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 int addrconf_sysctl_register(struct inet6_dev *idev);
 112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 113 #else
 114 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 115 {
 116         return 0;
 117 }
 118
 119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 120 {
 121 }
 122 #endif
 123
 124 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 125 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 126 static void ipv6_regen_rndid(unsigned long data);
 127
 128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 129 static int ipv6_count_addresses(struct inet6_dev *idev);
 130
 131 /*
 132  *      Configured unicast address hash table
 133  */
 134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 135 static DEFINE_SPINLOCK(addrconf_hash_lock);
 136
 137 static void addrconf_verify(void);
 138 static void addrconf_verify_rtnl(void);
 139 static void addrconf_verify_work(struct work_struct *);
 140
 141 static struct workqueue_struct *addrconf_wq;
 142 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 143
 144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 146
 147 static void addrconf_type_change(struct net_device *dev,
 148                                  unsigned long event);
 149 static int addrconf_ifdown(struct net_device *dev, int how);
 150
 151 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 152                                                   int plen,
 153                                                   const struct net_device *dev,
 154                                                   u32 flags, u32 noflags);
 155
 156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 157 static void addrconf_dad_work(struct work_struct *w);
 158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 159 static void addrconf_dad_run(struct inet6_dev *idev);
 160 static void addrconf_rs_timer(unsigned long data);
 161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 163
 164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 165                                 struct prefix_info *pinfo);
 166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 167                                struct net_device *dev);
 168
 169 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 170         .forwarding             = 0,
 171         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 172         .mtu6                   = IPV6_MIN_MTU,
 173         .accept_ra              = 1,
 174         .accept_redirects       = 1,
 175         .autoconf               = 1,
 176         .force_mld_version      = 0,
 177         .mldv1_unsolicited_report_interval = 10 * HZ,
 178         .mldv2_unsolicited_report_interval = HZ,
 179         .dad_transmits          = 1,
 180         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 181         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 182         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 183         .use_tempaddr           = 0,
 184         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 185         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 186         .regen_max_retry        = REGEN_MAX_RETRY,
 187         .max_desync_factor      = MAX_DESYNC_FACTOR,
 188         .max_addresses          = IPV6_MAX_ADDRESSES,
 189         .accept_ra_defrtr       = 1,
 190         .accept_ra_from_local   = 0,
 191         .accept_ra_pinfo        = 1,
 192 #ifdef CONFIG_IPV6_ROUTER_PREF
 193         .accept_ra_rtr_pref     = 1,
 194         .rtr_probe_interval     = 60 * HZ,
 195 #ifdef CONFIG_IPV6_ROUTE_INFO
 196         .accept_ra_rt_info_max_plen = 0,
 197 #endif
 198 #endif
 199         .proxy_ndp              = 0,
 200         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 201         .disable_ipv6           = 0,
 202         .accept_dad             = 1,
 203         .suppress_frag_ndisc    = 1,
 204         .accept_ra_mtu          = 1,
 205 };
 206
 207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 208         .forwarding             = 0,
 209         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 210         .mtu6                   = IPV6_MIN_MTU,
 211         .accept_ra              = 1,
 212         .accept_redirects       = 1,
 213         .autoconf               = 1,
 214         .force_mld_version      = 0,
 215         .mldv1_unsolicited_report_interval = 10 * HZ,
 216         .mldv2_unsolicited_report_interval = HZ,
 217         .dad_transmits          = 1,
 218         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 219         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 220         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 221         .use_tempaddr           = 0,
 222         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 223         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 224         .regen_max_retry        = REGEN_MAX_RETRY,
 225         .max_desync_factor      = MAX_DESYNC_FACTOR,
 226         .max_addresses          = IPV6_MAX_ADDRESSES,
 227         .accept_ra_defrtr       = 1,
 228         .accept_ra_from_local   = 0,
 229         .accept_ra_pinfo        = 1,
 230 #ifdef CONFIG_IPV6_ROUTER_PREF
 231         .accept_ra_rtr_pref     = 1,
 232         .rtr_probe_interval     = 60 * HZ,
 233 #ifdef CONFIG_IPV6_ROUTE_INFO
 234         .accept_ra_rt_info_max_plen = 0,
 235 #endif
 236 #endif
 237         .proxy_ndp              = 0,
 238         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 239         .disable_ipv6           = 0,
 240         .accept_dad             = 1,
 241         .suppress_frag_ndisc    = 1,
 242         .accept_ra_mtu          = 1,
 243 };
 244
 245 /* Check if a valid qdisc is available */
 246 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 247 {
 248         return !qdisc_tx_is_noop(dev);
 249 }
 250
 251 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 252 {
 253         if (del_timer(&idev->rs_timer))
 254                 __in6_dev_put(idev);
 255 }
 256
 257 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 258 {
 259         if (cancel_delayed_work(&ifp->dad_work))
 260                 __in6_ifa_put(ifp);
 261 }
 262
 263 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 264                                   unsigned long when)
 265 {
 266         if (!timer_pending(&idev->rs_timer))
 267                 in6_dev_hold(idev);
 268         mod_timer(&idev->rs_timer, jiffies + when);
 269 }
 270
 271 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 272                                    unsigned long delay)
 273 {
 274         if (!delayed_work_pending(&ifp->dad_work))
 275                 in6_ifa_hold(ifp);
 276         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 277 }
 278
 279 static int snmp6_alloc_dev(struct inet6_dev *idev)
 280 {
 281         int i;
 282
 283         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 284         if (!idev->stats.ipv6)
 285                 goto err_ip;
 286
 287         for_each_possible_cpu(i) {
 288                 struct ipstats_mib *addrconf_stats;
 289                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 290                 u64_stats_init(&addrconf_stats->syncp);
 291         }
 292
 293
 294         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 295                                         GFP_KERNEL);
 296         if (!idev->stats.icmpv6dev)
 297                 goto err_icmp;
 298         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 299                                            GFP_KERNEL);
 300         if (!idev->stats.icmpv6msgdev)
 301                 goto err_icmpmsg;
 302
 303         return 0;
 304
 305 err_icmpmsg:
 306         kfree(idev->stats.icmpv6dev);
 307 err_icmp:
 308         free_percpu(idev->stats.ipv6);
 309 err_ip:
 310         return -ENOMEM;
 311 }
 312
 313 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 314 {
 315         struct inet6_dev *ndev;
 316         int err = -ENOMEM;
 317
 318         ASSERT_RTNL();
 319
 320         if (dev->mtu < IPV6_MIN_MTU)
 321                 return ERR_PTR(-EINVAL);
 322
 323         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 324         if (ndev == NULL)
 325                 return ERR_PTR(err);
 326
 327         rwlock_init(&ndev->lock);
 328         ndev->dev = dev;
 329         INIT_LIST_HEAD(&ndev->addr_list);
 330         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 331                     (unsigned long)ndev);
 332         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 333         ndev->cnf.mtu6 = dev->mtu;
 334         ndev->cnf.sysctl = NULL;
 335         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 336         if (ndev->nd_parms == NULL) {
 337                 kfree(ndev);
 338                 return ERR_PTR(err);
 339         }
 340         if (ndev->cnf.forwarding)
 341                 dev_disable_lro(dev);
 342         /* We refer to the device */
 343         dev_hold(dev);
 344
 345         if (snmp6_alloc_dev(ndev) < 0) {
 346                 ADBG(KERN_WARNING
 347                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 348                         __func__, dev->name);
 349                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 350                 dev_put(dev);
 351                 kfree(ndev);
 352                 return ERR_PTR(err);
 353         }
 354
 355         if (snmp6_register_dev(ndev) < 0) {
 356                 ADBG(KERN_WARNING
 357                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 358                         __func__, dev->name);
 359                 goto err_release;
 360         }
 361
 362         /* One reference from device.  We must do this before
 363          * we invoke __ipv6_regen_rndid().
 364          */
 365         in6_dev_hold(ndev);
 366
 367         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 368                 ndev->cnf.accept_dad = -1;
 369
 370 #if IS_ENABLED(CONFIG_IPV6_SIT)
 371         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 372                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 373                 ndev->cnf.rtr_solicits = 0;
 374         }
 375 #endif
 376
 377         INIT_LIST_HEAD(&ndev->tempaddr_list);
 378         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 379         if ((dev->flags&IFF_LOOPBACK) ||
 380             dev->type == ARPHRD_TUNNEL ||
 381             dev->type == ARPHRD_TUNNEL6 ||
 382             dev->type == ARPHRD_SIT ||
 383             dev->type == ARPHRD_NONE) {
 384                 ndev->cnf.use_tempaddr = -1;
 385         } else {
 386                 in6_dev_hold(ndev);
 387                 ipv6_regen_rndid((unsigned long) ndev);
 388         }
 389
 390         ndev->token = in6addr_any;
 391
 392         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 393                 ndev->if_flags |= IF_READY;
 394
 395         ipv6_mc_init_dev(ndev);
 396         ndev->tstamp = jiffies;
 397         err = addrconf_sysctl_register(ndev);
 398         if (err) {
 399                 ipv6_mc_destroy_dev(ndev);
 400                 del_timer(&ndev->regen_timer);
 401                 goto err_release;
 402         }
 403         /* protected by rtnl_lock */
 404         rcu_assign_pointer(dev->ip6_ptr, ndev);
 405
 406         /* Join interface-local all-node multicast group */
 407         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 408
 409         /* Join all-node multicast group */
 410         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 411
 412         /* Join all-router multicast group if forwarding is set */
 413         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 414                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 415
 416         return ndev;
 417
 418 err_release:
 419         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 420         ndev->dead = 1;
 421         in6_dev_finish_destroy(ndev);
 422         return ERR_PTR(err);
 423 }
 424
 425 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 426 {
 427         struct inet6_dev *idev;
 428
 429         ASSERT_RTNL();
 430
 431         idev = __in6_dev_get(dev);
 432         if (!idev) {
 433                 idev = ipv6_add_dev(dev);
 434                 if (IS_ERR(idev))
 435                         return NULL;
 436         }
 437
 438         if (dev->flags&IFF_UP)
 439                 ipv6_mc_up(idev);
 440         return idev;
 441 }
 442
 443 static int inet6_netconf_msgsize_devconf(int type)
 444 {
 445         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 446                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 447
 448         /* type -1 is used for ALL */
 449         if (type == -1 || type == NETCONFA_FORWARDING)
 450                 size += nla_total_size(4);
 451 #ifdef CONFIG_IPV6_MROUTE
 452         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 453                 size += nla_total_size(4);
 454 #endif
 455         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 456                 size += nla_total_size(4);
 457
 458         return size;
 459 }
 460
 461 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 462                                       struct ipv6_devconf *devconf, u32 portid,
 463                                       u32 seq, int event, unsigned int flags,
 464                                       int type)
 465 {
 466         struct nlmsghdr  *nlh;
 467         struct netconfmsg *ncm;
 468
 469         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 470                         flags);
 471         if (nlh == NULL)
 472                 return -EMSGSIZE;
 473
 474         ncm = nlmsg_data(nlh);
 475         ncm->ncm_family = AF_INET6;
 476
 477         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 478                 goto nla_put_failure;
 479
 480         /* type -1 is used for ALL */
 481         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 482             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 483                 goto nla_put_failure;
 484 #ifdef CONFIG_IPV6_MROUTE
 485         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 486             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 487                         devconf->mc_forwarding) < 0)
 488                 goto nla_put_failure;
 489 #endif
 490         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 491             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 492                 goto nla_put_failure;
 493
 494         nlmsg_end(skb, nlh);
 495         return 0;
 496
 497 nla_put_failure:
 498         nlmsg_cancel(skb, nlh);
 499         return -EMSGSIZE;
 500 }
 501
 502 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 503                                   struct ipv6_devconf *devconf)
 504 {
 505         struct sk_buff *skb;
 506         int err = -ENOBUFS;
 507
 508         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 509         if (skb == NULL)
 510                 goto errout;
 511
 512         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 513                                          RTM_NEWNETCONF, 0, type);
 514         if (err < 0) {
 515                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 516                 WARN_ON(err == -EMSGSIZE);
 517                 kfree_skb(skb);
 518                 goto errout;
 519         }
 520         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 521         return;
 522 errout:
 523         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 524 }
 525
 526 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 527         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 528         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 529         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 530 };
 531
 532 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 533                                      struct nlmsghdr *nlh)
