763         memcpy(eui, dev->dev_addr + 12, 8);
1764         eui[0] |= 2;
1765         return 0;
1766 }
1767
1768 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1769 {
1770         if (addr == 0)
1771                 return -1;
1772         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1773                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1774                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1775                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1776                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1777                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1778         eui[1] = 0;
1779         eui[2] = 0x5E;
1780         eui[3] = 0xFE;
1781         memcpy(eui + 4, &addr, 4);
1782         return 0;
1783 }
1784
1785 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1786 {
1787         if (dev->priv_flags & IFF_ISATAP)
1788                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1789         return -1;
1790 }
1791
1792 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1793 {
1794         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1795 }
1796
1797 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1798 {
1799         memcpy(eui, dev->perm_addr, 3);
1800         memcpy(eui + 5, dev->perm_addr + 3, 3);
1801         eui[3] = 0xFF;
1802         eui[4] = 0xFE;
1803         eui[0] ^= 2;
1804         return 0;
1805 }
1806
1807 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1808 {
1809         switch (dev->type) {
1810         case ARPHRD_ETHER:
1811         case ARPHRD_FDDI:
1812                 return addrconf_ifid_eui48(eui, dev);
1813         case ARPHRD_ARCNET:
1814                 return addrconf_ifid_arcnet(eui, dev);
1815         case ARPHRD_INFINIBAND:
1816                 return addrconf_ifid_infiniband(eui, dev);
1817         case ARPHRD_SIT:
1818                 return addrconf_ifid_sit(eui, dev);
1819         case ARPHRD_IPGRE:
1820                 return addrconf_ifid_gre(eui, dev);
1821         case ARPHRD_IEEE802154:
1822                 return addrconf_ifid_eui64(eui, dev);
1823         case ARPHRD_IEEE1394:
1824                 return addrconf_ifid_ieee1394(eui, dev);
1825         case ARPHRD_TUNNEL6:
1826                 return addrconf_ifid_ip6tnl(eui, dev);
1827         }
1828         return -1;
1829 }
1830
1831 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1832 {
1833         int err = -1;
1834         struct inet6_ifaddr *ifp;
1835
1836         read_lock_bh(&idev->lock);
1837         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1838                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1839                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1840                         err = 0;
1841                         break;
1842                 }
1843         }
1844         read_unlock_bh(&idev->lock);
1845         return err;
1846 }
1847
1848 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1849 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1850 {
1851 regen:
1852         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1853         idev->rndid[0] &= ~0x02;
1854
1855         /*
1856          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1857          * check if generated address is not inappropriate
1858          *
1859          *  - Reserved subnet anycast (RFC 2526)
1860          *      11111101 11....11 1xxxxxxx
1861          *  - ISATAP (RFC4214) 6.1
1862          *      00-00-5E-FE-xx-xx-xx-xx
1863          *  - value 0
1864          *  - XXX: already assigned to an address on the device
1865          */
1866         if (idev->rndid[0] == 0xfd &&
1867             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1868             (idev->rndid[7]&0x80))
1869                 goto regen;
1870         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1871                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1872                         goto regen;
1873                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1874                         goto regen;
1875         }
1876 }
1877
1878 static void ipv6_regen_rndid(unsigned long data)
1879 {
1880         struct inet6_dev *idev = (struct inet6_dev *) data;
1881         unsigned long expires;
1882
1883         rcu_read_lock_bh();
1884         write_lock_bh(&idev->lock);
1885
1886         if (idev->dead)
1887                 goto out;
1888
1889         __ipv6_regen_rndid(idev);
1890
1891         expires = jiffies +
1892                 idev->cnf.temp_prefered_lft * HZ -
1893                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1894                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1895                 idev->cnf.max_desync_factor * HZ;
1896         if (time_before(expires, jiffies)) {
1897                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1898                         __func__, idev->dev->name);
1899                 goto out;
1900         }
1901
1902         if (!mod_timer(&idev->regen_timer, expires))
1903                 in6_dev_hold(idev);
1904
1905 out:
1906         write_unlock_bh(&idev->lock);
1907         rcu_read_unlock_bh();
1908         in6_dev_put(idev);
1909 }
1910
1911 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1912 {
1913         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1914                 __ipv6_regen_rndid(idev);
1915 }
1916
1917 /*
1918  *      Add prefix route.
1919  */
1920
1921 static void
1922 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1923                       unsigned long expires, u32 flags)
1924 {
1925         struct fib6_config cfg = {
1926                 .fc_table = RT6_TABLE_PREFIX,
1927                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1928                 .fc_ifindex = dev->ifindex,
1929                 .fc_expires = expires,
1930                 .fc_dst_len = plen,
1931                 .fc_flags = RTF_UP | flags,
1932                 .fc_nlinfo.nl_net = dev_net(dev),
1933                 .fc_protocol = RTPROT_KERNEL,
1934         };
1935
1936         cfg.fc_dst = *pfx;
1937
1938         /* Prevent useless cloning on PtP SIT.
1939            This thing is done here expecting that the whole
1940            class of non-broadcast devices need not cloning.
1941          */
1942 #if IS_ENABLED(CONFIG_IPV6_SIT)
1943         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1944                 cfg.fc_flags |= RTF_NONEXTHOP;
1945 #endif
1946
1947         ip6_route_add(&cfg);
1948 }
1949
1950
1951 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1952                                                   int plen,
1953                                                   const struct net_device *dev,
1954                                                   u32 flags, u32 noflags)
1955 {
1956         struct fib6_node *fn;
1957         struct rt6_info *rt = NULL;
1958         struct fib6_table *table;
1959
1960         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1961         if (table == NULL)
1962                 return NULL;
1963
1964         read_lock_bh(&table->tb6_lock);
1965         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1966         if (!fn)
1967                 goto out;
1968         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1969                 if (rt->dst.dev->ifindex != dev->ifindex)
1970                         continue;
1971                 if ((rt->rt6i_flags & flags) != flags)
1972                         continue;
1973                 if ((rt->rt6i_flags & noflags) != 0)
1974                         continue;
1975                 dst_hold(&rt->dst);
1976                 break;
1977         }
1978 out:
1979         read_unlock_bh(&table->tb6_lock);
1980         return rt;
1981 }
1982
1983
1984 /* Create "default" multicast route to the interface */
1985
1986 static void addrconf_add_mroute(struct net_device *dev)
1987 {
1988         struct fib6_config cfg = {
1989                 .fc_table = RT6_TABLE_LOCAL,
1990                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1991                 .fc_ifindex = dev->ifindex,
1992                 .fc_dst_len = 8,
1993                 .fc_flags = RTF_UP,
1994                 .fc_nlinfo.nl_net = dev_net(dev),
1995         };
1996
1997         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1998
1999         ip6_route_add(&cfg);
2000 }
2001
2002 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2003 {
2004         struct inet6_dev *idev;
2005
2006         ASSERT_RTNL();
2007
2008         idev = ipv6_find_idev(dev);
2009         if (!idev)
2010                 return ERR_PTR(-ENOBUFS);
2011
2012         if (idev->cnf.disable_ipv6)
2013                 return ERR_PTR(-EACCES);
2014
2015         /* Add default multicast route */
2016         if (!(dev->flags & IFF_LOOPBACK))
2017                 addrconf_add_mroute(dev);
2018
2019         return idev;
2020 }
2021
2022 static void manage_tempaddrs(struct inet6_dev *idev,
2023                              struct inet6_ifaddr *ifp,
2024                              __u32 valid_lft, __u32 prefered_lft,
2025                              bool create, unsigned long now)
2026 {
2027         u32 flags;
2028         struct inet6_ifaddr *ift;
2029
2030         read_lock_bh(&idev->lock);
2031         /* update all temporary addresses in the list */
2032         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2033                 int age, max_valid, max_prefered;
2034
2035                 if (ifp != ift->ifpub)
2036                         continue;
2037
2038                 /* RFC 4941 section 3.3:
2039                  * If a received option will extend the lifetime of a public
2040                  * address, the lifetimes of temporary addresses should
2041                  * be extended, subject to the overall constraint that no
2042                  * temporary addresses should ever remain "valid" or "preferred"
2043                  * for a time longer than (TEMP_VALID_LIFETIME) or
2044                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2045                  */
2046                 age = (now - ift->cstamp) / HZ;
2047                 max_valid = idev->cnf.temp_valid_lft - age;
2048                 if (max_valid < 0)
2049                         max_valid = 0;
2050
2051                 max_prefered = idev->cnf.temp_prefered_lft -
2052                                idev->cnf.max_desync_factor - age;
2053                 if (max_prefered < 0)
2054                         max_prefered = 0;
2055
2056                 if (valid_lft > max_valid)
2057                         valid_lft = max_valid;
2058
2059                 if (prefered_lft > max_prefered)
2060                         prefered_lft = max_prefered;
2061
2062                 spin_lock(&ift->lock);
2063                 flags = ift->flags;
2064                 ift->valid_lft = valid_lft;
2065                 ift->prefered_lft = prefered_lft;
2066                 ift->tstamp = now;
2067                 if (prefered_lft > 0)
2068                         ift->flags &= ~IFA_F_DEPRECATED;
2069
2070                 spin_unlock(&ift->lock);
2071                 if (!(flags&IFA_F_TENTATIVE))
2072                         ipv6_ifa_notify(0, ift);
2073         }
2074
2075         if ((create || list_empty(&idev->tempaddr_list)) &&
2076             idev->cnf.use_tempaddr > 0) {
2077                 /* When a new public address is created as described
2078                  * in [ADDRCONF], also create a new temporary address.
2079                  * Also create a temporary address if it's enabled but
2080                  * no temporary address currently exists.
2081                  */
2082                 read_unlock_bh(&idev->lock);
2083                 ipv6_create_tempaddr(ifp, NULL);
2084         } else {
2085                 read_unlock_bh(&idev->lock);
2086         }
2087 }
2088
2089 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2090 {
2091         struct prefix_info *pinfo;
2092         __u32 valid_lft;
2093         __u32 prefered_lft;
2094         int addr_type;
2095         struct inet6_dev *in6_dev;
2096         struct net *net = dev_net(dev);
2097
2098         pinfo = (struct prefix_info *) opt;
2099
2100         if (len < sizeof(struct prefix_info)) {
2101                 ADBG("addrconf: prefix option too short\n");
2102                 return;
2103         }
2104
2105         /*
2106          *      Validation checks ([ADDRCONF], page 19)
2107          */
2108
2109         addr_type = ipv6_addr_type(&pinfo->prefix);
2110
2111         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2112                 return;
2113
2114         valid_lft = ntohl(pinfo->valid);
2115         prefered_lft = ntohl(pinfo->prefered);
2116
2117         if (prefered_lft > valid_lft) {
2118                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2119                 return;
2120         }
2121
2122         in6_dev = in6_dev_get(dev);
2123
2124         if (in6_dev == NULL) {
2125                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2126                                     dev->name);
2127                 return;
2128         }
2129
2130         /*
2131          *      Two things going on here:
2132          *      1) Add routes for on-link prefixes
2133          *      2) Configure prefixes with the auto flag set
2134          */
2135
2136         if (pinfo->onlink) {
2137                 struct rt6_info *rt;
2138                 unsigned long rt_expires;
2139
2140                 /* Avoid arithmetic overflow. Really, we could
2141                  * save rt_expires in seconds, likely valid_lft,
2142                  * but it would require division in fib gc, that it
2143                  * not good.
