Pileus Git - ~andy/linux/blob - 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         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 446                 size += nla_total_size(4);
 447
 448         return size;
 449 }
 450
 451 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 452                                       struct ipv6_devconf *devconf, u32 portid,
 453                                       u32 seq, int event, unsigned int flags,
 454                                       int type)
 455 {
 456         struct nlmsghdr  *nlh;
 457         struct netconfmsg *ncm;
 458
 459         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 460                         flags);
 461         if (nlh == NULL)
 462                 return -EMSGSIZE;
 463
 464         ncm = nlmsg_data(nlh);
 465         ncm->ncm_family = AF_INET6;
 466
 467         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 468                 goto nla_put_failure;
 469
 470         /* type -1 is used for ALL */
 471         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 472             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 473                 goto nla_put_failure;
 474 #ifdef CONFIG_IPV6_MROUTE
 475         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 476             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 477                         devconf->mc_forwarding) < 0)
 478                 goto nla_put_failure;
 479 #endif
 480         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 481             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 482                 goto nla_put_failure;
 483
 484         return nlmsg_end(skb, nlh);
 485
 486 nla_put_failure:
 487         nlmsg_cancel(skb, nlh);
 488         return -EMSGSIZE;
 489 }
 490
 491 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 492                                   struct ipv6_devconf *devconf)
 493 {
 494         struct sk_buff *skb;
 495         int err = -ENOBUFS;
 496
 497         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 498         if (skb == NULL)
 499                 goto errout;
 500
 501         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 502                                          RTM_NEWNETCONF, 0, type);
 503         if (err < 0) {
 504                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 505                 WARN_ON(err == -EMSGSIZE);
 506                 kfree_skb(skb);
 507                 goto errout;
 508         }
 509         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 510         return;
 511 errout:
 512         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 513 }
 514
 515 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 516         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 517         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 518         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 519 };
 520
 521 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 522                                      struct nlmsghdr *nlh)
 523 {
 524         struct net *net = sock_net(in_skb->sk);
 525         struct nlattr *tb[NETCONFA_MAX+1];
 526         struct netconfmsg *ncm;
 527         struct sk_buff *skb;
 528         struct ipv6_devconf *devconf;
 529         struct inet6_dev *in6_dev;
 530         struct net_device *dev;
 531         int ifindex;
 532         int err;
 533
 534         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 535                           devconf_ipv6_policy);
 536         if (err < 0)
 537                 goto errout;
 538
 539         err = EINVAL;
 540         if (!tb[NETCONFA_IFINDEX])
 541                 goto errout;
 542
 543         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 544         switch (ifindex) {
 545         case NETCONFA_IFINDEX_ALL:
 546                 devconf = net->ipv6.devconf_all;
 547                 break;
 548         case NETCONFA_IFINDEX_DEFAULT:
 549                 devconf = net->ipv6.devconf_dflt;
 550                 break;
 551         default:
 552                 dev = __dev_get_by_index(net, ifindex);
 553                 if (dev == NULL)
 554                         goto errout;
 555                 in6_dev = __in6_dev_get(dev);
 556                 if (in6_dev == NULL)
 557                         goto errout;
 558                 devconf = &in6_dev->cnf;
 559                 break;
 560         }
 561
 562         err = -ENOBUFS;
 563         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 564         if (skb == NULL)
 565                 goto errout;
 566
 567         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 568                                          NETLINK_CB(in_skb).portid,
 569                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 570                                          -1);
 571         if (err < 0) {
 572                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 573                 WARN_ON(err == -EMSGSIZE);
 574                 kfree_skb(skb);
 575                 goto errout;
 576         }
 577         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 578 errout:
 579         return err;
 580 }
 581
 582 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 583                                       struct netlink_callback *cb)
 584 {
 585         struct net *net = sock_net(skb->sk);
 586         int h, s_h;
 587         int idx, s_idx;
 588         struct net_device *dev;
 589         struct inet6_dev *idev;
 590         struct hlist_head *head;
 591
 592         s_h = cb->args[0];
 593         s_idx = idx = cb->args[1];
 594
 595         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 596                 idx = 0;
 597                 head = &net->dev_index_head[h];
 598                 rcu_read_lock();
 599                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 600                           net->dev_base_seq;
 601                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 602                         if (idx < s_idx)
 603                                 goto cont;
 604                         idev = __in6_dev_get(dev);
 605                         if (!idev)
 606                                 goto cont;
 607
 608                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 609                                                        &idev->cnf,
 610                                                        NETLINK_CB(cb->skb).portid,
 611                                                        cb->nlh->nlmsg_seq,
 612                                                        RTM_NEWNETCONF,
 613                                                        NLM_F_MULTI,
 614                                                        -1) <= 0) {
 615                                 rcu_read_unlock();
 616                                 goto done;
 617                         }
 618                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 619 cont:
 620                         idx++;
 621                 }
 622                 rcu_read_unlock();
 623         }
 624         if (h == NETDEV_HASHENTRIES) {
 625                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 626                                                net->ipv6.devconf_all,
 627                                                NETLINK_CB(cb->skb).portid,
 628                                                cb->nlh->nlmsg_seq,
 629                                                RTM_NEWNETCONF, NLM_F_MULTI,
 630                                                -1) <= 0)
 631                         goto done;
 632                 else
 633                         h++;
 634         }
 635         if (h == NETDEV_HASHENTRIES + 1) {
 636                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 637                                                net->ipv6.devconf_dflt,
 638                                                NETLINK_CB(cb->skb).portid,
 639                                                cb->nlh->nlmsg_seq,
 640                                                RTM_NEWNETCONF, NLM_F_MULTI,
 641                                                -1) <= 0)
 642                         goto done;
 643                 else
 644                         h++;
 645         }
 646 done:
 647         cb->args[0] = h;
 648         cb->args[1] = idx;
 649
 650         return skb->len;
 651 }
 652
 653 #ifdef CONFIG_SYSCTL
 654 static void dev_forward_change(struct inet6_dev *idev)
 655 {
 656         struct net_device *dev;
 657         struct inet6_ifaddr *ifa;
 658
 659         if (!idev)
 660                 return;
 661         dev = idev->dev;
 662         if (idev->cnf.forwarding)
 663                 dev_disable_lro(dev);
 664         if (dev->flags & IFF_MULTICAST) {
 665                 if (idev->cnf.forwarding) {
 666                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 667                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 668                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 669                 } else {
 670                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 671                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 672                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 673                 }
 674         }
 675
 676         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 677                 if (ifa->flags&IFA_F_TENTATIVE)
 678                         continue;
 679                 if (idev->cnf.forwarding)
 680                         addrconf_join_anycast(ifa);
 681                 else
 682                         addrconf_leave_anycast(ifa);
 683         }
 684         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 685                                      dev->ifindex, &idev->cnf);
 686 }
 687
 688
 689 static void addrconf_forward_change(struct net *net, __s32 newf)
 690 {
 691         struct net_device *dev;
 692         struct inet6_dev *idev;
 693
 694         for_each_netdev(net, dev) {
 695                 idev = __in6_dev_get(dev);
 696                 if (idev) {
 697                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 698                         idev->cnf.forwarding = newf;
 699                         if (changed)
 700                                 dev_forward_change(idev);
 701                 }
 702         }
 703 }
 704
 705 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 706 {
 707         struct net *net;
 708         int old;
 709
 710         if (!rtnl_trylock())
 711                 return restart_syscall();
 712
 713         net = (struct net *)table->extra2;
 714         old = *p;
 715         *p = newf;
 716
 717         if (p == &net->ipv6.devconf_dflt->forwarding) {
 718                 if ((!newf) ^ (!old))
 719                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 720                                                      NETCONFA_IFINDEX_DEFAULT,
 721                                                      net->ipv6.devconf_dflt);
 722                 rtnl_unlock();
 723                 return 0;
 724         }
 725
 726         if (p == &net->ipv6.devconf_all->forwarding) {
 727                 net->ipv6.devconf_dflt->forwarding = newf;
 728                 addrconf_forward_change(net, newf);
 729                 if ((!newf) ^ (!old))
 730                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 731                                                      NETCONFA_IFINDEX_ALL,
 732                                                      net->ipv6.devconf_all);
 733         } else if ((!newf) ^ (!old))
 734                 dev_forward_change((struct inet6_dev *)table->extra1);
 735         rtnl_unlock();
 736
 737         if (newf)
 738                 rt6_purge_dflt_routers(net);
 739         return 1;
 740 }
 741 #endif
 742
 743 /* Nobody refers to this ifaddr, destroy it */
 744 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 745 {
 746         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 747
 748 #ifdef NET_REFCNT_DEBUG
 749         pr_debug("%s\n", __func__);
 750 #endif
 751
 752         in6_dev_put(ifp->idev);
 753
 754         if (del_timer(&ifp->dad_timer))
 755                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 756
 757         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 758                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 759                 return;
 760         }
 761         ip6_rt_put(ifp->rt);
 762
 763         kfree_rcu(ifp, rcu);
 764 }
 765
 766 static void
 767 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 768 {
 769         struct list_head *p;
 770         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 771
 772         /*
 773          * Each device address list is sorted in order of scope -
 774          * global before linklocal.
 775          */
 776         list_for_each(p, &idev->addr_list) {
 777                 struct inet6_ifaddr *ifa
 778                         = list_entry(p, struct inet6_ifaddr, if_list);
 779                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 780                         break;
 781         }
 782
 783         list_add_tail(&ifp->if_list, p);
 784 }
 785
 786 static u32 inet6_addr_hash(const struct in6_addr *addr)
 787 {
 788         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 789 }
 790
 791 /* On success it returns ifp with increased reference count */
 792
 793 static struct inet6_ifaddr *
 794 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 795               const struct in6_addr *peer_addr, int pfxlen,
 796               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 797 {
 798         struct inet6_ifaddr *ifa = NULL;
 799         struct rt6_info *rt;
 800         unsigned int hash;
 801         int err = 0;
 802         int addr_type = ipv6_addr_type(addr);
 803
 804         if (addr_type == IPV6_ADDR_ANY ||
 805             addr_type & IPV6_ADDR_MULTICAST ||
 806             (!(idev->dev->flags & IFF_LOOPBACK) &&
 807              addr_type & IPV6_ADDR_LOOPBACK))
 808                 return ERR_PTR(-EADDRNOTAVAIL);
 809
 810         rcu_read_lock_bh();
 811         if (idev->dead) {
 812                 err = -ENODEV;                  /*XXX*/
 813                 goto out2;
 814         }
 815
 816         if (idev->cnf.disable_ipv6) {
 817                 err = -EACCES;
 818                 goto out2;
 819         }
 820
 821         spin_lock(&addrconf_hash_lock);
 822
 823         /* Ignore adding duplicate addresses on an interface */
 824         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 825                 ADBG("ipv6_add_addr: already assigned\n");
 826                 err = -EEXIST;
 827                 goto out;
 828         }
 829
 830         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 831
 832         if (ifa == NULL) {
 833                 ADBG("ipv6_add_addr: malloc failed\n");
 834                 err = -ENOBUFS;
 835                 goto out;
 836         }
 837
 838         rt = addrconf_dst_alloc(idev, addr, false);
 839         if (IS_ERR(rt)) {
 840                 err = PTR_ERR(rt);
 841                 goto out;
 842         }
 843
 844         neigh_parms_data_state_setall(idev->nd_parms);
 845
 846         ifa->addr = *addr;
 847         if (peer_addr)
 848                 ifa->peer_addr = *peer_addr;
 849
 850         spin_lock_init(&ifa->lock);
 851         spin_lock_init(&ifa->state_lock);
 852         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 853                     (unsigned long)ifa);
 854         INIT_HLIST_NODE(&ifa->addr_lst);
 855         ifa->scope = scope;
 856         ifa->prefix_len = pfxlen;
 857         ifa->flags = flags | IFA_F_TENTATIVE;
 858         ifa->valid_lft = valid_lft;
 859         ifa->prefered_lft = prefered_lft;
 860         ifa->cstamp = ifa->tstamp = jiffies;
 861         ifa->tokenized = false;
 862
 863         ifa->rt = rt;
 864
 865         ifa->idev = idev;
 866         in6_dev_hold(idev);
 867         /* For caller */
 868         in6_ifa_hold(ifa);
 869
 870         /* Add to big hash table */
 871         hash = inet6_addr_hash(addr);
 872
 873         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 874         spin_unlock(&addrconf_hash_lock);
 875
 876         write_lock(&idev->lock);
 877         /* Add to inet6_dev unicast addr list. */
 878         ipv6_link_dev_addr(idev, ifa);
 879
 880         if (ifa->flags&IFA_F_TEMPORARY) {
 881                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 882                 in6_ifa_hold(ifa);
 883         }
 884
 885         in6_ifa_hold(ifa);
 886         write_unlock(&idev->lock);
 887 out2:
 888         rcu_read_unlock_bh();
 889
 890         if (likely(err == 0))
 891                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 892         else {
 893                 kfree(ifa);
 894                 ifa = ERR_PTR(err);
 895         }
 896
 897         return ifa;
 898 out:
 899         spin_unlock(&addrconf_hash_lock);
 900         goto out2;
 901 }
 902
 903 /* This function wants to get referenced ifp and releases it before return */
 904
 905 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 906 {
 907         struct inet6_ifaddr *ifa, *ifn;
 908         struct inet6_dev *idev = ifp->idev;
 909         int state;
 910         int deleted = 0, onlink = 0;
 911         unsigned long expires = jiffies;
 912
 913         spin_lock_bh(&ifp->state_lock);
 914         state = ifp->state;
 915         ifp->state = INET6_IFADDR_STATE_DEAD;
 916         spin_unlock_bh(&ifp->state_lock);
 917
 918         if (state == INET6_IFADDR_STATE_DEAD)
 919                 goto out;
 920
 921         spin_lock_bh(&addrconf_hash_lock);
 922         hlist_del_init_rcu(&ifp->addr_lst);
 923         spin_unlock_bh(&addrconf_hash_lock);
 924
 925         write_lock_bh(&idev->lock);
 926
 927         if (ifp->flags&IFA_F_TEMPORARY) {
 928                 list_del(&ifp->tmp_list);
 929                 if (ifp->ifpub) {
 930                         in6_ifa_put(ifp->ifpub);
 931                         ifp->ifpub = NULL;
 932                 }
 933                 __in6_ifa_put(ifp);
 934         }
 935
 936         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 937                 if (ifa == ifp) {
 938                         list_del_init(&ifp->if_list);
 939                         __in6_ifa_put(ifp);
 940
 941                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 942                                 break;
 943                         deleted = 1;
 944                         continue;
 945                 } else if (ifp->flags & IFA_F_PERMANENT) {
 946                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 947                                               ifp->prefix_len)) {
 948                                 if (ifa->flags & IFA_F_PERMANENT) {
 949                                         onlink = 1;
 950                                         if (deleted)
 951                                                 break;
 952                                 } else {
 953                                         unsigned long lifetime;
 954
 955                                         if (!onlink)
 956                                                 onlink = -1;
 957
 958                                         spin_lock(&ifa->lock);
 959
 960                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 961                                         /*
 962                                          * Note: Because this address is
 963                                          * not permanent, lifetime <
 964                                          * LONG_MAX / HZ here.
 965                                          */
 966                                         if (time_before(expires,
 967                                                         ifa->tstamp + lifetime * HZ))
 968                                                 expires = ifa->tstamp + lifetime * HZ;
 969                                         spin_unlock(&ifa->lock);
 970                                 }
 971                         }
 972                 }
 973         }
 974         write_unlock_bh(&idev->lock);
 975
 976         addrconf_del_dad_timer(ifp);
 977
 978         ipv6_ifa_notify(RTM_DELADDR, ifp);
 979
 980         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
 981
 982         /*
 983          * Purge or update corresponding prefix
 984          *
 985          * 1) we don't purge prefix here if address was not permanent.
 986          *    prefix is managed by its own lifetime.
 987          * 2) if there're no addresses, delete prefix.
 988          * 3) if there're still other permanent address(es),
 989          *    corresponding prefix is still permanent.
 990          * 4) otherwise, update prefix lifetime to the
 991          *    longest valid lifetime among the corresponding
 992          *    addresses on the device.
 993          *    Note: subsequent RA will update lifetime.
 994          *
 995          * --yoshfuji
 996          */
 997         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 998                 struct rt6_info *rt;
 999
1000                 rt = addrconf_get_prefix_route(&ifp->addr,
1001                                                ifp->prefix_len,
1002                                                ifp->idev->dev,
1003                                                0, RTF_GATEWAY | RTF_DEFAULT);
1004
1005                 if (rt) {
1006                         if (onlink == 0) {
1007                                 ip6_del_rt(rt);
1008                                 rt = NULL;
1009                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1010                                 rt6_set_expires(rt, expires);
1011                         }
1012                 }
1013                 ip6_rt_put(rt);
1014         }
1015
1016         /* clean up prefsrc entries */
1017         rt6_remove_prefsrc(ifp);
1018 out:
1019         in6_ifa_put(ifp);
1020 }
1021
1022 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1023 {
1024         struct inet6_dev *idev = ifp->idev;
1025         struct in6_addr addr, *tmpaddr;
1026         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1027         unsigned long regen_advance;
1028         int tmp_plen;
1029         int ret = 0;
1030         u32 addr_flags;
1031         unsigned long now = jiffies;
1032
1033         write_lock_bh(&idev->lock);
1034         if (ift) {
1035                 spin_lock_bh(&ift->lock);
1036                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1037                 spin_unlock_bh(&ift->lock);
1038                 tmpaddr = &addr;
1039         } else {
1040                 tmpaddr = NULL;
1041         }
1042 retry:
1043         in6_dev_hold(idev);
1044         if (idev->cnf.use_tempaddr <= 0) {
1045                 write_unlock_bh(&idev->lock);
1046                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1047                 in6_dev_put(idev);
1048                 ret = -1;
1049                 goto out;
1050         }
1051         spin_lock_bh(&ifp->lock);
1052         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1053                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1054                 spin_unlock_bh(&ifp->lock);
1055                 write_unlock_bh(&idev->lock);
1056                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1057                         __func__);
1058                 in6_dev_put(idev);
1059                 ret = -1;
1060                 goto out;
1061         }
1062         in6_ifa_hold(ifp);
1063         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1064         __ipv6_try_regen_rndid(idev, tmpaddr);
1065         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1066         age = (now - ifp->tstamp) / HZ;
1067         tmp_valid_lft = min_t(__u32,
1068                               ifp->valid_lft,
1069                               idev->cnf.temp_valid_lft + age);
1070         tmp_prefered_lft = min_t(__u32,
1071                                  ifp->prefered_lft,
1072                                  idev->cnf.temp_prefered_lft + age -
1073                                  idev->cnf.max_desync_factor);
1074         tmp_plen = ifp->prefix_len;
1075         tmp_tstamp = ifp->tstamp;
1076         spin_unlock_bh(&ifp->lock);
1077
1078         regen_advance = idev->cnf.regen_max_retry *
1079                         idev->cnf.dad_transmits *
1080                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1081         write_unlock_bh(&idev->lock);
1082
1083         /* A temporary address is created only if this calculated Preferred
1084          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1085          * an implementation must not create a temporary address with a zero
1086          * Preferred Lifetime.