 534 {
 535         struct net *net = sock_net(in_skb->sk);
 536         struct nlattr *tb[NETCONFA_MAX+1];
 537         struct netconfmsg *ncm;
 538         struct sk_buff *skb;
 539         struct ipv6_devconf *devconf;
 540         struct inet6_dev *in6_dev;
 541         struct net_device *dev;
 542         int ifindex;
 543         int err;
 544
 545         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 546                           devconf_ipv6_policy);
 547         if (err < 0)
 548                 goto errout;
 549
 550         err = EINVAL;
 551         if (!tb[NETCONFA_IFINDEX])
 552                 goto errout;
 553
 554         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 555         switch (ifindex) {
 556         case NETCONFA_IFINDEX_ALL:
 557                 devconf = net->ipv6.devconf_all;
 558                 break;
 559         case NETCONFA_IFINDEX_DEFAULT:
 560                 devconf = net->ipv6.devconf_dflt;
 561                 break;
 562         default:
 563                 dev = __dev_get_by_index(net, ifindex);
 564                 if (dev == NULL)
 565                         goto errout;
 566                 in6_dev = __in6_dev_get(dev);
 567                 if (in6_dev == NULL)
 568                         goto errout;
 569                 devconf = &in6_dev->cnf;
 570                 break;
 571         }
 572
 573         err = -ENOBUFS;
 574         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 575         if (skb == NULL)
 576                 goto errout;
 577
 578         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 579                                          NETLINK_CB(in_skb).portid,
 580                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 581                                          -1);
 582         if (err < 0) {
 583                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 584                 WARN_ON(err == -EMSGSIZE);
 585                 kfree_skb(skb);
 586                 goto errout;
 587         }
 588         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 589 errout:
 590         return err;
 591 }
 592
 593 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 594                                       struct netlink_callback *cb)
 595 {
 596         struct net *net = sock_net(skb->sk);
 597         int h, s_h;
 598         int idx, s_idx;
 599         struct net_device *dev;
 600         struct inet6_dev *idev;
 601         struct hlist_head *head;
 602
 603         s_h = cb->args[0];
 604         s_idx = idx = cb->args[1];
 605
 606         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 607                 idx = 0;
 608                 head = &net->dev_index_head[h];
 609                 rcu_read_lock();
 610                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 611                           net->dev_base_seq;
 612                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 613                         if (idx < s_idx)
 614                                 goto cont;
 615                         idev = __in6_dev_get(dev);
 616                         if (!idev)
 617                                 goto cont;
 618
 619                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 620                                                        &idev->cnf,
 621                                                        NETLINK_CB(cb->skb).portid,
 622                                                        cb->nlh->nlmsg_seq,
 623                                                        RTM_NEWNETCONF,
 624                                                        NLM_F_MULTI,
 625                                                        -1) < 0) {
 626                                 rcu_read_unlock();
 627                                 goto done;
 628                         }
 629                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 630 cont:
 631                         idx++;
 632                 }
 633                 rcu_read_unlock();
 634         }
 635         if (h == NETDEV_HASHENTRIES) {
 636                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 637                                                net->ipv6.devconf_all,
 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         if (h == NETDEV_HASHENTRIES + 1) {
 647                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 648                                                net->ipv6.devconf_dflt,
 649                                                NETLINK_CB(cb->skb).portid,
 650                                                cb->nlh->nlmsg_seq,
 651                                                RTM_NEWNETCONF, NLM_F_MULTI,
 652                                                -1) < 0)
 653                         goto done;
 654                 else
 655                         h++;
 656         }
 657 done:
 658         cb->args[0] = h;
 659         cb->args[1] = idx;
 660
 661         return skb->len;
 662 }
 663
 664 #ifdef CONFIG_SYSCTL
 665 static void dev_forward_change(struct inet6_dev *idev)
 666 {
 667         struct net_device *dev;
 668         struct inet6_ifaddr *ifa;
 669
 670         if (!idev)
 671                 return;
 672         dev = idev->dev;
 673         if (idev->cnf.forwarding)
 674                 dev_disable_lro(dev);
 675         if (dev->flags & IFF_MULTICAST) {
 676                 if (idev->cnf.forwarding) {
 677                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 678                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 679                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 680                 } else {
 681                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 682                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 683                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 684                 }
 685         }
 686
 687         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 688                 if (ifa->flags&IFA_F_TENTATIVE)
 689                         continue;
 690                 if (idev->cnf.forwarding)
 691                         addrconf_join_anycast(ifa);
 692                 else
 693                         addrconf_leave_anycast(ifa);
 694         }
 695         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 696                                      dev->ifindex, &idev->cnf);
 697 }
 698
 699
 700 static void addrconf_forward_change(struct net *net, __s32 newf)
 701 {
 702         struct net_device *dev;
 703         struct inet6_dev *idev;
 704
 705         for_each_netdev(net, dev) {
 706                 idev = __in6_dev_get(dev);
 707                 if (idev) {
 708                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 709                         idev->cnf.forwarding = newf;
 710                         if (changed)
 711                                 dev_forward_change(idev);
 712                 }
 713         }
 714 }
 715
 716 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 717 {
 718         struct net *net;
 719         int old;
 720
 721         if (!rtnl_trylock())
 722                 return restart_syscall();
 723
 724         net = (struct net *)table->extra2;
 725         old = *p;
 726         *p = newf;
 727
 728         if (p == &net->ipv6.devconf_dflt->forwarding) {
 729                 if ((!newf) ^ (!old))
 730                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 731                                                      NETCONFA_IFINDEX_DEFAULT,
 732                                                      net->ipv6.devconf_dflt);
 733                 rtnl_unlock();
 734                 return 0;
 735         }
 736
 737         if (p == &net->ipv6.devconf_all->forwarding) {
 738                 net->ipv6.devconf_dflt->forwarding = newf;
 739                 addrconf_forward_change(net, newf);
 740                 if ((!newf) ^ (!old))
 741                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 742                                                      NETCONFA_IFINDEX_ALL,
 743                                                      net->ipv6.devconf_all);
 744         } else if ((!newf) ^ (!old))
 745                 dev_forward_change((struct inet6_dev *)table->extra1);
 746         rtnl_unlock();
 747
 748         if (newf)
 749                 rt6_purge_dflt_routers(net);
 750         return 1;
 751 }
 752 #endif
 753
 754 /* Nobody refers to this ifaddr, destroy it */
 755 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 756 {
 757         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 758
 759 #ifdef NET_REFCNT_DEBUG
 760         pr_debug("%s\n", __func__);
 761 #endif
 762
 763         in6_dev_put(ifp->idev);
 764
 765         if (cancel_delayed_work(&ifp->dad_work))
 766                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 767                           ifp);
 768
 769         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 770                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 771                 return;
 772         }
 773         ip6_rt_put(ifp->rt);
 774
 775         kfree_rcu(ifp, rcu);
 776 }
 777
 778 static void
 779 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 780 {
 781         struct list_head *p;
 782         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 783
 784         /*
 785          * Each device address list is sorted in order of scope -
 786          * global before linklocal.
 787          */
 788         list_for_each(p, &idev->addr_list) {
 789                 struct inet6_ifaddr *ifa
 790                         = list_entry(p, struct inet6_ifaddr, if_list);
 791                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 792                         break;
 793         }
 794
 795         list_add_tail(&ifp->if_list, p);
 796 }
 797
 798 static u32 inet6_addr_hash(const struct in6_addr *addr)
 799 {
 800         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 801 }
 802
 803 /* On success it returns ifp with increased reference count */
 804
 805 static struct inet6_ifaddr *
 806 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 807               const struct in6_addr *peer_addr, int pfxlen,
 808               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 809 {
 810         struct inet6_ifaddr *ifa = NULL;
 811         struct rt6_info *rt;
 812         unsigned int hash;
 813         int err = 0;
 814         int addr_type = ipv6_addr_type(addr);
 815
 816         if (addr_type == IPV6_ADDR_ANY ||
 817             addr_type & IPV6_ADDR_MULTICAST ||
 818             (!(idev->dev->flags & IFF_LOOPBACK) &&
 819              addr_type & IPV6_ADDR_LOOPBACK))
 820                 return ERR_PTR(-EADDRNOTAVAIL);
 821
 822         rcu_read_lock_bh();
 823         if (idev->dead) {
 824                 err = -ENODEV;                  /*XXX*/
 825                 goto out2;
 826         }
 827
 828         if (idev->cnf.disable_ipv6) {
 829                 err = -EACCES;
 830                 goto out2;
 831         }
 832
 833         spin_lock(&addrconf_hash_lock);
 834
 835         /* Ignore adding duplicate addresses on an interface */
 836         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 837                 ADBG("ipv6_add_addr: already assigned\n");
 838                 err = -EEXIST;
 839                 goto out;
 840         }
 841
 842         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 843
 844         if (ifa == NULL) {
 845                 ADBG("ipv6_add_addr: malloc failed\n");
 846                 err = -ENOBUFS;
 847                 goto out;
 848         }
 849
 850         rt = addrconf_dst_alloc(idev, addr, false);
 851         if (IS_ERR(rt)) {
 852                 err = PTR_ERR(rt);
 853                 goto out;
 854         }
 855
 856         neigh_parms_data_state_setall(idev->nd_parms);
 857
 858         ifa->addr = *addr;
 859         if (peer_addr)
 860                 ifa->peer_addr = *peer_addr;
 861
 862         spin_lock_init(&ifa->lock);
 863         spin_lock_init(&ifa->state_lock);
 864         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 865         INIT_HLIST_NODE(&ifa->addr_lst);
 866         ifa->scope = scope;
 867         ifa->prefix_len = pfxlen;
 868         ifa->flags = flags | IFA_F_TENTATIVE;
 869         ifa->valid_lft = valid_lft;
 870         ifa->prefered_lft = prefered_lft;
 871         ifa->cstamp = ifa->tstamp = jiffies;
 872         ifa->tokenized = false;
 873
 874         ifa->rt = rt;
 875
 876         ifa->idev = idev;
 877         in6_dev_hold(idev);
 878         /* For caller */
 879         in6_ifa_hold(ifa);
 880
 881         /* Add to big hash table */
 882         hash = inet6_addr_hash(addr);
 883
 884         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 885         spin_unlock(&addrconf_hash_lock);
 886
 887         write_lock(&idev->lock);
 888         /* Add to inet6_dev unicast addr list. */
 889         ipv6_link_dev_addr(idev, ifa);
 890
 891         if (ifa->flags&IFA_F_TEMPORARY) {
 892                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 893                 in6_ifa_hold(ifa);
 894         }
 895
 896         in6_ifa_hold(ifa);
 897         write_unlock(&idev->lock);
 898 out2:
 899         rcu_read_unlock_bh();
 900
 901         if (likely(err == 0))
 902                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 903         else {
 904                 kfree(ifa);
 905                 ifa = ERR_PTR(err);
 906         }
 907
 908         return ifa;
 909 out:
 910         spin_unlock(&addrconf_hash_lock);
 911         goto out2;
 912 }
 913
 914 enum cleanup_prefix_rt_t {
 915         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
 916         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
 917         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
 918 };
 919
 920 /*
 921  * Check, whether the prefix for ifp would still need a prefix route
 922  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
 923  * constants.
 924  *
 925  * 1) we don't purge prefix if address was not permanent.
 926  *    prefix is managed by its own lifetime.
 927  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
 928  * 3) if there are no addresses, delete prefix.
 929  * 4) if there are still other permanent address(es),
 930  *    corresponding prefix is still permanent.
 931  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
 932  *    don't purge the prefix, assume user space is managing it.
 933  * 6) otherwise, update prefix lifetime to the
 934  *    longest valid lifetime among the corresponding
 935  *    addresses on the device.
 936  *    Note: subsequent RA will update lifetime.
 937  **/
 938 static enum cleanup_prefix_rt_t
 939 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
 940 {
 941         struct inet6_ifaddr *ifa;
 942         struct inet6_dev *idev = ifp->idev;
 943         unsigned long lifetime;
 944         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
 945
 946         *expires = jiffies;
 947
 948         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 949                 if (ifa == ifp)
 950                         continue;
 951                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 952                                        ifp->prefix_len))
 953                         continue;
 954                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
 955                         return CLEANUP_PREFIX_RT_NOP;
 956
 957                 action = CLEANUP_PREFIX_RT_EXPIRE;
 958
 959                 spin_lock(&ifa->lock);
 960
 961                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 962                 /*
 963                  * Note: Because this address is
 964                  * not permanent, lifetime <
 965                  * LONG_MAX / HZ here.