2144                  */
2145                 if (HZ > USER_HZ)
2146                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2147                 else
2148                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2149
2150                 if (addrconf_finite_timeout(rt_expires))
2151                         rt_expires *= HZ;
2152
2153                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2154                                                pinfo->prefix_len,
2155                                                dev,
2156                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2157                                                RTF_GATEWAY | RTF_DEFAULT);
2158
2159                 if (rt) {
2160                         /* Autoconf prefix route */
2161                         if (valid_lft == 0) {
2162                                 ip6_del_rt(rt);
2163                                 rt = NULL;
2164                         } else if (addrconf_finite_timeout(rt_expires)) {
2165                                 /* not infinity */
2166                                 rt6_set_expires(rt, jiffies + rt_expires);
2167                         } else {
2168                                 rt6_clean_expires(rt);
2169                         }
2170                 } else if (valid_lft) {
2171                         clock_t expires = 0;
2172                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2173                         if (addrconf_finite_timeout(rt_expires)) {
2174                                 /* not infinity */
2175                                 flags |= RTF_EXPIRES;
2176                                 expires = jiffies_to_clock_t(rt_expires);
2177                         }
2178                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2179                                               dev, expires, flags);
2180                 }
2181                 ip6_rt_put(rt);
2182         }
2183
2184         /* Try to figure out our local address for this prefix */
2185
2186         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2187                 struct inet6_ifaddr *ifp;
2188                 struct in6_addr addr;
2189                 int create = 0, update_lft = 0;
2190                 bool tokenized = false;
2191
2192                 if (pinfo->prefix_len == 64) {
2193                         memcpy(&addr, &pinfo->prefix, 8);
2194
2195                         if (!ipv6_addr_any(&in6_dev->token)) {
2196                                 read_lock_bh(&in6_dev->lock);
2197                                 memcpy(addr.s6_addr + 8,
2198                                        in6_dev->token.s6_addr + 8, 8);
2199                                 read_unlock_bh(&in6_dev->lock);
2200                                 tokenized = true;
2201                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2202                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2203                                 in6_dev_put(in6_dev);
2204                                 return;
2205                         }
2206                         goto ok;
2207                 }
2208                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2209                                     pinfo->prefix_len);
2210                 in6_dev_put(in6_dev);
2211                 return;
2212
2213 ok:
2214
2215                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2216
2217                 if (ifp == NULL && valid_lft) {
2218                         int max_addresses = in6_dev->cnf.max_addresses;
2219                         u32 addr_flags = 0;
2220
2221 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2222                         if (in6_dev->cnf.optimistic_dad &&
2223                             !net->ipv6.devconf_all->forwarding && sllao)
2224                                 addr_flags = IFA_F_OPTIMISTIC;
2225 #endif
2226
2227                         /* Do not allow to create too much of autoconfigured
2228                          * addresses; this would be too easy way to crash kernel.
2229                          */
2230                         if (!max_addresses ||
2231                             ipv6_count_addresses(in6_dev) < max_addresses)
2232                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2233                                                     pinfo->prefix_len,
2234                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2235                                                     addr_flags, valid_lft,
2236                                                     prefered_lft);
2237
2238                         if (IS_ERR_OR_NULL(ifp)) {
2239                                 in6_dev_put(in6_dev);
2240                                 return;
2241                         }
2242
2243                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2244                         update_lft = 0;
2245                         create = 1;
2246                         ifp->cstamp = jiffies;
2247                         ifp->tokenized = tokenized;
2248                         addrconf_dad_start(ifp);
2249                 }
2250
2251                 if (ifp) {
2252                         u32 flags;
2253                         unsigned long now;
2254                         u32 stored_lft;
2255
2256                         /* update lifetime (RFC2462 5.5.3 e) */
2257                         spin_lock(&ifp->lock);
2258                         now = jiffies;
2259                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2260                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2261                         else
2262                                 stored_lft = 0;
2263                         if (!update_lft && !create && stored_lft) {
2264                                 const u32 minimum_lft = min(
2265                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2266                                 valid_lft = max(valid_lft, minimum_lft);
2267
2268                                 /* RFC4862 Section 5.5.3e:
2269                                  * "Note that the preferred lifetime of the
2270                                  *  corresponding address is always reset to
2271                                  *  the Preferred Lifetime in the received
2272                                  *  Prefix Information option, regardless of
2273                                  *  whether the valid lifetime is also reset or
2274                                  *  ignored."
2275                                  *
2276                                  * So we should always update prefered_lft here.
2277                                  */
2278                                 update_lft = 1;
2279                         }
2280
2281                         if (update_lft) {
2282                                 ifp->valid_lft = valid_lft;
2283                                 ifp->prefered_lft = prefered_lft;
2284                                 ifp->tstamp = now;
2285                                 flags = ifp->flags;
2286                                 ifp->flags &= ~IFA_F_DEPRECATED;
2287                                 spin_unlock(&ifp->lock);
2288
2289                                 if (!(flags&IFA_F_TENTATIVE))
2290                                         ipv6_ifa_notify(0, ifp);
2291                         } else
2292                                 spin_unlock(&ifp->lock);
2293
2294                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2295                                          create, now);
2296
2297                         in6_ifa_put(ifp);
2298                         addrconf_verify(0);
2299                 }
2300         }
2301         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2302         in6_dev_put(in6_dev);
2303 }
2304
2305 /*
2306  *      Set destination address.
2307  *      Special case for SIT interfaces where we create a new "virtual"
2308  *      device.
2309  */
2310 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2311 {
2312         struct in6_ifreq ireq;
2313         struct net_device *dev;
2314         int err = -EINVAL;
2315
2316         rtnl_lock();
2317
2318         err = -EFAULT;
2319         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2320                 goto err_exit;
2321
2322         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2323
2324         err = -ENODEV;
2325         if (dev == NULL)
2326                 goto err_exit;
2327
2328 #if IS_ENABLED(CONFIG_IPV6_SIT)
2329         if (dev->type == ARPHRD_SIT) {
2330                 const struct net_device_ops *ops = dev->netdev_ops;
2331                 struct ifreq ifr;
2332                 struct ip_tunnel_parm p;
2333
2334                 err = -EADDRNOTAVAIL;
2335                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2336                         goto err_exit;
2337
2338                 memset(&p, 0, sizeof(p));
2339                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2340                 p.iph.saddr = 0;
2341                 p.iph.version = 4;
2342                 p.iph.ihl = 5;
2343                 p.iph.protocol = IPPROTO_IPV6;
2344                 p.iph.ttl = 64;
2345                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2346
2347                 if (ops->ndo_do_ioctl) {
2348                         mm_segment_t oldfs = get_fs();
2349
2350                         set_fs(KERNEL_DS);
2351                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2352                         set_fs(oldfs);
2353                 } else
2354                         err = -EOPNOTSUPP;
2355
2356                 if (err == 0) {
2357                         err = -ENOBUFS;
2358                         dev = __dev_get_by_name(net, p.name);
2359                         if (!dev)
2360                                 goto err_exit;
2361                         err = dev_open(dev);
2362                 }
2363         }
2364 #endif
2365
2366 err_exit:
2367         rtnl_unlock();
2368         return err;
2369 }
2370
2371 /*
2372  *      Manual configuration of address on an interface
2373  */
2374 static int inet6_addr_add(struct net *net, int ifindex,
2375                           const struct in6_addr *pfx,
2376                           const struct in6_addr *peer_pfx,
2377                           unsigned int plen, __u32 ifa_flags,
2378                           __u32 prefered_lft, __u32 valid_lft)
2379 {
2380         struct inet6_ifaddr *ifp;
2381         struct inet6_dev *idev;
2382         struct net_device *dev;
2383         int scope;
2384         u32 flags;
2385         clock_t expires;
2386         unsigned long timeout;
2387
2388         ASSERT_RTNL();
2389
2390         if (plen > 128)
2391                 return -EINVAL;
2392
2393         /* check the lifetime */
2394         if (!valid_lft || prefered_lft > valid_lft)
2395                 return -EINVAL;
2396
2397         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2398                 return -EINVAL;
2399
2400         dev = __dev_get_by_index(net, ifindex);
2401         if (!dev)
2402                 return -ENODEV;
2403
2404         idev = addrconf_add_dev(dev);
2405         if (IS_ERR(idev))
2406                 return PTR_ERR(idev);
2407
2408         scope = ipv6_addr_scope(pfx);
2409
2410         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2411         if (addrconf_finite_timeout(timeout)) {
2412                 expires = jiffies_to_clock_t(timeout * HZ);
2413                 valid_lft = timeout;
2414                 flags = RTF_EXPIRES;
2415         } else {
2416                 expires = 0;
2417                 flags = 0;
2418                 ifa_flags |= IFA_F_PERMANENT;
2419         }
2420
2421         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2422         if (addrconf_finite_timeout(timeout)) {
2423                 if (timeout == 0)
2424                         ifa_flags |= IFA_F_DEPRECATED;
2425                 prefered_lft = timeout;
2426         }
2427
2428         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2429                             valid_lft, prefered_lft);
2430
2431         if (!IS_ERR(ifp)) {
2432                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2433                                       expires, flags);
2434                 /*
2435                  * Note that section 3.1 of RFC 4429 indicates
2436                  * that the Optimistic flag should not be set for
2437                  * manually configured addresses
2438                  */
2439                 addrconf_dad_start(ifp);
2440                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2441                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2442                                          true, jiffies);
2443                 in6_ifa_put(ifp);
2444                 addrconf_verify(0);
2445                 return 0;
2446         }
2447
2448         return PTR_ERR(ifp);
2449 }
2450
2451 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2452                           unsigned int plen)
2453 {
2454         struct inet6_ifaddr *ifp;
2455         struct inet6_dev *idev;
2456         struct net_device *dev;
2457
2458         if (plen > 128)
2459                 return -EINVAL;
2460
2461         dev = __dev_get_by_index(net, ifindex);
2462         if (!dev)
2463                 return -ENODEV;
2464
2465         if ((idev = __in6_dev_get(dev)) == NULL)
2466                 return -ENXIO;
2467
2468         read_lock_bh(&idev->lock);
2469         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2470                 if (ifp->prefix_len == plen &&
2471                     ipv6_addr_equal(pfx, &ifp->addr)) {
2472                         in6_ifa_hold(ifp);
2473                         read_unlock_bh(&idev->lock);
2474
2475                         ipv6_del_addr(ifp);
2476                         return 0;
2477                 }
2478         }
2479         read_unlock_bh(&idev->lock);
2480         return -EADDRNOTAVAIL;
2481 }
2482
2483
2484 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2485 {
2486         struct in6_ifreq ireq;
2487         int err;
2488
2489         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2490                 return -EPERM;
2491
2492         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2493                 return -EFAULT;
2494
2495         rtnl_lock();
2496         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2497                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2498                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2499         rtnl_unlock();
2500         return err;
2501 }
2502
2503 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2504 {
2505         struct in6_ifreq ireq;
2506         int err;
2507
2508         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2509                 return -EPERM;
2510
2511         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2512                 return -EFAULT;
2513
2514         rtnl_lock();
2515         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2516                              ireq.ifr6_prefixlen);
2517         rtnl_unlock();
2518         return err;
2519 }
2520
2521 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2522                      int plen, int scope)
2523 {
2524         struct inet6_ifaddr *ifp;
2525
2526         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2527                             scope, IFA_F_PERMANENT, 0, 0);
2528         if (!IS_ERR(ifp)) {
2529                 spin_lock_bh(&ifp->lock);
2530                 ifp->flags &= ~IFA_F_TENTATIVE;
2531                 spin_unlock_bh(&ifp->lock);
2532                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2533                 in6_ifa_put(ifp);
2534         }
2535 }
2536
2537 #if IS_ENABLED(CONFIG_IPV6_SIT)
2538 static void sit_add_v4_addrs(struct inet6_dev *idev)
2539 {
2540         struct in6_addr addr;
2541         struct net_device *dev;
2542         struct net *net = dev_net(idev->dev);
2543         int scope, plen;
2544         u32 pflags = 0;
2545
2546         ASSERT_RTNL();
2547
2548         memset(&addr, 0, sizeof(struct in6_addr));
2549         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2550
2551         if (idev->dev->flags&IFF_POINTOPOINT) {
2552                 addr.s6_addr32[0] = htonl(0xfe800000);
2553                 scope = IFA_LINK;
2554                 plen = 64;
2555         } else {
2556                 scope = IPV6_ADDR_COMPATv4;
2557                 plen = 96;
2558                 pflags |= RTF_NONEXTHOP;
2559         }
2560
2561         if (addr.s6_addr32[3]) {
2562                 add_addr(idev, &addr, plen, scope);
2563                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2564                 return;
2565         }
2566
2567         for_each_netdev(net, dev) {
2568                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2569                 if (in_dev && (dev->flags & IFF_UP)) {
2570                         struct in_ifaddr *ifa;
2571
2572                         int flag = scope;
2573
2574                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2575
2576                                 addr.s6_addr32[3] = ifa->ifa_local;
2577
2578                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2579                                         continue;
2580                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2581                                         if (idev->dev->flags&IFF_POINTOPOINT)
2582                                                 continue;
2583                                         flag |= IFA_HOST;
2584                                 }
2585
2586                                 add_addr(idev, &addr, plen, flag);
2587                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2588                                                       pflags);
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, false);
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                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2715                         addrconf_add_linklocal(idev, &addr);
2716                 return;
2717         }
2718
2719         sit_add_v4_addrs(idev);
2720
2721         if (dev->flags&IFF_POINTOPOINT)
2722                 addrconf_add_mroute(dev);
2723 }
2724 #endif
2725
2726 #if IS_ENABLED(CONFIG_NET_IPGRE)
2727 static void addrconf_gre_config(struct net_device *dev)
2728 {
2729         struct inet6_dev *idev;
2730         struct in6_addr addr;
2731
2732         ASSERT_RTNL();
2733
2734         if ((idev = ipv6_find_idev(dev)) == NULL) {
2735                 pr_debug("%s: add_dev failed\n", __func__);
2736                 return;
2737         }
2738
2739         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2740         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2741                 addrconf_add_linklocal(idev, &addr);
2742 }
2743 #endif
2744
2745 static inline int
2746 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2747 {
2748         struct in6_addr lladdr;
2749
2750         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2751                 addrconf_add_linklocal(idev, &lladdr);
2752                 return 0;
2753         }
2754         return -1;
2755 }
2756
2757 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2758                            void *ptr)
2759 {
2760         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2761         struct inet6_dev *idev = __in6_dev_get(dev);
2762         int run_pending = 0;
2763         int err;
2764
2765         switch (event) {
2766         case NETDEV_REGISTER:
2767                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2768                         idev = ipv6_add_dev(dev);
2769                         if (!idev)
2770                                 return notifier_from_errno(-ENOMEM);
2771                 }
2772                 break;
2773
2774         case NETDEV_UP:
2775         case NETDEV_CHANGE:
2776                 if (dev->flags & IFF_SLAVE)
2777                         break;
2778
2779                 if (event == NETDEV_UP) {
2780                         if (!addrconf_qdisc_ok(dev)) {
2781                                 /* device is not ready yet. */
2782                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2783                                         dev->name);
2784                                 break;
2785                         }
2786
2787                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2788                                 idev = ipv6_add_dev(dev);
2789
2790                         if (idev) {
2791                                 idev->if_flags |= IF_READY;
2792                                 run_pending = 1;
2793                         }
2794                 } else {
2795                         if (!addrconf_qdisc_ok(dev)) {
2796                                 /* device is still not ready. */
2797                                 break;
2798                         }
2799
2800                         if (idev) {
2801                                 if (idev->if_flags & IF_READY)
2802                                         /* device is already configured. */
2803                                         break;
2804                                 idev->if_flags |= IF_READY;
2805                         }
2806
2807                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2808                                 dev->name);
2809
2810                         run_pending = 1;
2811                 }
2812
2813                 switch (dev->type) {
2814 #if IS_ENABLED(CONFIG_IPV6_SIT)
2815                 case ARPHRD_SIT:
2816                         addrconf_sit_config(dev);
2817                         break;
2818 #endif
2819 #if IS_ENABLED(CONFIG_NET_IPGRE)
2820                 case ARPHRD_IPGRE:
2821                         addrconf_gre_config(dev);
2822                         break;
2823 #endif
2824                 case ARPHRD_LOOPBACK:
2825                         init_loopback(dev);
2826                         break;
2827
2828                 default:
2829                         addrconf_dev_config(dev);
2830                         break;
2831                 }
2832
2833                 if (idev) {
2834                         if (run_pending)
2835                                 addrconf_dad_run(idev);
2836
2837                         /*
2838                          * If the MTU changed during the interface down,
2839                          * when the interface up, the changed MTU must be
2840                          * reflected in the idev as well as routers.