1087          */
1088         if (tmp_prefered_lft <= regen_advance) {
1089                 in6_ifa_put(ifp);
1090                 in6_dev_put(idev);
1091                 ret = -1;
1092                 goto out;
1093         }
1094
1095         addr_flags = IFA_F_TEMPORARY;
1096         /* set in addrconf_prefix_rcv() */
1097         if (ifp->flags & IFA_F_OPTIMISTIC)
1098                 addr_flags |= IFA_F_OPTIMISTIC;
1099
1100         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1101                             ipv6_addr_scope(&addr), addr_flags,
1102                             tmp_valid_lft, tmp_prefered_lft);
1103         if (IS_ERR(ift)) {
1104                 in6_ifa_put(ifp);
1105                 in6_dev_put(idev);
1106                 pr_info("%s: retry temporary address regeneration\n", __func__);
1107                 tmpaddr = &addr;
1108                 write_lock_bh(&idev->lock);
1109                 goto retry;
1110         }
1111
1112         spin_lock_bh(&ift->lock);
1113         ift->ifpub = ifp;
1114         ift->cstamp = now;
1115         ift->tstamp = tmp_tstamp;
1116         spin_unlock_bh(&ift->lock);
1117
1118         addrconf_dad_start(ift);
1119         in6_ifa_put(ift);
1120         in6_dev_put(idev);
1121 out:
1122         return ret;
1123 }
1124
1125 /*
1126  *      Choose an appropriate source address (RFC3484)
1127  */
1128 enum {
1129         IPV6_SADDR_RULE_INIT = 0,
1130         IPV6_SADDR_RULE_LOCAL,
1131         IPV6_SADDR_RULE_SCOPE,
1132         IPV6_SADDR_RULE_PREFERRED,
1133 #ifdef CONFIG_IPV6_MIP6
1134         IPV6_SADDR_RULE_HOA,
1135 #endif
1136         IPV6_SADDR_RULE_OIF,
1137         IPV6_SADDR_RULE_LABEL,
1138         IPV6_SADDR_RULE_PRIVACY,
1139         IPV6_SADDR_RULE_ORCHID,
1140         IPV6_SADDR_RULE_PREFIX,
1141         IPV6_SADDR_RULE_MAX
1142 };
1143
1144 struct ipv6_saddr_score {
1145         int                     rule;
1146         int                     addr_type;
1147         struct inet6_ifaddr     *ifa;
1148         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1149         int                     scopedist;
1150         int                     matchlen;
1151 };
1152
1153 struct ipv6_saddr_dst {
1154         const struct in6_addr *addr;
1155         int ifindex;
1156         int scope;
1157         int label;
1158         unsigned int prefs;
1159 };
1160
1161 static inline int ipv6_saddr_preferred(int type)
1162 {
1163         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1164                 return 1;
1165         return 0;
1166 }
1167
1168 static int ipv6_get_saddr_eval(struct net *net,
1169                                struct ipv6_saddr_score *score,
1170                                struct ipv6_saddr_dst *dst,
1171                                int i)
1172 {
1173         int ret;
1174
1175         if (i <= score->rule) {
1176                 switch (i) {
1177                 case IPV6_SADDR_RULE_SCOPE:
1178                         ret = score->scopedist;
1179                         break;
1180                 case IPV6_SADDR_RULE_PREFIX:
1181                         ret = score->matchlen;
1182                         break;
1183                 default:
1184                         ret = !!test_bit(i, score->scorebits);
1185                 }
1186                 goto out;
1187         }
1188
1189         switch (i) {
1190         case IPV6_SADDR_RULE_INIT:
1191                 /* Rule 0: remember if hiscore is not ready yet */
1192                 ret = !!score->ifa;
1193                 break;
1194         case IPV6_SADDR_RULE_LOCAL:
1195                 /* Rule 1: Prefer same address */
1196                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1197                 break;
1198         case IPV6_SADDR_RULE_SCOPE:
1199                 /* Rule 2: Prefer appropriate scope
1200                  *
1201                  *      ret
1202                  *       ^
1203                  *    -1 |  d 15
1204                  *    ---+--+-+---> scope
1205                  *       |
1206                  *       |             d is scope of the destination.
1207                  *  B-d  |  \
1208                  *       |   \      <- smaller scope is better if
1209                  *  B-15 |    \        if scope is enough for destinaion.
1210                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1211                  * d-C-1 | /
1212                  *       |/         <- greater is better
1213                  *   -C  /             if scope is not enough for destination.
1214                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1215                  *
1216                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1217                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1218                  * Assume B = 0 and we get C > 29.
1219                  */
1220                 ret = __ipv6_addr_src_scope(score->addr_type);
1221                 if (ret >= dst->scope)
1222                         ret = -ret;
1223                 else
1224                         ret -= 128;     /* 30 is enough */
1225                 score->scopedist = ret;
1226                 break;
1227         case IPV6_SADDR_RULE_PREFERRED:
1228                 /* Rule 3: Avoid deprecated and optimistic addresses */
1229                 ret = ipv6_saddr_preferred(score->addr_type) ||
1230                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1231                 break;
1232 #ifdef CONFIG_IPV6_MIP6
1233         case IPV6_SADDR_RULE_HOA:
1234             {
1235                 /* Rule 4: Prefer home address */
1236                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1237                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1238                 break;
1239             }
1240 #endif
1241         case IPV6_SADDR_RULE_OIF:
1242                 /* Rule 5: Prefer outgoing interface */
1243                 ret = (!dst->ifindex ||
1244                        dst->ifindex == score->ifa->idev->dev->ifindex);
1245                 break;
1246         case IPV6_SADDR_RULE_LABEL:
1247                 /* Rule 6: Prefer matching label */
1248                 ret = ipv6_addr_label(net,
1249                                       &score->ifa->addr, score->addr_type,
1250                                       score->ifa->idev->dev->ifindex) == dst->label;
1251                 break;
1252         case IPV6_SADDR_RULE_PRIVACY:
1253             {
1254                 /* Rule 7: Prefer public address
1255                  * Note: prefer temporary address if use_tempaddr >= 2
1256                  */
1257                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1258                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1259                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1260                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1261                 break;
1262             }
1263         case IPV6_SADDR_RULE_ORCHID:
1264                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1265                  *          non-ORCHID vs non-ORCHID
1266                  */
1267                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1268                         ipv6_addr_orchid(dst->addr));
1269                 break;
1270         case IPV6_SADDR_RULE_PREFIX:
1271                 /* Rule 8: Use longest matching prefix */
1272                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1273                 if (ret > score->ifa->prefix_len)
1274                         ret = score->ifa->prefix_len;
1275                 score->matchlen = ret;
1276                 break;
1277         default:
1278                 ret = 0;
1279         }
1280
1281         if (ret)
1282                 __set_bit(i, score->scorebits);
1283         score->rule = i;
1284 out:
1285         return ret;
1286 }
1287
1288 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1289                        const struct in6_addr *daddr, unsigned int prefs,
1290                        struct in6_addr *saddr)
1291 {
1292         struct ipv6_saddr_score scores[2],
1293                                 *score = &scores[0], *hiscore = &scores[1];
1294         struct ipv6_saddr_dst dst;
1295         struct net_device *dev;
1296         int dst_type;
1297
1298         dst_type = __ipv6_addr_type(daddr);
1299         dst.addr = daddr;
1300         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1301         dst.scope = __ipv6_addr_src_scope(dst_type);
1302         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1303         dst.prefs = prefs;
1304
1305         hiscore->rule = -1;
1306         hiscore->ifa = NULL;
1307
1308         rcu_read_lock();
1309
1310         for_each_netdev_rcu(net, dev) {
1311                 struct inet6_dev *idev;
1312
1313                 /* Candidate Source Address (section 4)
1314                  *  - multicast and link-local destination address,
1315                  *    the set of candidate source address MUST only
1316                  *    include addresses assigned to interfaces
1317                  *    belonging to the same link as the outgoing
1318                  *    interface.
1319                  * (- For site-local destination addresses, the
1320                  *    set of candidate source addresses MUST only
1321                  *    include addresses assigned to interfaces
1322                  *    belonging to the same site as the outgoing
1323                  *    interface.)
1324                  */
1325                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1326                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1327                     dst.ifindex && dev->ifindex != dst.ifindex)
1328                         continue;
1329
1330                 idev = __in6_dev_get(dev);
1331                 if (!idev)
1332                         continue;
1333
1334                 read_lock_bh(&idev->lock);
1335                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1336                         int i;
1337
1338                         /*
1339                          * - Tentative Address (RFC2462 section 5.4)
1340                          *  - A tentative address is not considered
1341                          *    "assigned to an interface" in the traditional
1342                          *    sense, unless it is also flagged as optimistic.
1343                          * - Candidate Source Address (section 4)
1344                          *  - In any case, anycast addresses, multicast
1345                          *    addresses, and the unspecified address MUST
1346                          *    NOT be included in a candidate set.
1347                          */
1348                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1349                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1350                                 continue;
1351
1352                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1353
1354                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1355                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1356                                 LIMIT_NETDEBUG(KERN_DEBUG
1357                                                "ADDRCONF: unspecified / multicast address "
1358                                                "assigned as unicast address on %s",
1359                                                dev->name);
1360                                 continue;
1361                         }
1362
1363                         score->rule = -1;
1364                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1365
1366                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1367                                 int minihiscore, miniscore;
1368
1369                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1370                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1371
1372                                 if (minihiscore > miniscore) {
1373                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1374                                             score->scopedist > 0) {
1375                                                 /*
1376                                                  * special case:
1377                                                  * each remaining entry
1378                                                  * has too small (not enough)
1379                                                  * scope, because ifa entries
1380                                                  * are sorted by their scope
1381                                                  * values.
1382                                                  */
1383                                                 goto try_nextdev;
1384                                         }
1385                                         break;
1386                                 } else if (minihiscore < miniscore) {
1387                                         if (hiscore->ifa)
1388                                                 in6_ifa_put(hiscore->ifa);
1389
1390                                         in6_ifa_hold(score->ifa);
1391
1392                                         swap(hiscore, score);
1393
1394                                         /* restore our iterator */
1395                                         score->ifa = hiscore->ifa;
1396
1397                                         break;
1398                                 }
1399                         }
1400                 }
1401 try_nextdev:
1402                 read_unlock_bh(&idev->lock);
1403         }
1404         rcu_read_unlock();
1405
1406         if (!hiscore->ifa)
1407                 return -EADDRNOTAVAIL;
1408
1409         *saddr = hiscore->ifa->addr;
1410         in6_ifa_put(hiscore->ifa);
1411         return 0;
1412 }
1413 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1414
1415 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1416                       u32 banned_flags)
1417 {
1418         struct inet6_ifaddr *ifp;
1419         int err = -EADDRNOTAVAIL;
1420
1421         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1422                 if (ifp->scope == IFA_LINK &&
1423                     !(ifp->flags & banned_flags)) {
1424                         *addr = ifp->addr;
1425                         err = 0;
1426                         break;
1427                 }
1428         }
1429         return err;
1430 }
1431
1432 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1433                     u32 banned_flags)
1434 {
1435         struct inet6_dev *idev;
1436         int err = -EADDRNOTAVAIL;
1437
1438         rcu_read_lock();
1439         idev = __in6_dev_get(dev);
1440         if (idev) {
1441                 read_lock_bh(&idev->lock);
1442                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1443                 read_unlock_bh(&idev->lock);
1444         }
1445         rcu_read_unlock();
1446         return err;
1447 }
1448
1449 static int ipv6_count_addresses(struct inet6_dev *idev)
1450 {
1451         int cnt = 0;
1452         struct inet6_ifaddr *ifp;
1453
1454         read_lock_bh(&idev->lock);
1455         list_for_each_entry(ifp, &idev->addr_list, if_list)
1456                 cnt++;
1457         read_unlock_bh(&idev->lock);
1458         return cnt;
1459 }
1460
1461 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1462                   const struct net_device *dev, int strict)
1463 {
1464         struct inet6_ifaddr *ifp;
1465         unsigned int hash = inet6_addr_hash(addr);
1466
1467         rcu_read_lock_bh();
1468         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1469                 if (!net_eq(dev_net(ifp->idev->dev), net))
1470                         continue;
1471                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1472                     !(ifp->flags&IFA_F_TENTATIVE) &&
1473                     (dev == NULL || ifp->idev->dev == dev ||
1474                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1475                         rcu_read_unlock_bh();
1476                         return 1;
1477                 }
1478         }
1479
1480         rcu_read_unlock_bh();
1481         return 0;
1482 }
1483 EXPORT_SYMBOL(ipv6_chk_addr);
1484
1485 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1486                                struct net_device *dev)
1487 {
1488         unsigned int hash = inet6_addr_hash(addr);
1489         struct inet6_ifaddr *ifp;
1490
1491         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1492                 if (!net_eq(dev_net(ifp->idev->dev), net))
1493                         continue;
1494                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1495                         if (dev == NULL || ifp->idev->dev == dev)
1496                                 return true;
1497                 }
1498         }
1499         return false;
1500 }
1501
1502 /* Compares an address/prefix_len with addresses on device @dev.
1503  * If one is found it returns true.
1504  */
1505 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1506         const unsigned int prefix_len, struct net_device *dev)
1507 {
1508         struct inet6_dev *idev;
1509         struct inet6_ifaddr *ifa;
1510         bool ret = false;
1511
1512         rcu_read_lock();
1513         idev = __in6_dev_get(dev);
1514         if (idev) {
1515                 read_lock_bh(&idev->lock);
1516                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1517                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1518                         if (ret)
1519                                 break;
1520                 }
1521                 read_unlock_bh(&idev->lock);
1522         }
1523         rcu_read_unlock();
1524
1525         return ret;
1526 }
1527 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1528
1529 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1530 {
1531         struct inet6_dev *idev;
1532         struct inet6_ifaddr *ifa;
1533         int     onlink;
1534
1535         onlink = 0;
1536         rcu_read_lock();
1537         idev = __in6_dev_get(dev);
1538         if (idev) {
1539                 read_lock_bh(&idev->lock);
1540                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1541                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1542                                                    ifa->prefix_len);
1543                         if (onlink)
1544                                 break;
1545                 }
1546                 read_unlock_bh(&idev->lock);
1547         }
1548         rcu_read_unlock();
1549         return onlink;
1550 }
1551 EXPORT_SYMBOL(ipv6_chk_prefix);
1552
1553 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1554                                      struct net_device *dev, int strict)
1555 {
1556         struct inet6_ifaddr *ifp, *result = NULL;
1557         unsigned int hash = inet6_addr_hash(addr);
1558
1559         rcu_read_lock_bh();
1560         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1561                 if (!net_eq(dev_net(ifp->idev->dev), net))
1562                         continue;
1563                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1564                         if (dev == NULL || ifp->idev->dev == dev ||
1565                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1566                                 result = ifp;
1567                                 in6_ifa_hold(ifp);
1568                                 break;
1569                         }
1570                 }
1571         }
1572         rcu_read_unlock_bh();
1573
1574         return result;
1575 }
1576
1577 /* Gets referenced address, destroys ifaddr */
1578
1579 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1580 {
1581         if (ifp->flags&IFA_F_PERMANENT) {
1582                 spin_lock_bh(&ifp->lock);
1583                 addrconf_del_dad_timer(ifp);
1584                 ifp->flags |= IFA_F_TENTATIVE;
1585                 if (dad_failed)
1586                         ifp->flags |= IFA_F_DADFAILED;
1587                 spin_unlock_bh(&ifp->lock);
1588                 if (dad_failed)
1589                         ipv6_ifa_notify(0, ifp);
1590                 in6_ifa_put(ifp);
1591         } else if (ifp->flags&IFA_F_TEMPORARY) {
1592                 struct inet6_ifaddr *ifpub;
1593                 spin_lock_bh(&ifp->lock);
1594                 ifpub = ifp->ifpub;
1595                 if (ifpub) {
1596                         in6_ifa_hold(ifpub);
1597                         spin_unlock_bh(&ifp->lock);
1598                         ipv6_create_tempaddr(ifpub, ifp);
1599                         in6_ifa_put(ifpub);
1600                 } else {
1601                         spin_unlock_bh(&ifp->lock);
1602                 }
1603                 ipv6_del_addr(ifp);
1604         } else
1605                 ipv6_del_addr(ifp);
1606 }
1607
1608 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1609 {
1610         int err = -ENOENT;
1611
1612         spin_lock(&ifp->state_lock);
1613         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1614                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1615                 err = 0;
1616         }
1617         spin_unlock(&ifp->state_lock);
1618
1619         return err;
1620 }
1621
1622 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1623 {
1624         struct inet6_dev *idev = ifp->idev;
1625
1626         if (addrconf_dad_end(ifp)) {
1627                 in6_ifa_put(ifp);
1628                 return;
1629         }
1630
1631         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1632                              ifp->idev->dev->name, &ifp->addr);
1633
1634         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1635                 struct in6_addr addr;
1636
1637                 addr.s6_addr32[0] = htonl(0xfe800000);
1638                 addr.s6_addr32[1] = 0;
1639
1640                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1641                     ipv6_addr_equal(&ifp->addr, &addr)) {
1642                         /* DAD failed for link-local based on MAC address */
1643                         idev->cnf.disable_ipv6 = 1;
1644
1645                         pr_info("%s: IPv6 being disabled!\n",
1646                                 ifp->idev->dev->name);
1647                 }
1648         }
1649
1650         addrconf_dad_stop(ifp, 1);
1651 }
1652
1653 /* Join to solicited addr multicast group. */
1654
1655 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1656 {
1657         struct in6_addr maddr;
1658
1659         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1660                 return;
1661
1662         addrconf_addr_solict_mult(addr, &maddr);
1663         ipv6_dev_mc_inc(dev, &maddr);
1664 }
1665
1666 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1667 {
1668         struct in6_addr maddr;
1669
1670         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1671                 return;
1672
1673         addrconf_addr_solict_mult(addr, &maddr);
1674         __ipv6_dev_mc_dec(idev, &maddr);
1675 }
1676
1677 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1678 {
1679         struct in6_addr addr;
1680         if (ifp->prefix_len >= 127) /* RFC 6164 */
1681                 return;
1682         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1683         if (ipv6_addr_any(&addr))
1684                 return;
1685         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1686 }
1687
1688 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1689 {
1690         struct in6_addr addr;
1691         if (ifp->prefix_len >= 127) /* RFC 6164 */
1692                 return;
1693         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1694         if (ipv6_addr_any(&addr))
1695                 return;
1696         __ipv6_dev_ac_dec(ifp->idev, &addr);
1697 }
1698
1699 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1700 {
1701         if (dev->addr_len != ETH_ALEN)
1702                 return -1;
1703         memcpy(eui, dev->dev_addr, 3);
1704         memcpy(eui + 5, dev->dev_addr + 3, 3);
1705
1706         /*
1707          * The zSeries OSA network cards can be shared among various
1708          * OS instances, but the OSA cards have only one MAC address.