 966                  */
 967                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
 968                         *expires = ifa->tstamp + lifetime * HZ;
 969                 spin_unlock(&ifa->lock);
 970         }
 971
 972         return action;
 973 }
 974
 975 static void
 976 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
 977 {
 978         struct rt6_info *rt;
 979
 980         rt = addrconf_get_prefix_route(&ifp->addr,
 981                                        ifp->prefix_len,
 982                                        ifp->idev->dev,
 983                                        0, RTF_GATEWAY | RTF_DEFAULT);
 984         if (rt) {
 985                 if (del_rt)
 986                         ip6_del_rt(rt);
 987                 else {
 988                         if (!(rt->rt6i_flags & RTF_EXPIRES))
 989                                 rt6_set_expires(rt, expires);
 990                         ip6_rt_put(rt);
 991                 }
 992         }
 993 }
 994
 995
 996 /* This function wants to get referenced ifp and releases it before return */
 997
 998 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 999 {
1000         int state;
1001         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1002         unsigned long expires;
1003
1004         ASSERT_RTNL();
1005
1006         spin_lock_bh(&ifp->state_lock);
1007         state = ifp->state;
1008         ifp->state = INET6_IFADDR_STATE_DEAD;
1009         spin_unlock_bh(&ifp->state_lock);
1010
1011         if (state == INET6_IFADDR_STATE_DEAD)
1012                 goto out;
1013
1014         spin_lock_bh(&addrconf_hash_lock);
1015         hlist_del_init_rcu(&ifp->addr_lst);
1016         spin_unlock_bh(&addrconf_hash_lock);
1017
1018         write_lock_bh(&ifp->idev->lock);
1019
1020         if (ifp->flags&IFA_F_TEMPORARY) {
1021                 list_del(&ifp->tmp_list);
1022                 if (ifp->ifpub) {
1023                         in6_ifa_put(ifp->ifpub);
1024                         ifp->ifpub = NULL;
1025                 }
1026                 __in6_ifa_put(ifp);
1027         }
1028
1029         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1030                 action = check_cleanup_prefix_route(ifp, &expires);
1031
1032         list_del_init(&ifp->if_list);
1033         __in6_ifa_put(ifp);
1034
1035         write_unlock_bh(&ifp->idev->lock);
1036
1037         addrconf_del_dad_work(ifp);
1038
1039         ipv6_ifa_notify(RTM_DELADDR, ifp);
1040
1041         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1042
1043         if (action != CLEANUP_PREFIX_RT_NOP) {
1044                 cleanup_prefix_route(ifp, expires,
1045                         action == CLEANUP_PREFIX_RT_DEL);
1046         }
1047
1048         /* clean up prefsrc entries */
1049         rt6_remove_prefsrc(ifp);
1050 out:
1051         in6_ifa_put(ifp);
1052 }
1053
1054 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1055 {
1056         struct inet6_dev *idev = ifp->idev;
1057         struct in6_addr addr, *tmpaddr;
1058         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1059         unsigned long regen_advance;
1060         int tmp_plen;
1061         int ret = 0;
1062         u32 addr_flags;
1063         unsigned long now = jiffies;
1064
1065         write_lock_bh(&idev->lock);
1066         if (ift) {
1067                 spin_lock_bh(&ift->lock);
1068                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1069                 spin_unlock_bh(&ift->lock);
1070                 tmpaddr = &addr;
1071         } else {
1072                 tmpaddr = NULL;
1073         }
1074 retry:
1075         in6_dev_hold(idev);
1076         if (idev->cnf.use_tempaddr <= 0) {
1077                 write_unlock_bh(&idev->lock);
1078                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1079                 in6_dev_put(idev);
1080                 ret = -1;
1081                 goto out;
1082         }
1083         spin_lock_bh(&ifp->lock);
1084         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1085                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1086                 spin_unlock_bh(&ifp->lock);
1087                 write_unlock_bh(&idev->lock);
1088                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1089                         __func__);
1090                 in6_dev_put(idev);
1091                 ret = -1;
1092                 goto out;
1093         }
1094         in6_ifa_hold(ifp);
1095         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1096         __ipv6_try_regen_rndid(idev, tmpaddr);
1097         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1098         age = (now - ifp->tstamp) / HZ;
1099         tmp_valid_lft = min_t(__u32,
1100                               ifp->valid_lft,
1101                               idev->cnf.temp_valid_lft + age);
1102         tmp_prefered_lft = min_t(__u32,
1103                                  ifp->prefered_lft,
1104                                  idev->cnf.temp_prefered_lft + age -
1105                                  idev->cnf.max_desync_factor);
1106         tmp_plen = ifp->prefix_len;
1107         tmp_tstamp = ifp->tstamp;
1108         spin_unlock_bh(&ifp->lock);
1109
1110         regen_advance = idev->cnf.regen_max_retry *
1111                         idev->cnf.dad_transmits *
1112                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1113         write_unlock_bh(&idev->lock);
1114
1115         /* A temporary address is created only if this calculated Preferred
1116          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1117          * an implementation must not create a temporary address with a zero
1118          * Preferred Lifetime.
1119          * Use age calculation as in addrconf_verify to avoid unnecessary
1120          * temporary addresses being generated.
1121          */
1122         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1123         if (tmp_prefered_lft <= regen_advance + age) {
1124                 in6_ifa_put(ifp);
1125                 in6_dev_put(idev);
1126                 ret = -1;
1127                 goto out;
1128         }
1129
1130         addr_flags = IFA_F_TEMPORARY;
1131         /* set in addrconf_prefix_rcv() */
1132         if (ifp->flags & IFA_F_OPTIMISTIC)
1133                 addr_flags |= IFA_F_OPTIMISTIC;
1134
1135         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1136                             ipv6_addr_scope(&addr), addr_flags,
1137                             tmp_valid_lft, tmp_prefered_lft);
1138         if (IS_ERR(ift)) {
1139                 in6_ifa_put(ifp);
1140                 in6_dev_put(idev);
1141                 pr_info("%s: retry temporary address regeneration\n", __func__);
1142                 tmpaddr = &addr;
1143                 write_lock_bh(&idev->lock);
1144                 goto retry;
1145         }
1146
1147         spin_lock_bh(&ift->lock);
1148         ift->ifpub = ifp;
1149         ift->cstamp = now;
1150         ift->tstamp = tmp_tstamp;
1151         spin_unlock_bh(&ift->lock);
1152
1153         addrconf_dad_start(ift);
1154         in6_ifa_put(ift);
1155         in6_dev_put(idev);
1156 out:
1157         return ret;
1158 }
1159
1160 /*
1161  *      Choose an appropriate source address (RFC3484)
1162  */
1163 enum {
1164         IPV6_SADDR_RULE_INIT = 0,
1165         IPV6_SADDR_RULE_LOCAL,
1166         IPV6_SADDR_RULE_SCOPE,
1167         IPV6_SADDR_RULE_PREFERRED,
1168 #ifdef CONFIG_IPV6_MIP6
1169         IPV6_SADDR_RULE_HOA,
1170 #endif
1171         IPV6_SADDR_RULE_OIF,
1172         IPV6_SADDR_RULE_LABEL,
1173         IPV6_SADDR_RULE_PRIVACY,
1174         IPV6_SADDR_RULE_ORCHID,
1175         IPV6_SADDR_RULE_PREFIX,
1176 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1177         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1178 #endif
1179         IPV6_SADDR_RULE_MAX
1180 };
1181
1182 struct ipv6_saddr_score {
1183         int                     rule;
1184         int                     addr_type;
1185         struct inet6_ifaddr     *ifa;
1186         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1187         int                     scopedist;
1188         int                     matchlen;
1189 };
1190
1191 struct ipv6_saddr_dst {
1192         const struct in6_addr *addr;
1193         int ifindex;
1194         int scope;
1195         int label;
1196         unsigned int prefs;
1197 };
1198
1199 static inline int ipv6_saddr_preferred(int type)
1200 {
1201         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1202                 return 1;
1203         return 0;
1204 }
1205
1206 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1207 {
1208 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1209         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1210 #else
1211         return false;
1212 #endif
1213 }
1214
1215 static int ipv6_get_saddr_eval(struct net *net,
1216                                struct ipv6_saddr_score *score,
1217                                struct ipv6_saddr_dst *dst,
1218                                int i)
1219 {
1220         int ret;
1221
1222         if (i <= score->rule) {
1223                 switch (i) {
1224                 case IPV6_SADDR_RULE_SCOPE:
1225                         ret = score->scopedist;
1226                         break;
1227                 case IPV6_SADDR_RULE_PREFIX:
1228                         ret = score->matchlen;
1229                         break;
1230                 default:
1231                         ret = !!test_bit(i, score->scorebits);
1232                 }
1233                 goto out;
1234         }
1235
1236         switch (i) {
1237         case IPV6_SADDR_RULE_INIT:
1238                 /* Rule 0: remember if hiscore is not ready yet */
1239                 ret = !!score->ifa;
1240                 break;
1241         case IPV6_SADDR_RULE_LOCAL:
1242                 /* Rule 1: Prefer same address */
1243                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1244                 break;
1245         case IPV6_SADDR_RULE_SCOPE:
1246                 /* Rule 2: Prefer appropriate scope
1247                  *
1248                  *      ret
1249                  *       ^
1250                  *    -1 |  d 15
1251                  *    ---+--+-+---> scope
1252                  *       |
1253                  *       |             d is scope of the destination.
1254                  *  B-d  |  \
1255                  *       |   \      <- smaller scope is better if
1256                  *  B-15 |    \        if scope is enough for destination.
1257                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1258                  * d-C-1 | /
1259                  *       |/         <- greater is better
1260                  *   -C  /             if scope is not enough for destination.
1261                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1262                  *
1263                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1264                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1265                  * Assume B = 0 and we get C > 29.
1266                  */
1267                 ret = __ipv6_addr_src_scope(score->addr_type);
1268                 if (ret >= dst->scope)
1269                         ret = -ret;
1270                 else
1271                         ret -= 128;     /* 30 is enough */
1272                 score->scopedist = ret;
1273                 break;
1274         case IPV6_SADDR_RULE_PREFERRED:
1275             {
1276                 /* Rule 3: Avoid deprecated and optimistic addresses */
1277                 u8 avoid = IFA_F_DEPRECATED;
1278
1279                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1280                         avoid |= IFA_F_OPTIMISTIC;
1281                 ret = ipv6_saddr_preferred(score->addr_type) ||
1282                       !(score->ifa->flags & avoid);
1283                 break;
1284             }
1285 #ifdef CONFIG_IPV6_MIP6
1286         case IPV6_SADDR_RULE_HOA:
1287             {
1288                 /* Rule 4: Prefer home address */
1289                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1290                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1291                 break;
1292             }
1293 #endif
1294         case IPV6_SADDR_RULE_OIF:
1295                 /* Rule 5: Prefer outgoing interface */
1296                 ret = (!dst->ifindex ||
1297                        dst->ifindex == score->ifa->idev->dev->ifindex);
1298                 break;
1299         case IPV6_SADDR_RULE_LABEL:
1300                 /* Rule 6: Prefer matching label */
1301                 ret = ipv6_addr_label(net,
1302                                       &score->ifa->addr, score->addr_type,
1303                                       score->ifa->idev->dev->ifindex) == dst->label;
1304                 break;
1305         case IPV6_SADDR_RULE_PRIVACY:
1306             {
1307                 /* Rule 7: Prefer public address
1308                  * Note: prefer temporary address if use_tempaddr >= 2
1309                  */
1310                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1311                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1312                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1313                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1314                 break;
1315             }
1316         case IPV6_SADDR_RULE_ORCHID:
1317                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1318                  *          non-ORCHID vs non-ORCHID
1319                  */
1320                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1321                         ipv6_addr_orchid(dst->addr));
1322                 break;
1323         case IPV6_SADDR_RULE_PREFIX:
1324                 /* Rule 8: Use longest matching prefix */
1325                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1326                 if (ret > score->ifa->prefix_len)
1327                         ret = score->ifa->prefix_len;
1328                 score->matchlen = ret;
1329                 break;
1330 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1331         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1332                 /* Optimistic addresses still have lower precedence than other
1333                  * preferred addresses.
1334                  */
1335                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1336                 break;
1337 #endif
1338         default:
1339                 ret = 0;
1340         }
1341
1342         if (ret)
1343                 __set_bit(i, score->scorebits);
1344         score->rule = i;
1345 out:
1346         return ret;
1347 }
1348
1349 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1350                        const struct in6_addr *daddr, unsigned int prefs,
1351                        struct in6_addr *saddr)
1352 {
1353         struct ipv6_saddr_score scores[2],
1354                                 *score = &scores[0], *hiscore = &scores[1];
1355         struct ipv6_saddr_dst dst;
1356         struct net_device *dev;
1357         int dst_type;
1358
1359         dst_type = __ipv6_addr_type(daddr);
1360         dst.addr = daddr;
1361         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1362         dst.scope = __ipv6_addr_src_scope(dst_type);
1363         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1364         dst.prefs = prefs;
1365
1366         hiscore->rule = -1;
1367         hiscore->ifa = NULL;
1368
1369         rcu_read_lock();
1370
1371         for_each_netdev_rcu(net, dev) {
1372                 struct inet6_dev *idev;
1373
1374                 /* Candidate Source Address (section 4)
1375                  *  - multicast and link-local destination address,
1376                  *    the set of candidate source address MUST only
1377                  *    include addresses assigned to interfaces
1378                  *    belonging to the same link as the outgoing
1379                  *    interface.
1380                  * (- For site-local destination addresses, the
1381                  *    set of candidate source addresses MUST only
1382                  *    include addresses assigned to interfaces
1383                  *    belonging to the same site as the outgoing
1384                  *    interface.)
1385                  */
1386                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1387                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1388                     dst.ifindex && dev->ifindex != dst.ifindex)
1389                         continue;
1390
1391                 idev = __in6_dev_get(dev);
1392                 if (!idev)
1393                         continue;
1394
1395                 read_lock_bh(&idev->lock);
1396                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1397                         int i;
1398
1399                         /*
1400                          * - Tentative Address (RFC2462 section 5.4)
1401                          *  - A tentative address is not considered
1402                          *    "assigned to an interface" in the traditional
1403                          *    sense, unless it is also flagged as optimistic.