2841                          */
2842                         if (idev->cnf.mtu6 != dev->mtu &&
2843                             dev->mtu >= IPV6_MIN_MTU) {
2844                                 rt6_mtu_change(dev, dev->mtu);
2845                                 idev->cnf.mtu6 = dev->mtu;
2846                         }
2847                         idev->tstamp = jiffies;
2848                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2849
2850                         /*
2851                          * If the changed mtu during down is lower than
2852                          * IPV6_MIN_MTU stop IPv6 on this interface.
2853                          */
2854                         if (dev->mtu < IPV6_MIN_MTU)
2855                                 addrconf_ifdown(dev, 1);
2856                 }
2857                 break;
2858
2859         case NETDEV_CHANGEMTU:
2860                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2861                         rt6_mtu_change(dev, dev->mtu);
2862                         idev->cnf.mtu6 = dev->mtu;
2863                         break;
2864                 }
2865
2866                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2867                         idev = ipv6_add_dev(dev);
2868                         if (idev)
2869                                 break;
2870                 }
2871
2872                 /*
2873                  * MTU falled under IPV6_MIN_MTU.
2874                  * Stop IPv6 on this interface.
2875                  */
2876
2877         case NETDEV_DOWN:
2878         case NETDEV_UNREGISTER:
2879                 /*
2880                  *      Remove all addresses from this interface.
2881                  */
2882                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2883                 break;
2884
2885         case NETDEV_CHANGENAME:
2886                 if (idev) {
2887                         snmp6_unregister_dev(idev);
2888                         addrconf_sysctl_unregister(idev);
2889                         addrconf_sysctl_register(idev);
2890                         err = snmp6_register_dev(idev);
2891                         if (err)
2892                                 return notifier_from_errno(err);
2893                 }
2894                 break;
2895
2896         case NETDEV_PRE_TYPE_CHANGE:
2897         case NETDEV_POST_TYPE_CHANGE:
2898                 addrconf_type_change(dev, event);
2899                 break;
2900         }
2901
2902         return NOTIFY_OK;
2903 }
2904
2905 /*
2906  *      addrconf module should be notified of a device going up
2907  */
2908 static struct notifier_block ipv6_dev_notf = {
2909         .notifier_call = addrconf_notify,
2910 };
2911
2912 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2913 {
2914         struct inet6_dev *idev;
2915         ASSERT_RTNL();
2916
2917         idev = __in6_dev_get(dev);
2918
2919         if (event == NETDEV_POST_TYPE_CHANGE)
2920                 ipv6_mc_remap(idev);
2921         else if (event == NETDEV_PRE_TYPE_CHANGE)
2922                 ipv6_mc_unmap(idev);
2923 }
2924
2925 static int addrconf_ifdown(struct net_device *dev, int how)
2926 {
2927         struct net *net = dev_net(dev);
2928         struct inet6_dev *idev;
2929         struct inet6_ifaddr *ifa;
2930         int state, i;
2931
2932         ASSERT_RTNL();
2933
2934         rt6_ifdown(net, dev);
2935         neigh_ifdown(&nd_tbl, dev);
2936
2937         idev = __in6_dev_get(dev);
2938         if (idev == NULL)
2939                 return -ENODEV;
2940
2941         /*
2942          * Step 1: remove reference to ipv6 device from parent device.
2943          *         Do not dev_put!
2944          */
2945         if (how) {
2946                 idev->dead = 1;
2947
2948                 /* protected by rtnl_lock */
2949                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2950
2951                 /* Step 1.5: remove snmp6 entry */
2952                 snmp6_unregister_dev(idev);
2953
2954         }
2955
2956         /* Step 2: clear hash table */
2957         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2958                 struct hlist_head *h = &inet6_addr_lst[i];
2959
2960                 spin_lock_bh(&addrconf_hash_lock);
2961         restart:
2962                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2963                         if (ifa->idev == idev) {
2964                                 hlist_del_init_rcu(&ifa->addr_lst);
2965                                 addrconf_del_dad_timer(ifa);
2966                                 goto restart;
2967                         }
2968                 }
2969                 spin_unlock_bh(&addrconf_hash_lock);
2970         }
2971
2972         write_lock_bh(&idev->lock);
2973
2974         addrconf_del_rs_timer(idev);
2975
2976         /* Step 2: clear flags for stateless addrconf */
2977         if (!how)
2978                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2979
2980         if (how && del_timer(&idev->regen_timer))
2981                 in6_dev_put(idev);
2982
2983         /* Step 3: clear tempaddr list */
2984         while (!list_empty(&idev->tempaddr_list)) {
2985                 ifa = list_first_entry(&idev->tempaddr_list,
2986                                        struct inet6_ifaddr, tmp_list);
2987                 list_del(&ifa->tmp_list);
2988                 write_unlock_bh(&idev->lock);
2989                 spin_lock_bh(&ifa->lock);
2990
2991                 if (ifa->ifpub) {
2992                         in6_ifa_put(ifa->ifpub);
2993                         ifa->ifpub = NULL;
2994                 }
2995                 spin_unlock_bh(&ifa->lock);
2996                 in6_ifa_put(ifa);
2997                 write_lock_bh(&idev->lock);
2998         }
2999
3000         while (!list_empty(&idev->addr_list)) {
3001                 ifa = list_first_entry(&idev->addr_list,
3002                                        struct inet6_ifaddr, if_list);
3003                 addrconf_del_dad_timer(ifa);
3004
3005                 list_del(&ifa->if_list);
3006
3007                 write_unlock_bh(&idev->lock);
3008
3009                 spin_lock_bh(&ifa->state_lock);
3010                 state = ifa->state;
3011                 ifa->state = INET6_IFADDR_STATE_DEAD;
3012                 spin_unlock_bh(&ifa->state_lock);
3013
3014                 if (state != INET6_IFADDR_STATE_DEAD) {
3015                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3016                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3017                 }
3018                 in6_ifa_put(ifa);
3019
3020                 write_lock_bh(&idev->lock);
3021         }
3022
3023         write_unlock_bh(&idev->lock);
3024
3025         /* Step 5: Discard multicast list */
3026         if (how)
3027                 ipv6_mc_destroy_dev(idev);
3028         else
3029                 ipv6_mc_down(idev);
3030
3031         idev->tstamp = jiffies;
3032
3033         /* Last: Shot the device (if unregistered) */
3034         if (how) {
3035                 addrconf_sysctl_unregister(idev);
3036                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3037                 neigh_ifdown(&nd_tbl, dev);
3038                 in6_dev_put(idev);
3039         }
3040         return 0;
3041 }
3042
3043 static void addrconf_rs_timer(unsigned long data)
3044 {
3045         struct inet6_dev *idev = (struct inet6_dev *)data;
3046         struct net_device *dev = idev->dev;
3047         struct in6_addr lladdr;
3048
3049         write_lock(&idev->lock);
3050         if (idev->dead || !(idev->if_flags & IF_READY))
3051                 goto out;
3052
3053         if (!ipv6_accept_ra(idev))
3054                 goto out;
3055
3056         /* Announcement received after solicitation was sent */
3057         if (idev->if_flags & IF_RA_RCVD)
3058                 goto out;
3059
3060         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3061                 write_unlock(&idev->lock);
3062                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3063                         ndisc_send_rs(dev, &lladdr,
3064                                       &in6addr_linklocal_allrouters);
3065                 else
3066                         goto put;
3067
3068                 write_lock(&idev->lock);
3069                 /* The wait after the last probe can be shorter */
3070                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3071                                              idev->cnf.rtr_solicits) ?
3072                                       idev->cnf.rtr_solicit_delay :
3073                                       idev->cnf.rtr_solicit_interval);
3074         } else {
3075                 /*
3076                  * Note: we do not support deprecated "all on-link"
3077                  * assumption any longer.
3078                  */
3079                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3080         }
3081
3082 out:
3083         write_unlock(&idev->lock);
3084 put:
3085         in6_dev_put(idev);
3086 }
3087
3088 /*
3089  *      Duplicate Address Detection
3090  */
3091 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3092 {
3093         unsigned long rand_num;
3094         struct inet6_dev *idev = ifp->idev;
3095
3096         if (ifp->flags & IFA_F_OPTIMISTIC)
3097                 rand_num = 0;
3098         else
3099                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3100
3101         ifp->dad_probes = idev->cnf.dad_transmits;
3102         addrconf_mod_dad_timer(ifp, rand_num);
3103 }
3104
3105 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3106 {
3107         struct inet6_dev *idev = ifp->idev;
3108         struct net_device *dev = idev->dev;
3109
3110         addrconf_join_solict(dev, &ifp->addr);
3111
3112         net_srandom(ifp->addr.s6_addr32[3]);
3113
3114         read_lock_bh(&idev->lock);
3115         spin_lock(&ifp->lock);
3116         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3117                 goto out;
3118
3119         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3120             idev->cnf.accept_dad < 1 ||
3121             !(ifp->flags&IFA_F_TENTATIVE) ||
3122             ifp->flags & IFA_F_NODAD) {
3123                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3124                 spin_unlock(&ifp->lock);
3125                 read_unlock_bh(&idev->lock);
3126
3127                 addrconf_dad_completed(ifp);
3128                 return;
3129         }
3130
3131         if (!(idev->if_flags & IF_READY)) {
3132                 spin_unlock(&ifp->lock);
3133                 read_unlock_bh(&idev->lock);
3134                 /*
3135                  * If the device is not ready:
3136                  * - keep it tentative if it is a permanent address.