1709          * This leads to duplicate address conflicts in conjunction
1710          * with IPv6 if more than one instance uses the same card.
1711          *
1712          * The driver for these cards can deliver a unique 16-bit
1713          * identifier for each instance sharing the same card.  It is
1714          * placed instead of 0xFFFE in the interface identifier.  The
1715          * "u" bit of the interface identifier is not inverted in this
1716          * case.  Hence the resulting interface identifier has local
1717          * scope according to RFC2373.
1718          */
1719         if (dev->dev_id) {
1720                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1721                 eui[4] = dev->dev_id & 0xFF;
1722         } else {
1723                 eui[3] = 0xFF;
1724                 eui[4] = 0xFE;
1725                 eui[0] ^= 2;
1726         }
1727         return 0;
1728 }
1729
1730 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1731 {
1732         if (dev->addr_len != IEEE802154_ADDR_LEN)
1733                 return -1;
1734         memcpy(eui, dev->dev_addr, 8);
1735         eui[0] ^= 2;
1736         return 0;
1737 }
1738
1739 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1740 {
1741         union fwnet_hwaddr *ha;
1742
1743         if (dev->addr_len != FWNET_ALEN)
1744                 return -1;
1745
1746         ha = (union fwnet_hwaddr *)dev->dev_addr;
1747
1748         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1749         eui[0] ^= 2;
1750         return 0;
1751 }
1752
1753 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1754 {
1755         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1756         if (dev->addr_len != ARCNET_ALEN)
1757                 return -1;
1758         memset(eui, 0, 7);
1759         eui[7] = *(u8 *)dev->dev_addr;
1760         return 0;
1761 }
1762
1763 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1764 {
1765         if (dev->addr_len != INFINIBAND_ALEN)
1766                 return -1;
1767         memcpy(eui, dev->dev_addr + 12, 8);
1768         eui[0] |= 2;
1769         return 0;
1770 }
1771
1772 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1773 {
1774         if (addr == 0)
1775                 return -1;
1776         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1777                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1778                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1779                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1780                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1781                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1782         eui[1] = 0;
1783         eui[2] = 0x5E;
1784         eui[3] = 0xFE;
1785         memcpy(eui + 4, &addr, 4);
1786         return 0;
1787 }
1788
1789 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1790 {
1791         if (dev->priv_flags & IFF_ISATAP)
1792                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1793         return -1;
1794 }
1795
1796 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1797 {
1798         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1799 }
1800
1801 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1802 {
1803         memcpy(eui, dev->perm_addr, 3);
1804         memcpy(eui + 5, dev->perm_addr + 3, 3);
1805         eui[3] = 0xFF;
1806         eui[4] = 0xFE;
1807         eui[0] ^= 2;
1808         return 0;
1809 }
1810
1811 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1812 {
1813         switch (dev->type) {
1814         case ARPHRD_ETHER:
1815         case ARPHRD_FDDI:
1816                 return addrconf_ifid_eui48(eui, dev);
1817         case ARPHRD_ARCNET:
1818                 return addrconf_ifid_arcnet(eui, dev);
1819         case ARPHRD_INFINIBAND:
1820                 return addrconf_ifid_infiniband(eui, dev);
1821         case ARPHRD_SIT:
1822                 return addrconf_ifid_sit(eui, dev);
1823         case ARPHRD_IPGRE:
1824                 return addrconf_ifid_gre(eui, dev);
1825         case ARPHRD_6LOWPAN:
1826         case ARPHRD_IEEE802154:
1827                 return addrconf_ifid_eui64(eui, dev);
1828         case ARPHRD_IEEE1394:
1829                 return addrconf_ifid_ieee1394(eui, dev);
1830         case ARPHRD_TUNNEL6:
1831                 return addrconf_ifid_ip6tnl(eui, dev);
1832         }
1833         return -1;
1834 }
1835
1836 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1837 {
1838         int err = -1;
1839         struct inet6_ifaddr *ifp;
1840
1841         read_lock_bh(&idev->lock);
1842         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1843                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1844                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1845                         err = 0;
1846                         break;
1847                 }
1848         }
1849         read_unlock_bh(&idev->lock);
1850         return err;
1851 }
1852
1853 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1854 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1855 {
1856 regen:
1857         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1858         idev->rndid[0] &= ~0x02;
1859
1860         /*
1861          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1862          * check if generated address is not inappropriate
1863          *
1864          *  - Reserved subnet anycast (RFC 2526)
1865          *      11111101 11....11 1xxxxxxx
1866          *  - ISATAP (RFC4214) 6.1
1867          *      00-00-5E-FE-xx-xx-xx-xx
1868          *  - value 0
1869          *  - XXX: already assigned to an address on the device
1870          */
1871         if (idev->rndid[0] == 0xfd &&
1872             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1873             (idev->rndid[7]&0x80))
1874                 goto regen;
1875         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1876                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1877                         goto regen;
1878                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1879                         goto regen;
1880         }
1881 }
1882
1883 static void ipv6_regen_rndid(unsigned long data)
1884 {
1885         struct inet6_dev *idev = (struct inet6_dev *) data;
1886         unsigned long expires;
1887
1888         rcu_read_lock_bh();
1889         write_lock_bh(&idev->lock);
1890
1891         if (idev->dead)
1892                 goto out;
1893
1894         __ipv6_regen_rndid(idev);
1895
1896         expires = jiffies +
1897                 idev->cnf.temp_prefered_lft * HZ -
1898                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1899                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1900                 idev->cnf.max_desync_factor * HZ;
1901         if (time_before(expires, jiffies)) {
1902                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1903                         __func__, idev->dev->name);
1904                 goto out;
1905         }
1906
1907         if (!mod_timer(&idev->regen_timer, expires))
1908                 in6_dev_hold(idev);
1909
1910 out:
1911         write_unlock_bh(&idev->lock);
1912         rcu_read_unlock_bh();
1913         in6_dev_put(idev);
1914 }
1915
1916 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1917 {
1918         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1919                 __ipv6_regen_rndid(idev);
1920 }
1921
1922 /*
1923  *      Add prefix route.
1924  */
1925
1926 static void
1927 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1928                       unsigned long expires, u32 flags)
1929 {
1930         struct fib6_config cfg = {
1931                 .fc_table = RT6_TABLE_PREFIX,
1932                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1933                 .fc_ifindex = dev->ifindex,
1934                 .fc_expires = expires,
1935                 .fc_dst_len = plen,
1936                 .fc_flags = RTF_UP | flags,
1937                 .fc_nlinfo.nl_net = dev_net(dev),
1938                 .fc_protocol = RTPROT_KERNEL,
1939         };
1940
1941         cfg.fc_dst = *pfx;
1942
1943         /* Prevent useless cloning on PtP SIT.
1944            This thing is done here expecting that the whole
1945            class of non-broadcast devices need not cloning.
1946          */
1947 #if IS_ENABLED(CONFIG_IPV6_SIT)
1948         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1949                 cfg.fc_flags |= RTF_NONEXTHOP;
1950 #endif
1951
1952         ip6_route_add(&cfg);
1953 }
1954
1955
1956 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1957                                                   int plen,
1958                                                   const struct net_device *dev,
1959                                                   u32 flags, u32 noflags)
1960 {
1961         struct fib6_node *fn;
1962         struct rt6_info *rt = NULL;
1963         struct fib6_table *table;
1964
1965         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1966         if (table == NULL)
1967                 return NULL;
1968
1969         read_lock_bh(&table->tb6_lock);
1970         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1971         if (!fn)
1972                 goto out;
1973         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1974                 if (rt->dst.dev->ifindex != dev->ifindex)
1975                         continue;
1976                 if ((rt->rt6i_flags & flags) != flags)
1977                         continue;
1978                 if ((rt->rt6i_flags & noflags) != 0)
1979                         continue;
1980                 dst_hold(&rt->dst);
1981                 break;
1982         }
1983 out:
1984         read_unlock_bh(&table->tb6_lock);
1985         return rt;
1986 }
1987
1988
1989 /* Create "default" multicast route to the interface */
1990
1991 static void addrconf_add_mroute(struct net_device *dev)
1992 {
1993         struct fib6_config cfg = {
1994                 .fc_table = RT6_TABLE_LOCAL,
1995                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1996                 .fc_ifindex = dev->ifindex,
1997                 .fc_dst_len = 8,
1998                 .fc_flags = RTF_UP,
1999                 .fc_nlinfo.nl_net = dev_net(dev),
2000         };
2001
2002         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2003
2004         ip6_route_add(&cfg);
2005 }
2006
2007 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2008 {
2009         struct inet6_dev *idev;
2010
2011         ASSERT_RTNL();
2012
2013         idev = ipv6_find_idev(dev);
2014         if (!idev)
2015                 return ERR_PTR(-ENOBUFS);
2016
2017         if (idev->cnf.disable_ipv6)
2018                 return ERR_PTR(-EACCES);
2019
2020         /* Add default multicast route */
2021         if (!(dev->flags & IFF_LOOPBACK))
2022                 addrconf_add_mroute(dev);
2023
2024         return idev;
2025 }
2026
2027 static void manage_tempaddrs(struct inet6_dev *idev,
2028                              struct inet6_ifaddr *ifp,
2029                              __u32 valid_lft, __u32 prefered_lft,
2030                              bool create, unsigned long now)
2031 {
2032         u32 flags;
2033         struct inet6_ifaddr *ift;
2034
2035         read_lock_bh(&idev->lock);
2036         /* update all temporary addresses in the list */
2037         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2038                 int age, max_valid, max_prefered;
2039
2040                 if (ifp != ift->ifpub)
2041                         continue;
2042
2043                 /* RFC 4941 section 3.3:
2044                  * If a received option will extend the lifetime of a public
2045                  * address, the lifetimes of temporary addresses should
2046                  * be extended, subject to the overall constraint that no
2047                  * temporary addresses should ever remain "valid" or "preferred"
2048                  * for a time longer than (TEMP_VALID_LIFETIME) or
2049                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2050                  */
2051                 age = (now - ift->cstamp) / HZ;
2052                 max_valid = idev->cnf.temp_valid_lft - age;
2053                 if (max_valid < 0)
2054                         max_valid = 0;
2055
2056                 max_prefered = idev->cnf.temp_prefered_lft -
2057                                idev->cnf.max_desync_factor - age;
2058                 if (max_prefered < 0)
2059                         max_prefered = 0;
2060
2061                 if (valid_lft > max_valid)
2062                         valid_lft = max_valid;
2063
2064                 if (prefered_lft > max_prefered)
2065                         prefered_lft = max_prefered;
2066
2067                 spin_lock(&ift->lock);
2068                 flags = ift->flags;
2069                 ift->valid_lft = valid_lft;
2070                 ift->prefered_lft = prefered_lft;
2071                 ift->tstamp = now;
2072                 if (prefered_lft > 0)
2073                         ift->flags &= ~IFA_F_DEPRECATED;
2074
2075                 spin_unlock(&ift->lock);
2076                 if (!(flags&IFA_F_TENTATIVE))
2077                         ipv6_ifa_notify(0, ift);
2078         }
2079
2080         if ((create || list_empty(&idev->tempaddr_list)) &&
2081             idev->cnf.use_tempaddr > 0) {
2082                 /* When a new public address is created as described
2083                  * in [ADDRCONF], also create a new temporary address.
2084                  * Also create a temporary address if it's enabled but
2085                  * no temporary address currently exists.
2086                  */
2087                 read_unlock_bh(&idev->lock);
2088                 ipv6_create_tempaddr(ifp, NULL);
2089         } else {
2090                 read_unlock_bh(&idev->lock);
2091         }
2092 }
2093
2094 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2095 {
2096         struct prefix_info *pinfo;
2097         __u32 valid_lft;
2098         __u32 prefered_lft;
2099         int addr_type;
2100         struct inet6_dev *in6_dev;
2101         struct net *net = dev_net(dev);
2102
2103         pinfo = (struct prefix_info *) opt;
2104
2105         if (len < sizeof(struct prefix_info)) {
2106                 ADBG("addrconf: prefix option too short\n");
2107                 return;
2108         }
2109
2110         /*
2111          *      Validation checks ([ADDRCONF], page 19)
2112          */
2113
2114         addr_type = ipv6_addr_type(&pinfo->prefix);
2115
2116         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2117                 return;
2118
2119         valid_lft = ntohl(pinfo->valid);
2120         prefered_lft = ntohl(pinfo->prefered);
2121
2122         if (prefered_lft > valid_lft) {
2123                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2124                 return;
2125         }
2126
2127         in6_dev = in6_dev_get(dev);
2128
2129         if (in6_dev == NULL) {
2130                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2131                                     dev->name);
2132                 return;
2133         }
2134
2135         /*
2136          *      Two things going on here:
2137          *      1) Add routes for on-link prefixes
2138          *      2) Configure prefixes with the auto flag set
2139          */
2140
2141         if (pinfo->onlink) {
2142                 struct rt6_info *rt;
2143                 unsigned long rt_expires;
2144
2145                 /* Avoid arithmetic overflow. Really, we could
2146                  * save rt_expires in seconds, likely valid_lft,
2147                  * but it would require division in fib gc, that it
2148                  * not good.
2149                  */
2150                 if (HZ > USER_HZ)
2151                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2152                 else
2153                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2154
2155                 if (addrconf_finite_timeout(rt_expires))
2156                         rt_expires *= HZ;
2157
2158                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2159                                                pinfo->prefix_len,
2160                                                dev,
2161                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2162                                                RTF_GATEWAY | RTF_DEFAULT);
2163
2164                 if (rt) {
2165                         /* Autoconf prefix route */
2166                         if (valid_lft == 0) {
2167                                 ip6_del_rt(rt);
2168                                 rt = NULL;
2169                         } else if (addrconf_finite_timeout(rt_expires)) {
2170                                 /* not infinity */
2171                                 rt6_set_expires(rt, jiffies + rt_expires);
2172                         } else {
2173                                 rt6_clean_expires(rt);
2174                         }
2175                 } else if (valid_lft) {
2176                         clock_t expires = 0;
2177                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2178                         if (addrconf_finite_timeout(rt_expires)) {
2179                                 /* not infinity */
2180                                 flags |= RTF_EXPIRES;
2181                                 expires = jiffies_to_clock_t(rt_expires);
2182                         }
2183                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2184                                               dev, expires, flags);
2185                 }
2186                 ip6_rt_put(rt);
2187         }
2188
2189         /* Try to figure out our local address for this prefix */
2190
2191         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2192                 struct inet6_ifaddr *ifp;
2193                 struct in6_addr addr;
2194                 int create = 0, update_lft = 0;
2195                 bool tokenized = false;
2196
2197                 if (pinfo->prefix_len == 64) {
2198                         memcpy(&addr, &pinfo->prefix, 8);
2199
2200                         if (!ipv6_addr_any(&in6_dev->token)) {
2201                                 read_lock_bh(&in6_dev->lock);
2202                                 memcpy(addr.s6_addr + 8,
2203                                        in6_dev->token.s6_addr + 8, 8);
2204                                 read_unlock_bh(&in6_dev->lock);
2205                                 tokenized = true;
2206                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2207                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2208                                 in6_dev_put(in6_dev);
2209                                 return;
2210                         }
2211                         goto ok;
2212                 }
2213                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2214                                     pinfo->prefix_len);
2215                 in6_dev_put(in6_dev);
2216                 return;
2217
2218 ok:
2219
2220                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2221
2222                 if (ifp == NULL && valid_lft) {
2223                         int max_addresses = in6_dev->cnf.max_addresses;
2224                         u32 addr_flags = 0;
2225
2226 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2227                         if (in6_dev->cnf.optimistic_dad &&
2228                             !net->ipv6.devconf_all->forwarding && sllao)
2229                                 addr_flags = IFA_F_OPTIMISTIC;
2230 #endif
2231
2232                         /* Do not allow to create too much of autoconfigured
2233                          * addresses; this would be too easy way to crash kernel.