1404                          * - Candidate Source Address (section 4)
1405                          *  - In any case, anycast addresses, multicast
1406                          *    addresses, and the unspecified address MUST
1407                          *    NOT be included in a candidate set.
1408                          */
1409                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1410                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1411                                 continue;
1412
1413                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1414
1415                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1416                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1417                                 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1418                                                     dev->name);
1419                                 continue;
1420                         }
1421
1422                         score->rule = -1;
1423                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1424
1425                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1426                                 int minihiscore, miniscore;
1427
1428                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1429                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1430
1431                                 if (minihiscore > miniscore) {
1432                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1433                                             score->scopedist > 0) {
1434                                                 /*
1435                                                  * special case:
1436                                                  * each remaining entry
1437                                                  * has too small (not enough)
1438                                                  * scope, because ifa entries
1439                                                  * are sorted by their scope
1440                                                  * values.
1441                                                  */
1442                                                 goto try_nextdev;
1443                                         }
1444                                         break;
1445                                 } else if (minihiscore < miniscore) {
1446                                         if (hiscore->ifa)
1447                                                 in6_ifa_put(hiscore->ifa);
1448
1449                                         in6_ifa_hold(score->ifa);
1450
1451                                         swap(hiscore, score);
1452
1453                                         /* restore our iterator */
1454                                         score->ifa = hiscore->ifa;
1455
1456                                         break;
1457                                 }
1458                         }
1459                 }
1460 try_nextdev:
1461                 read_unlock_bh(&idev->lock);
1462         }
1463         rcu_read_unlock();
1464
1465         if (!hiscore->ifa)
1466                 return -EADDRNOTAVAIL;
1467
1468         *saddr = hiscore->ifa->addr;
1469         in6_ifa_put(hiscore->ifa);
1470         return 0;
1471 }
1472 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1473
1474 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1475                       u32 banned_flags)
1476 {
1477         struct inet6_ifaddr *ifp;
1478         int err = -EADDRNOTAVAIL;
1479
1480         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1481                 if (ifp->scope > IFA_LINK)
1482                         break;
1483                 if (ifp->scope == IFA_LINK &&
1484                     !(ifp->flags & banned_flags)) {
1485                         *addr = ifp->addr;
1486                         err = 0;
1487                         break;
1488                 }
1489         }
1490         return err;
1491 }
1492
1493 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1494                     u32 banned_flags)
1495 {
1496         struct inet6_dev *idev;
1497         int err = -EADDRNOTAVAIL;
1498
1499         rcu_read_lock();
1500         idev = __in6_dev_get(dev);
1501         if (idev) {
1502                 read_lock_bh(&idev->lock);
1503                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1504                 read_unlock_bh(&idev->lock);
1505         }
1506         rcu_read_unlock();
1507         return err;
1508 }
1509
1510 static int ipv6_count_addresses(struct inet6_dev *idev)
1511 {
1512         int cnt = 0;
1513         struct inet6_ifaddr *ifp;
1514
1515         read_lock_bh(&idev->lock);
1516         list_for_each_entry(ifp, &idev->addr_list, if_list)
1517                 cnt++;
1518         read_unlock_bh(&idev->lock);
1519         return cnt;
1520 }
1521
1522 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1523                   const struct net_device *dev, int strict)
1524 {
1525         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1526 }
1527 EXPORT_SYMBOL(ipv6_chk_addr);
1528
1529 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1530                             const struct net_device *dev, int strict,
1531                             u32 banned_flags)
1532 {
1533         struct inet6_ifaddr *ifp;
1534         unsigned int hash = inet6_addr_hash(addr);
1535         u32 ifp_flags;
1536
1537         rcu_read_lock_bh();
1538         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1539                 if (!net_eq(dev_net(ifp->idev->dev), net))
1540                         continue;
1541                 /* Decouple optimistic from tentative for evaluation here.
1542                  * Ban optimistic addresses explicitly, when required.
1543                  */
1544                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1545                             ? (ifp->flags&~IFA_F_TENTATIVE)
1546                             : ifp->flags;
1547                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1548                     !(ifp_flags&banned_flags) &&
1549                     (dev == NULL || ifp->idev->dev == dev ||
1550                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1551                         rcu_read_unlock_bh();
1552                         return 1;
1553                 }
1554         }
1555
1556         rcu_read_unlock_bh();
1557         return 0;
1558 }
1559 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1560
1561 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1562                                struct net_device *dev)
1563 {
1564         unsigned int hash = inet6_addr_hash(addr);
1565         struct inet6_ifaddr *ifp;
1566
1567         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1568                 if (!net_eq(dev_net(ifp->idev->dev), net))
1569                         continue;
1570                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1571                         if (dev == NULL || ifp->idev->dev == dev)
1572                                 return true;
1573                 }
1574         }
1575         return false;
1576 }
1577
1578 /* Compares an address/prefix_len with addresses on device @dev.
1579  * If one is found it returns true.
1580  */
1581 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1582         const unsigned int prefix_len, struct net_device *dev)
1583 {
1584         struct inet6_dev *idev;
1585         struct inet6_ifaddr *ifa;
1586         bool ret = false;
1587
1588         rcu_read_lock();
1589         idev = __in6_dev_get(dev);
1590         if (idev) {
1591                 read_lock_bh(&idev->lock);
1592                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1593                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1594                         if (ret)
1595                                 break;
1596                 }
1597                 read_unlock_bh(&idev->lock);
1598         }
1599         rcu_read_unlock();
1600
1601         return ret;
1602 }
1603 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1604
1605 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1606 {
1607         struct inet6_dev *idev;
1608         struct inet6_ifaddr *ifa;
1609         int     onlink;
1610
1611         onlink = 0;
1612         rcu_read_lock();
1613         idev = __in6_dev_get(dev);
1614         if (idev) {
1615                 read_lock_bh(&idev->lock);
1616                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1617                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1618                                                    ifa->prefix_len);
1619                         if (onlink)
1620                                 break;
1621                 }
1622                 read_unlock_bh(&idev->lock);
1623         }
1624         rcu_read_unlock();
1625         return onlink;
1626 }
1627 EXPORT_SYMBOL(ipv6_chk_prefix);
1628
1629 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1630                                      struct net_device *dev, int strict)
1631 {
1632         struct inet6_ifaddr *ifp, *result = NULL;
1633         unsigned int hash = inet6_addr_hash(addr);
1634
1635         rcu_read_lock_bh();
1636         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1637                 if (!net_eq(dev_net(ifp->idev->dev), net))
1638                         continue;
1639                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1640                         if (dev == NULL || ifp->idev->dev == dev ||
1641                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1642                                 result = ifp;
1643                                 in6_ifa_hold(ifp);
1644                                 break;
1645                         }
1646                 }
1647         }
1648         rcu_read_unlock_bh();
1649
1650         return result;
1651 }
1652
1653 /* Gets referenced address, destroys ifaddr */
1654
1655 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1656 {
1657         if (ifp->flags&IFA_F_PERMANENT) {
1658                 spin_lock_bh(&ifp->lock);
1659                 addrconf_del_dad_work(ifp);
1660                 ifp->flags |= IFA_F_TENTATIVE;
1661                 if (dad_failed)
1662                         ifp->flags |= IFA_F_DADFAILED;
1663                 spin_unlock_bh(&ifp->lock);
1664                 if (dad_failed)
1665                         ipv6_ifa_notify(0, ifp);
1666                 in6_ifa_put(ifp);
1667         } else if (ifp->flags&IFA_F_TEMPORARY) {
1668                 struct inet6_ifaddr *ifpub;
1669                 spin_lock_bh(&ifp->lock);
1670                 ifpub = ifp->ifpub;
1671                 if (ifpub) {
1672                         in6_ifa_hold(ifpub);
1673                         spin_unlock_bh(&ifp->lock);
1674                         ipv6_create_tempaddr(ifpub, ifp);
1675                         in6_ifa_put(ifpub);
1676                 } else {
1677                         spin_unlock_bh(&ifp->lock);
1678                 }
1679                 ipv6_del_addr(ifp);
1680         } else {
1681                 ipv6_del_addr(ifp);
1682         }
1683 }
1684
1685 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1686 {
1687         int err = -ENOENT;
1688
1689         spin_lock_bh(&ifp->state_lock);
1690         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1691                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1692                 err = 0;
1693         }
1694         spin_unlock_bh(&ifp->state_lock);
1695
1696         return err;
1697 }
1698
1699 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1700 {
1701         struct inet6_dev *idev = ifp->idev;
1702
1703         if (addrconf_dad_end(ifp)) {
1704                 in6_ifa_put(ifp);
1705                 return;
1706         }
1707
1708         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1709                              ifp->idev->dev->name, &ifp->addr);
1710
1711         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1712                 struct in6_addr addr;
1713
1714                 addr.s6_addr32[0] = htonl(0xfe800000);
1715                 addr.s6_addr32[1] = 0;
1716
1717                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1718                     ipv6_addr_equal(&ifp->addr, &addr)) {
1719                         /* DAD failed for link-local based on MAC address */
1720                         idev->cnf.disable_ipv6 = 1;
1721
1722                         pr_info("%s: IPv6 being disabled!\n",
1723                                 ifp->idev->dev->name);
1724                 }
1725         }
1726
1727         spin_lock_bh(&ifp->state_lock);
1728         /* transition from _POSTDAD to _ERRDAD */
1729         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1730         spin_unlock_bh(&ifp->state_lock);
1731
1732         addrconf_mod_dad_work(ifp, 0);
1733 }
1734
1735 /* Join to solicited addr multicast group.
1736  * caller must hold RTNL */
1737 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1738 {
1739         struct in6_addr maddr;
1740
1741         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1742                 return;
1743
1744         addrconf_addr_solict_mult(addr, &maddr);
1745         ipv6_dev_mc_inc(dev, &maddr);
1746 }
1747
1748 /* caller must hold RTNL */
1749 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1750 {
1751         struct in6_addr maddr;
1752
1753         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1754                 return;
1755
1756         addrconf_addr_solict_mult(addr, &maddr);
1757         __ipv6_dev_mc_dec(idev, &maddr);
1758 }
1759
1760 /* caller must hold RTNL */
1761 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1762 {
1763         struct in6_addr addr;
1764
1765         if (ifp->prefix_len >= 127) /* RFC 6164 */
1766                 return;
1767         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1768         if (ipv6_addr_any(&addr))
1769                 return;
1770         __ipv6_dev_ac_inc(ifp->idev, &addr);
1771 }
1772
1773 /* caller must hold RTNL */
1774 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1775 {
1776         struct in6_addr addr;
1777
1778         if (ifp->prefix_len >= 127) /* RFC 6164 */
1779                 return;
1780         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1781         if (ipv6_addr_any(&addr))
1782                 return;
1783         __ipv6_dev_ac_dec(ifp->idev, &addr);
1784 }
1785
1786 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1787 {
1788         if (dev->addr_len != ETH_ALEN)
1789                 return -1;
1790         memcpy(eui, dev->dev_addr, 3);
1791         memcpy(eui + 5, dev->dev_addr + 3, 3);
1792
1793         /*
1794          * The zSeries OSA network cards can be shared among various
1795          * OS instances, but the OSA cards have only one MAC address.
1796          * This leads to duplicate address conflicts in conjunction
1797          * with IPv6 if more than one instance uses the same card.
1798          *
1799          * The driver for these cards can deliver a unique 16-bit
1800          * identifier for each instance sharing the same card.  It is
1801          * placed instead of 0xFFFE in the interface identifier.  The
1802          * "u" bit of the interface identifier is not inverted in this
1803          * case.  Hence the resulting interface identifier has local
1804          * scope according to RFC2373.