3137                  * - otherwise, kill it.
3138                  */
3139                 in6_ifa_hold(ifp);
3140                 addrconf_dad_stop(ifp, 0);
3141                 return;
3142         }
3143
3144         /*
3145          * Optimistic nodes can start receiving
3146          * Frames right away
3147          */
3148         if (ifp->flags & IFA_F_OPTIMISTIC)
3149                 ip6_ins_rt(ifp->rt);
3150
3151         addrconf_dad_kick(ifp);
3152 out:
3153         spin_unlock(&ifp->lock);
3154         read_unlock_bh(&idev->lock);
3155 }
3156
3157 static void addrconf_dad_timer(unsigned long data)
3158 {
3159         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3160         struct inet6_dev *idev = ifp->idev;
3161         struct in6_addr mcaddr;
3162
3163         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3164                 goto out;
3165
3166         write_lock(&idev->lock);
3167         if (idev->dead || !(idev->if_flags & IF_READY)) {
3168                 write_unlock(&idev->lock);
3169                 goto out;
3170         }
3171
3172         spin_lock(&ifp->lock);
3173         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3174                 spin_unlock(&ifp->lock);
3175                 write_unlock(&idev->lock);
3176                 goto out;
3177         }
3178
3179         if (ifp->dad_probes == 0) {
3180                 /*
3181                  * DAD was successful
3182                  */
3183
3184                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3185                 spin_unlock(&ifp->lock);
3186                 write_unlock(&idev->lock);
3187
3188                 addrconf_dad_completed(ifp);
3189
3190                 goto out;
3191         }
3192
3193         ifp->dad_probes--;
3194         addrconf_mod_dad_timer(ifp,
3195                                NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3196         spin_unlock(&ifp->lock);
3197         write_unlock(&idev->lock);
3198
3199         /* send a neighbour solicitation for our addr */
3200         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3201         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3202 out:
3203         in6_ifa_put(ifp);
3204 }
3205
3206 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3207 {
3208         struct net_device *dev = ifp->idev->dev;
3209         struct in6_addr lladdr;
3210         bool send_rs, send_mld;
3211
3212         addrconf_del_dad_timer(ifp);
3213
3214         /*
3215          *      Configure the address for reception. Now it is valid.
3216          */
3217
3218         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3219
3220         /* If added prefix is link local and we are prepared to process
3221            router advertisements, start sending router solicitations.
3222          */
3223
3224         read_lock_bh(&ifp->idev->lock);
3225         spin_lock(&ifp->lock);
3226         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3227                    ifp->idev->valid_ll_addr_cnt == 1;
3228         send_rs = send_mld &&
3229                   ipv6_accept_ra(ifp->idev) &&
3230                   ifp->idev->cnf.rtr_solicits > 0 &&
3231                   (dev->flags&IFF_LOOPBACK) == 0;
3232         spin_unlock(&ifp->lock);
3233         read_unlock_bh(&ifp->idev->lock);
3234
3235         /* While dad is in progress mld report's source address is in6_addrany.
3236          * Resend with proper ll now.
3237          */
3238         if (send_mld)
3239                 ipv6_mc_dad_complete(ifp->idev);
3240
3241         if (send_rs) {
3242                 /*
3243                  *      If a host as already performed a random delay
3244                  *      [...] as part of DAD [...] there is no need
3245                  *      to delay again before sending the first RS
3246                  */
3247                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3248                         return;
3249                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3250
3251                 write_lock_bh(&ifp->idev->lock);
3252                 spin_lock(&ifp->lock);
3253                 ifp->idev->rs_probes = 1;
3254                 ifp->idev->if_flags |= IF_RS_SENT;
3255                 addrconf_mod_rs_timer(ifp->idev,
3256                                       ifp->idev->cnf.rtr_solicit_interval);
3257                 spin_unlock(&ifp->lock);
3258                 write_unlock_bh(&ifp->idev->lock);
3259         }
3260 }
3261
3262 static void addrconf_dad_run(struct inet6_dev *idev)
3263 {
3264         struct inet6_ifaddr *ifp;
3265
3266         read_lock_bh(&idev->lock);
3267         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3268                 spin_lock(&ifp->lock);
3269                 if (ifp->flags & IFA_F_TENTATIVE &&
3270                     ifp->state == INET6_IFADDR_STATE_DAD)
3271                         addrconf_dad_kick(ifp);
3272                 spin_unlock(&ifp->lock);
3273         }
3274         read_unlock_bh(&idev->lock);
3275 }
3276
3277 #ifdef CONFIG_PROC_FS
3278 struct if6_iter_state {
3279         struct seq_net_private p;
3280         int bucket;
3281         int offset;
3282 };
3283
3284 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3285 {
3286         struct inet6_ifaddr *ifa = NULL;
3287         struct if6_iter_state *state = seq->private;
3288         struct net *net = seq_file_net(seq);
3289         int p = 0;
3290
3291         /* initial bucket if pos is 0 */
3292         if (pos == 0) {
3293                 state->bucket = 0;
3294                 state->offset = 0;
3295         }
3296
3297         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3298                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3299                                          addr_lst) {
3300                         if (!net_eq(dev_net(ifa->idev->dev), net))
3301                                 continue;
3302                         /* sync with offset */
3303                         if (p < state->offset) {
3304                                 p++;
3305                                 continue;
3306                         }
3307                         state->offset++;
3308                         return ifa;
3309                 }
3310
3311                 /* prepare for next bucket */
3312                 state->offset = 0;
3313                 p = 0;
3314         }
3315         return NULL;
3316 }
3317
3318 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3319                                          struct inet6_ifaddr *ifa)
3320 {
3321         struct if6_iter_state *state = seq->private;
3322         struct net *net = seq_file_net(seq);
3323
3324         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3325                 if (!net_eq(dev_net(ifa->idev->dev), net))
3326                         continue;
3327                 state->offset++;
3328                 return ifa;
3329         }
3330
3331         while (++state->bucket < IN6_ADDR_HSIZE) {
3332                 state->offset = 0;
3333                 hlist_for_each_entry_rcu_bh(ifa,
3334                                      &inet6_addr_lst[state->bucket], addr_lst) {
3335                         if (!net_eq(dev_net(ifa->idev->dev), net))
3336                                 continue;
3337                         state->offset++;
3338                         return ifa;
3339                 }
3340         }
3341
3342         return NULL;
3343 }
3344
3345 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3346         __acquires(rcu_bh)
3347 {
3348         rcu_read_lock_bh();
3349         return if6_get_first(seq, *pos);
3350 }
3351
3352 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3353 {
3354         struct inet6_ifaddr *ifa;
3355
3356         ifa = if6_get_next(seq, v);
3357         ++*pos;
3358         return ifa;
3359 }
3360
3361 static void if6_seq_stop(struct seq_file *seq, void *v)
3362         __releases(rcu_bh)
3363 {
3364         rcu_read_unlock_bh();
3365 }
3366
3367 static int if6_seq_show(struct seq_file *seq, void *v)
3368 {
3369         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3370         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3371                    &ifp->addr,
3372                    ifp->idev->dev->ifindex,
3373                    ifp->prefix_len,
3374                    ifp->scope,
3375                    (u8) ifp->flags,
3376                    ifp->idev->dev->name);
3377         return 0;
3378 }
3379
3380 static const struct seq_operations if6_seq_ops = {
3381         .start  = if6_seq_start,
3382         .next   = if6_seq_next,
3383         .show   = if6_seq_show,
3384         .stop   = if6_seq_stop,
3385 };
3386
3387 static int if6_seq_open(struct inode *inode, struct file *file)
3388 {
3389         return seq_open_net(inode, file, &if6_seq_ops,
3390                             sizeof(struct if6_iter_state));
3391 }
3392
3393 static const struct file_operations if6_fops = {
3394         .owner          = THIS_MODULE,
3395         .open           = if6_seq_open,
3396         .read           = seq_read,
3397         .llseek         = seq_lseek,
3398         .release        = seq_release_net,
3399 };
3400
3401 static int __net_init if6_proc_net_init(struct net *net)
3402 {
3403         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3404                 return -ENOMEM;
3405         return 0;
3406 }
3407
3408 static void __net_exit if6_proc_net_exit(struct net *net)
3409 {
3410         remove_proc_entry("if_inet6", net->proc_net);
3411 }
3412
3413 static struct pernet_operations if6_proc_net_ops = {
3414        .init = if6_proc_net_init,
3415        .exit = if6_proc_net_exit,
3416 };
3417
3418 int __init if6_proc_init(void)
3419 {
3420         return register_pernet_subsys(&if6_proc_net_ops);
3421 }
3422
3423 void if6_proc_exit(void)
3424 {
3425         unregister_pernet_subsys(&if6_proc_net_ops);
3426 }
3427 #endif  /* CONFIG_PROC_FS */
3428
3429 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3430 /* Check if address is a home address configured on any interface. */
3431 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3432 {
3433         int ret = 0;
3434         struct inet6_ifaddr *ifp = NULL;
3435         unsigned int hash = inet6_addr_hash(addr);
3436
3437         rcu_read_lock_bh();
3438         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3439                 if (!net_eq(dev_net(ifp->idev->dev), net))
3440                         continue;
3441                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3442                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3443                         ret = 1;
3444                         break;
3445                 }
3446         }
3447         rcu_read_unlock_bh();
3448         return ret;
3449 }
3450 #endif
3451
3452 /*
3453  *      Periodic address status verification
3454  */
3455
3456 static void addrconf_verify(unsigned long foo)
3457 {
3458         unsigned long now, next, next_sec, next_sched;
3459         struct inet6_ifaddr *ifp;
3460         int i;
3461
3462         rcu_read_lock_bh();
3463         spin_lock(&addrconf_verify_lock);
3464         now = jiffies;
3465         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3466
3467         del_timer(&addr_chk_timer);
3468
3469         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3470 restart:
3471                 hlist_for_each_entry_rcu_bh(ifp,
3472                                          &inet6_addr_lst[i], addr_lst) {
3473                         unsigned long age;
3474
3475                         if (ifp->flags & IFA_F_PERMANENT)
3476                                 continue;
3477
3478                         spin_lock(&ifp->lock);
3479                         /* We try to batch several events at once. */
3480                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3481
3482                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3483                             age >= ifp->valid_lft) {
3484                                 spin_unlock(&ifp->lock);
3485                                 in6_ifa_hold(ifp);
3486                                 ipv6_del_addr(ifp);
3487                                 goto restart;
3488                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3489                                 spin_unlock(&ifp->lock);
3490                                 continue;
3491                         } else if (age >= ifp->prefered_lft) {
3492                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3493                                 int deprecate = 0;
3494
3495                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3496                                         deprecate = 1;
3497                                         ifp->flags |= IFA_F_DEPRECATED;
3498                                 }
3499
3500                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3501                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3502
3503                                 spin_unlock(&ifp->lock);
3504
3505                                 if (deprecate) {
3506                                         in6_ifa_hold(ifp);
3507
3508                                         ipv6_ifa_notify(0, ifp);
3509                                         in6_ifa_put(ifp);
3510                                         goto restart;
3511                                 }
3512                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3513                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3514                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3515                                         ifp->idev->cnf.dad_transmits *
3516                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3517
3518                                 if (age >= ifp->prefered_lft - regen_advance) {
3519                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3520                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3521                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3522                                         if (!ifp->regen_count && ifpub) {
3523                                                 ifp->regen_count++;
3524                                                 in6_ifa_hold(ifp);
3525                                                 in6_ifa_hold(ifpub);
3526                                                 spin_unlock(&ifp->lock);
3527
3528                                                 spin_lock(&ifpub->lock);
3529                                                 ifpub->regen_count = 0;
3530                                                 spin_unlock(&ifpub->lock);
3531                                                 ipv6_create_tempaddr(ifpub, ifp);
3532                                                 in6_ifa_put(ifpub);
3533                                                 in6_ifa_put(ifp);
3534                                                 goto restart;
3535                                         }