2234                          */
2235                         if (!max_addresses ||
2236                             ipv6_count_addresses(in6_dev) < max_addresses)
2237                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2238                                                     pinfo->prefix_len,
2239                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2240                                                     addr_flags, valid_lft,
2241                                                     prefered_lft);
2242
2243                         if (IS_ERR_OR_NULL(ifp)) {
2244                                 in6_dev_put(in6_dev);
2245                                 return;
2246                         }
2247
2248                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2249                         update_lft = 0;
2250                         create = 1;
2251                         ifp->cstamp = jiffies;
2252                         ifp->tokenized = tokenized;
2253                         addrconf_dad_start(ifp);
2254                 }
2255
2256                 if (ifp) {
2257                         u32 flags;
2258                         unsigned long now;
2259                         u32 stored_lft;
2260
2261                         /* update lifetime (RFC2462 5.5.3 e) */
2262                         spin_lock(&ifp->lock);
2263                         now = jiffies;
2264                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2265                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2266                         else
2267                                 stored_lft = 0;
2268                         if (!update_lft && !create && stored_lft) {
2269                                 const u32 minimum_lft = min(
2270                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2271                                 valid_lft = max(valid_lft, minimum_lft);
2272
2273                                 /* RFC4862 Section 5.5.3e:
2274                                  * "Note that the preferred lifetime of the
2275                                  *  corresponding address is always reset to
2276                                  *  the Preferred Lifetime in the received
2277                                  *  Prefix Information option, regardless of
2278                                  *  whether the valid lifetime is also reset or
2279                                  *  ignored."
2280                                  *
2281                                  * So we should always update prefered_lft here.
2282                                  */
2283                                 update_lft = 1;
2284                         }
2285
2286                         if (update_lft) {
2287                                 ifp->valid_lft = valid_lft;
2288                                 ifp->prefered_lft = prefered_lft;
2289                                 ifp->tstamp = now;
2290                                 flags = ifp->flags;
2291                                 ifp->flags &= ~IFA_F_DEPRECATED;
2292                                 spin_unlock(&ifp->lock);
2293
2294                                 if (!(flags&IFA_F_TENTATIVE))
2295                                         ipv6_ifa_notify(0, ifp);
2296                         } else
2297                                 spin_unlock(&ifp->lock);
2298
2299                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2300                                          create, now);
2301
2302                         in6_ifa_put(ifp);
2303                         addrconf_verify(0);
2304                 }
2305         }
2306         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2307         in6_dev_put(in6_dev);
2308 }
2309
2310 /*
2311  *      Set destination address.
2312  *      Special case for SIT interfaces where we create a new "virtual"
2313  *      device.
2314  */
2315 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2316 {
2317         struct in6_ifreq ireq;
2318         struct net_device *dev;
2319         int err = -EINVAL;
2320
2321         rtnl_lock();
2322
2323         err = -EFAULT;
2324         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2325                 goto err_exit;
2326
2327         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2328
2329         err = -ENODEV;
2330         if (dev == NULL)
2331                 goto err_exit;
2332
2333 #if IS_ENABLED(CONFIG_IPV6_SIT)
2334         if (dev->type == ARPHRD_SIT) {
2335                 const struct net_device_ops *ops = dev->netdev_ops;
2336                 struct ifreq ifr;
2337                 struct ip_tunnel_parm p;
2338
2339                 err = -EADDRNOTAVAIL;
2340                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2341                         goto err_exit;
2342
2343                 memset(&p, 0, sizeof(p));
2344                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2345                 p.iph.saddr = 0;
2346                 p.iph.version = 4;
2347                 p.iph.ihl = 5;
2348                 p.iph.protocol = IPPROTO_IPV6;
2349                 p.iph.ttl = 64;
2350                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2351
2352                 if (ops->ndo_do_ioctl) {
2353                         mm_segment_t oldfs = get_fs();
2354
2355                         set_fs(KERNEL_DS);
2356                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2357                         set_fs(oldfs);
2358                 } else
2359                         err = -EOPNOTSUPP;
2360
2361                 if (err == 0) {
2362                         err = -ENOBUFS;
2363                         dev = __dev_get_by_name(net, p.name);
2364                         if (!dev)
2365                                 goto err_exit;
2366                         err = dev_open(dev);
2367                 }
2368         }
2369 #endif
2370
2371 err_exit:
2372         rtnl_unlock();
2373         return err;
2374 }
2375
2376 /*
2377  *      Manual configuration of address on an interface
2378  */
2379 static int inet6_addr_add(struct net *net, int ifindex,
2380                           const struct in6_addr *pfx,
2381                           const struct in6_addr *peer_pfx,
2382                           unsigned int plen, __u32 ifa_flags,
2383                           __u32 prefered_lft, __u32 valid_lft)
2384 {
2385         struct inet6_ifaddr *ifp;
2386         struct inet6_dev *idev;
2387         struct net_device *dev;
2388         int scope;
2389         u32 flags;
2390         clock_t expires;
2391         unsigned long timeout;
2392
2393         ASSERT_RTNL();
2394
2395         if (plen > 128)
2396                 return -EINVAL;
2397
2398         /* check the lifetime */
2399         if (!valid_lft || prefered_lft > valid_lft)
2400                 return -EINVAL;
2401
2402         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2403                 return -EINVAL;
2404
2405         dev = __dev_get_by_index(net, ifindex);
2406         if (!dev)
2407                 return -ENODEV;
2408
2409         idev = addrconf_add_dev(dev);
2410         if (IS_ERR(idev))
2411                 return PTR_ERR(idev);
2412
2413         scope = ipv6_addr_scope(pfx);
2414
2415         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2416         if (addrconf_finite_timeout(timeout)) {
2417                 expires = jiffies_to_clock_t(timeout * HZ);
2418                 valid_lft = timeout;
2419                 flags = RTF_EXPIRES;
2420         } else {
2421                 expires = 0;
2422                 flags = 0;
2423                 ifa_flags |= IFA_F_PERMANENT;
2424         }
2425
2426         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2427         if (addrconf_finite_timeout(timeout)) {
2428                 if (timeout == 0)
2429                         ifa_flags |= IFA_F_DEPRECATED;
2430                 prefered_lft = timeout;
2431         }
2432
2433         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2434                             valid_lft, prefered_lft);
2435
2436         if (!IS_ERR(ifp)) {
2437                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2438                                       expires, flags);
2439                 /*
2440                  * Note that section 3.1 of RFC 4429 indicates
2441                  * that the Optimistic flag should not be set for
2442                  * manually configured addresses
2443                  */
2444                 addrconf_dad_start(ifp);
2445                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2446                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2447                                          true, jiffies);
2448                 in6_ifa_put(ifp);
2449                 addrconf_verify(0);
2450                 return 0;
2451         }
2452
2453         return PTR_ERR(ifp);
2454 }
2455
2456 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2457                           unsigned int plen)
2458 {
2459         struct inet6_ifaddr *ifp;
2460         struct inet6_dev *idev;
2461         struct net_device *dev;
2462
2463         if (plen > 128)
2464                 return -EINVAL;
2465
2466         dev = __dev_get_by_index(net, ifindex);
2467         if (!dev)
2468                 return -ENODEV;
2469
2470         if ((idev = __in6_dev_get(dev)) == NULL)
2471                 return -ENXIO;
2472
2473         read_lock_bh(&idev->lock);
2474         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2475                 if (ifp->prefix_len == plen &&
2476                     ipv6_addr_equal(pfx, &ifp->addr)) {
2477                         in6_ifa_hold(ifp);
2478                         read_unlock_bh(&idev->lock);
2479
2480                         ipv6_del_addr(ifp);
2481                         return 0;
2482                 }
2483         }
2484         read_unlock_bh(&idev->lock);
2485         return -EADDRNOTAVAIL;
2486 }
2487
2488
2489 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2490 {
2491         struct in6_ifreq ireq;
2492         int err;
2493
2494         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2495                 return -EPERM;
2496
2497         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2498                 return -EFAULT;
2499
2500         rtnl_lock();
2501         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2502                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2503                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2504         rtnl_unlock();
2505         return err;
2506 }
2507
2508 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2509 {
2510         struct in6_ifreq ireq;
2511         int err;
2512
2513         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2514                 return -EPERM;
2515
2516         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2517                 return -EFAULT;
2518
2519         rtnl_lock();
2520         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2521                              ireq.ifr6_prefixlen);
2522         rtnl_unlock();
2523         return err;
2524 }
2525
2526 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2527                      int plen, int scope)
2528 {
2529         struct inet6_ifaddr *ifp;
2530
2531         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2532                             scope, IFA_F_PERMANENT,
2533                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2534         if (!IS_ERR(ifp)) {
2535                 spin_lock_bh(&ifp->lock);
2536                 ifp->flags &= ~IFA_F_TENTATIVE;
2537                 spin_unlock_bh(&ifp->lock);
2538                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2539                 in6_ifa_put(ifp);
2540         }
2541 }
2542
2543 #if IS_ENABLED(CONFIG_IPV6_SIT)
2544 static void sit_add_v4_addrs(struct inet6_dev *idev)
2545 {
2546         struct in6_addr addr;
2547         struct net_device *dev;
2548         struct net *net = dev_net(idev->dev);
2549         int scope, plen;
2550         u32 pflags = 0;
2551
2552         ASSERT_RTNL();
2553
2554         memset(&addr, 0, sizeof(struct in6_addr));
2555         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2556
2557         if (idev->dev->flags&IFF_POINTOPOINT) {
2558                 addr.s6_addr32[0] = htonl(0xfe800000);
2559                 scope = IFA_LINK;
2560                 plen = 64;
2561         } else {
2562                 scope = IPV6_ADDR_COMPATv4;
2563                 plen = 96;
2564                 pflags |= RTF_NONEXTHOP;
2565         }
2566
2567         if (addr.s6_addr32[3]) {
2568                 add_addr(idev, &addr, plen, scope);
2569                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2570                 return;
2571         }
2572
2573         for_each_netdev(net, dev) {
2574                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2575                 if (in_dev && (dev->flags & IFF_UP)) {
2576                         struct in_ifaddr *ifa;
2577
2578                         int flag = scope;
2579
2580                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2581
2582                                 addr.s6_addr32[3] = ifa->ifa_local;
2583
2584                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2585                                         continue;
2586                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2587                                         if (idev->dev->flags&IFF_POINTOPOINT)
2588                                                 continue;
2589                                         flag |= IFA_HOST;
2590                                 }
2591
2592                                 add_addr(idev, &addr, plen, flag);
2593                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2594                                                       pflags);
2595                         }
2596                 }
2597         }
2598 }
2599 #endif
2600
2601 static void init_loopback(struct net_device *dev)
2602 {
2603         struct inet6_dev  *idev;
2604         struct net_device *sp_dev;
2605         struct inet6_ifaddr *sp_ifa;
2606         struct rt6_info *sp_rt;
2607
2608         /* ::1 */
2609
2610         ASSERT_RTNL();
2611
2612         if ((idev = ipv6_find_idev(dev)) == NULL) {
2613                 pr_debug("%s: add_dev failed\n", __func__);
2614                 return;
2615         }
2616
2617         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2618
2619         /* Add routes to other interface's IPv6 addresses */
2620         for_each_netdev(dev_net(dev), sp_dev) {
2621                 if (!strcmp(sp_dev->name, dev->name))
2622                         continue;
2623
2624                 idev = __in6_dev_get(sp_dev);
2625                 if (!idev)
2626                         continue;
2627
2628                 read_lock_bh(&idev->lock);
2629                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2630
2631                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2632                                 continue;
2633
2634                         if (sp_ifa->rt)
2635                                 continue;
2636
2637                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2638
2639                         /* Failure cases are ignored */
2640                         if (!IS_ERR(sp_rt)) {
2641                                 sp_ifa->rt = sp_rt;
2642                                 ip6_ins_rt(sp_rt);
2643                         }
2644                 }
2645                 read_unlock_bh(&idev->lock);
2646         }
2647 }
2648
2649 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2650 {
2651         struct inet6_ifaddr *ifp;
2652         u32 addr_flags = IFA_F_PERMANENT;
2653
2654 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2655         if (idev->cnf.optimistic_dad &&
2656             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2657                 addr_flags |= IFA_F_OPTIMISTIC;
2658 #endif
2659
2660
2661         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2662                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2663         if (!IS_ERR(ifp)) {
2664                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2665                 addrconf_dad_start(ifp);
2666                 in6_ifa_put(ifp);
2667         }
2668 }
2669
2670 static void addrconf_dev_config(struct net_device *dev)
2671 {
2672         struct in6_addr addr;
2673         struct inet6_dev *idev;
2674
2675         ASSERT_RTNL();
2676
2677         if ((dev->type != ARPHRD_ETHER) &&
2678             (dev->type != ARPHRD_FDDI) &&
2679             (dev->type != ARPHRD_ARCNET) &&
2680             (dev->type != ARPHRD_INFINIBAND) &&
2681             (dev->type != ARPHRD_IEEE802154) &&
2682             (dev->type != ARPHRD_IEEE1394) &&
2683             (dev->type != ARPHRD_TUNNEL6) &&
2684             (dev->type != ARPHRD_6LOWPAN)) {
2685                 /* Alas, we support only Ethernet autoconfiguration. */
2686                 return;
2687         }
2688
2689         idev = addrconf_add_dev(dev);
2690         if (IS_ERR(idev))
2691                 return;
2692
2693         memset(&addr, 0, sizeof(struct in6_addr));
2694         addr.s6_addr32[0] = htonl(0xFE800000);
2695
2696         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2697                 addrconf_add_linklocal(idev, &addr);
2698 }
2699
2700 #if IS_ENABLED(CONFIG_IPV6_SIT)
2701 static void addrconf_sit_config(struct net_device *dev)
2702 {
2703         struct inet6_dev *idev;
2704
2705         ASSERT_RTNL();
2706
2707         /*
2708          * Configure the tunnel with one of our IPv4
2709          * addresses... we should configure all of
2710          * our v4 addrs in the tunnel
2711          */
2712
2713         if ((idev = ipv6_find_idev(dev)) == NULL) {
2714                 pr_debug("%s: add_dev failed\n", __func__);
2715                 return;
2716         }
2717
2718         if (dev->priv_flags & IFF_ISATAP) {
2719                 struct in6_addr addr;
2720
2721                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2722                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2723                         addrconf_add_linklocal(idev, &addr);
2724                 return;
2725         }
2726
2727         sit_add_v4_addrs(idev);
2728
2729         if (dev->flags&IFF_POINTOPOINT)
2730                 addrconf_add_mroute(dev);
2731 }
2732 #endif
2733
2734 #if IS_ENABLED(CONFIG_NET_IPGRE)
2735 static void addrconf_gre_config(struct net_device *dev)
2736 {
2737         struct inet6_dev *idev;
2738         struct in6_addr addr;
2739
2740         ASSERT_RTNL();
2741
2742         if ((idev = ipv6_find_idev(dev)) == NULL) {
2743                 pr_debug("%s: add_dev failed\n", __func__);
2744                 return;
2745         }
2746
2747         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2748         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2749                 addrconf_add_linklocal(idev, &addr);
2750 }
2751 #endif
2752
2753 static inline int
2754 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2755 {
2756         struct in6_addr lladdr;
2757
2758         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2759                 addrconf_add_linklocal(idev, &lladdr);
2760                 return 0;
2761         }
2762         return -1;
2763 }
2764
2765 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2766                            void *ptr)
2767 {
2768         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2769         struct inet6_dev *idev = __in6_dev_get(dev);
2770         int run_pending = 0;
2771         int err;
2772
2773         switch (event) {
2774         case NETDEV_REGISTER:
2775                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2776                         idev = ipv6_add_dev(dev);
2777                         if (!idev)
2778                                 return notifier_from_errno(-ENOMEM);
2779                 }
2780                 break;
2781
2782         case NETDEV_UP:
2783         case NETDEV_CHANGE:
2784                 if (dev->flags & IFF_SLAVE)
2785                         break;
2786
2787                 if (event == NETDEV_UP) {
2788                         if (!addrconf_qdisc_ok(dev)) {
2789                                 /* device is not ready yet. */
2790                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2791                                         dev->name);
2792                                 break;
2793                         }
2794
2795                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2796                                 idev = ipv6_add_dev(dev);
2797
2798                         if (idev) {
2799                                 idev->if_flags |= IF_READY;
2800                                 run_pending = 1;
2801                         }
2802                 } else {
2803                         if (!addrconf_qdisc_ok(dev)) {
2804                                 /* device is still not ready. */
2805                                 break;
2806                         }
2807
2808                         if (idev) {
2809                                 if (idev->if_flags & IF_READY)
2810                                         /* device is already configured. */
2811                                         break;
2812                                 idev->if_flags |= IF_READY;
2813                         }
2814
2815                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2816                                 dev->name);
2817
2818                         run_pending = 1;
2819                 }
2820
2821                 switch (dev->type) {
2822 #if IS_ENABLED(CONFIG_IPV6_SIT)
2823                 case ARPHRD_SIT:
2824                         addrconf_sit_config(dev);
2825                         break;
2826 #endif
2827 #if IS_ENABLED(CONFIG_NET_IPGRE)
2828                 case ARPHRD_IPGRE:
2829                         addrconf_gre_config(dev);
2830                         break;
2831 #endif
2832                 case ARPHRD_LOOPBACK:
2833                         init_loopback(dev);
2834                         break;
2835
2836                 default:
2837                         addrconf_dev_config(dev);
2838                         break;
2839                 }
2840
2841                 if (idev) {
2842                         if (run_pending)
2843                                 addrconf_dad_run(idev);
2844
2845                         /*
2846                          * If the MTU changed during the interface down,
2847                          * when the interface up, the changed MTU must be
2848                          * reflected in the idev as well as routers.