1805          */
1806         if (dev->dev_id) {
1807                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1808                 eui[4] = dev->dev_id & 0xFF;
1809         } else {
1810                 eui[3] = 0xFF;
1811                 eui[4] = 0xFE;
1812                 eui[0] ^= 2;
1813         }
1814         return 0;
1815 }
1816
1817 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1818 {
1819         if (dev->addr_len != IEEE802154_ADDR_LEN)
1820                 return -1;
1821         memcpy(eui, dev->dev_addr, 8);
1822         eui[0] ^= 2;
1823         return 0;
1824 }
1825
1826 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1827 {
1828         union fwnet_hwaddr *ha;
1829
1830         if (dev->addr_len != FWNET_ALEN)
1831                 return -1;
1832
1833         ha = (union fwnet_hwaddr *)dev->dev_addr;
1834
1835         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1836         eui[0] ^= 2;
1837         return 0;
1838 }
1839
1840 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1841 {
1842         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1843         if (dev->addr_len != ARCNET_ALEN)
1844                 return -1;
1845         memset(eui, 0, 7);
1846         eui[7] = *(u8 *)dev->dev_addr;
1847         return 0;
1848 }
1849
1850 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1851 {
1852         if (dev->addr_len != INFINIBAND_ALEN)
1853                 return -1;
1854         memcpy(eui, dev->dev_addr + 12, 8);
1855         eui[0] |= 2;
1856         return 0;
1857 }
1858
1859 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1860 {
1861         if (addr == 0)
1862                 return -1;
1863         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1864                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1865                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1866                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1867                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1868                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1869         eui[1] = 0;
1870         eui[2] = 0x5E;
1871         eui[3] = 0xFE;
1872         memcpy(eui + 4, &addr, 4);
1873         return 0;
1874 }
1875
1876 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1877 {
1878         if (dev->priv_flags & IFF_ISATAP)
1879                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1880         return -1;
1881 }
1882
1883 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1884 {
1885         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1886 }
1887
1888 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1889 {
1890         memcpy(eui, dev->perm_addr, 3);
1891         memcpy(eui + 5, dev->perm_addr + 3, 3);
1892         eui[3] = 0xFF;
1893         eui[4] = 0xFE;
1894         eui[0] ^= 2;
1895         return 0;
1896 }
1897
1898 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1899 {
1900         switch (dev->type) {
1901         case ARPHRD_ETHER:
1902         case ARPHRD_FDDI:
1903                 return addrconf_ifid_eui48(eui, dev);
1904         case ARPHRD_ARCNET:
1905                 return addrconf_ifid_arcnet(eui, dev);
1906         case ARPHRD_INFINIBAND:
1907                 return addrconf_ifid_infiniband(eui, dev);
1908         case ARPHRD_SIT:
1909                 return addrconf_ifid_sit(eui, dev);
1910         case ARPHRD_IPGRE:
1911                 return addrconf_ifid_gre(eui, dev);
1912         case ARPHRD_6LOWPAN:
1913         case ARPHRD_IEEE802154:
1914                 return addrconf_ifid_eui64(eui, dev);
1915         case ARPHRD_IEEE1394:
1916                 return addrconf_ifid_ieee1394(eui, dev);
1917         case ARPHRD_TUNNEL6:
1918                 return addrconf_ifid_ip6tnl(eui, dev);
1919         }
1920         return -1;
1921 }
1922
1923 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1924 {
1925         int err = -1;
1926         struct inet6_ifaddr *ifp;
1927
1928         read_lock_bh(&idev->lock);
1929         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1930                 if (ifp->scope > IFA_LINK)
1931                         break;
1932                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1933                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1934                         err = 0;
1935                         break;
1936                 }
1937         }
1938         read_unlock_bh(&idev->lock);
1939         return err;
1940 }
1941
1942 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1943 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1944 {
1945 regen:
1946         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1947         idev->rndid[0] &= ~0x02;
1948
1949         /*
1950          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1951          * check if generated address is not inappropriate
1952          *
1953          *  - Reserved subnet anycast (RFC 2526)
1954          *      11111101 11....11 1xxxxxxx
1955          *  - ISATAP (RFC4214) 6.1
1956          *      00-00-5E-FE-xx-xx-xx-xx
1957          *  - value 0
1958          *  - XXX: already assigned to an address on the device
1959          */
1960         if (idev->rndid[0] == 0xfd &&
1961             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1962             (idev->rndid[7]&0x80))
1963                 goto regen;
1964         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1965                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1966                         goto regen;
1967                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1968                         goto regen;
1969         }
1970 }
1971
1972 static void ipv6_regen_rndid(unsigned long data)
1973 {
1974         struct inet6_dev *idev = (struct inet6_dev *) data;
1975         unsigned long expires;
1976
1977         rcu_read_lock_bh();
1978         write_lock_bh(&idev->lock);
1979
1980         if (idev->dead)
1981                 goto out;
1982
1983         __ipv6_regen_rndid(idev);
1984
1985         expires = jiffies +
1986                 idev->cnf.temp_prefered_lft * HZ -
1987                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1988                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1989                 idev->cnf.max_desync_factor * HZ;
1990         if (time_before(expires, jiffies)) {
1991                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1992                         __func__, idev->dev->name);
1993                 goto out;
1994         }
1995
1996         if (!mod_timer(&idev->regen_timer, expires))
1997                 in6_dev_hold(idev);
1998
1999 out:
2000         write_unlock_bh(&idev->lock);
2001         rcu_read_unlock_bh();
2002         in6_dev_put(idev);
2003 }
2004
2005 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2006 {
2007         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2008                 __ipv6_regen_rndid(idev);
2009 }
2010
2011 /*
2012  *      Add prefix route.
2013  */
2014
2015 static void
2016 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2017                       unsigned long expires, u32 flags)
2018 {
2019         struct fib6_config cfg = {
2020                 .fc_table = RT6_TABLE_PREFIX,
2021                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2022                 .fc_ifindex = dev->ifindex,
2023                 .fc_expires = expires,
2024                 .fc_dst_len = plen,
2025                 .fc_flags = RTF_UP | flags,
2026                 .fc_nlinfo.nl_net = dev_net(dev),
2027                 .fc_protocol = RTPROT_KERNEL,
2028         };
2029
2030         cfg.fc_dst = *pfx;
2031
2032         /* Prevent useless cloning on PtP SIT.
2033            This thing is done here expecting that the whole
2034            class of non-broadcast devices need not cloning.
2035          */
2036 #if IS_ENABLED(CONFIG_IPV6_SIT)
2037         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2038                 cfg.fc_flags |= RTF_NONEXTHOP;
2039 #endif
2040
2041         ip6_route_add(&cfg);
2042 }
2043
2044
2045 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2046                                                   int plen,
2047                                                   const struct net_device *dev,
2048                                                   u32 flags, u32 noflags)
2049 {
2050         struct fib6_node *fn;
2051         struct rt6_info *rt = NULL;
2052         struct fib6_table *table;
2053
2054         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2055         if (table == NULL)
2056                 return NULL;
2057
2058         read_lock_bh(&table->tb6_lock);
2059         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2060         if (!fn)
2061                 goto out;
2062         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2063                 if (rt->dst.dev->ifindex != dev->ifindex)
2064                         continue;
2065                 if ((rt->rt6i_flags & flags) != flags)
2066                         continue;
2067                 if ((rt->rt6i_flags & noflags) != 0)
2068                         continue;
2069                 dst_hold(&rt->dst);
2070                 break;
2071         }
2072 out:
2073         read_unlock_bh(&table->tb6_lock);
2074         return rt;
2075 }
2076
2077
2078 /* Create "default" multicast route to the interface */
2079
2080 static void addrconf_add_mroute(struct net_device *dev)
2081 {
2082         struct fib6_config cfg = {
2083                 .fc_table = RT6_TABLE_LOCAL,
2084                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2085                 .fc_ifindex = dev->ifindex,
2086                 .fc_dst_len = 8,
2087                 .fc_flags = RTF_UP,
2088                 .fc_nlinfo.nl_net = dev_net(dev),
2089         };
2090
2091         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2092
2093         ip6_route_add(&cfg);
2094 }
2095
2096 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2097 {
2098         struct inet6_dev *idev;
2099
2100         ASSERT_RTNL();
2101
2102         idev = ipv6_find_idev(dev);
2103         if (!idev)
2104                 return ERR_PTR(-ENOBUFS);
2105
2106         if (idev->cnf.disable_ipv6)
2107                 return ERR_PTR(-EACCES);
2108
2109         /* Add default multicast route */
2110         if (!(dev->flags & IFF_LOOPBACK))
2111                 addrconf_add_mroute(dev);
2112
2113         return idev;
2114 }
2115
2116 static void manage_tempaddrs(struct inet6_dev *idev,
2117                              struct inet6_ifaddr *ifp,
2118                              __u32 valid_lft, __u32 prefered_lft,
2119                              bool create, unsigned long now)
2120 {
2121         u32 flags;
2122         struct inet6_ifaddr *ift;
2123
2124         read_lock_bh(&idev->lock);
2125         /* update all temporary addresses in the list */
2126         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2127                 int age, max_valid, max_prefered;
2128
2129                 if (ifp != ift->ifpub)
2130                         continue;
2131
2132                 /* RFC 4941 section 3.3:
2133                  * If a received option will extend the lifetime of a public
2134                  * address, the lifetimes of temporary addresses should
2135                  * be extended, subject to the overall constraint that no
2136                  * temporary addresses should ever remain "valid" or "preferred"
2137                  * for a time longer than (TEMP_VALID_LIFETIME) or
2138                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2139                  */
2140                 age = (now - ift->cstamp) / HZ;
2141                 max_valid = idev->cnf.temp_valid_lft - age;
2142                 if (max_valid < 0)
2143                         max_valid = 0;
2144
2145                 max_prefered = idev->cnf.temp_prefered_lft -
2146                                idev->cnf.max_desync_factor - age;
2147                 if (max_prefered < 0)
2148                         max_prefered = 0;
2149
2150                 if (valid_lft > max_valid)
2151                         valid_lft = max_valid;
2152
2153                 if (prefered_lft > max_prefered)
2154                         prefered_lft = max_prefered;
2155
2156                 spin_lock(&ift->lock);
2157                 flags = ift->flags;
2158                 ift->valid_lft = valid_lft;
2159                 ift->prefered_lft = prefered_lft;
2160                 ift->tstamp = now;
2161                 if (prefered_lft > 0)
2162                         ift->flags &= ~IFA_F_DEPRECATED;
2163
2164                 spin_unlock(&ift->lock);
2165                 if (!(flags&IFA_F_TENTATIVE))
2166                         ipv6_ifa_notify(0, ift);
2167         }
2168
2169         if ((create || list_empty(&idev->tempaddr_list)) &&
2170             idev->cnf.use_tempaddr > 0) {
2171                 /* When a new public address is created as described
2172                  * in [ADDRCONF], also create a new temporary address.
2173                  * Also create a temporary address if it's enabled but
2174                  * no temporary address currently exists.
2175                  */
2176                 read_unlock_bh(&idev->lock);
2177                 ipv6_create_tempaddr(ifp, NULL);
2178         } else {
2179                 read_unlock_bh(&idev->lock);
2180         }
2181 }
2182
2183 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2184 {
2185         struct prefix_info *pinfo;
2186         __u32 valid_lft;
2187         __u32 prefered_lft;
2188         int addr_type;
2189         struct inet6_dev *in6_dev;
2190         struct net *net = dev_net(dev);
2191
2192         pinfo = (struct prefix_info *) opt;
2193
2194         if (len < sizeof(struct prefix_info)) {
2195                 ADBG("addrconf: prefix option too short\n");
2196                 return;
2197         }
2198
2199         /*
2200          *      Validation checks ([ADDRCONF], page 19)
2201          */
2202
2203         addr_type = ipv6_addr_type(&pinfo->prefix);
2204
2205         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2206                 return;
2207
2208         valid_lft = ntohl(pinfo->valid);
2209         prefered_lft = ntohl(pinfo->prefered);
2210
2211         if (prefered_lft > valid_lft) {
2212                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2213                 return;
2214         }
2215
2216         in6_dev = in6_dev_get(dev);
2217
2218         if (in6_dev == NULL) {
2219                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2220                                     dev->name);
2221                 return;
2222         }
2223
2224         /*
2225          *      Two things going on here:
2226          *      1) Add routes for on-link prefixes
2227          *      2) Configure prefixes with the auto flag set
2228          */
2229
2230         if (pinfo->onlink) {
2231                 struct rt6_info *rt;
2232                 unsigned long rt_expires;
2233
2234                 /* Avoid arithmetic overflow. Really, we could
2235                  * save rt_expires in seconds, likely valid_lft,
2236                  * but it would require division in fib gc, that it
2237                  * not good.