3536                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3537                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3538                                 spin_unlock(&ifp->lock);
3539                         } else {
3540                                 /* ifp->prefered_lft <= ifp->valid_lft */
3541                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3542                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3543                                 spin_unlock(&ifp->lock);
3544                         }
3545                 }
3546         }
3547
3548         next_sec = round_jiffies_up(next);
3549         next_sched = next;
3550
3551         /* If rounded timeout is accurate enough, accept it. */
3552         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3553                 next_sched = next_sec;
3554
3555         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3556         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3557                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3558
3559         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3560               now, next, next_sec, next_sched);
3561
3562         addr_chk_timer.expires = next_sched;
3563         add_timer(&addr_chk_timer);
3564         spin_unlock(&addrconf_verify_lock);
3565         rcu_read_unlock_bh();
3566 }
3567
3568 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3569                                      struct in6_addr **peer_pfx)
3570 {
3571         struct in6_addr *pfx = NULL;
3572
3573         *peer_pfx = NULL;
3574
3575         if (addr)
3576                 pfx = nla_data(addr);
3577
3578         if (local) {
3579                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3580                         *peer_pfx = pfx;
3581                 pfx = nla_data(local);
3582         }
3583
3584         return pfx;
3585 }
3586
3587 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3588         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3589         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3590         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3591         [IFA_FLAGS]             = { .len = sizeof(u32) },
3592 };
3593
3594 static int
3595 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3596 {
3597         struct net *net = sock_net(skb->sk);
3598         struct ifaddrmsg *ifm;
3599         struct nlattr *tb[IFA_MAX+1];
3600         struct in6_addr *pfx, *peer_pfx;
3601         int err;
3602
3603         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3604         if (err < 0)
3605                 return err;
3606
3607         ifm = nlmsg_data(nlh);
3608         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3609         if (pfx == NULL)
3610                 return -EINVAL;
3611
3612         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3613 }
3614
3615 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3616                              u32 prefered_lft, u32 valid_lft)
3617 {
3618         u32 flags;
3619         clock_t expires;
3620         unsigned long timeout;
3621         bool was_managetempaddr;
3622
3623         if (!valid_lft || (prefered_lft > valid_lft))
3624                 return -EINVAL;
3625
3626         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3627             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3628                 return -EINVAL;
3629
3630         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3631         if (addrconf_finite_timeout(timeout)) {
3632                 expires = jiffies_to_clock_t(timeout * HZ);
3633                 valid_lft = timeout;
3634                 flags = RTF_EXPIRES;
3635         } else {
3636                 expires = 0;
3637                 flags = 0;
3638                 ifa_flags |= IFA_F_PERMANENT;
3639         }
3640
3641         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3642         if (addrconf_finite_timeout(timeout)) {
3643                 if (timeout == 0)
3644                         ifa_flags |= IFA_F_DEPRECATED;
3645                 prefered_lft = timeout;
3646         }
3647
3648         spin_lock_bh(&ifp->lock);
3649         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3650         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3651                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR);
3652         ifp->flags |= ifa_flags;
3653         ifp->tstamp = jiffies;
3654         ifp->valid_lft = valid_lft;
3655         ifp->prefered_lft = prefered_lft;
3656
3657         spin_unlock_bh(&ifp->lock);
3658         if (!(ifp->flags&IFA_F_TENTATIVE))
3659                 ipv6_ifa_notify(0, ifp);
3660
3661         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3662                               expires, flags);
3663
3664         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3665                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3666                         valid_lft = prefered_lft = 0;
3667                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3668                                  !was_managetempaddr, jiffies);
3669         }
3670
3671         addrconf_verify(0);
3672
3673         return 0;
3674 }
3675
3676 static int
3677 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3678 {
3679         struct net *net = sock_net(skb->sk);
3680         struct ifaddrmsg *ifm;
3681         struct nlattr *tb[IFA_MAX+1];
3682         struct in6_addr *pfx, *peer_pfx;
3683         struct inet6_ifaddr *ifa;
3684         struct net_device *dev;
3685         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3686         u32 ifa_flags;
3687         int err;
3688
3689         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3690         if (err < 0)
3691                 return err;
3692
3693         ifm = nlmsg_data(nlh);
3694         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3695         if (pfx == NULL)
3696                 return -EINVAL;
3697
3698         if (tb[IFA_CACHEINFO]) {
3699                 struct ifa_cacheinfo *ci;
3700
3701                 ci = nla_data(tb[IFA_CACHEINFO]);
3702                 valid_lft = ci->ifa_valid;
3703                 preferred_lft = ci->ifa_prefered;
3704         } else {
3705                 preferred_lft = INFINITY_LIFE_TIME;
3706                 valid_lft = INFINITY_LIFE_TIME;
3707         }
3708
3709         dev =  __dev_get_by_index(net, ifm->ifa_index);
3710         if (dev == NULL)
3711                 return -ENODEV;
3712
3713         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3714
3715         /* We ignore other flags so far. */
3716         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR;
3717
3718         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3719         if (ifa == NULL) {
3720                 /*
3721                  * It would be best to check for !NLM_F_CREATE here but
3722                  * userspace alreay relies on not having to provide this.
3723                  */
3724                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3725                                       ifm->ifa_prefixlen, ifa_flags,
3726                                       preferred_lft, valid_lft);
3727         }
3728
3729         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3730             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3731                 err = -EEXIST;
3732         else
3733                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3734
3735         in6_ifa_put(ifa);
3736
3737         return err;
3738 }
3739
3740 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3741                           u8 scope, int ifindex)
3742 {
3743         struct ifaddrmsg *ifm;
3744
3745         ifm = nlmsg_data(nlh);
3746         ifm->ifa_family = AF_INET6;
3747         ifm->ifa_prefixlen = prefixlen;
3748         ifm->ifa_flags = flags;
3749         ifm->ifa_scope = scope;
3750         ifm->ifa_index = ifindex;
3751 }
3752
3753 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3754                          unsigned long tstamp, u32 preferred, u32 valid)
3755 {
3756         struct ifa_cacheinfo ci;
3757
3758         ci.cstamp = cstamp_delta(cstamp);
3759         ci.tstamp = cstamp_delta(tstamp);
3760         ci.ifa_prefered = preferred;
3761         ci.ifa_valid = valid;
3762
3763         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3764 }
3765
3766 static inline int rt_scope(int ifa_scope)
3767 {
3768         if (ifa_scope & IFA_HOST)
3769                 return RT_SCOPE_HOST;
3770         else if (ifa_scope & IFA_LINK)
3771                 return RT_SCOPE_LINK;
3772         else if (ifa_scope & IFA_SITE)
3773                 return RT_SCOPE_SITE;
3774         else
3775                 return RT_SCOPE_UNIVERSE;
3776 }
3777
3778 static inline int inet6_ifaddr_msgsize(void)
3779 {
3780         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3781                + nla_total_size(16) /* IFA_LOCAL */
3782                + nla_total_size(16) /* IFA_ADDRESS */
3783                + nla_total_size(sizeof(struct ifa_cacheinfo))
3784                + nla_total_size(4)  /* IFA_FLAGS */;
3785 }
3786
3787 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3788                              u32 portid, u32 seq, int event, unsigned int flags)
3789 {
3790         struct nlmsghdr  *nlh;
3791         u32 preferred, valid;
3792
3793         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3794         if (nlh == NULL)
3795                 return -EMSGSIZE;
3796
3797         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3798                       ifa->idev->dev->ifindex);
3799
3800         if (!(ifa->flags&IFA_F_PERMANENT)) {
3801                 preferred = ifa->prefered_lft;
3802                 valid = ifa->valid_lft;
3803                 if (preferred != INFINITY_LIFE_TIME) {
3804                         long tval = (jiffies - ifa->tstamp)/HZ;
3805                         if (preferred > tval)
3806                                 preferred -= tval;
3807                         else
3808                                 preferred = 0;
3809                         if (valid != INFINITY_LIFE_TIME) {
3810                                 if (valid > tval)
3811                                         valid -= tval;
3812                                 else
3813                                         valid = 0;
3814                         }
3815                 }
3816         } else {
3817                 preferred = INFINITY_LIFE_TIME;
3818                 valid = INFINITY_LIFE_TIME;
3819         }
3820
3821         if (!ipv6_addr_any(&ifa->peer_addr)) {
3822                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3823                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3824                         goto error;
3825         } else
3826                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3827                         goto error;
3828
3829         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3830                 goto error;
3831
3832         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
3833                 goto error;
3834
3835         return nlmsg_end(skb, nlh);
3836
3837 error:
3838         nlmsg_cancel(skb, nlh);
3839         return -EMSGSIZE;
3840 }
3841
3842 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3843                                 u32 portid, u32 seq, int event, u16 flags)
3844 {
3845         struct nlmsghdr  *nlh;
3846         u8 scope = RT_SCOPE_UNIVERSE;
3847         int ifindex = ifmca->idev->dev->ifindex;
3848
3849         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3850                 scope = RT_SCOPE_SITE;
3851
3852         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3853         if (nlh == NULL)
3854                 return -EMSGSIZE;
3855
3856         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3857         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3858             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3859                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3860                 nlmsg_cancel(skb, nlh);
3861                 return -EMSGSIZE;
3862         }
3863
3864         return nlmsg_end(skb, nlh);
3865 }
3866
3867 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3868                                 u32 portid, u32 seq, int event, unsigned int flags)
3869 {
3870         struct nlmsghdr  *nlh;
3871         u8 scope = RT_SCOPE_UNIVERSE;
3872         int ifindex = ifaca->aca_idev->dev->ifindex;
3873
3874         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3875                 scope = RT_SCOPE_SITE;
3876
3877         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3878         if (nlh == NULL)
3879                 return -EMSGSIZE;
3880
3881         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3882         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3883             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3884                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3885                 nlmsg_cancel(skb, nlh);
3886                 return -EMSGSIZE;
3887         }
3888
3889         return nlmsg_end(skb, nlh);
3890 }
3891
3892 enum addr_type_t {
3893         UNICAST_ADDR,
3894         MULTICAST_ADDR,
3895         ANYCAST_ADDR,
3896 };
3897
3898 /* called with rcu_read_lock() */
3899 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3900                           struct netlink_callback *cb, enum addr_type_t type,
3901                           int s_ip_idx, int *p_ip_idx)
3902 {
3903         struct ifmcaddr6 *ifmca;
3904         struct ifacaddr6 *ifaca;
3905         int err = 1;
3906         int ip_idx = *p_ip_idx;
3907
3908         read_lock_bh(&idev->lock);
3909         switch (type) {
3910         case UNICAST_ADDR: {
3911                 struct inet6_ifaddr *ifa;
3912
3913                 /* unicast address incl. temp addr */
3914                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3915                         if (++ip_idx < s_ip_idx)
3916                                 continue;
3917                         err = inet6_fill_ifaddr(skb, ifa,
3918                                                 NETLINK_CB(cb->skb).portid,
3919                                                 cb->nlh->nlmsg_seq,
3920                                                 RTM_NEWADDR,
3921                                                 NLM_F_MULTI);
3922                         if (err <= 0)
3923                                 break;
3924                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3925                 }
3926                 break;
3927         }
3928         case MULTICAST_ADDR:
3929                 /* multicast address */
3930                 for (ifmca = idev->mc_list; ifmca;
3931                      ifmca = ifmca->next, ip_idx++) {
3932                         if (ip_idx < s_ip_idx)
3933                                 continue;
3934                         err = inet6_fill_ifmcaddr(skb, ifmca,