2849                          */
2850                         if (idev->cnf.mtu6 != dev->mtu &&
2851                             dev->mtu >= IPV6_MIN_MTU) {
2852                                 rt6_mtu_change(dev, dev->mtu);
2853                                 idev->cnf.mtu6 = dev->mtu;
2854                         }
2855                         idev->tstamp = jiffies;
2856                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2857
2858                         /*
2859                          * If the changed mtu during down is lower than
2860                          * IPV6_MIN_MTU stop IPv6 on this interface.
2861                          */
2862                         if (dev->mtu < IPV6_MIN_MTU)
2863                                 addrconf_ifdown(dev, 1);
2864                 }
2865                 break;
2866
2867         case NETDEV_CHANGEMTU:
2868                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2869                         rt6_mtu_change(dev, dev->mtu);
2870                         idev->cnf.mtu6 = dev->mtu;
2871                         break;
2872                 }
2873
2874                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2875                         idev = ipv6_add_dev(dev);
2876                         if (idev)
2877                                 break;
2878                 }
2879
2880                 /*
2881                  * MTU falled under IPV6_MIN_MTU.
2882                  * Stop IPv6 on this interface.
2883                  */
2884
2885         case NETDEV_DOWN:
2886         case NETDEV_UNREGISTER:
2887                 /*
2888                  *      Remove all addresses from this interface.
2889                  */
2890                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2891                 break;
2892
2893         case NETDEV_CHANGENAME:
2894                 if (idev) {
2895                         snmp6_unregister_dev(idev);
2896                         addrconf_sysctl_unregister(idev);
2897                         addrconf_sysctl_register(idev);
2898                         err = snmp6_register_dev(idev);
2899                         if (err)
2900                                 return notifier_from_errno(err);
2901                 }
2902                 break;
2903
2904         case NETDEV_PRE_TYPE_CHANGE:
2905         case NETDEV_POST_TYPE_CHANGE:
2906                 addrconf_type_change(dev, event);
2907                 break;
2908         }
2909
2910         return NOTIFY_OK;
2911 }
2912
2913 /*
2914  *      addrconf module should be notified of a device going up
2915  */
2916 static struct notifier_block ipv6_dev_notf = {
2917         .notifier_call = addrconf_notify,
2918 };
2919
2920 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2921 {
2922         struct inet6_dev *idev;
2923         ASSERT_RTNL();
2924
2925         idev = __in6_dev_get(dev);
2926
2927         if (event == NETDEV_POST_TYPE_CHANGE)
2928                 ipv6_mc_remap(idev);
2929         else if (event == NETDEV_PRE_TYPE_CHANGE)
2930                 ipv6_mc_unmap(idev);
2931 }
2932
2933 static int addrconf_ifdown(struct net_device *dev, int how)
2934 {
2935         struct net *net = dev_net(dev);
2936         struct inet6_dev *idev;
2937         struct inet6_ifaddr *ifa;
2938         int state, i;
2939
2940         ASSERT_RTNL();
2941
2942         rt6_ifdown(net, dev);
2943         neigh_ifdown(&nd_tbl, dev);
2944
2945         idev = __in6_dev_get(dev);
2946         if (idev == NULL)
2947                 return -ENODEV;
2948
2949         /*
2950          * Step 1: remove reference to ipv6 device from parent device.
2951          *         Do not dev_put!
2952          */
2953         if (how) {
2954                 idev->dead = 1;
2955
2956                 /* protected by rtnl_lock */
2957                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2958
2959                 /* Step 1.5: remove snmp6 entry */
2960                 snmp6_unregister_dev(idev);
2961
2962         }
2963
2964         /* Step 2: clear hash table */
2965         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2966                 struct hlist_head *h = &inet6_addr_lst[i];
2967
2968                 spin_lock_bh(&addrconf_hash_lock);
2969         restart:
2970                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2971                         if (ifa->idev == idev) {
2972                                 hlist_del_init_rcu(&ifa->addr_lst);
2973                                 addrconf_del_dad_timer(ifa);
2974                                 goto restart;
2975                         }
2976                 }
2977                 spin_unlock_bh(&addrconf_hash_lock);
2978         }
2979
2980         write_lock_bh(&idev->lock);
2981
2982         addrconf_del_rs_timer(idev);
2983
2984         /* Step 2: clear flags for stateless addrconf */
2985         if (!how)
2986                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2987
2988         if (how && del_timer(&idev->regen_timer))
2989                 in6_dev_put(idev);
2990
2991         /* Step 3: clear tempaddr list */
2992         while (!list_empty(&idev->tempaddr_list)) {
2993                 ifa = list_first_entry(&idev->tempaddr_list,
2994                                        struct inet6_ifaddr, tmp_list);
2995                 list_del(&ifa->tmp_list);
2996                 write_unlock_bh(&idev->lock);
2997                 spin_lock_bh(&ifa->lock);
2998
2999                 if (ifa->ifpub) {
3000                         in6_ifa_put(ifa->ifpub);
3001                         ifa->ifpub = NULL;
3002                 }
3003                 spin_unlock_bh(&ifa->lock);
3004                 in6_ifa_put(ifa);
3005                 write_lock_bh(&idev->lock);
3006         }
3007
3008         while (!list_empty(&idev->addr_list)) {
3009                 ifa = list_first_entry(&idev->addr_list,
3010                                        struct inet6_ifaddr, if_list);
3011                 addrconf_del_dad_timer(ifa);
3012
3013                 list_del(&ifa->if_list);
3014
3015                 write_unlock_bh(&idev->lock);
3016
3017                 spin_lock_bh(&ifa->state_lock);
3018                 state = ifa->state;
3019                 ifa->state = INET6_IFADDR_STATE_DEAD;
3020                 spin_unlock_bh(&ifa->state_lock);
3021
3022                 if (state != INET6_IFADDR_STATE_DEAD) {
3023                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3024                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3025                 }
3026                 in6_ifa_put(ifa);
3027
3028                 write_lock_bh(&idev->lock);
3029         }
3030
3031         write_unlock_bh(&idev->lock);
3032
3033         /* Step 5: Discard multicast list */
3034         if (how)
3035                 ipv6_mc_destroy_dev(idev);
3036         else
3037                 ipv6_mc_down(idev);
3038
3039         idev->tstamp = jiffies;
3040
3041         /* Last: Shot the device (if unregistered) */
3042         if (how) {
3043                 addrconf_sysctl_unregister(idev);
3044                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3045                 neigh_ifdown(&nd_tbl, dev);
3046                 in6_dev_put(idev);
3047         }
3048         return 0;
3049 }
3050
3051 static void addrconf_rs_timer(unsigned long data)
3052 {
3053         struct inet6_dev *idev = (struct inet6_dev *)data;
3054         struct net_device *dev = idev->dev;
3055         struct in6_addr lladdr;
3056
3057         write_lock(&idev->lock);
3058         if (idev->dead || !(idev->if_flags & IF_READY))
3059                 goto out;
3060
3061         if (!ipv6_accept_ra(idev))
3062                 goto out;
3063
3064         /* Announcement received after solicitation was sent */
3065         if (idev->if_flags & IF_RA_RCVD)
3066                 goto out;
3067
3068         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3069                 write_unlock(&idev->lock);
3070                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3071                         ndisc_send_rs(dev, &lladdr,
3072                                       &in6addr_linklocal_allrouters);
3073                 else
3074                         goto put;
3075
3076                 write_lock(&idev->lock);
3077                 /* The wait after the last probe can be shorter */
3078                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3079                                              idev->cnf.rtr_solicits) ?
3080                                       idev->cnf.rtr_solicit_delay :
3081                                       idev->cnf.rtr_solicit_interval);
3082         } else {
3083                 /*
3084                  * Note: we do not support deprecated "all on-link"
3085                  * assumption any longer.
3086                  */
3087                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3088         }
3089
3090 out:
3091         write_unlock(&idev->lock);
3092 put:
3093         in6_dev_put(idev);
3094 }
3095
3096 /*
3097  *      Duplicate Address Detection
3098  */
3099 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3100 {
3101         unsigned long rand_num;
3102         struct inet6_dev *idev = ifp->idev;
3103
3104         if (ifp->flags & IFA_F_OPTIMISTIC)
3105                 rand_num = 0;
3106         else
3107                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3108
3109         ifp->dad_probes = idev->cnf.dad_transmits;
3110         addrconf_mod_dad_timer(ifp, rand_num);
3111 }
3112
3113 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3114 {
3115         struct inet6_dev *idev = ifp->idev;
3116         struct net_device *dev = idev->dev;
3117
3118         addrconf_join_solict(dev, &ifp->addr);
3119
3120         net_srandom(ifp->addr.s6_addr32[3]);
3121
3122         read_lock_bh(&idev->lock);
3123         spin_lock(&ifp->lock);
3124         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3125                 goto out;
3126
3127         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3128             idev->cnf.accept_dad < 1 ||
3129             !(ifp->flags&IFA_F_TENTATIVE) ||
3130             ifp->flags & IFA_F_NODAD) {
3131                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3132                 spin_unlock(&ifp->lock);
3133                 read_unlock_bh(&idev->lock);
3134
3135                 addrconf_dad_completed(ifp);
3136                 return;
3137         }
3138
3139         if (!(idev->if_flags & IF_READY)) {
3140                 spin_unlock(&ifp->lock);
3141                 read_unlock_bh(&idev->lock);
3142                 /*
3143                  * If the device is not ready:
3144                  * - keep it tentative if it is a permanent address.
3145                  * - otherwise, kill it.
3146                  */
3147                 in6_ifa_hold(ifp);
3148                 addrconf_dad_stop(ifp, 0);
3149                 return;
3150         }
3151
3152         /*
3153          * Optimistic nodes can start receiving
3154          * Frames right away
3155          */
3156         if (ifp->flags & IFA_F_OPTIMISTIC)
3157                 ip6_ins_rt(ifp->rt);
3158
3159         addrconf_dad_kick(ifp);
3160 out:
3161         spin_unlock(&ifp->lock);
3162         read_unlock_bh(&idev->lock);
3163 }
3164
3165 static void addrconf_dad_timer(unsigned long data)
3166 {
3167         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3168         struct inet6_dev *idev = ifp->idev;
3169         struct in6_addr mcaddr;
3170
3171         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3172                 goto out;
3173
3174         write_lock(&idev->lock);
3175         if (idev->dead || !(idev->if_flags & IF_READY)) {
3176                 write_unlock(&idev->lock);
3177                 goto out;
3178         }
3179
3180         spin_lock(&ifp->lock);
3181         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3182                 spin_unlock(&ifp->lock);
3183                 write_unlock(&idev->lock);
3184                 goto out;
3185         }
3186
3187         if (ifp->dad_probes == 0) {
3188                 /*
3189                  * DAD was successful
3190                  */
3191
3192                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3193                 spin_unlock(&ifp->lock);
3194                 write_unlock(&idev->lock);
3195
3196                 addrconf_dad_completed(ifp);
3197
3198                 goto out;
3199         }
3200
3201         ifp->dad_probes--;
3202         addrconf_mod_dad_timer(ifp,
3203                                NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3204         spin_unlock(&ifp->lock);
3205         write_unlock(&idev->lock);
3206
3207         /* send a neighbour solicitation for our addr */
3208         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3209         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3210 out:
3211         in6_ifa_put(ifp);
3212 }
3213
3214 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3215 {
3216         struct net_device *dev = ifp->idev->dev;
3217         struct in6_addr lladdr;
3218         bool send_rs, send_mld;
3219
3220         addrconf_del_dad_timer(ifp);
3221
3222         /*
3223          *      Configure the address for reception. Now it is valid.
3224          */
3225
3226         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3227
3228         /* If added prefix is link local and we are prepared to process
3229            router advertisements, start sending router solicitations.
3230          */
3231
3232         read_lock_bh(&ifp->idev->lock);
3233         spin_lock(&ifp->lock);
3234         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3235                    ifp->idev->valid_ll_addr_cnt == 1;
3236         send_rs = send_mld &&
3237                   ipv6_accept_ra(ifp->idev) &&
3238                   ifp->idev->cnf.rtr_solicits > 0 &&
3239                   (dev->flags&IFF_LOOPBACK) == 0;
3240         spin_unlock(&ifp->lock);
3241         read_unlock_bh(&ifp->idev->lock);
3242
3243         /* While dad is in progress mld report's source address is in6_addrany.
3244          * Resend with proper ll now.
3245          */
3246         if (send_mld)
3247                 ipv6_mc_dad_complete(ifp->idev);
3248
3249         if (send_rs) {
3250                 /*
3251                  *      If a host as already performed a random delay
3252                  *      [...] as part of DAD [...] there is no need
3253                  *      to delay again before sending the first RS
3254                  */
3255                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3256                         return;
3257                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3258
3259                 write_lock_bh(&ifp->idev->lock);
3260                 spin_lock(&ifp->lock);
3261                 ifp->idev->rs_probes = 1;
3262                 ifp->idev->if_flags |= IF_RS_SENT;
3263                 addrconf_mod_rs_timer(ifp->idev,
3264                                       ifp->idev->cnf.rtr_solicit_interval);
3265                 spin_unlock(&ifp->lock);
3266                 write_unlock_bh(&ifp->idev->lock);
3267         }
3268 }
3269
3270 static void addrconf_dad_run(struct inet6_dev *idev)
3271 {
3272         struct inet6_ifaddr *ifp;
3273
3274         read_lock_bh(&idev->lock);
3275         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3276                 spin_lock(&ifp->lock);
3277                 if (ifp->flags & IFA_F_TENTATIVE &&
3278                     ifp->state == INET6_IFADDR_STATE_DAD)
3279                         addrconf_dad_kick(ifp);
3280                 spin_unlock(&ifp->lock);
3281         }
3282         read_unlock_bh(&idev->lock);
3283 }
3284
3285 #ifdef CONFIG_PROC_FS
3286 struct if6_iter_state {
3287         struct seq_net_private p;
3288         int bucket;
3289         int offset;
3290 };
3291
3292 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3293 {
3294         struct inet6_ifaddr *ifa = NULL;
3295         struct if6_iter_state *state = seq->private;
3296         struct net *net = seq_file_net(seq);
3297         int p = 0;
3298
3299         /* initial bucket if pos is 0 */
3300         if (pos == 0) {
3301                 state->bucket = 0;
3302                 state->offset = 0;
3303         }
3304
3305         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3306                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3307                                          addr_lst) {
3308                         if (!net_eq(dev_net(ifa->idev->dev), net))
3309                                 continue;
3310                         /* sync with offset */
3311                         if (p < state->offset) {
3312                                 p++;
3313                                 continue;
3314                         }
3315                         state->offset++;
3316                         return ifa;
3317                 }
3318
3319                 /* prepare for next bucket */
3320                 state->offset = 0;
3321                 p = 0;
3322         }
3323         return NULL;
3324 }
3325
3326 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3327                                          struct inet6_ifaddr *ifa)
3328 {
3329         struct if6_iter_state *state = seq->private;
3330         struct net *net = seq_file_net(seq);
3331
3332         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3333                 if (!net_eq(dev_net(ifa->idev->dev), net))
3334                         continue;
3335                 state->offset++;
3336                 return ifa;
3337         }
3338
3339         while (++state->bucket < IN6_ADDR_HSIZE) {
3340                 state->offset = 0;
3341                 hlist_for_each_entry_rcu_bh(ifa,
3342                                      &inet6_addr_lst[state->bucket], addr_lst) {
3343                         if (!net_eq(dev_net(ifa->idev->dev), net))
3344                                 continue;
3345                         state->offset++;
3346                         return ifa;
3347                 }
3348         }
3349
3350         return NULL;
3351 }
3352
3353 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3354         __acquires(rcu_bh)
3355 {
3356         rcu_read_lock_bh();
3357         return if6_get_first(seq, *pos);
3358 }
3359
3360 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3361 {
3362         struct inet6_ifaddr *ifa;
3363
3364         ifa = if6_get_next(seq, v);
3365         ++*pos;
3366         return ifa;
3367 }
3368
3369 static void if6_seq_stop(struct seq_file *seq, void *v)
3370         __releases(rcu_bh)
3371 {
3372         rcu_read_unlock_bh();
3373 }
3374
3375 static int if6_seq_show(struct seq_file *seq, void *v)
3376 {
3377         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3378         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3379                    &ifp->addr,
3380                    ifp->idev->dev->ifindex,
3381                    ifp->prefix_len,
3382                    ifp->scope,
3383                    (u8) ifp->flags,
3384                    ifp->idev->dev->name);
3385         return 0;
3386 }
3387
3388 static const struct seq_operations if6_seq_ops = {
3389         .start  = if6_seq_start,
3390         .next   = if6_seq_next,
3391         .show   = if6_seq_show,
3392         .stop   = if6_seq_stop,
3393 };
3394
3395 static int if6_seq_open(struct inode *inode, struct file *file)
3396 {
3397         return seq_open_net(inode, file, &if6_seq_ops,
3398                             sizeof(struct if6_iter_state));
3399 }
3400
3401 static const struct file_operations if6_fops = {
3402         .owner          = THIS_MODULE,
3403         .open           = if6_seq_open,
3404         .read           = seq_read,
3405         .llseek         = seq_lseek,
3406         .release        = seq_release_net,
3407 };
3408
3409 static int __net_init if6_proc_net_init(struct net *net)
3410 {
3411         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3412                 return -ENOMEM;
3413         return 0;
3414 }
3415
3416 static void __net_exit if6_proc_net_exit(struct net *net)
3417 {
3418         remove_proc_entry("if_inet6", net->proc_net);
3419 }
3420
3421 static struct pernet_operations if6_proc_net_ops = {
3422        .init = if6_proc_net_init,
3423        .exit = if6_proc_net_exit,
3424 };
3425
3426 int __init if6_proc_init(void)
3427 {
3428         return register_pernet_subsys(&if6_proc_net_ops);
3429 }
3430
3431 void if6_proc_exit(void)
3432 {
3433         unregister_pernet_subsys(&if6_proc_net_ops);
3434 }
3435 #endif  /* CONFIG_PROC_FS */
3436
3437 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3438 /* Check if address is a home address configured on any interface. */
3439 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3440 {
3441         int ret = 0;
3442         struct inet6_ifaddr *ifp = NULL;
3443         unsigned int hash = inet6_addr_hash(addr);
3444
3445         rcu_read_lock_bh();
3446         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3447                 if (!net_eq(dev_net(ifp->idev->dev), net))
3448                         continue;
3449                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3450                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3451                         ret = 1;
3452                         break;
3453                 }
3454         }
3455         rcu_read_unlock_bh();
3456         return ret;
3457 }
3458 #endif
3459
3460 /*
3461  *      Periodic address status verification
3462  */
3463
3464 static void addrconf_verify(unsigned long foo)
3465 {
3466         unsigned long now, next, next_sec, next_sched;
3467         struct inet6_ifaddr *ifp;
3468         int i;
3469
3470         rcu_read_lock_bh();
3471         spin_lock(&addrconf_verify_lock);
3472         now = jiffies;
3473         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3474
3475         del_timer(&addr_chk_timer);
3476
3477         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3478 restart:
3479                 hlist_for_each_entry_rcu_bh(ifp,
3480                                          &inet6_addr_lst[i], addr_lst) {
3481                         unsigned long age;
3482
3483                         /* When setting preferred_lft to a value not zero or
3484                          * infinity, while valid_lft is infinity
3485                          * IFA_F_PERMANENT has a non-infinity life time.