2238                  */
2239                 if (HZ > USER_HZ)
2240                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2241                 else
2242                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2243
2244                 if (addrconf_finite_timeout(rt_expires))
2245                         rt_expires *= HZ;
2246
2247                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2248                                                pinfo->prefix_len,
2249                                                dev,
2250                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2251                                                RTF_GATEWAY | RTF_DEFAULT);
2252
2253                 if (rt) {
2254                         /* Autoconf prefix route */
2255                         if (valid_lft == 0) {
2256                                 ip6_del_rt(rt);
2257                                 rt = NULL;
2258                         } else if (addrconf_finite_timeout(rt_expires)) {
2259                                 /* not infinity */
2260                                 rt6_set_expires(rt, jiffies + rt_expires);
2261                         } else {
2262                                 rt6_clean_expires(rt);
2263                         }
2264                 } else if (valid_lft) {
2265                         clock_t expires = 0;
2266                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2267                         if (addrconf_finite_timeout(rt_expires)) {
2268                                 /* not infinity */
2269                                 flags |= RTF_EXPIRES;
2270                                 expires = jiffies_to_clock_t(rt_expires);
2271                         }
2272                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2273                                               dev, expires, flags);
2274                 }
2275                 ip6_rt_put(rt);
2276         }
2277
2278         /* Try to figure out our local address for this prefix */
2279
2280         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2281                 struct inet6_ifaddr *ifp;
2282                 struct in6_addr addr;
2283                 int create = 0, update_lft = 0;
2284                 bool tokenized = false;
2285
2286                 if (pinfo->prefix_len == 64) {
2287                         memcpy(&addr, &pinfo->prefix, 8);
2288
2289                         if (!ipv6_addr_any(&in6_dev->token)) {
2290                                 read_lock_bh(&in6_dev->lock);
2291                                 memcpy(addr.s6_addr + 8,
2292                                        in6_dev->token.s6_addr + 8, 8);
2293                                 read_unlock_bh(&in6_dev->lock);
2294                                 tokenized = true;
2295                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2296                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2297                                 in6_dev_put(in6_dev);
2298                                 return;
2299                         }
2300                         goto ok;
2301                 }
2302                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2303                                     pinfo->prefix_len);
2304                 in6_dev_put(in6_dev);
2305                 return;
2306
2307 ok:
2308
2309                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2310
2311                 if (ifp == NULL && valid_lft) {
2312                         int max_addresses = in6_dev->cnf.max_addresses;
2313                         u32 addr_flags = 0;
2314
2315 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2316                         if (in6_dev->cnf.optimistic_dad &&
2317                             !net->ipv6.devconf_all->forwarding && sllao)
2318                                 addr_flags = IFA_F_OPTIMISTIC;
2319 #endif
2320
2321                         /* Do not allow to create too much of autoconfigured
2322                          * addresses; this would be too easy way to crash kernel.
2323                          */
2324                         if (!max_addresses ||
2325                             ipv6_count_addresses(in6_dev) < max_addresses)
2326                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2327                                                     pinfo->prefix_len,
2328                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2329                                                     addr_flags, valid_lft,
2330                                                     prefered_lft);
2331
2332                         if (IS_ERR_OR_NULL(ifp)) {
2333                                 in6_dev_put(in6_dev);
2334                                 return;
2335                         }
2336
2337                         update_lft = 0;
2338                         create = 1;
2339                         spin_lock_bh(&ifp->lock);
2340                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2341                         ifp->cstamp = jiffies;
2342                         ifp->tokenized = tokenized;
2343                         spin_unlock_bh(&ifp->lock);
2344                         addrconf_dad_start(ifp);
2345                 }
2346
2347                 if (ifp) {
2348                         u32 flags;
2349                         unsigned long now;
2350                         u32 stored_lft;
2351
2352                         /* update lifetime (RFC2462 5.5.3 e) */
2353                         spin_lock(&ifp->lock);
2354                         now = jiffies;
2355                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2356                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2357                         else
2358                                 stored_lft = 0;
2359                         if (!update_lft && !create && stored_lft) {
2360                                 const u32 minimum_lft = min_t(u32,
2361                                         stored_lft, MIN_VALID_LIFETIME);
2362                                 valid_lft = max(valid_lft, minimum_lft);
2363
2364                                 /* RFC4862 Section 5.5.3e:
2365                                  * "Note that the preferred lifetime of the
2366                                  *  corresponding address is always reset to
2367                                  *  the Preferred Lifetime in the received
2368                                  *  Prefix Information option, regardless of
2369                                  *  whether the valid lifetime is also reset or
2370                                  *  ignored."
2371                                  *
2372                                  * So we should always update prefered_lft here.
2373                                  */
2374                                 update_lft = 1;
2375                         }
2376
2377                         if (update_lft) {
2378                                 ifp->valid_lft = valid_lft;
2379                                 ifp->prefered_lft = prefered_lft;
2380                                 ifp->tstamp = now;
2381                                 flags = ifp->flags;
2382                                 ifp->flags &= ~IFA_F_DEPRECATED;
2383                                 spin_unlock(&ifp->lock);
2384
2385                                 if (!(flags&IFA_F_TENTATIVE))
2386                                         ipv6_ifa_notify(0, ifp);
2387                         } else
2388                                 spin_unlock(&ifp->lock);
2389
2390                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2391                                          create, now);
2392
2393                         in6_ifa_put(ifp);
2394                         addrconf_verify();
2395                 }
2396         }
2397         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2398         in6_dev_put(in6_dev);
2399 }
2400
2401 /*
2402  *      Set destination address.
2403  *      Special case for SIT interfaces where we create a new "virtual"
2404  *      device.
2405  */
2406 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2407 {
2408         struct in6_ifreq ireq;
2409         struct net_device *dev;
2410         int err = -EINVAL;
2411
2412         rtnl_lock();
2413
2414         err = -EFAULT;
2415         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2416                 goto err_exit;
2417
2418         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2419
2420         err = -ENODEV;
2421         if (dev == NULL)
2422                 goto err_exit;
2423
2424 #if IS_ENABLED(CONFIG_IPV6_SIT)
2425         if (dev->type == ARPHRD_SIT) {
2426                 const struct net_device_ops *ops = dev->netdev_ops;
2427                 struct ifreq ifr;
2428                 struct ip_tunnel_parm p;
2429
2430                 err = -EADDRNOTAVAIL;
2431                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2432                         goto err_exit;
2433
2434                 memset(&p, 0, sizeof(p));
2435                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2436                 p.iph.saddr = 0;
2437                 p.iph.version = 4;
2438                 p.iph.ihl = 5;
2439                 p.iph.protocol = IPPROTO_IPV6;
2440                 p.iph.ttl = 64;
2441                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2442
2443                 if (ops->ndo_do_ioctl) {
2444                         mm_segment_t oldfs = get_fs();
2445
2446                         set_fs(KERNEL_DS);
2447                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2448                         set_fs(oldfs);
2449                 } else
2450                         err = -EOPNOTSUPP;
2451
2452                 if (err == 0) {
2453                         err = -ENOBUFS;
2454                         dev = __dev_get_by_name(net, p.name);
2455                         if (!dev)
2456                                 goto err_exit;
2457                         err = dev_open(dev);
2458                 }
2459         }
2460 #endif
2461
2462 err_exit:
2463         rtnl_unlock();
2464         return err;
2465 }
2466
2467 static int ipv6_mc_config(struct sock *sk, bool join,
2468                           const struct in6_addr *addr, int ifindex)
2469 {
2470         int ret;
2471
2472         ASSERT_RTNL();
2473
2474         lock_sock(sk);
2475         if (join)
2476                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2477         else
2478                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2479         release_sock(sk);
2480
2481         return ret;
2482 }
2483
2484 /*
2485  *      Manual configuration of address on an interface
2486  */
2487 static int inet6_addr_add(struct net *net, int ifindex,
2488                           const struct in6_addr *pfx,
2489                           const struct in6_addr *peer_pfx,
2490                           unsigned int plen, __u32 ifa_flags,
2491                           __u32 prefered_lft, __u32 valid_lft)
2492 {
2493         struct inet6_ifaddr *ifp;
2494         struct inet6_dev *idev;
2495         struct net_device *dev;
2496         unsigned long timeout;
2497         clock_t expires;
2498         int scope;
2499         u32 flags;
2500
2501         ASSERT_RTNL();
2502
2503         if (plen > 128)
2504                 return -EINVAL;
2505
2506         /* check the lifetime */
2507         if (!valid_lft || prefered_lft > valid_lft)
2508                 return -EINVAL;
2509
2510         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2511                 return -EINVAL;
2512
2513         dev = __dev_get_by_index(net, ifindex);
2514         if (!dev)
2515                 return -ENODEV;
2516
2517         idev = addrconf_add_dev(dev);
2518         if (IS_ERR(idev))
2519                 return PTR_ERR(idev);
2520
2521         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2522                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2523                                          true, pfx, ifindex);
2524
2525                 if (ret < 0)
2526                         return ret;
2527         }
2528
2529         scope = ipv6_addr_scope(pfx);
2530
2531         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2532         if (addrconf_finite_timeout(timeout)) {
2533                 expires = jiffies_to_clock_t(timeout * HZ);
2534                 valid_lft = timeout;
2535                 flags = RTF_EXPIRES;
2536         } else {
2537                 expires = 0;
2538                 flags = 0;
2539                 ifa_flags |= IFA_F_PERMANENT;
2540         }
2541
2542         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2543         if (addrconf_finite_timeout(timeout)) {
2544                 if (timeout == 0)
2545                         ifa_flags |= IFA_F_DEPRECATED;
2546                 prefered_lft = timeout;
2547         }
2548
2549         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2550                             valid_lft, prefered_lft);
2551
2552         if (!IS_ERR(ifp)) {
2553                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2554                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2555                                               expires, flags);
2556                 }
2557
2558                 /*
2559                  * Note that section 3.1 of RFC 4429 indicates
2560                  * that the Optimistic flag should not be set for
2561                  * manually configured addresses
2562                  */
2563                 addrconf_dad_start(ifp);
2564                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2565                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2566                                          true, jiffies);
2567                 in6_ifa_put(ifp);
2568                 addrconf_verify_rtnl();
2569                 return 0;
2570         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2571                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2572                                false, pfx, ifindex);
2573         }
2574
2575         return PTR_ERR(ifp);
2576 }
2577
2578 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2579                           const struct in6_addr *pfx, unsigned int plen)
2580 {
2581         struct inet6_ifaddr *ifp;
2582         struct inet6_dev *idev;
2583         struct net_device *dev;
2584
2585         if (plen > 128)
2586                 return -EINVAL;
2587
2588         dev = __dev_get_by_index(net, ifindex);
2589         if (!dev)
2590                 return -ENODEV;
2591
2592         idev = __in6_dev_get(dev);
2593         if (idev == NULL)
2594                 return -ENXIO;
2595
2596         read_lock_bh(&idev->lock);
2597         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2598                 if (ifp->prefix_len == plen &&
2599                     ipv6_addr_equal(pfx, &ifp->addr)) {
2600                         in6_ifa_hold(ifp);
2601                         read_unlock_bh(&idev->lock);
2602
2603                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2604                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2605                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2606                                                  jiffies);
2607                         ipv6_del_addr(ifp);
2608                         addrconf_verify_rtnl();
2609                         if (ipv6_addr_is_multicast(pfx)) {
2610                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2611                                                false, pfx, dev->ifindex);
2612                         }
2613                         return 0;
2614                 }
2615         }
2616         read_unlock_bh(&idev->lock);
2617         return -EADDRNOTAVAIL;
2618 }
2619
2620
2621 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2622 {
2623         struct in6_ifreq ireq;
2624         int err;
2625
2626         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2627                 return -EPERM;
2628
2629         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2630                 return -EFAULT;
2631
2632         rtnl_lock();
2633         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2634                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2635                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2636         rtnl_unlock();
2637         return err;
2638 }
2639
2640 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2641 {
2642         struct in6_ifreq ireq;
2643         int err;
2644
2645         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2646                 return -EPERM;
2647
2648         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2649                 return -EFAULT;
2650
2651         rtnl_lock();
2652         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2653                              ireq.ifr6_prefixlen);
2654         rtnl_unlock();
2655         return err;
2656 }
2657
2658 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2659                      int plen, int scope)
2660 {
2661         struct inet6_ifaddr *ifp;
2662
2663         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2664                             scope, IFA_F_PERMANENT,
2665                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2666         if (!IS_ERR(ifp)) {
2667                 spin_lock_bh(&ifp->lock);
2668                 ifp->flags &= ~IFA_F_TENTATIVE;
2669                 spin_unlock_bh(&ifp->lock);
2670                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2671                 in6_ifa_put(ifp);
2672         }
2673 }
2674
2675 #if IS_ENABLED(CONFIG_IPV6_SIT)
2676 static void sit_add_v4_addrs(struct inet6_dev *idev)
2677 {
2678         struct in6_addr addr;
2679         struct net_device *dev;
2680         struct net *net = dev_net(idev->dev);
2681         int scope, plen;
2682         u32 pflags = 0;
2683
2684         ASSERT_RTNL();
2685
2686         memset(&addr, 0, sizeof(struct in6_addr));
2687         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2688
2689         if (idev->dev->flags&IFF_POINTOPOINT) {
2690                 addr.s6_addr32[0] = htonl(0xfe800000);
2691                 scope = IFA_LINK;
2692                 plen = 64;
2693         } else {
2694                 scope = IPV6_ADDR_COMPATv4;
2695                 plen = 96;
2696                 pflags |= RTF_NONEXTHOP;
2697         }
2698
2699         if (addr.s6_addr32[3]) {
2700                 add_addr(idev, &addr, plen, scope);
2701                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2702                 return;
2703         }
2704
2705         for_each_netdev(net, dev) {
2706                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2707                 if (in_dev && (dev->flags & IFF_UP)) {
2708                         struct in_ifaddr *ifa;
2709
2710                         int flag = scope;
2711
2712                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2713
2714                                 addr.s6_addr32[3] = ifa->ifa_local;
2715
2716                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2717                                         continue;
2718                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2719                                         if (idev->dev->flags&IFF_POINTOPOINT)
2720                                                 continue;
2721                                         flag |= IFA_HOST;
2722                                 }
2723
2724                                 add_addr(idev, &addr, plen, flag);
2725                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2726                                                       pflags);
2727                         }
2728                 }
2729         }
2730 }
2731 #endif
2732
2733 static void init_loopback(struct net_device *dev)
2734 {
2735         struct inet6_dev  *idev;
2736         struct net_device *sp_dev;
2737         struct inet6_ifaddr *sp_ifa;
2738         struct rt6_info *sp_rt;
2739
2740         /* ::1 */
2741
2742         ASSERT_RTNL();
2743
2744         idev = ipv6_find_idev(dev);
2745         if (idev == NULL) {
2746                 pr_debug("%s: add_dev failed\n", __func__);
2747                 return;
2748         }
2749
2750         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2751
2752         /* Add routes to other interface's IPv6 addresses */
2753         for_each_netdev(dev_net(dev), sp_dev) {
2754                 if (!strcmp(sp_dev->name, dev->name))
2755                         continue;
2756
2757                 idev = __in6_dev_get(sp_dev);
2758                 if (!idev)
2759                         continue;
2760
2761                 read_lock_bh(&idev->lock);
2762                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2763
2764                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2765                                 continue;
2766
2767                         if (sp_ifa->rt) {
2768                                 /* This dst has been added to garbage list when
2769                                  * lo device down, release this obsolete dst and
2770                                  * reallocate a new router for ifa.