3935                                                   NETLINK_CB(cb->skb).portid,
3936                                                   cb->nlh->nlmsg_seq,
3937                                                   RTM_GETMULTICAST,
3938                                                   NLM_F_MULTI);
3939                         if (err <= 0)
3940                                 break;
3941                 }
3942                 break;
3943         case ANYCAST_ADDR:
3944                 /* anycast address */
3945                 for (ifaca = idev->ac_list; ifaca;
3946                      ifaca = ifaca->aca_next, ip_idx++) {
3947                         if (ip_idx < s_ip_idx)
3948                                 continue;
3949                         err = inet6_fill_ifacaddr(skb, ifaca,
3950                                                   NETLINK_CB(cb->skb).portid,
3951                                                   cb->nlh->nlmsg_seq,
3952                                                   RTM_GETANYCAST,
3953                                                   NLM_F_MULTI);
3954                         if (err <= 0)
3955                                 break;
3956                 }
3957                 break;
3958         default:
3959                 break;
3960         }
3961         read_unlock_bh(&idev->lock);
3962         *p_ip_idx = ip_idx;
3963         return err;
3964 }
3965
3966 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3967                            enum addr_type_t type)
3968 {
3969         struct net *net = sock_net(skb->sk);
3970         int h, s_h;
3971         int idx, ip_idx;
3972         int s_idx, s_ip_idx;
3973         struct net_device *dev;
3974         struct inet6_dev *idev;
3975         struct hlist_head *head;
3976
3977         s_h = cb->args[0];
3978         s_idx = idx = cb->args[1];
3979         s_ip_idx = ip_idx = cb->args[2];
3980
3981         rcu_read_lock();
3982         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3983         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3984                 idx = 0;
3985                 head = &net->dev_index_head[h];
3986                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3987                         if (idx < s_idx)
3988                                 goto cont;
3989                         if (h > s_h || idx > s_idx)
3990                                 s_ip_idx = 0;
3991                         ip_idx = 0;
3992                         idev = __in6_dev_get(dev);
3993                         if (!idev)
3994                                 goto cont;
3995
3996                         if (in6_dump_addrs(idev, skb, cb, type,
3997                                            s_ip_idx, &ip_idx) <= 0)
3998                                 goto done;
3999 cont:
4000                         idx++;
4001                 }
4002         }
4003 done:
4004         rcu_read_unlock();
4005         cb->args[0] = h;
4006         cb->args[1] = idx;
4007         cb->args[2] = ip_idx;
4008
4009         return skb->len;
4010 }
4011
4012 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4013 {
4014         enum addr_type_t type = UNICAST_ADDR;
4015
4016         return inet6_dump_addr(skb, cb, type);
4017 }
4018
4019 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4020 {
4021         enum addr_type_t type = MULTICAST_ADDR;
4022
4023         return inet6_dump_addr(skb, cb, type);
4024 }
4025
4026
4027 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4028 {
4029         enum addr_type_t type = ANYCAST_ADDR;
4030
4031         return inet6_dump_addr(skb, cb, type);
4032 }
4033
4034 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4035 {
4036         struct net *net = sock_net(in_skb->sk);
4037         struct ifaddrmsg *ifm;
4038         struct nlattr *tb[IFA_MAX+1];
4039         struct in6_addr *addr = NULL, *peer;
4040         struct net_device *dev = NULL;
4041         struct inet6_ifaddr *ifa;
4042         struct sk_buff *skb;
4043         int err;
4044
4045         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4046         if (err < 0)
4047                 goto errout;
4048
4049         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4050         if (addr == NULL) {
4051                 err = -EINVAL;
4052                 goto errout;
4053         }
4054
4055         ifm = nlmsg_data(nlh);
4056         if (ifm->ifa_index)
4057                 dev = __dev_get_by_index(net, ifm->ifa_index);
4058
4059         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4060         if (!ifa) {
4061                 err = -EADDRNOTAVAIL;
4062                 goto errout;
4063         }
4064
4065         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4066         if (!skb) {
4067                 err = -ENOBUFS;
4068                 goto errout_ifa;
4069         }
4070
4071         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4072                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4073         if (err < 0) {
4074                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4075                 WARN_ON(err == -EMSGSIZE);
4076                 kfree_skb(skb);
4077                 goto errout_ifa;
4078         }
4079         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4080 errout_ifa:
4081         in6_ifa_put(ifa);
4082 errout:
4083         return err;
4084 }
4085
4086 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4087 {
4088         struct sk_buff *skb;
4089         struct net *net = dev_net(ifa->idev->dev);
4090         int err = -ENOBUFS;
4091
4092         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4093         if (skb == NULL)
4094                 goto errout;
4095
4096         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4097         if (err < 0) {
4098                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4099                 WARN_ON(err == -EMSGSIZE);
4100                 kfree_skb(skb);
4101                 goto errout;
4102         }
4103         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4104         return;
4105 errout:
4106         if (err < 0)
4107                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4108 }
4109
4110 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4111                                 __s32 *array, int bytes)
4112 {
4113         BUG_ON(bytes < (DEVCONF_MAX * 4));
4114
4115         memset(array, 0, bytes);
4116         array[DEVCONF_FORWARDING] = cnf->forwarding;
4117         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4118         array[DEVCONF_MTU6] = cnf->mtu6;
4119         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4120         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4121         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4122         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4123         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4124         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4125                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4126         array[DEVCONF_RTR_SOLICIT_DELAY] =
4127                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4128         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4129         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4130                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4131         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4132                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4133         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4134         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4135         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4136         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4137         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4138         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4139         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4140         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4141 #ifdef CONFIG_IPV6_ROUTER_PREF
4142         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4143         array[DEVCONF_RTR_PROBE_INTERVAL] =
4144                 jiffies_to_msecs(cnf->rtr_probe_interval);
4145 #ifdef CONFIG_IPV6_ROUTE_INFO
4146         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4147 #endif
4148 #endif
4149         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4150         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4151 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4152         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4153 #endif
4154 #ifdef CONFIG_IPV6_MROUTE
4155         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4156 #endif
4157         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4158         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4159         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4160         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4161         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4162 }
4163
4164 static inline size_t inet6_ifla6_size(void)
4165 {
4166         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4167              + nla_total_size(sizeof(struct ifla_cacheinfo))
4168              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4169              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4170              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4171              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4172 }
4173
4174 static inline size_t inet6_if_nlmsg_size(void)
4175 {
4176         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4177                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4178                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4179                + nla_total_size(4) /* IFLA_MTU */
4180                + nla_total_size(4) /* IFLA_LINK */
4181                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4182 }
4183
4184 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4185                                       int items, int bytes)
4186 {
4187         int i;
4188         int pad = bytes - sizeof(u64) * items;
4189         BUG_ON(pad < 0);
4190
4191         /* Use put_unaligned() because stats may not be aligned for u64. */
4192         put_unaligned(items, &stats[0]);
4193         for (i = 1; i < items; i++)
4194                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4195
4196         memset(&stats[items], 0, pad);
4197 }
4198
4199 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4200                                       int items, int bytes, size_t syncpoff)
4201 {
4202         int i;
4203         int pad = bytes - sizeof(u64) * items;
4204         BUG_ON(pad < 0);
4205
4206         /* Use put_unaligned() because stats may not be aligned for u64. */
4207         put_unaligned(items, &stats[0]);
4208         for (i = 1; i < items; i++)
4209                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4210
4211         memset(&stats[items], 0, pad);
4212 }
4213
4214 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4215                              int bytes)
4216 {
4217         switch (attrtype) {
4218         case IFLA_INET6_STATS:
4219                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4220                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4221                 break;
4222         case IFLA_INET6_ICMP6STATS:
4223                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4224                 break;
4225         }
4226 }
4227
4228 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4229 {
4230         struct nlattr *nla;
4231         struct ifla_cacheinfo ci;
4232
4233         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4234                 goto nla_put_failure;
4235         ci.max_reasm_len = IPV6_MAXPLEN;
4236         ci.tstamp = cstamp_delta(idev->tstamp);
4237         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4238         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4239         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4240                 goto nla_put_failure;
4241         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4242         if (nla == NULL)
4243                 goto nla_put_failure;
4244         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4245
4246         /* XXX - MC not implemented */
4247
4248         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4249         if (nla == NULL)
4250                 goto nla_put_failure;
4251         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4252
4253         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4254         if (nla == NULL)
4255                 goto nla_put_failure;
4256         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4257
4258         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4259         if (nla == NULL)
4260                 goto nla_put_failure;
4261         read_lock_bh(&idev->lock);
4262         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4263         read_unlock_bh(&idev->lock);
4264
4265         return 0;
4266
4267 nla_put_failure:
4268         return -EMSGSIZE;
4269 }
4270
4271 static size_t inet6_get_link_af_size(const struct net_device *dev)
4272 {
4273         if (!__in6_dev_get(dev))
4274                 return 0;
4275
4276         return inet6_ifla6_size();
4277 }
4278
4279 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4280 {
4281         struct inet6_dev *idev = __in6_dev_get(dev);
4282
4283         if (!idev)
4284                 return -ENODATA;
4285
4286         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4287                 return -EMSGSIZE;
4288
4289         return 0;
4290 }
4291
4292 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4293 {
4294         struct inet6_ifaddr *ifp;
4295         struct net_device *dev = idev->dev;
4296         bool update_rs = false;
4297         struct in6_addr ll_addr;
4298
4299         if (token == NULL)
4300                 return -EINVAL;
4301         if (ipv6_addr_any(token))
4302                 return -EINVAL;
4303         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4304                 return -EINVAL;
4305         if (!ipv6_accept_ra(idev))
4306                 return -EINVAL;
4307         if (idev->cnf.rtr_solicits <= 0)
4308                 return -EINVAL;
4309
4310         write_lock_bh(&idev->lock);
4311
4312         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4313         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4314
4315         write_unlock_bh(&idev->lock);
4316
4317         if (!idev->dead && (idev->if_flags & IF_READY) &&
4318             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4319                              IFA_F_OPTIMISTIC)) {
4320
4321                 /* If we're not ready, then normal ifup will take care
4322                  * of this. Otherwise, we need to request our rs here.