3486                          */
3487                         if ((ifp->flags & IFA_F_PERMANENT) &&
3488                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3489                                 continue;
3490
3491                         spin_lock(&ifp->lock);
3492                         /* We try to batch several events at once. */
3493                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3494
3495                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3496                             age >= ifp->valid_lft) {
3497                                 spin_unlock(&ifp->lock);
3498                                 in6_ifa_hold(ifp);
3499                                 ipv6_del_addr(ifp);
3500                                 goto restart;
3501                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3502                                 spin_unlock(&ifp->lock);
3503                                 continue;
3504                         } else if (age >= ifp->prefered_lft) {
3505                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3506                                 int deprecate = 0;
3507
3508                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3509                                         deprecate = 1;
3510                                         ifp->flags |= IFA_F_DEPRECATED;
3511                                 }
3512
3513                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3514                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3515                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3516
3517                                 spin_unlock(&ifp->lock);
3518
3519                                 if (deprecate) {
3520                                         in6_ifa_hold(ifp);
3521
3522                                         ipv6_ifa_notify(0, ifp);
3523                                         in6_ifa_put(ifp);
3524                                         goto restart;
3525                                 }
3526                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3527                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3528                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3529                                         ifp->idev->cnf.dad_transmits *
3530                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3531
3532                                 if (age >= ifp->prefered_lft - regen_advance) {
3533                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3534                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3535                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3536                                         if (!ifp->regen_count && ifpub) {
3537                                                 ifp->regen_count++;
3538                                                 in6_ifa_hold(ifp);
3539                                                 in6_ifa_hold(ifpub);
3540                                                 spin_unlock(&ifp->lock);
3541
3542                                                 spin_lock(&ifpub->lock);
3543                                                 ifpub->regen_count = 0;
3544                                                 spin_unlock(&ifpub->lock);
3545                                                 ipv6_create_tempaddr(ifpub, ifp);
3546                                                 in6_ifa_put(ifpub);
3547                                                 in6_ifa_put(ifp);
3548                                                 goto restart;
3549                                         }
3550                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3551                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3552                                 spin_unlock(&ifp->lock);
3553                         } else {
3554                                 /* ifp->prefered_lft <= ifp->valid_lft */
3555                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3556                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3557                                 spin_unlock(&ifp->lock);
3558                         }
3559                 }
3560         }
3561
3562         next_sec = round_jiffies_up(next);
3563         next_sched = next;
3564
3565         /* If rounded timeout is accurate enough, accept it. */
3566         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3567                 next_sched = next_sec;
3568
3569         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3570         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3571                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3572
3573         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3574               now, next, next_sec, next_sched);
3575
3576         addr_chk_timer.expires = next_sched;
3577         add_timer(&addr_chk_timer);
3578         spin_unlock(&addrconf_verify_lock);
3579         rcu_read_unlock_bh();
3580 }
3581
3582 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3583                                      struct in6_addr **peer_pfx)
3584 {
3585         struct in6_addr *pfx = NULL;
3586
3587         *peer_pfx = NULL;
3588
3589         if (addr)
3590                 pfx = nla_data(addr);
3591
3592         if (local) {
3593                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3594                         *peer_pfx = pfx;
3595                 pfx = nla_data(local);
3596         }
3597
3598         return pfx;
3599 }
3600
3601 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3602         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3603         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3604         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3605         [IFA_FLAGS]             = { .len = sizeof(u32) },
3606 };
3607
3608 static int
3609 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3610 {
3611         struct net *net = sock_net(skb->sk);
3612         struct ifaddrmsg *ifm;
3613         struct nlattr *tb[IFA_MAX+1];
3614         struct in6_addr *pfx, *peer_pfx;
3615         int err;
3616
3617         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3618         if (err < 0)
3619                 return err;
3620
3621         ifm = nlmsg_data(nlh);
3622         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3623         if (pfx == NULL)
3624                 return -EINVAL;
3625
3626         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3627 }
3628
3629 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3630                              u32 prefered_lft, u32 valid_lft)
3631 {
3632         u32 flags;
3633         clock_t expires;
3634         unsigned long timeout;
3635         bool was_managetempaddr;
3636
3637         if (!valid_lft || (prefered_lft > valid_lft))
3638                 return -EINVAL;
3639
3640         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3641             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3642                 return -EINVAL;
3643
3644         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3645         if (addrconf_finite_timeout(timeout)) {
3646                 expires = jiffies_to_clock_t(timeout * HZ);
3647                 valid_lft = timeout;
3648                 flags = RTF_EXPIRES;
3649         } else {
3650                 expires = 0;
3651                 flags = 0;
3652                 ifa_flags |= IFA_F_PERMANENT;
3653         }
3654
3655         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3656         if (addrconf_finite_timeout(timeout)) {
3657                 if (timeout == 0)
3658                         ifa_flags |= IFA_F_DEPRECATED;
3659                 prefered_lft = timeout;
3660         }
3661
3662         spin_lock_bh(&ifp->lock);
3663         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3664         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3665                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR);
3666         ifp->flags |= ifa_flags;
3667         ifp->tstamp = jiffies;
3668         ifp->valid_lft = valid_lft;
3669         ifp->prefered_lft = prefered_lft;
3670
3671         spin_unlock_bh(&ifp->lock);
3672         if (!(ifp->flags&IFA_F_TENTATIVE))
3673                 ipv6_ifa_notify(0, ifp);
3674
3675         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3676                               expires, flags);
3677
3678         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3679                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3680                         valid_lft = prefered_lft = 0;
3681                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3682                                  !was_managetempaddr, jiffies);
3683         }
3684
3685         addrconf_verify(0);
3686
3687         return 0;
3688 }
3689
3690 static int
3691 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3692 {
3693         struct net *net = sock_net(skb->sk);
3694         struct ifaddrmsg *ifm;
3695         struct nlattr *tb[IFA_MAX+1];
3696         struct in6_addr *pfx, *peer_pfx;
3697         struct inet6_ifaddr *ifa;
3698         struct net_device *dev;
3699         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3700         u32 ifa_flags;
3701         int err;
3702
3703         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3704         if (err < 0)
3705                 return err;
3706
3707         ifm = nlmsg_data(nlh);
3708         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3709         if (pfx == NULL)
3710                 return -EINVAL;
3711
3712         if (tb[IFA_CACHEINFO]) {
3713                 struct ifa_cacheinfo *ci;
3714
3715                 ci = nla_data(tb[IFA_CACHEINFO]);
3716                 valid_lft = ci->ifa_valid;
3717                 preferred_lft = ci->ifa_prefered;
3718         } else {
3719                 preferred_lft = INFINITY_LIFE_TIME;
3720                 valid_lft = INFINITY_LIFE_TIME;
3721         }
3722
3723         dev =  __dev_get_by_index(net, ifm->ifa_index);
3724         if (dev == NULL)
3725                 return -ENODEV;
3726
3727         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3728
3729         /* We ignore other flags so far. */
3730         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR;
3731
3732         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3733         if (ifa == NULL) {
3734                 /*
3735                  * It would be best to check for !NLM_F_CREATE here but
3736                  * userspace alreay relies on not having to provide this.
3737                  */
3738                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3739                                       ifm->ifa_prefixlen, ifa_flags,
3740                                       preferred_lft, valid_lft);
3741         }
3742
3743         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3744             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3745                 err = -EEXIST;
3746         else
3747                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3748
3749         in6_ifa_put(ifa);
3750
3751         return err;
3752 }
3753
3754 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3755                           u8 scope, int ifindex)
3756 {
3757         struct ifaddrmsg *ifm;
3758
3759         ifm = nlmsg_data(nlh);
3760         ifm->ifa_family = AF_INET6;
3761         ifm->ifa_prefixlen = prefixlen;
3762         ifm->ifa_flags = flags;
3763         ifm->ifa_scope = scope;
3764         ifm->ifa_index = ifindex;
3765 }
3766
3767 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3768                          unsigned long tstamp, u32 preferred, u32 valid)
3769 {
3770         struct ifa_cacheinfo ci;
3771
3772         ci.cstamp = cstamp_delta(cstamp);
3773         ci.tstamp = cstamp_delta(tstamp);
3774         ci.ifa_prefered = preferred;
3775         ci.ifa_valid = valid;
3776
3777         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3778 }
3779
3780 static inline int rt_scope(int ifa_scope)
3781 {
3782         if (ifa_scope & IFA_HOST)
3783                 return RT_SCOPE_HOST;
3784         else if (ifa_scope & IFA_LINK)
3785                 return RT_SCOPE_LINK;
3786         else if (ifa_scope & IFA_SITE)
3787                 return RT_SCOPE_SITE;
3788         else
3789                 return RT_SCOPE_UNIVERSE;
3790 }
3791
3792 static inline int inet6_ifaddr_msgsize(void)
3793 {
3794         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3795                + nla_total_size(16) /* IFA_LOCAL */
3796                + nla_total_size(16) /* IFA_ADDRESS */
3797                + nla_total_size(sizeof(struct ifa_cacheinfo))
3798                + nla_total_size(4)  /* IFA_FLAGS */;
3799 }
3800
3801 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3802                              u32 portid, u32 seq, int event, unsigned int flags)
3803 {
3804         struct nlmsghdr  *nlh;
3805         u32 preferred, valid;
3806
3807         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3808         if (nlh == NULL)
3809                 return -EMSGSIZE;
3810
3811         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3812                       ifa->idev->dev->ifindex);
3813
3814         if (!((ifa->flags&IFA_F_PERMANENT) &&
3815               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3816                 preferred = ifa->prefered_lft;
3817                 valid = ifa->valid_lft;
3818                 if (preferred != INFINITY_LIFE_TIME) {
3819                         long tval = (jiffies - ifa->tstamp)/HZ;
3820                         if (preferred > tval)
3821                                 preferred -= tval;
3822                         else
3823                                 preferred = 0;
3824                         if (valid != INFINITY_LIFE_TIME) {
3825                                 if (valid > tval)
3826                                         valid -= tval;
3827                                 else
3828                                         valid = 0;
3829                         }
3830                 }
3831         } else {
3832                 preferred = INFINITY_LIFE_TIME;
3833                 valid = INFINITY_LIFE_TIME;
3834         }
3835
3836         if (!ipv6_addr_any(&ifa->peer_addr)) {
3837                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3838                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3839                         goto error;
3840         } else
3841                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3842                         goto error;
3843
3844         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3845                 goto error;
3846
3847         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
3848                 goto error;
3849
3850         return nlmsg_end(skb, nlh);
3851
3852 error:
3853         nlmsg_cancel(skb, nlh);
3854         return -EMSGSIZE;
3855 }
3856
3857 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3858                                 u32 portid, u32 seq, int event, u16 flags)
3859 {
3860         struct nlmsghdr  *nlh;
3861         u8 scope = RT_SCOPE_UNIVERSE;
3862         int ifindex = ifmca->idev->dev->ifindex;
3863
3864         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3865                 scope = RT_SCOPE_SITE;
3866
3867         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3868         if (nlh == NULL)
3869                 return -EMSGSIZE;
3870
3871         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3872         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3873             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3874                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3875                 nlmsg_cancel(skb, nlh);
3876                 return -EMSGSIZE;
3877         }
3878
3879         return nlmsg_end(skb, nlh);
3880 }
3881
3882 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3883                                 u32 portid, u32 seq, int event, unsigned int flags)
3884 {
3885         struct nlmsghdr  *nlh;
3886         u8 scope = RT_SCOPE_UNIVERSE;
3887         int ifindex = ifaca->aca_idev->dev->ifindex;
3888
3889         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3890                 scope = RT_SCOPE_SITE;
3891
3892         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3893         if (nlh == NULL)
3894                 return -EMSGSIZE;
3895
3896         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3897         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3898             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3899                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3900                 nlmsg_cancel(skb, nlh);
3901                 return -EMSGSIZE;
3902         }
3903
3904         return nlmsg_end(skb, nlh);
3905 }
3906
3907 enum addr_type_t {
3908         UNICAST_ADDR,
3909         MULTICAST_ADDR,
3910         ANYCAST_ADDR,
3911 };
3912
3913 /* called with rcu_read_lock() */
3914 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3915                           struct netlink_callback *cb, enum addr_type_t type,
3916                           int s_ip_idx, int *p_ip_idx)
3917 {
3918         struct ifmcaddr6 *ifmca;
3919         struct ifacaddr6 *ifaca;
3920         int err = 1;
3921         int ip_idx = *p_ip_idx;
3922
3923         read_lock_bh(&idev->lock);
3924         switch (type) {
3925         case UNICAST_ADDR: {
3926                 struct inet6_ifaddr *ifa;
3927
3928                 /* unicast address incl. temp addr */
3929                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3930                         if (++ip_idx < s_ip_idx)
3931                                 continue;
3932                         err = inet6_fill_ifaddr(skb, ifa,
3933                                                 NETLINK_CB(cb->skb).portid,
3934                                                 cb->nlh->nlmsg_seq,
3935                                                 RTM_NEWADDR,
3936                                                 NLM_F_MULTI);
3937                         if (err <= 0)
3938                                 break;
3939                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3940                 }
3941                 break;
3942         }
3943         case MULTICAST_ADDR:
3944                 /* multicast address */
3945                 for (ifmca = idev->mc_list; ifmca;
3946                      ifmca = ifmca->next, ip_idx++) {
3947                         if (ip_idx < s_ip_idx)
3948                                 continue;
3949                         err = inet6_fill_ifmcaddr(skb, ifmca,
3950                                                   NETLINK_CB(cb->skb).portid,
3951                                                   cb->nlh->nlmsg_seq,
3952                                                   RTM_GETMULTICAST,
3953                                                   NLM_F_MULTI);
3954                         if (err <= 0)
3955                                 break;
3956                 }
3957                 break;
3958         case ANYCAST_ADDR:
3959                 /* anycast address */
3960                 for (ifaca = idev->ac_list; ifaca;
3961                      ifaca = ifaca->aca_next, ip_idx++) {
3962                         if (ip_idx < s_ip_idx)
3963                                 continue;
3964                         err = inet6_fill_ifacaddr(skb, ifaca,
3965                                                   NETLINK_CB(cb->skb).portid,
3966                                                   cb->nlh->nlmsg_seq,
3967                                                   RTM_GETANYCAST,
3968                                                   NLM_F_MULTI);
3969                         if (err <= 0)
3970                                 break;
3971                 }
3972                 break;
3973         default:
3974                 break;
3975         }
3976         read_unlock_bh(&idev->lock);
3977         *p_ip_idx = ip_idx;
3978         return err;
3979 }
3980
3981 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3982                            enum addr_type_t type)
3983 {
3984         struct net *net = sock_net(skb->sk);
3985         int h, s_h;
3986         int idx, ip_idx;
3987         int s_idx, s_ip_idx;
3988         struct net_device *dev;
3989         struct inet6_dev *idev;
3990         struct hlist_head *head;
3991
3992         s_h = cb->args[0];
3993         s_idx = idx = cb->args[1];
3994         s_ip_idx = ip_idx = cb->args[2];
3995
3996         rcu_read_lock();
3997         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3998         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3999                 idx = 0;
4000                 head = &net->dev_index_head[h];
4001                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4002                         if (idx < s_idx)
4003                                 goto cont;
4004                         if (h > s_h || idx > s_idx)
4005                                 s_ip_idx = 0;
4006                         ip_idx = 0;
4007                         idev = __in6_dev_get(dev);
4008                         if (!idev)
4009                                 goto cont;
4010
4011                         if (in6_dump_addrs(idev, skb, cb, type,
4012                                            s_ip_idx, &ip_idx) <= 0)
4013                                 goto done;
4014 cont:
4015                         idx++;
4016                 }
4017         }
4018 done:
4019         rcu_read_unlock();
4020         cb->args[0] = h;
4021         cb->args[1] = idx;
4022         cb->args[2] = ip_idx;
4023
4024         return skb->len;
4025 }
4026
4027 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4028 {
4029         enum addr_type_t type = UNICAST_ADDR;
4030
4031         return inet6_dump_addr(skb, cb, type);
4032 }
4033
4034 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4035 {
4036         enum addr_type_t type = MULTICAST_ADDR;
4037
4038         return inet6_dump_addr(skb, cb, type);
4039 }
4040
4041
4042 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4043 {
4044         enum addr_type_t type = ANYCAST_ADDR;
4045
4046         return inet6_dump_addr(skb, cb, type);
4047 }
4048
4049 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4050 {
4051         struct net *net = sock_net(in_skb->sk);
4052         struct ifaddrmsg *ifm;
4053         struct nlattr *tb[IFA_MAX+1];
4054         struct in6_addr *addr = NULL, *peer;
4055         struct net_device *dev = NULL;
4056         struct inet6_ifaddr *ifa;
4057         struct sk_buff *skb;
4058         int err;
4059
4060         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4061         if (err < 0)
4062                 goto errout;
4063
4064         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4065         if (addr == NULL) {
4066                 err = -EINVAL;
4067                 goto errout;
4068         }
4069
4070         ifm = nlmsg_data(nlh);
4071         if (ifm->ifa_index)
4072                 dev = __dev_get_by_index(net, ifm->ifa_index);
4073
4074         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4075         if (!ifa) {
4076                 err = -EADDRNOTAVAIL;
4077                 goto errout;
4078         }
4079
4080         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4081         if (!skb) {
4082                 err = -ENOBUFS;
4083                 goto errout_ifa;
4084         }
4085
4086         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4087                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4088         if (err < 0) {