2771                                  */
2772                                 if (sp_ifa->rt->dst.obsolete > 0) {
2773                                         ip6_rt_put(sp_ifa->rt);
2774                                         sp_ifa->rt = NULL;
2775                                 } else {
2776                                         continue;
2777                                 }
2778                         }
2779
2780                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2781
2782                         /* Failure cases are ignored */
2783                         if (!IS_ERR(sp_rt)) {
2784                                 sp_ifa->rt = sp_rt;
2785                                 ip6_ins_rt(sp_rt);
2786                         }
2787                 }
2788                 read_unlock_bh(&idev->lock);
2789         }
2790 }
2791
2792 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2793 {
2794         struct inet6_ifaddr *ifp;
2795         u32 addr_flags = IFA_F_PERMANENT;
2796
2797 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2798         if (idev->cnf.optimistic_dad &&
2799             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2800                 addr_flags |= IFA_F_OPTIMISTIC;
2801 #endif
2802
2803
2804         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2805                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2806         if (!IS_ERR(ifp)) {
2807                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2808                 addrconf_dad_start(ifp);
2809                 in6_ifa_put(ifp);
2810         }
2811 }
2812
2813 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2814 {
2815         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2816                 struct in6_addr addr;
2817
2818                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2819                 /* addrconf_add_linklocal also adds a prefix_route and we
2820                  * only need to care about prefix routes if ipv6_generate_eui64
2821                  * couldn't generate one.
2822                  */
2823                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2824                         addrconf_add_linklocal(idev, &addr);
2825                 else if (prefix_route)
2826                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2827         }
2828 }
2829
2830 static void addrconf_dev_config(struct net_device *dev)
2831 {
2832         struct inet6_dev *idev;
2833
2834         ASSERT_RTNL();
2835
2836         if ((dev->type != ARPHRD_ETHER) &&
2837             (dev->type != ARPHRD_FDDI) &&
2838             (dev->type != ARPHRD_ARCNET) &&
2839             (dev->type != ARPHRD_INFINIBAND) &&
2840             (dev->type != ARPHRD_IEEE802154) &&
2841             (dev->type != ARPHRD_IEEE1394) &&
2842             (dev->type != ARPHRD_TUNNEL6) &&
2843             (dev->type != ARPHRD_6LOWPAN)) {
2844                 /* Alas, we support only Ethernet autoconfiguration. */
2845                 return;
2846         }
2847
2848         idev = addrconf_add_dev(dev);
2849         if (IS_ERR(idev))
2850                 return;
2851
2852         addrconf_addr_gen(idev, false);
2853 }
2854
2855 #if IS_ENABLED(CONFIG_IPV6_SIT)
2856 static void addrconf_sit_config(struct net_device *dev)
2857 {
2858         struct inet6_dev *idev;
2859
2860         ASSERT_RTNL();
2861
2862         /*
2863          * Configure the tunnel with one of our IPv4
2864          * addresses... we should configure all of
2865          * our v4 addrs in the tunnel
2866          */
2867
2868         idev = ipv6_find_idev(dev);
2869         if (idev == NULL) {
2870                 pr_debug("%s: add_dev failed\n", __func__);
2871                 return;
2872         }
2873
2874         if (dev->priv_flags & IFF_ISATAP) {
2875                 addrconf_addr_gen(idev, false);
2876                 return;
2877         }
2878
2879         sit_add_v4_addrs(idev);
2880
2881         if (dev->flags&IFF_POINTOPOINT)
2882                 addrconf_add_mroute(dev);
2883 }
2884 #endif
2885
2886 #if IS_ENABLED(CONFIG_NET_IPGRE)
2887 static void addrconf_gre_config(struct net_device *dev)
2888 {
2889         struct inet6_dev *idev;
2890
2891         ASSERT_RTNL();
2892
2893         idev = ipv6_find_idev(dev);
2894         if (idev == NULL) {
2895                 pr_debug("%s: add_dev failed\n", __func__);
2896                 return;
2897         }
2898
2899         addrconf_addr_gen(idev, true);
2900 }
2901 #endif
2902
2903 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2904                            void *ptr)
2905 {
2906         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2907         struct inet6_dev *idev = __in6_dev_get(dev);
2908         int run_pending = 0;
2909         int err;
2910
2911         switch (event) {
2912         case NETDEV_REGISTER:
2913                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2914                         idev = ipv6_add_dev(dev);
2915                         if (IS_ERR(idev))
2916                                 return notifier_from_errno(PTR_ERR(idev));
2917                 }
2918                 break;
2919
2920         case NETDEV_UP:
2921         case NETDEV_CHANGE:
2922                 if (dev->flags & IFF_SLAVE)
2923                         break;
2924
2925                 if (idev && idev->cnf.disable_ipv6)
2926                         break;
2927
2928                 if (event == NETDEV_UP) {
2929                         if (!addrconf_qdisc_ok(dev)) {
2930                                 /* device is not ready yet. */
2931                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2932                                         dev->name);
2933                                 break;
2934                         }
2935
2936                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2937                                 idev = ipv6_add_dev(dev);
2938
2939                         if (!IS_ERR_OR_NULL(idev)) {
2940                                 idev->if_flags |= IF_READY;
2941                                 run_pending = 1;
2942                         }
2943                 } else {
2944                         if (!addrconf_qdisc_ok(dev)) {
2945                                 /* device is still not ready. */
2946                                 break;
2947                         }
2948
2949                         if (idev) {
2950                                 if (idev->if_flags & IF_READY)
2951                                         /* device is already configured. */
2952                                         break;
2953                                 idev->if_flags |= IF_READY;
2954                         }
2955
2956                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2957                                 dev->name);
2958
2959                         run_pending = 1;
2960                 }
2961
2962                 switch (dev->type) {
2963 #if IS_ENABLED(CONFIG_IPV6_SIT)
2964                 case ARPHRD_SIT:
2965                         addrconf_sit_config(dev);
2966                         break;
2967 #endif
2968 #if IS_ENABLED(CONFIG_NET_IPGRE)
2969                 case ARPHRD_IPGRE:
2970                         addrconf_gre_config(dev);
2971                         break;
2972 #endif
2973                 case ARPHRD_LOOPBACK:
2974                         init_loopback(dev);
2975                         break;
2976
2977                 default:
2978                         addrconf_dev_config(dev);
2979                         break;
2980                 }
2981
2982                 if (!IS_ERR_OR_NULL(idev)) {
2983                         if (run_pending)
2984                                 addrconf_dad_run(idev);
2985
2986                         /*
2987                          * If the MTU changed during the interface down,
2988                          * when the interface up, the changed MTU must be
2989                          * reflected in the idev as well as routers.
2990                          */
2991                         if (idev->cnf.mtu6 != dev->mtu &&
2992                             dev->mtu >= IPV6_MIN_MTU) {
2993                                 rt6_mtu_change(dev, dev->mtu);
2994                                 idev->cnf.mtu6 = dev->mtu;
2995                         }
2996                         idev->tstamp = jiffies;
2997                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2998
2999                         /*
3000                          * If the changed mtu during down is lower than
3001                          * IPV6_MIN_MTU stop IPv6 on this interface.
3002                          */
3003                         if (dev->mtu < IPV6_MIN_MTU)
3004                                 addrconf_ifdown(dev, 1);
3005                 }
3006                 break;
3007
3008         case NETDEV_CHANGEMTU:
3009                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3010                         rt6_mtu_change(dev, dev->mtu);
3011                         idev->cnf.mtu6 = dev->mtu;
3012                         break;
3013                 }
3014
3015                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3016                         idev = ipv6_add_dev(dev);
3017                         if (!IS_ERR(idev))
3018                                 break;
3019                 }
3020
3021                 /*
3022                  * if MTU under IPV6_MIN_MTU.
3023                  * Stop IPv6 on this interface.
3024                  */
3025
3026         case NETDEV_DOWN:
3027         case NETDEV_UNREGISTER:
3028                 /*
3029                  *      Remove all addresses from this interface.
3030                  */
3031                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3032                 break;
3033
3034         case NETDEV_CHANGENAME:
3035                 if (idev) {
3036                         snmp6_unregister_dev(idev);
3037                         addrconf_sysctl_unregister(idev);
3038                         err = addrconf_sysctl_register(idev);
3039                         if (err)
3040                                 return notifier_from_errno(err);
3041                         err = snmp6_register_dev(idev);
3042                         if (err) {
3043                                 addrconf_sysctl_unregister(idev);
3044                                 return notifier_from_errno(err);
3045                         }
3046                 }
3047                 break;
3048
3049         case NETDEV_PRE_TYPE_CHANGE:
3050         case NETDEV_POST_TYPE_CHANGE:
3051                 addrconf_type_change(dev, event);
3052                 break;
3053         }
3054
3055         return NOTIFY_OK;
3056 }
3057
3058 /*
3059  *      addrconf module should be notified of a device going up
3060  */
3061 static struct notifier_block ipv6_dev_notf = {
3062         .notifier_call = addrconf_notify,
3063 };
3064
3065 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3066 {
3067         struct inet6_dev *idev;
3068         ASSERT_RTNL();
3069
3070         idev = __in6_dev_get(dev);
3071
3072         if (event == NETDEV_POST_TYPE_CHANGE)
3073                 ipv6_mc_remap(idev);
3074         else if (event == NETDEV_PRE_TYPE_CHANGE)
3075                 ipv6_mc_unmap(idev);
3076 }
3077
3078 static int addrconf_ifdown(struct net_device *dev, int how)
3079 {
3080         struct net *net = dev_net(dev);
3081         struct inet6_dev *idev;
3082         struct inet6_ifaddr *ifa;
3083         int state, i;
3084
3085         ASSERT_RTNL();
3086
3087         rt6_ifdown(net, dev);
3088         neigh_ifdown(&nd_tbl, dev);
3089
3090         idev = __in6_dev_get(dev);
3091         if (idev == NULL)
3092                 return -ENODEV;
3093
3094         /*
3095          * Step 1: remove reference to ipv6 device from parent device.
3096          *         Do not dev_put!