4323                  */
4324                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4325                 update_rs = true;
4326         }
4327
4328         write_lock_bh(&idev->lock);
4329
4330         if (update_rs) {
4331                 idev->if_flags |= IF_RS_SENT;
4332                 idev->rs_probes = 1;
4333                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4334         }
4335
4336         /* Well, that's kinda nasty ... */
4337         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4338                 spin_lock(&ifp->lock);
4339                 if (ifp->tokenized) {
4340                         ifp->valid_lft = 0;
4341                         ifp->prefered_lft = 0;
4342                 }
4343                 spin_unlock(&ifp->lock);
4344         }
4345
4346         write_unlock_bh(&idev->lock);
4347         addrconf_verify(0);
4348         return 0;
4349 }
4350
4351 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4352 {
4353         int err = -EINVAL;
4354         struct inet6_dev *idev = __in6_dev_get(dev);
4355         struct nlattr *tb[IFLA_INET6_MAX + 1];
4356
4357         if (!idev)
4358                 return -EAFNOSUPPORT;
4359
4360         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4361                 BUG();
4362
4363         if (tb[IFLA_INET6_TOKEN])
4364                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4365
4366         return err;
4367 }
4368
4369 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4370                              u32 portid, u32 seq, int event, unsigned int flags)
4371 {
4372         struct net_device *dev = idev->dev;
4373         struct ifinfomsg *hdr;
4374         struct nlmsghdr *nlh;
4375         void *protoinfo;
4376
4377         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4378         if (nlh == NULL)
4379                 return -EMSGSIZE;
4380
4381         hdr = nlmsg_data(nlh);
4382         hdr->ifi_family = AF_INET6;
4383         hdr->__ifi_pad = 0;
4384         hdr->ifi_type = dev->type;
4385         hdr->ifi_index = dev->ifindex;
4386         hdr->ifi_flags = dev_get_flags(dev);
4387         hdr->ifi_change = 0;
4388
4389         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4390             (dev->addr_len &&
4391              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4392             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4393             (dev->ifindex != dev->iflink &&
4394              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4395                 goto nla_put_failure;
4396         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4397         if (protoinfo == NULL)
4398                 goto nla_put_failure;
4399
4400         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4401                 goto nla_put_failure;
4402
4403         nla_nest_end(skb, protoinfo);
4404         return nlmsg_end(skb, nlh);
4405
4406 nla_put_failure:
4407         nlmsg_cancel(skb, nlh);
4408         return -EMSGSIZE;
4409 }
4410
4411 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4412 {
4413         struct net *net = sock_net(skb->sk);
4414         int h, s_h;
4415         int idx = 0, s_idx;
4416         struct net_device *dev;
4417         struct inet6_dev *idev;
4418         struct hlist_head *head;
4419
4420         s_h = cb->args[0];
4421         s_idx = cb->args[1];
4422
4423         rcu_read_lock();
4424         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4425                 idx = 0;
4426                 head = &net->dev_index_head[h];
4427                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4428                         if (idx < s_idx)
4429                                 goto cont;
4430                         idev = __in6_dev_get(dev);
4431                         if (!idev)
4432                                 goto cont;
4433                         if (inet6_fill_ifinfo(skb, idev,
4434                                               NETLINK_CB(cb->skb).portid,
4435                                               cb->nlh->nlmsg_seq,
4436                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4437                                 goto out;
4438 cont:
4439                         idx++;
4440                 }
4441         }
4442 out:
4443         rcu_read_unlock();
4444         cb->args[1] = idx;
4445         cb->args[0] = h;
4446
4447         return skb->len;
4448 }
4449
4450 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4451 {
4452         struct sk_buff *skb;
4453         struct net *net = dev_net(idev->dev);
4454         int err = -ENOBUFS;
4455
4456         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4457         if (skb == NULL)
4458                 goto errout;
4459
4460         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4461         if (err < 0) {
4462                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4463                 WARN_ON(err == -EMSGSIZE);
4464                 kfree_skb(skb);
4465                 goto errout;
4466         }
4467         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4468         return;
4469 errout:
4470         if (err < 0)
4471                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4472 }
4473
4474 static inline size_t inet6_prefix_nlmsg_size(void)
4475 {
4476         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4477                + nla_total_size(sizeof(struct in6_addr))
4478                + nla_total_size(sizeof(struct prefix_cacheinfo));
4479 }
4480
4481 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4482                              struct prefix_info *pinfo, u32 portid, u32 seq,
4483                              int event, unsigned int flags)
4484 {
4485         struct prefixmsg *pmsg;
4486         struct nlmsghdr *nlh;
4487         struct prefix_cacheinfo ci;
4488
4489         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4490         if (nlh == NULL)
4491                 return -EMSGSIZE;
4492
4493         pmsg = nlmsg_data(nlh);
4494         pmsg->prefix_family = AF_INET6;
4495         pmsg->prefix_pad1 = 0;
4496         pmsg->prefix_pad2 = 0;
4497         pmsg->prefix_ifindex = idev->dev->ifindex;
4498         pmsg->prefix_len = pinfo->prefix_len;
4499         pmsg->prefix_type = pinfo->type;
4500         pmsg->prefix_pad3 = 0;
4501         pmsg->prefix_flags = 0;
4502         if (pinfo->onlink)
4503                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4504         if (pinfo->autoconf)
4505                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4506
4507         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4508                 goto nla_put_failure;
4509         ci.preferred_time = ntohl(pinfo->prefered);
4510         ci.valid_time = ntohl(pinfo->valid);
4511         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4512                 goto nla_put_failure;
4513         return nlmsg_end(skb, nlh);
4514
4515 nla_put_failure:
4516         nlmsg_cancel(skb, nlh);
4517         return -EMSGSIZE;
4518 }
4519
4520 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4521                          struct prefix_info *pinfo)
4522 {
4523         struct sk_buff *skb;
4524         struct net *net = dev_net(idev->dev);
4525         int err = -ENOBUFS;
4526
4527         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4528         if (skb == NULL)
4529                 goto errout;
4530
4531         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4532         if (err < 0) {
4533                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4534                 WARN_ON(err == -EMSGSIZE);
4535                 kfree_skb(skb);
4536                 goto errout;
4537         }
4538         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4539         return;
4540 errout:
4541         if (err < 0)
4542                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4543 }
4544
4545 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4546 {
4547         write_lock_bh(&ifp->idev->lock);
4548         spin_lock(&ifp->lock);
4549         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4550                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4551             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4552                 ifp->idev->valid_ll_addr_cnt += count;
4553         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4554         spin_unlock(&ifp->lock);
4555         write_unlock_bh(&ifp->idev->lock);
4556 }
4557
4558 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4559 {
4560         struct net *net = dev_net(ifp->idev->dev);
4561
4562         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4563
4564         switch (event) {
4565         case RTM_NEWADDR:
4566                 update_valid_ll_addr_cnt(ifp, 1);
4567
4568                 /*
4569                  * If the address was optimistic
4570                  * we inserted the route at the start of
4571                  * our DAD process, so we don't need
4572                  * to do it again
4573                  */
4574                 if (!(ifp->rt->rt6i_node))
4575                         ip6_ins_rt(ifp->rt);
4576                 if (ifp->idev->cnf.forwarding)
4577                         addrconf_join_anycast(ifp);
4578                 if (!ipv6_addr_any(&ifp->peer_addr))
4579                         addrconf_prefix_route(&ifp->peer_addr, 128,
4580                                               ifp->idev->dev, 0, 0);
4581                 break;
4582         case RTM_DELADDR:
4583                 update_valid_ll_addr_cnt(ifp, -1);
4584
4585                 if (ifp->idev->cnf.forwarding)
4586                         addrconf_leave_anycast(ifp);
4587                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4588                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4589                         struct rt6_info *rt;
4590                         struct net_device *dev = ifp->idev->dev;
4591
4592                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4593                                         dev->ifindex, 1);
4594                         if (rt) {
4595                                 dst_hold(&rt->dst);
4596                                 if (ip6_del_rt(rt))
4597                                         dst_free(&rt->dst);
4598                         }
4599                 }
4600                 dst_hold(&ifp->rt->dst);
4601
4602                 if (ip6_del_rt(ifp->rt))
4603                         dst_free(&ifp->rt->dst);
4604                 break;
4605         }
4606         atomic_inc(&net->ipv6.dev_addr_genid);
4607         rt_genid_bump_ipv6(net);
4608 }
4609
4610 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4611 {
4612         rcu_read_lock_bh();
4613         if (likely(ifp->idev->dead == 0))
4614                 __ipv6_ifa_notify(event, ifp);
4615         rcu_read_unlock_bh();
4616 }
4617
4618 #ifdef CONFIG_SYSCTL
4619
4620 static
4621 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4622                            void __user *buffer, size_t *lenp, loff_t *ppos)
4623 {
4624         int *valp = ctl->data;
4625         int val = *valp;
4626         loff_t pos = *ppos;
4627         struct ctl_table lctl;
4628         int ret;
4629
4630         /*
4631          * ctl->data points to idev->cnf.forwarding, we should
4632          * not modify it until we get the rtnl lock.
4633          */
4634         lctl = *ctl;
4635         lctl.data = &val;
4636
4637         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4638
4639         if (write)
4640                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4641         if (ret)
4642                 *ppos = pos;
4643         return ret;
4644 }
4645
4646 static void dev_disable_change(struct inet6_dev *idev)
4647 {
4648         struct netdev_notifier_info info;
4649
4650         if (!idev || !idev->dev)
4651                 return;
4652
4653         netdev_notifier_info_init(&info, idev->dev);
4654         if (idev->cnf.disable_ipv6)
4655                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4656         else
4657                 addrconf_notify(NULL, NETDEV_UP, &info);
4658 }
4659
4660 static void addrconf_disable_change(struct net *net, __s32 newf)
4661 {
4662         struct net_device *dev;
4663         struct inet6_dev *idev;
4664
4665         rcu_read_lock();
4666         for_each_netdev_rcu(net, dev) {
4667                 idev = __in6_dev_get(dev);
4668                 if (idev) {
4669                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4670                         idev->cnf.disable_ipv6 = newf;
4671                         if (changed)
4672                                 dev_disable_change(idev);
4673                 }
4674         }
4675         rcu_read_unlock();
4676 }
4677
4678 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4679 {
4680         struct net *net;
4681         int old;
4682
4683         if (!rtnl_trylock())
4684                 return restart_syscall();
4685
4686         net = (struct net *)table->extra2;
4687         old = *p;
4688         *p = newf;
4689
4690         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4691                 rtnl_unlock();
4692                 return 0;
4693         }
4694
4695         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4696                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4697                 addrconf_disable_change(net, newf);
4698         } else if ((!newf) ^ (!old))
4699                 dev_disable_change((struct inet6_dev *)table->extra1);
4700
4701         rtnl_unlock();
4702         return 0;
4703 }
4704
4705 static
4706 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4707                             void __user *buffer, size_t *lenp, loff_t *ppos)
4708 {
4709         int *valp = ctl->data;
4710         int val = *valp;
4711         loff_t pos = *ppos;
4712         struct ctl_table lctl;
4713         int ret;
4714
4715         /*
4716          * ctl->data points to idev->cnf.disable_ipv6, we should
4717          * not modify it until we get the rtnl lock.
4718          */
4719         lctl = *ctl;
4720         lctl.data = &val;
4721
4722         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4723
4724         if (write)
4725                 ret = addrconf_disable_ipv6(ctl, valp, val);
4726         if (ret)
4727                 *ppos = pos;
4728         return ret;
4729 }
4730
4731 static struct addrconf_sysctl_table
4732 {
4733         struct ctl_table_header *sysctl_header;
4734         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4735 } addrconf_sysctl __read_mostly = {
4736         .sysctl_header = NULL,
4737         .addrconf_vars = {
4738                 {
4739                         .procname       = "forwarding",
4740                         .data           = &ipv6_devconf.forwarding,
4741                         .maxlen         = sizeof(int),
4742                         .mode           = 0644,
4743                         .proc_handler   = addrconf_sysctl_forward,
4744                 },
4745                 {
4746                         .procname       = "hop_limit",
4747                         .data           = &ipv6_devconf.hop_limit,
4748                         .maxlen         = sizeof(int),
4749                         .mode           = 0644,
4750                         .proc_handler   = proc_dointvec,
4751                 },
4752                 {
4753                         .procname       = "mtu",
4754                         .data           = &ipv6_devconf.mtu6,
4755                         .maxlen         = sizeof(int),
4756                         .mode           = 0644,
4757                         .proc_handler   = proc_dointvec,
4758                 },
4759                 {
4760                         .procname       = "accept_ra",
4761                         .data           = &ipv6_devconf.accept_ra,
4762                         .maxlen         = sizeof(int),
4763                         .mode           = 0644,
4764                         .proc_handler   = proc_dointvec,
4765                 },
4766                 {
4767                         .procname       = "accept_redirects",
4768                         .data           = &ipv6_devconf.accept_redirects,
4769                         .maxlen         = sizeof(int),
4770                         .mode           = 0644,
4771                         .proc_handler   = proc_dointvec,
4772                 },
4773                 {
4774                         .procname       = "autoconf",
4775                         .data           = &ipv6_devconf.autoconf,
4776                         .maxlen         = sizeof(int),
4777                         .mode           = 0644,
4778                         .proc_handler   = proc_dointvec,
4779                 },
4780                 {
4781                         .procname       = "dad_transmits",
4782                         .data           = &ipv6_devconf.dad_transmits,
4783                         .maxlen         = sizeof(int),
4784                         .mode           = 0644,
4785                         .proc_handler   = proc_dointvec,
4786                 },
4787                 {
4788                         .procname       = "router_solicitations",
4789                         .data           = &ipv6_devconf.rtr_solicits,
4790                         .maxlen         = sizeof(int),
4791                         .mode           = 0644,
4792                         .proc_handler   = proc_dointvec,
4793                 },
4794                 {
4795                         .procname       = "router_solicitation_interval",
4796                         .data           = &ipv6_devconf.rtr_solicit_interval,