4089                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4090                 WARN_ON(err == -EMSGSIZE);
4091                 kfree_skb(skb);
4092                 goto errout_ifa;
4093         }
4094         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4095 errout_ifa:
4096         in6_ifa_put(ifa);
4097 errout:
4098         return err;
4099 }
4100
4101 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4102 {
4103         struct sk_buff *skb;
4104         struct net *net = dev_net(ifa->idev->dev);
4105         int err = -ENOBUFS;
4106
4107         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4108         if (skb == NULL)
4109                 goto errout;
4110
4111         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4112         if (err < 0) {
4113                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4114                 WARN_ON(err == -EMSGSIZE);
4115                 kfree_skb(skb);
4116                 goto errout;
4117         }
4118         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4119         return;
4120 errout:
4121         if (err < 0)
4122                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4123 }
4124
4125 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4126                                 __s32 *array, int bytes)
4127 {
4128         BUG_ON(bytes < (DEVCONF_MAX * 4));
4129
4130         memset(array, 0, bytes);
4131         array[DEVCONF_FORWARDING] = cnf->forwarding;
4132         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4133         array[DEVCONF_MTU6] = cnf->mtu6;
4134         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4135         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4136         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4137         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4138         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4139         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4140                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4141         array[DEVCONF_RTR_SOLICIT_DELAY] =
4142                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4143         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4144         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4145                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4146         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4147                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4148         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4149         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4150         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4151         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4152         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4153         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4154         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4155         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4156 #ifdef CONFIG_IPV6_ROUTER_PREF
4157         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4158         array[DEVCONF_RTR_PROBE_INTERVAL] =
4159                 jiffies_to_msecs(cnf->rtr_probe_interval);
4160 #ifdef CONFIG_IPV6_ROUTE_INFO
4161         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4162 #endif
4163 #endif
4164         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4165         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4166 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4167         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4168 #endif
4169 #ifdef CONFIG_IPV6_MROUTE
4170         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4171 #endif
4172         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4173         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4174         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4175         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4176         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4177 }
4178
4179 static inline size_t inet6_ifla6_size(void)
4180 {
4181         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4182              + nla_total_size(sizeof(struct ifla_cacheinfo))
4183              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4184              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4185              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4186              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4187 }
4188
4189 static inline size_t inet6_if_nlmsg_size(void)
4190 {
4191         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4192                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4193                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4194                + nla_total_size(4) /* IFLA_MTU */
4195                + nla_total_size(4) /* IFLA_LINK */
4196                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4197 }
4198
4199 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4200                                       int items, int bytes)
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(atomic_long_read(&mib[i]), &stats[i]);
4210
4211         memset(&stats[items], 0, pad);
4212 }
4213
4214 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4215                                       int items, int bytes, size_t syncpoff)
4216 {
4217         int i;
4218         int pad = bytes - sizeof(u64) * items;
4219         BUG_ON(pad < 0);
4220
4221         /* Use put_unaligned() because stats may not be aligned for u64. */
4222         put_unaligned(items, &stats[0]);
4223         for (i = 1; i < items; i++)
4224                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4225
4226         memset(&stats[items], 0, pad);
4227 }
4228
4229 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4230                              int bytes)
4231 {
4232         switch (attrtype) {
4233         case IFLA_INET6_STATS:
4234                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4235                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4236                 break;
4237         case IFLA_INET6_ICMP6STATS:
4238                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4239                 break;
4240         }
4241 }
4242
4243 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4244 {
4245         struct nlattr *nla;
4246         struct ifla_cacheinfo ci;
4247
4248         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4249                 goto nla_put_failure;
4250         ci.max_reasm_len = IPV6_MAXPLEN;
4251         ci.tstamp = cstamp_delta(idev->tstamp);
4252         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4253         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4254         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4255                 goto nla_put_failure;
4256         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4257         if (nla == NULL)
4258                 goto nla_put_failure;
4259         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4260
4261         /* XXX - MC not implemented */
4262
4263         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4264         if (nla == NULL)
4265                 goto nla_put_failure;
4266         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4267
4268         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4269         if (nla == NULL)
4270                 goto nla_put_failure;
4271         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4272
4273         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4274         if (nla == NULL)
4275                 goto nla_put_failure;
4276         read_lock_bh(&idev->lock);
4277         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4278         read_unlock_bh(&idev->lock);
4279
4280         return 0;
4281
4282 nla_put_failure:
4283         return -EMSGSIZE;
4284 }
4285
4286 static size_t inet6_get_link_af_size(const struct net_device *dev)
4287 {
4288         if (!__in6_dev_get(dev))
4289                 return 0;
4290
4291         return inet6_ifla6_size();
4292 }
4293
4294 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4295 {
4296         struct inet6_dev *idev = __in6_dev_get(dev);
4297
4298         if (!idev)
4299                 return -ENODATA;
4300
4301         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4302                 return -EMSGSIZE;
4303
4304         return 0;
4305 }
4306
4307 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4308 {
4309         struct inet6_ifaddr *ifp;
4310         struct net_device *dev = idev->dev;
4311         bool update_rs = false;
4312         struct in6_addr ll_addr;
4313
4314         if (token == NULL)
4315                 return -EINVAL;
4316         if (ipv6_addr_any(token))
4317                 return -EINVAL;
4318         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4319                 return -EINVAL;
4320         if (!ipv6_accept_ra(idev))
4321                 return -EINVAL;
4322         if (idev->cnf.rtr_solicits <= 0)
4323                 return -EINVAL;
4324
4325         write_lock_bh(&idev->lock);
4326
4327         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4328         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4329
4330         write_unlock_bh(&idev->lock);
4331
4332         if (!idev->dead && (idev->if_flags & IF_READY) &&
4333             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4334                              IFA_F_OPTIMISTIC)) {
4335
4336                 /* If we're not ready, then normal ifup will take care
4337                  * of this. Otherwise, we need to request our rs here.
4338                  */
4339                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4340                 update_rs = true;
4341         }
4342
4343         write_lock_bh(&idev->lock);
4344
4345         if (update_rs) {
4346                 idev->if_flags |= IF_RS_SENT;
4347                 idev->rs_probes = 1;
4348                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4349         }
4350
4351         /* Well, that's kinda nasty ... */
4352         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4353                 spin_lock(&ifp->lock);
4354                 if (ifp->tokenized) {
4355                         ifp->valid_lft = 0;
4356                         ifp->prefered_lft = 0;
4357                 }
4358                 spin_unlock(&ifp->lock);
4359         }
4360
4361         write_unlock_bh(&idev->lock);
4362         addrconf_verify(0);
4363         return 0;
4364 }
4365
4366 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4367 {
4368         int err = -EINVAL;
4369         struct inet6_dev *idev = __in6_dev_get(dev);
4370         struct nlattr *tb[IFLA_INET6_MAX + 1];
4371
4372         if (!idev)
4373                 return -EAFNOSUPPORT;
4374
4375         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4376                 BUG();
4377
4378         if (tb[IFLA_INET6_TOKEN])
4379                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4380
4381         return err;
4382 }
4383
4384 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4385                              u32 portid, u32 seq, int event, unsigned int flags)
4386 {
4387         struct net_device *dev = idev->dev;
4388         struct ifinfomsg *hdr;
4389         struct nlmsghdr *nlh;
4390         void *protoinfo;
4391
4392         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4393         if (nlh == NULL)
4394                 return -EMSGSIZE;
4395
4396         hdr = nlmsg_data(nlh);
4397         hdr->ifi_family = AF_INET6;
4398         hdr->__ifi_pad = 0;
4399         hdr->ifi_type = dev->type;
4400         hdr->ifi_index = dev->ifindex;
4401         hdr->ifi_flags = dev_get_flags(dev);
4402         hdr->ifi_change = 0;
4403
4404         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4405             (dev->addr_len &&
4406              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4407             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4408             (dev->ifindex != dev->iflink &&
4409              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4410                 goto nla_put_failure;
4411         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4412         if (protoinfo == NULL)
4413                 goto nla_put_failure;
4414
4415         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4416                 goto nla_put_failure;
4417
4418         nla_nest_end(skb, protoinfo);
4419         return nlmsg_end(skb, nlh);
4420
4421 nla_put_failure:
4422         nlmsg_cancel(skb, nlh);
4423         return -EMSGSIZE;
4424 }
4425
4426 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4427 {
4428         struct net *net = sock_net(skb->sk);
4429         int h, s_h;
4430         int idx = 0, s_idx;
4431         struct net_device *dev;
4432         struct inet6_dev *idev;
4433         struct hlist_head *head;
4434
4435         s_h = cb->args[0];
4436         s_idx = cb->args[1];
4437
4438         rcu_read_lock();
4439         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4440                 idx = 0;
4441                 head = &net->dev_index_head[h];
4442                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4443                         if (idx < s_idx)
4444                                 goto cont;
4445                         idev = __in6_dev_get(dev);
4446                         if (!idev)
4447                                 goto cont;
4448                         if (inet6_fill_ifinfo(skb, idev,
4449                                               NETLINK_CB(cb->skb).portid,
4450                                               cb->nlh->nlmsg_seq,
4451                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4452                                 goto out;
4453 cont:
4454                         idx++;
4455                 }
4456         }
4457 out:
4458         rcu_read_unlock();
4459         cb->args[1] = idx;
4460         cb->args[0] = h;
4461
4462         return skb->len;
4463 }
4464
4465 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4466 {
4467         struct sk_buff *skb;
4468         struct net *net = dev_net(idev->dev);
4469         int err = -ENOBUFS;
4470
4471         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4472         if (skb == NULL)
4473                 goto errout;
4474
4475         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4476         if (err < 0) {
4477                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4478                 WARN_ON(err == -EMSGSIZE);
4479                 kfree_skb(skb);
4480                 goto errout;
4481         }
4482         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4483         return;
4484 errout:
4485         if (err < 0)
4486                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4487 }
4488
4489 static inline size_t inet6_prefix_nlmsg_size(void)
4490 {
4491         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4492                + nla_total_size(sizeof(struct in6_addr))
4493                + nla_total_size(sizeof(struct prefix_cacheinfo));
4494 }
4495
4496 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4497                              struct prefix_info *pinfo, u32 portid, u32 seq,
4498                              int event, unsigned int flags)
4499 {
4500         struct prefixmsg *pmsg;
4501         struct nlmsghdr *nlh;
4502         struct prefix_cacheinfo ci;
4503
4504         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4505         if (nlh == NULL)
4506                 return -EMSGSIZE;
4507
4508         pmsg = nlmsg_data(nlh);
4509         pmsg->prefix_family = AF_INET6;
4510         pmsg->prefix_pad1 = 0;
4511         pmsg->prefix_pad2 = 0;
4512         pmsg->prefix_ifindex = idev->dev->ifindex;
4513         pmsg->prefix_len = pinfo->prefix_len;
4514         pmsg->prefix_type = pinfo->type;
4515         pmsg->prefix_pad3 = 0;
4516         pmsg->prefix_flags = 0;
4517         if (pinfo->onlink)
4518                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4519         if (pinfo->autoconf)
4520                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4521
4522         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4523                 goto nla_put_failure;
4524         ci.preferred_time = ntohl(pinfo->prefered);
4525         ci.valid_time = ntohl(pinfo->valid);
4526         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4527                 goto nla_put_failure;
4528         return nlmsg_end(skb, nlh);
4529
4530 nla_put_failure:
4531         nlmsg_cancel(skb, nlh);
4532         return -EMSGSIZE;
4533 }
4534
4535 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4536                          struct prefix_info *pinfo)
4537 {
4538         struct sk_buff *skb;
4539         struct net *net = dev_net(idev->dev);
4540         int err = -ENOBUFS;
4541
4542         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4543         if (skb == NULL)
4544                 goto errout;
4545
4546         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4547         if (err < 0) {
4548                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4549                 WARN_ON(err == -EMSGSIZE);
4550                 kfree_skb(skb);
4551                 goto errout;
4552         }
4553         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4554         return;
4555 errout:
4556         if (err < 0)
4557                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4558 }
4559
4560 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4561 {
4562         write_lock_bh(&ifp->idev->lock);
4563         spin_lock(&ifp->lock);
4564         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4565                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4566             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4567                 ifp->idev->valid_ll_addr_cnt += count;
4568         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4569         spin_unlock(&ifp->lock);
4570         write_unlock_bh(&ifp->idev->lock);
4571 }
4572
4573 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4574 {
4575         struct net *net = dev_net(ifp->idev->dev);
4576
4577         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4578
4579         switch (event) {
4580         case RTM_NEWADDR:
4581                 update_valid_ll_addr_cnt(ifp, 1);
4582
4583                 /*
4584                  * If the address was optimistic
4585                  * we inserted the route at the start of
4586                  * our DAD process, so we don't need
4587                  * to do it again
4588                  */
4589                 if (!(ifp->rt->rt6i_node))
4590                         ip6_ins_rt(ifp->rt);
4591                 if (ifp->idev->cnf.forwarding)
4592                         addrconf_join_anycast(ifp);
4593                 if (!ipv6_addr_any(&ifp->peer_addr))
4594                         addrconf_prefix_route(&ifp->peer_addr, 128,
4595                                               ifp->idev->dev, 0, 0);
4596                 break;
4597         case RTM_DELADDR:
4598                 update_valid_ll_addr_cnt(ifp, -1);
4599
4600                 if (ifp->idev->cnf.forwarding)
4601                         addrconf_leave_anycast(ifp);
4602                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4603                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4604                         struct rt6_info *rt;
4605                         struct net_device *dev = ifp->idev->dev;
4606
4607                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4608                                         dev->ifindex, 1);
4609                         if (rt) {
4610                                 dst_hold(&rt->dst);
4611                                 if (ip6_del_rt(rt))
4612                                         dst_free(&rt->dst);
4613                         }
4614                 }
4615                 dst_hold(&ifp->rt->dst);
4616
4617                 if (ip6_del_rt(ifp->rt))
4618                         dst_free(&ifp->rt->dst);
4619                 break;
4620         }
4621         atomic_inc(&net->ipv6.dev_addr_genid);
4622         rt_genid_bump_ipv6(net);
4623 }
4624
4625 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4626 {
4627         rcu_read_lock_bh();
4628         if (likely(ifp->idev->dead == 0))
4629                 __ipv6_ifa_notify(event, ifp);
4630         rcu_read_unlock_bh();
4631 }
4632
4633 #ifdef CONFIG_SYSCTL
4634
4635 static
4636 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4637                            void __user *buffer, size_t *lenp, loff_t *ppos)
4638 {
4639         int *valp = ctl->data;
4640         int val = *valp;
4641         loff_t pos = *ppos;
4642         struct ctl_table lctl;
4643         int ret;
4644
4645         /*
4646          * ctl->data points to idev->cnf.forwarding, we should
4647          * not modify it until we get the rtnl lock.
4648          */
4649         lctl = *ctl;
4650         lctl.data = &val;
4651
4652         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4653
4654         if (write)
4655                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4656         if (ret)
4657                 *ppos = pos;
4658         return ret;
4659 }
4660
4661 static void dev_disable_change(struct inet6_dev *idev)
4662 {
4663         struct netdev_notifier_info info;
4664
4665         if (!idev || !idev->dev)
4666                 return;
4667
4668         netdev_notifier_info_init(&info, idev->dev);
4669         if (idev->cnf.disable_ipv6)
4670                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4671         else
4672                 addrconf_notify(NULL, NETDEV_UP, &info);
4673 }
4674
4675 static void addrconf_disable_change(struct net *net, __s32 newf)
4676 {
4677         struct net_device *dev;
4678         struct inet6_dev *idev;
4679
4680         rcu_read_lock();
4681         for_each_netdev_rcu(net, dev) {
4682                 idev = __in6_dev_get(dev);
4683                 if (idev) {
4684                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4685                         idev->cnf.disable_ipv6 = newf;
4686                         if (changed)
4687                                 dev_disable_change(idev);
4688                 }
4689         }
4690         rcu_read_unlock();
4691 }
4692
4693 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4694 {
4695         struct net *net;
4696         int old;
4697
4698         if (!rtnl_trylock())
4699                 return restart_syscall();
4700
4701         net = (struct net *)table->extra2;
4702         old = *p;
4703         *p = newf;
4704
4705         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4706                 rtnl_unlock();
4707                 return 0;
4708         }
4709
4710         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4711                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4712                 addrconf_disable_change(net, newf);
4713         } else if ((!newf) ^ (!old))
4714                 dev_disable_change((struct inet6_dev *)table->extra1);
4715
4716         rtnl_unlock();
4717         return 0;
4718 }
4719
4720 static
4721 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4722                             void __user *buffer, size_t *lenp, loff_t *ppos)
4723 {
4724         int *valp = ctl->data;
4725         int val = *valp;
4726         loff_t pos = *ppos;
4727         struct ctl_table lctl;
4728         int ret;
4729
4730         /*
4731          * ctl->data points to idev->cnf.disable_ipv6, we should
4732          * not modify it until we get the rtnl lock.