4797                         .maxlen         = sizeof(int),
4798                         .mode           = 0644,
4799                         .proc_handler   = proc_dointvec_jiffies,
4800                 },
4801                 {
4802                         .procname       = "router_solicitation_delay",
4803                         .data           = &ipv6_devconf.rtr_solicit_delay,
4804                         .maxlen         = sizeof(int),
4805                         .mode           = 0644,
4806                         .proc_handler   = proc_dointvec_jiffies,
4807                 },
4808                 {
4809                         .procname       = "force_mld_version",
4810                         .data           = &ipv6_devconf.force_mld_version,
4811                         .maxlen         = sizeof(int),
4812                         .mode           = 0644,
4813                         .proc_handler   = proc_dointvec,
4814                 },
4815                 {
4816                         .procname       = "mldv1_unsolicited_report_interval",
4817                         .data           =
4818                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4819                         .maxlen         = sizeof(int),
4820                         .mode           = 0644,
4821                         .proc_handler   = proc_dointvec_ms_jiffies,
4822                 },
4823                 {
4824                         .procname       = "mldv2_unsolicited_report_interval",
4825                         .data           =
4826                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4827                         .maxlen         = sizeof(int),
4828                         .mode           = 0644,
4829                         .proc_handler   = proc_dointvec_ms_jiffies,
4830                 },
4831                 {
4832                         .procname       = "use_tempaddr",
4833                         .data           = &ipv6_devconf.use_tempaddr,
4834                         .maxlen         = sizeof(int),
4835                         .mode           = 0644,
4836                         .proc_handler   = proc_dointvec,
4837                 },
4838                 {
4839                         .procname       = "temp_valid_lft",
4840                         .data           = &ipv6_devconf.temp_valid_lft,
4841                         .maxlen         = sizeof(int),
4842                         .mode           = 0644,
4843                         .proc_handler   = proc_dointvec,
4844                 },
4845                 {
4846                         .procname       = "temp_prefered_lft",
4847                         .data           = &ipv6_devconf.temp_prefered_lft,
4848                         .maxlen         = sizeof(int),
4849                         .mode           = 0644,
4850                         .proc_handler   = proc_dointvec,
4851                 },
4852                 {
4853                         .procname       = "regen_max_retry",
4854                         .data           = &ipv6_devconf.regen_max_retry,
4855                         .maxlen         = sizeof(int),
4856                         .mode           = 0644,
4857                         .proc_handler   = proc_dointvec,
4858                 },
4859                 {
4860                         .procname       = "max_desync_factor",
4861                         .data           = &ipv6_devconf.max_desync_factor,
4862                         .maxlen         = sizeof(int),
4863                         .mode           = 0644,
4864                         .proc_handler   = proc_dointvec,
4865                 },
4866                 {
4867                         .procname       = "max_addresses",
4868                         .data           = &ipv6_devconf.max_addresses,
4869                         .maxlen         = sizeof(int),
4870                         .mode           = 0644,
4871                         .proc_handler   = proc_dointvec,
4872                 },
4873                 {
4874                         .procname       = "accept_ra_defrtr",
4875                         .data           = &ipv6_devconf.accept_ra_defrtr,
4876                         .maxlen         = sizeof(int),
4877                         .mode           = 0644,
4878                         .proc_handler   = proc_dointvec,
4879                 },
4880                 {
4881                         .procname       = "accept_ra_pinfo",
4882                         .data           = &ipv6_devconf.accept_ra_pinfo,
4883                         .maxlen         = sizeof(int),
4884                         .mode           = 0644,
4885                         .proc_handler   = proc_dointvec,
4886                 },
4887 #ifdef CONFIG_IPV6_ROUTER_PREF
4888                 {
4889                         .procname       = "accept_ra_rtr_pref",
4890                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4891                         .maxlen         = sizeof(int),
4892                         .mode           = 0644,
4893                         .proc_handler   = proc_dointvec,
4894                 },
4895                 {
4896                         .procname       = "router_probe_interval",
4897                         .data           = &ipv6_devconf.rtr_probe_interval,
4898                         .maxlen         = sizeof(int),
4899                         .mode           = 0644,
4900                         .proc_handler   = proc_dointvec_jiffies,
4901                 },
4902 #ifdef CONFIG_IPV6_ROUTE_INFO
4903                 {
4904                         .procname       = "accept_ra_rt_info_max_plen",
4905                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4906                         .maxlen         = sizeof(int),
4907                         .mode           = 0644,
4908                         .proc_handler   = proc_dointvec,
4909                 },
4910 #endif
4911 #endif
4912                 {
4913                         .procname       = "proxy_ndp",
4914                         .data           = &ipv6_devconf.proxy_ndp,
4915                         .maxlen         = sizeof(int),
4916                         .mode           = 0644,
4917                         .proc_handler   = proc_dointvec,
4918                 },
4919                 {
4920                         .procname       = "accept_source_route",
4921                         .data           = &ipv6_devconf.accept_source_route,
4922                         .maxlen         = sizeof(int),
4923                         .mode           = 0644,
4924                         .proc_handler   = proc_dointvec,
4925                 },
4926 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4927                 {
4928                         .procname       = "optimistic_dad",
4929                         .data           = &ipv6_devconf.optimistic_dad,
4930                         .maxlen         = sizeof(int),
4931                         .mode           = 0644,
4932                         .proc_handler   = proc_dointvec,
4933
4934                 },
4935 #endif
4936 #ifdef CONFIG_IPV6_MROUTE
4937                 {
4938                         .procname       = "mc_forwarding",
4939                         .data           = &ipv6_devconf.mc_forwarding,
4940                         .maxlen         = sizeof(int),
4941                         .mode           = 0444,
4942                         .proc_handler   = proc_dointvec,
4943                 },
4944 #endif
4945                 {
4946                         .procname       = "disable_ipv6",
4947                         .data           = &ipv6_devconf.disable_ipv6,
4948                         .maxlen         = sizeof(int),
4949                         .mode           = 0644,
4950                         .proc_handler   = addrconf_sysctl_disable,
4951                 },
4952                 {
4953                         .procname       = "accept_dad",
4954                         .data           = &ipv6_devconf.accept_dad,
4955                         .maxlen         = sizeof(int),
4956                         .mode           = 0644,
4957                         .proc_handler   = proc_dointvec,
4958                 },
4959                 {
4960                         .procname       = "force_tllao",
4961                         .data           = &ipv6_devconf.force_tllao,
4962                         .maxlen         = sizeof(int),
4963                         .mode           = 0644,
4964                         .proc_handler   = proc_dointvec
4965                 },
4966                 {
4967                         .procname       = "ndisc_notify",
4968                         .data           = &ipv6_devconf.ndisc_notify,
4969                         .maxlen         = sizeof(int),
4970                         .mode           = 0644,
4971                         .proc_handler   = proc_dointvec
4972                 },
4973                 {
4974                         .procname       = "suppress_frag_ndisc",
4975                         .data           = &ipv6_devconf.suppress_frag_ndisc,
4976                         .maxlen         = sizeof(int),
4977                         .mode           = 0644,
4978                         .proc_handler   = proc_dointvec
4979                 },
4980                 {
4981                         /* sentinel */
4982                 }
4983         },
4984 };
4985
4986 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4987                 struct inet6_dev *idev, struct ipv6_devconf *p)
4988 {
4989         int i;
4990         struct addrconf_sysctl_table *t;
4991         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4992
4993         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4994         if (t == NULL)
4995                 goto out;
4996
4997         for (i = 0; t->addrconf_vars[i].data; i++) {
4998                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4999                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5000                 t->addrconf_vars[i].extra2 = net;
5001         }
5002
5003         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5004
5005         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5006         if (t->sysctl_header == NULL)
5007                 goto free;
5008
5009         p->sysctl = t;
5010         return 0;
5011
5012 free:
5013         kfree(t);
5014 out:
5015         return -ENOBUFS;
5016 }
5017
5018 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5019 {
5020         struct addrconf_sysctl_table *t;
5021
5022         if (p->sysctl == NULL)
5023                 return;
5024
5025         t = p->sysctl;
5026         p->sysctl = NULL;
5027         unregister_net_sysctl_table(t->sysctl_header);
5028         kfree(t);
5029 }
5030
5031 static void addrconf_sysctl_register(struct inet6_dev *idev)
5032 {
5033         neigh_sysctl_register(idev->dev, idev->nd_parms,
5034                               &ndisc_ifinfo_sysctl_change);
5035         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5036                                         idev, &idev->cnf);
5037 }
5038
5039 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5040 {
5041         __addrconf_sysctl_unregister(&idev->cnf);
5042         neigh_sysctl_unregister(idev->nd_parms);
5043 }
5044
5045
5046 #endif
5047
5048 static int __net_init addrconf_init_net(struct net *net)
5049 {
5050         int err = -ENOMEM;
5051         struct ipv6_devconf *all, *dflt;
5052
5053         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5054         if (all == NULL)
5055                 goto err_alloc_all;
5056
5057         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5058         if (dflt == NULL)
5059                 goto err_alloc_dflt;
5060
5061         /* these will be inherited by all namespaces */
5062         dflt->autoconf = ipv6_defaults.autoconf;
5063         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5064
5065         net->ipv6.devconf_all = all;
5066         net->ipv6.devconf_dflt = dflt;
5067
5068 #ifdef CONFIG_SYSCTL
5069         err = __addrconf_sysctl_register(net, "all", NULL, all);
5070         if (err < 0)
5071                 goto err_reg_all;
5072
5073         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5074         if (err < 0)
5075                 goto err_reg_dflt;
5076 #endif
5077         return 0;
5078
5079 #ifdef CONFIG_SYSCTL
5080 err_reg_dflt:
5081         __addrconf_sysctl_unregister(all);
5082 err_reg_all:
5083         kfree(dflt);
5084 #endif
5085 err_alloc_dflt:
5086         kfree(all);
5087 err_alloc_all:
5088         return err;
5089 }
5090
5091 static void __net_exit addrconf_exit_net(struct net *net)
5092 {
5093 #ifdef CONFIG_SYSCTL
5094         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5095         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5096 #endif
5097         if (!net_eq(net, &init_net)) {
5098                 kfree(net->ipv6.devconf_dflt);
5099                 kfree(net->ipv6.devconf_all);
5100         }
5101 }
5102
5103 static struct pernet_operations addrconf_ops = {
5104         .init = addrconf_init_net,
5105         .exit = addrconf_exit_net,
5106 };
5107
5108 static struct rtnl_af_ops inet6_ops = {
5109         .family           = AF_INET6,
5110         .fill_link_af     = inet6_fill_link_af,
5111         .get_link_af_size = inet6_get_link_af_size,
5112         .set_link_af      = inet6_set_link_af,
5113 };
5114
5115 /*
5116  *      Init / cleanup code
5117  */
5118
5119 int __init addrconf_init(void)
5120 {
5121         int i, err;
5122
5123         err = ipv6_addr_label_init();
5124         if (err < 0) {
5125                 pr_crit("%s: cannot initialize default policy table: %d\n",
5126                         __func__, err);
5127                 goto out;
5128         }
5129
5130         err = register_pernet_subsys(&addrconf_ops);
5131         if (err < 0)
5132                 goto out_addrlabel;
5133
5134         /* The addrconf netdev notifier requires that loopback_dev
5135          * has it's ipv6 private information allocated and setup
5136          * before it can bring up and give link-local addresses
5137          * to other devices which are up.
5138          *
5139          * Unfortunately, loopback_dev is not necessarily the first
5140          * entry in the global dev_base list of net devices.  In fact,
5141          * it is likely to be the very last entry on that list.
5142          * So this causes the notifier registry below to try and
5143          * give link-local addresses to all devices besides loopback_dev
5144          * first, then loopback_dev, which cases all the non-loopback_dev
5145          * devices to fail to get a link-local address.
5146          *
5147          * So, as a temporary fix, allocate the ipv6 structure for
5148          * loopback_dev first by hand.
5149          * Longer term, all of the dependencies ipv6 has upon the loopback
5150          * device and it being up should be removed.
5151          */
5152         rtnl_lock();
5153         if (!ipv6_add_dev(init_net.loopback_dev))
5154                 err = -ENOMEM;
5155         rtnl_unlock();
5156         if (err)
5157                 goto errlo;
5158
5159         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5160                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5161
5162         register_netdevice_notifier(&ipv6_dev_notf);
5163
5164         addrconf_verify(0);
5165
5166         err = rtnl_af_register(&inet6_ops);
5167         if (err < 0)
5168                 goto errout_af;
5169
5170         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5171                               NULL);
5172         if (err < 0)
5173                 goto errout;
5174
5175         /* Only the first call to __rtnl_register can fail */
5176         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5177         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5178         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5179                         inet6_dump_ifaddr, NULL);
5180         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5181                         inet6_dump_ifmcaddr, NULL);
5182         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5183                         inet6_dump_ifacaddr, NULL);
5184         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5185                         inet6_netconf_dump_devconf, NULL);
5186
5187         ipv6_addr_label_rtnl_register();
5188
5189         return 0;
5190 errout:
5191         rtnl_af_unregister(&inet6_ops);
5192 errout_af:
5193         unregister_netdevice_notifier(&ipv6_dev_notf);
5194 errlo:
5195         unregister_pernet_subsys(&addrconf_ops);
5196 out_addrlabel:
5197         ipv6_addr_label_cleanup();
5198 out:
5199         return err;
5200 }
5201
5202 void addrconf_cleanup(void)
5203 {
5204         struct net_device *dev;
5205         int i;
5206
5207         unregister_netdevice_notifier(&ipv6_dev_notf);
5208         unregister_pernet_subsys(&addrconf_ops);
5209         ipv6_addr_label_cleanup();
5210
5211         rtnl_lock();
5212
5213         __rtnl_af_unregister(&inet6_ops);
5214
5215         /* clean dev list */
5216         for_each_netdev(&init_net, dev) {
5217                 if (__in6_dev_get(dev) == NULL)
5218                         continue;
5219                 addrconf_ifdown(dev, 1);
5220         }
5221         addrconf_ifdown(init_net.loopback_dev, 2);
5222
5223         /*
5224          *      Check hash table.
5225          */
5226         spin_lock_bh(&addrconf_hash_lock);
5227         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5228                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5229         spin_unlock_bh(&addrconf_hash_lock);
5230
5231         del_timer(&addr_chk_timer);
5232         rtnl_unlock();
5233 }