4733          */
4734         lctl = *ctl;
4735         lctl.data = &val;
4736
4737         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4738
4739         if (write)
4740                 ret = addrconf_disable_ipv6(ctl, valp, val);
4741         if (ret)
4742                 *ppos = pos;
4743         return ret;
4744 }
4745
4746 static
4747 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4748                               void __user *buffer, size_t *lenp, loff_t *ppos)
4749 {
4750         int *valp = ctl->data;
4751         int ret;
4752         int old, new;
4753
4754         old = *valp;
4755         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4756         new = *valp;
4757
4758         if (write && old != new) {
4759                 struct net *net = ctl->extra2;
4760
4761                 if (!rtnl_trylock())
4762                         return restart_syscall();
4763
4764                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4765                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4766                                                      NETCONFA_IFINDEX_DEFAULT,
4767                                                      net->ipv6.devconf_dflt);
4768                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4769                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4770                                                      NETCONFA_IFINDEX_ALL,
4771                                                      net->ipv6.devconf_all);
4772                 else {
4773                         struct inet6_dev *idev = ctl->extra1;
4774
4775                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4776                                                      idev->dev->ifindex,
4777                                                      &idev->cnf);
4778                 }
4779                 rtnl_unlock();
4780         }
4781
4782         return ret;
4783 }
4784
4785
4786 static struct addrconf_sysctl_table
4787 {
4788         struct ctl_table_header *sysctl_header;
4789         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4790 } addrconf_sysctl __read_mostly = {
4791         .sysctl_header = NULL,
4792         .addrconf_vars = {
4793                 {
4794                         .procname       = "forwarding",
4795                         .data           = &ipv6_devconf.forwarding,
4796                         .maxlen         = sizeof(int),
4797                         .mode           = 0644,
4798                         .proc_handler   = addrconf_sysctl_forward,
4799                 },
4800                 {
4801                         .procname       = "hop_limit",
4802                         .data           = &ipv6_devconf.hop_limit,
4803                         .maxlen         = sizeof(int),
4804                         .mode           = 0644,
4805                         .proc_handler   = proc_dointvec,
4806                 },
4807                 {
4808                         .procname       = "mtu",
4809                         .data           = &ipv6_devconf.mtu6,
4810                         .maxlen         = sizeof(int),
4811                         .mode           = 0644,
4812                         .proc_handler   = proc_dointvec,
4813                 },
4814                 {
4815                         .procname       = "accept_ra",
4816                         .data           = &ipv6_devconf.accept_ra,
4817                         .maxlen         = sizeof(int),
4818                         .mode           = 0644,
4819                         .proc_handler   = proc_dointvec,
4820                 },
4821                 {
4822                         .procname       = "accept_redirects",
4823                         .data           = &ipv6_devconf.accept_redirects,
4824                         .maxlen         = sizeof(int),
4825                         .mode           = 0644,
4826                         .proc_handler   = proc_dointvec,
4827                 },
4828                 {
4829                         .procname       = "autoconf",
4830                         .data           = &ipv6_devconf.autoconf,
4831                         .maxlen         = sizeof(int),
4832                         .mode           = 0644,
4833                         .proc_handler   = proc_dointvec,
4834                 },
4835                 {
4836                         .procname       = "dad_transmits",
4837                         .data           = &ipv6_devconf.dad_transmits,
4838                         .maxlen         = sizeof(int),
4839                         .mode           = 0644,
4840                         .proc_handler   = proc_dointvec,
4841                 },
4842                 {
4843                         .procname       = "router_solicitations",
4844                         .data           = &ipv6_devconf.rtr_solicits,
4845                         .maxlen         = sizeof(int),
4846                         .mode           = 0644,
4847                         .proc_handler   = proc_dointvec,
4848                 },
4849                 {
4850                         .procname       = "router_solicitation_interval",
4851                         .data           = &ipv6_devconf.rtr_solicit_interval,
4852                         .maxlen         = sizeof(int),
4853                         .mode           = 0644,
4854                         .proc_handler   = proc_dointvec_jiffies,
4855                 },
4856                 {
4857                         .procname       = "router_solicitation_delay",
4858                         .data           = &ipv6_devconf.rtr_solicit_delay,
4859                         .maxlen         = sizeof(int),
4860                         .mode           = 0644,
4861                         .proc_handler   = proc_dointvec_jiffies,
4862                 },
4863                 {
4864                         .procname       = "force_mld_version",
4865                         .data           = &ipv6_devconf.force_mld_version,
4866                         .maxlen         = sizeof(int),
4867                         .mode           = 0644,
4868                         .proc_handler   = proc_dointvec,
4869                 },
4870                 {
4871                         .procname       = "mldv1_unsolicited_report_interval",
4872                         .data           =
4873                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4874                         .maxlen         = sizeof(int),
4875                         .mode           = 0644,
4876                         .proc_handler   = proc_dointvec_ms_jiffies,
4877                 },
4878                 {
4879                         .procname       = "mldv2_unsolicited_report_interval",
4880                         .data           =
4881                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4882                         .maxlen         = sizeof(int),
4883                         .mode           = 0644,
4884                         .proc_handler   = proc_dointvec_ms_jiffies,
4885                 },
4886                 {
4887                         .procname       = "use_tempaddr",
4888                         .data           = &ipv6_devconf.use_tempaddr,
4889                         .maxlen         = sizeof(int),
4890                         .mode           = 0644,
4891                         .proc_handler   = proc_dointvec,
4892                 },
4893                 {
4894                         .procname       = "temp_valid_lft",
4895                         .data           = &ipv6_devconf.temp_valid_lft,
4896                         .maxlen         = sizeof(int),
4897                         .mode           = 0644,
4898                         .proc_handler   = proc_dointvec,
4899                 },
4900                 {
4901                         .procname       = "temp_prefered_lft",
4902                         .data           = &ipv6_devconf.temp_prefered_lft,
4903                         .maxlen         = sizeof(int),
4904                         .mode           = 0644,
4905                         .proc_handler   = proc_dointvec,
4906                 },
4907                 {
4908                         .procname       = "regen_max_retry",
4909                         .data           = &ipv6_devconf.regen_max_retry,
4910                         .maxlen         = sizeof(int),
4911                         .mode           = 0644,
4912                         .proc_handler   = proc_dointvec,
4913                 },
4914                 {
4915                         .procname       = "max_desync_factor",
4916                         .data           = &ipv6_devconf.max_desync_factor,
4917                         .maxlen         = sizeof(int),
4918                         .mode           = 0644,
4919                         .proc_handler   = proc_dointvec,
4920                 },
4921                 {
4922                         .procname       = "max_addresses",
4923                         .data           = &ipv6_devconf.max_addresses,
4924                         .maxlen         = sizeof(int),
4925                         .mode           = 0644,
4926                         .proc_handler   = proc_dointvec,
4927                 },
4928                 {
4929                         .procname       = "accept_ra_defrtr",
4930                         .data           = &ipv6_devconf.accept_ra_defrtr,
4931                         .maxlen         = sizeof(int),
4932                         .mode           = 0644,
4933                         .proc_handler   = proc_dointvec,
4934                 },
4935                 {
4936                         .procname       = "accept_ra_pinfo",
4937                         .data           = &ipv6_devconf.accept_ra_pinfo,
4938                         .maxlen         = sizeof(int),
4939                         .mode           = 0644,
4940                         .proc_handler   = proc_dointvec,
4941                 },
4942 #ifdef CONFIG_IPV6_ROUTER_PREF
4943                 {
4944                         .procname       = "accept_ra_rtr_pref",
4945                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4946                         .maxlen         = sizeof(int),
4947                         .mode           = 0644,
4948                         .proc_handler   = proc_dointvec,
4949                 },
4950                 {
4951                         .procname       = "router_probe_interval",
4952                         .data           = &ipv6_devconf.rtr_probe_interval,
4953                         .maxlen         = sizeof(int),
4954                         .mode           = 0644,
4955                         .proc_handler   = proc_dointvec_jiffies,
4956                 },
4957 #ifdef CONFIG_IPV6_ROUTE_INFO
4958                 {
4959                         .procname       = "accept_ra_rt_info_max_plen",
4960                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4961                         .maxlen         = sizeof(int),
4962                         .mode           = 0644,
4963                         .proc_handler   = proc_dointvec,
4964                 },
4965 #endif
4966 #endif
4967                 {
4968                         .procname       = "proxy_ndp",
4969                         .data           = &ipv6_devconf.proxy_ndp,
4970                         .maxlen         = sizeof(int),
4971                         .mode           = 0644,
4972                         .proc_handler   = addrconf_sysctl_proxy_ndp,
4973                 },
4974                 {
4975                         .procname       = "accept_source_route",
4976                         .data           = &ipv6_devconf.accept_source_route,
4977                         .maxlen         = sizeof(int),
4978                         .mode           = 0644,
4979                         .proc_handler   = proc_dointvec,
4980                 },
4981 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4982                 {
4983                         .procname       = "optimistic_dad",
4984                         .data           = &ipv6_devconf.optimistic_dad,
4985                         .maxlen         = sizeof(int),
4986                         .mode           = 0644,
4987                         .proc_handler   = proc_dointvec,
4988
4989                 },
4990 #endif
4991 #ifdef CONFIG_IPV6_MROUTE
4992                 {
4993                         .procname       = "mc_forwarding",
4994                         .data           = &ipv6_devconf.mc_forwarding,
4995                         .maxlen         = sizeof(int),
4996                         .mode           = 0444,
4997                         .proc_handler   = proc_dointvec,
4998                 },
4999 #endif
5000                 {
5001                         .procname       = "disable_ipv6",
5002                         .data           = &ipv6_devconf.disable_ipv6,
5003                         .maxlen         = sizeof(int),
5004                         .mode           = 0644,
5005                         .proc_handler   = addrconf_sysctl_disable,
5006                 },
5007                 {
5008                         .procname       = "accept_dad",
5009                         .data           = &ipv6_devconf.accept_dad,
5010                         .maxlen         = sizeof(int),
5011                         .mode           = 0644,
5012                         .proc_handler   = proc_dointvec,
5013                 },
5014                 {
5015                         .procname       = "force_tllao",
5016                         .data           = &ipv6_devconf.force_tllao,
5017                         .maxlen         = sizeof(int),
5018                         .mode           = 0644,
5019                         .proc_handler   = proc_dointvec
5020                 },
5021                 {
5022                         .procname       = "ndisc_notify",
5023                         .data           = &ipv6_devconf.ndisc_notify,
5024                         .maxlen         = sizeof(int),
5025                         .mode           = 0644,
5026                         .proc_handler   = proc_dointvec
5027                 },
5028                 {
5029                         .procname       = "suppress_frag_ndisc",
5030                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5031                         .maxlen         = sizeof(int),
5032                         .mode           = 0644,
5033                         .proc_handler   = proc_dointvec
5034                 },
5035                 {
5036                         /* sentinel */
5037                 }
5038         },
5039 };
5040
5041 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5042                 struct inet6_dev *idev, struct ipv6_devconf *p)
5043 {
5044         int i;
5045         struct addrconf_sysctl_table *t;
5046         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5047
5048         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5049         if (t == NULL)
5050                 goto out;
5051
5052         for (i = 0; t->addrconf_vars[i].data; i++) {
5053                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5054                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5055                 t->addrconf_vars[i].extra2 = net;
5056         }
5057
5058         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5059
5060         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5061         if (t->sysctl_header == NULL)
5062                 goto free;
5063
5064         p->sysctl = t;
5065         return 0;
5066
5067 free:
5068         kfree(t);
5069 out:
5070         return -ENOBUFS;
5071 }
5072
5073 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5074 {
5075         struct addrconf_sysctl_table *t;
5076
5077         if (p->sysctl == NULL)
5078                 return;
5079
5080         t = p->sysctl;
5081         p->sysctl = NULL;
5082         unregister_net_sysctl_table(t->sysctl_header);
5083         kfree(t);
5084 }
5085
5086 static void addrconf_sysctl_register(struct inet6_dev *idev)
5087 {
5088         neigh_sysctl_register(idev->dev, idev->nd_parms,
5089                               &ndisc_ifinfo_sysctl_change);
5090         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5091                                         idev, &idev->cnf);
5092 }
5093
5094 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5095 {
5096         __addrconf_sysctl_unregister(&idev->cnf);
5097         neigh_sysctl_unregister(idev->nd_parms);
5098 }
5099
5100
5101 #endif
5102
5103 static int __net_init addrconf_init_net(struct net *net)
5104 {
5105         int err = -ENOMEM;
5106         struct ipv6_devconf *all, *dflt;
5107
5108         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5109         if (all == NULL)
5110                 goto err_alloc_all;
5111
5112         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5113         if (dflt == NULL)
5114                 goto err_alloc_dflt;
5115
5116         /* these will be inherited by all namespaces */
5117         dflt->autoconf = ipv6_defaults.autoconf;
5118         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5119
5120         net->ipv6.devconf_all = all;
5121         net->ipv6.devconf_dflt = dflt;
5122
5123 #ifdef CONFIG_SYSCTL
5124         err = __addrconf_sysctl_register(net, "all", NULL, all);
5125         if (err < 0)
5126                 goto err_reg_all;
5127
5128         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5129         if (err < 0)
5130                 goto err_reg_dflt;
5131 #endif
5132         return 0;
5133
5134 #ifdef CONFIG_SYSCTL
5135 err_reg_dflt:
5136         __addrconf_sysctl_unregister(all);
5137 err_reg_all:
5138         kfree(dflt);
5139 #endif
5140 err_alloc_dflt:
5141         kfree(all);
5142 err_alloc_all:
5143         return err;
5144 }
5145
5146 static void __net_exit addrconf_exit_net(struct net *net)
5147 {
5148 #ifdef CONFIG_SYSCTL
5149         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5150         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5151 #endif
5152         if (!net_eq(net, &init_net)) {
5153                 kfree(net->ipv6.devconf_dflt);
5154                 kfree(net->ipv6.devconf_all);
5155         }
5156 }
5157
5158 static struct pernet_operations addrconf_ops = {
5159         .init = addrconf_init_net,
5160         .exit = addrconf_exit_net,
5161 };
5162
5163 static struct rtnl_af_ops inet6_ops = {
5164         .family           = AF_INET6,
5165         .fill_link_af     = inet6_fill_link_af,
5166         .get_link_af_size = inet6_get_link_af_size,
5167         .set_link_af      = inet6_set_link_af,
5168 };
5169
5170 /*
5171  *      Init / cleanup code
5172  */
5173
5174 int __init addrconf_init(void)
5175 {
5176         int i, err;
5177
5178         err = ipv6_addr_label_init();
5179         if (err < 0) {
5180                 pr_crit("%s: cannot initialize default policy table: %d\n",
5181                         __func__, err);
5182                 goto out;
5183         }
5184
5185         err = register_pernet_subsys(&addrconf_ops);
5186         if (err < 0)
5187                 goto out_addrlabel;
5188
5189         /* The addrconf netdev notifier requires that loopback_dev
5190          * has it's ipv6 private information allocated and setup
5191          * before it can bring up and give link-local addresses
5192          * to other devices which are up.
5193          *
5194          * Unfortunately, loopback_dev is not necessarily the first
5195          * entry in the global dev_base list of net devices.  In fact,
5196          * it is likely to be the very last entry on that list.
5197          * So this causes the notifier registry below to try and
5198          * give link-local addresses to all devices besides loopback_dev
5199          * first, then loopback_dev, which cases all the non-loopback_dev
5200          * devices to fail to get a link-local address.
5201          *
5202          * So, as a temporary fix, allocate the ipv6 structure for
5203          * loopback_dev first by hand.
5204          * Longer term, all of the dependencies ipv6 has upon the loopback
5205          * device and it being up should be removed.
5206          */
5207         rtnl_lock();
5208         if (!ipv6_add_dev(init_net.loopback_dev))
5209                 err = -ENOMEM;
5210         rtnl_unlock();
5211         if (err)
5212                 goto errlo;
5213
5214         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5215                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5216
5217         register_netdevice_notifier(&ipv6_dev_notf);
5218
5219         addrconf_verify(0);
5220
5221         rtnl_af_register(&inet6_ops);
5222
5223         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5224                               NULL);
5225         if (err < 0)
5226                 goto errout;
5227
5228         /* Only the first call to __rtnl_register can fail */
5229         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5230         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5231         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5232                         inet6_dump_ifaddr, NULL);
5233         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5234                         inet6_dump_ifmcaddr, NULL);
5235         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5236                         inet6_dump_ifacaddr, NULL);
5237         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5238                         inet6_netconf_dump_devconf, NULL);
5239
5240         ipv6_addr_label_rtnl_register();
5241
5242         return 0;
5243 errout:
5244         rtnl_af_unregister(&inet6_ops);
5245         unregister_netdevice_notifier(&ipv6_dev_notf);
5246 errlo:
5247         unregister_pernet_subsys(&addrconf_ops);
5248 out_addrlabel:
5249         ipv6_addr_label_cleanup();
5250 out:
5251         return err;
5252 }
5253
5254 void addrconf_cleanup(void)
5255 {
5256         struct net_device *dev;
5257         int i;
5258
5259         unregister_netdevice_notifier(&ipv6_dev_notf);
5260         unregister_pernet_subsys(&addrconf_ops);
5261         ipv6_addr_label_cleanup();
5262
5263         rtnl_lock();
5264
5265         __rtnl_af_unregister(&inet6_ops);
5266
5267         /* clean dev list */
5268         for_each_netdev(&init_net, dev) {
5269                 if (__in6_dev_get(dev) == NULL)
5270                         continue;
5271                 addrconf_ifdown(dev, 1);
5272         }
5273         addrconf_ifdown(init_net.loopback_dev, 2);
5274
5275         /*
5276          *      Check hash table.
5277          */
5278         spin_lock_bh(&addrconf_hash_lock);
5279         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5280                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5281         spin_unlock_bh(&addrconf_hash_lock);
5282
5283         del_timer(&addr_chk_timer);
5284         rtnl_unlock();
5285 }