3097          */
3098         if (how) {
3099                 idev->dead = 1;
3100
3101                 /* protected by rtnl_lock */
3102                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3103
3104                 /* Step 1.5: remove snmp6 entry */
3105                 snmp6_unregister_dev(idev);
3106
3107         }
3108
3109         /* Step 2: clear hash table */
3110         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3111                 struct hlist_head *h = &inet6_addr_lst[i];
3112
3113                 spin_lock_bh(&addrconf_hash_lock);
3114 restart:
3115                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3116                         if (ifa->idev == idev) {
3117                                 hlist_del_init_rcu(&ifa->addr_lst);
3118                                 addrconf_del_dad_work(ifa);
3119                                 goto restart;
3120                         }
3121                 }
3122                 spin_unlock_bh(&addrconf_hash_lock);
3123         }
3124
3125         write_lock_bh(&idev->lock);
3126
3127         addrconf_del_rs_timer(idev);
3128
3129         /* Step 2: clear flags for stateless addrconf */
3130         if (!how)
3131                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3132
3133         if (how && del_timer(&idev->regen_timer))
3134                 in6_dev_put(idev);
3135
3136         /* Step 3: clear tempaddr list */
3137         while (!list_empty(&idev->tempaddr_list)) {
3138                 ifa = list_first_entry(&idev->tempaddr_list,
3139                                        struct inet6_ifaddr, tmp_list);
3140                 list_del(&ifa->tmp_list);
3141                 write_unlock_bh(&idev->lock);
3142                 spin_lock_bh(&ifa->lock);
3143
3144                 if (ifa->ifpub) {
3145                         in6_ifa_put(ifa->ifpub);
3146                         ifa->ifpub = NULL;
3147                 }
3148                 spin_unlock_bh(&ifa->lock);
3149                 in6_ifa_put(ifa);
3150                 write_lock_bh(&idev->lock);
3151         }
3152
3153         while (!list_empty(&idev->addr_list)) {
3154                 ifa = list_first_entry(&idev->addr_list,
3155                                        struct inet6_ifaddr, if_list);
3156                 addrconf_del_dad_work(ifa);
3157
3158                 list_del(&ifa->if_list);
3159
3160                 write_unlock_bh(&idev->lock);
3161
3162                 spin_lock_bh(&ifa->state_lock);
3163                 state = ifa->state;
3164                 ifa->state = INET6_IFADDR_STATE_DEAD;
3165                 spin_unlock_bh(&ifa->state_lock);
3166
3167                 if (state != INET6_IFADDR_STATE_DEAD) {
3168                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3169                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3170                 }
3171                 in6_ifa_put(ifa);
3172
3173                 write_lock_bh(&idev->lock);
3174         }
3175
3176         write_unlock_bh(&idev->lock);
3177
3178         /* Step 5: Discard anycast and multicast list */
3179         if (how) {
3180                 ipv6_ac_destroy_dev(idev);
3181                 ipv6_mc_destroy_dev(idev);
3182         } else {
3183                 ipv6_mc_down(idev);
3184         }
3185
3186         idev->tstamp = jiffies;
3187
3188         /* Last: Shot the device (if unregistered) */
3189         if (how) {
3190                 addrconf_sysctl_unregister(idev);
3191                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3192                 neigh_ifdown(&nd_tbl, dev);
3193                 in6_dev_put(idev);
3194         }
3195         return 0;
3196 }
3197
3198 static void addrconf_rs_timer(unsigned long data)
3199 {
3200         struct inet6_dev *idev = (struct inet6_dev *)data;
3201         struct net_device *dev = idev->dev;
3202         struct in6_addr lladdr;
3203
3204         write_lock(&idev->lock);
3205         if (idev->dead || !(idev->if_flags & IF_READY))
3206                 goto out;
3207
3208         if (!ipv6_accept_ra(idev))
3209                 goto out;
3210
3211         /* Announcement received after solicitation was sent */
3212         if (idev->if_flags & IF_RA_RCVD)
3213                 goto out;
3214
3215         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3216                 write_unlock(&idev->lock);
3217                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3218                         ndisc_send_rs(dev, &lladdr,
3219                                       &in6addr_linklocal_allrouters);
3220                 else
3221                         goto put;
3222
3223                 write_lock(&idev->lock);
3224                 /* The wait after the last probe can be shorter */
3225                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3226                                              idev->cnf.rtr_solicits) ?
3227                                       idev->cnf.rtr_solicit_delay :
3228                                       idev->cnf.rtr_solicit_interval);
3229         } else {
3230                 /*
3231                  * Note: we do not support deprecated "all on-link"
3232                  * assumption any longer.
3233                  */
3234                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3235         }
3236
3237 out:
3238         write_unlock(&idev->lock);
3239 put:
3240         in6_dev_put(idev);
3241 }
3242
3243 /*
3244  *      Duplicate Address Detection
3245  */
3246 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3247 {
3248         unsigned long rand_num;
3249         struct inet6_dev *idev = ifp->idev;
3250
3251         if (ifp->flags & IFA_F_OPTIMISTIC)
3252                 rand_num = 0;
3253         else
3254                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3255
3256         ifp->dad_probes = idev->cnf.dad_transmits;
3257         addrconf_mod_dad_work(ifp, rand_num);
3258 }
3259
3260 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3261 {
3262         struct inet6_dev *idev = ifp->idev;
3263         struct net_device *dev = idev->dev;
3264
3265         addrconf_join_solict(dev, &ifp->addr);
3266
3267         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3268
3269         read_lock_bh(&idev->lock);
3270         spin_lock(&ifp->lock);
3271         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3272                 goto out;
3273
3274         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3275             idev->cnf.accept_dad < 1 ||
3276             !(ifp->flags&IFA_F_TENTATIVE) ||
3277             ifp->flags & IFA_F_NODAD) {
3278                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3279                 spin_unlock(&ifp->lock);
3280                 read_unlock_bh(&idev->lock);
3281
3282                 addrconf_dad_completed(ifp);
3283                 return;
3284         }
3285
3286         if (!(idev->if_flags & IF_READY)) {
3287                 spin_unlock(&ifp->lock);
3288                 read_unlock_bh(&idev->lock);
3289                 /*
3290                  * If the device is not ready:
3291                  * - keep it tentative if it is a permanent address.
3292                  * - otherwise, kill it.
3293                  */
3294                 in6_ifa_hold(ifp);
3295                 addrconf_dad_stop(ifp, 0);
3296                 return;
3297         }
3298
3299         /*
3300          * Optimistic nodes can start receiving
3301          * Frames right away
3302          */
3303         if (ifp->flags & IFA_F_OPTIMISTIC) {
3304                 ip6_ins_rt(ifp->rt);
3305                 if (ipv6_use_optimistic_addr(idev)) {
3306                         /* Because optimistic nodes can use this address,
3307                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3308                          */
3309                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3310                 }
3311         }
3312
3313         addrconf_dad_kick(ifp);
3314 out:
3315         spin_unlock(&ifp->lock);
3316         read_unlock_bh(&idev->lock);
3317 }
3318
3319 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3320 {
3321         bool begin_dad = false;
3322
3323         spin_lock_bh(&ifp->state_lock);
3324         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3325                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3326                 begin_dad = true;
3327         }
3328         spin_unlock_bh(&ifp->state_lock);
3329
3330         if (begin_dad)
3331                 addrconf_mod_dad_work(ifp, 0);
3332 }
3333
3334 static void addrconf_dad_work(struct work_struct *w)
3335 {
3336         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3337                                                 struct inet6_ifaddr,
3338                                                 dad_work);
3339         struct inet6_dev *idev = ifp->idev;
3340         struct in6_addr mcaddr;
3341
3342         enum {
3343                 DAD_PROCESS,
3344                 DAD_BEGIN,
3345                 DAD_ABORT,
3346         } action = DAD_PROCESS;
3347
3348         rtnl_lock();
3349
3350         spin_lock_bh(&ifp->state_lock);
3351         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3352                 action = DAD_BEGIN;
3353                 ifp->state = INET6_IFADDR_STATE_DAD;
3354         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3355                 action = DAD_ABORT;
3356                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3357         }
3358         spin_unlock_bh(&ifp->state_lock);
3359
3360         if (action == DAD_BEGIN) {
3361                 addrconf_dad_begin(ifp);
3362                 goto out;
3363         } else if (action == DAD_ABORT) {
3364                 addrconf_dad_stop(ifp, 1);
3365                 goto out;
3366         }
3367
3368         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3369                 goto out;
3370
3371         write_lock_bh(&idev->lock);
3372         if (idev->dead || !(idev->if_flags & IF_READY)) {
3373                 write_unlock_bh(&idev->lock);
3374                 goto out;
3375         }
3376
3377         spin_lock(&ifp->lock);
3378         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3379                 spin_unlock(&ifp->lock);
3380                 write_unlock_bh(&idev->lock);
3381                 goto out;
3382         }
3383
3384         if (ifp->dad_probes == 0) {
3385                 /*
3386                  * DAD was successful
3387                  */
3388
3389                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3390                 spin_unlock(&ifp->lock);
3391                 write_unlock_bh(&idev->lock);
3392
3393                 addrconf_dad_completed(ifp);
3394
3395                 goto out;
3396         }
3397
3398         ifp->dad_probes--;
3399         addrconf_mod_dad_work(ifp,
3400                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3401         spin_unlock(&ifp->lock);
3402         write_unlock_bh(&idev->lock);
3403
3404         /* send a neighbour solicitation for our addr */
3405         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3406         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3407 out:
3408         in6_ifa_put(ifp);
3409         rtnl_unlock();
3410 }
3411
3412 /* ifp->idev must be at least read locked */
3413 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3414 {
3415         struct inet6_ifaddr *ifpiter;
3416         struct inet6_dev *idev = ifp->idev;
3417
3418         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3419                 if (ifpiter->scope > IFA_LINK)
3420                         break;
3421                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3422                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3423                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3424                     IFA_F_PERMANENT)
3425                         return false;
3426         }
3427         return true;
3428 }
3429
3430 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3431 {
3432         struct net_device *dev = ifp->idev->dev;
3433         struct in6_addr lladdr;
3434         bool send_rs, send_mld;
3435
3436         addrconf_del_dad_work(ifp);
3437
3438         /*
3439          *      Configure the address for reception. Now it is valid.
3440          */
3441
3442         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3443
3444         /* If added prefix is link local and we are prepared to process
3445            router advertisements, start sending router solicitations.
3446          */
3447
3448         read_lock_bh(&ifp->idev->lock);
3449         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3450         send_rs = send_mld &&
3451                   ipv6_accept_ra(ifp->idev) &&
3452                   ifp->idev->cnf.rtr_solicits > 0 &&
3453                   (dev->flags&IFF_LOOPBACK) == 0;
3454         read_unlock_bh(&ifp->idev->lock);
3455
3456         /* While dad is in progress mld report's source address is in6_addrany.
3457          * Resend with proper ll now.
3458          */
3459         if (send_mld)
3460                 ipv6_mc_dad_complete(ifp->idev);
3461
3462         if (send_rs) {
3463                 /*
3464                  *      If a host as already performed a random delay
3465                  *      [...] as part of DAD [...] there is no need
3466                  *      to delay again before sending the first RS
3467                  */
3468                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3469                         return;
3470                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3471
3472                 write_lock_bh(&ifp->idev->lock);
3473                 spin_lock(&ifp->lock);
3474                 ifp->idev->rs_probes = 1;
3475                 ifp->idev->if_flags |= IF_RS_SENT;
3476                 addrconf_mod_rs_timer(ifp->idev,
3477                                       ifp->idev->cnf.rtr_solicit_interval);
3478                 spin_unlock(&ifp->lock);
3479                 write_unlock_bh(&ifp->idev->lock);
3480         }
3481 }
3482
3483 static void addrconf_dad_run(struct inet6_dev *idev)
3484 {
3485         struct inet6_ifaddr *ifp;
3486
3487         read_lock_bh(&idev->lock);
3488         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3489                 spin_lock(&ifp->lock);
3490                 if (ifp->flags & IFA_F_TENTATIVE &&
3491                     ifp->state == INET6_IFADDR_STATE_DAD)
3492                         addrconf_dad_kick(ifp);
3493                 spin_unlock(&ifp->lock);
3494         }
3495         read_unlock_bh(&idev->lock);
3496 }
3497
3498 #ifdef CONFIG_PROC_FS
3499 struct if6_iter_state {
3500         struct seq_net_private p;
3501         int bucket;
3502         int offset;
3503 };
3504
3505 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3506 {
3507         struct inet6_ifaddr *ifa = NULL;
3508         struct if6_iter_state *state = seq->private;
3509         struct net *net = seq_file_net(seq);
3510         int p = 0;
3511
3512         /* initial bucket if pos is 0 */
3513         if (pos == 0) {
3514                 state->bucket = 0;
3515                 state->offset = 0;
3516         }
3517
3518         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3519                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3520                                          addr_lst) {
3521                         if (!net_eq(dev_net(ifa->idev->dev), net))
3522                                 continue;
3523                         /* sync with offset */
3524                         if (p < state->offset) {
3525                                 p++;
3526                                 continue;
3527                         }
3528                         state->offset++;
3529                         return ifa;
3530                 }
3531
3532                 /* prepare for next bucket */
3533                 state->offset = 0;
3534                 p = 0;
3535         }
3536         return NULL;
3537 }
3538
3539 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3540                                          struct inet6_ifaddr *ifa)
3541 {
3542         struct if6_iter_state *state = seq->private;
3543         struct net *net = seq_file_net(seq);
3544
3545         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3546                 if (!net_eq(dev_net(ifa->idev->dev), net))
3547                         continue;
3548                 state->offset++;
3549                 return ifa;
3550         }
3551
3552         while (++state->bucket < IN6_ADDR_HSIZE) {
3553                 state->offset = 0;
3554                 hlist_for_each_entry_rcu_bh(ifa,
3555                                      &inet6_addr_lst[state->bucket], addr_lst) {
3556                         if (!net_eq(dev_net(ifa->idev->dev), net))
3557                                 continue;
3558                         state->offset++;
3559                         return ifa;
3560                 }
3561         }
3562
3563         return NULL;
3564 }
3565
3566 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3567         __acquires(rcu_bh)
3568 {
3569         rcu_read_lock_bh();
3570         return if6_get_first(seq, *pos);
3571 }
3572
3573 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3574 {
3575         struct inet6_ifaddr *ifa;
3576
3577         ifa = if6_get_next(seq, v);
3578         ++*pos;
3579         return ifa;
3580 }
3581
3582 static void if6_seq_stop(struct seq_file *seq, void *v)
3583         __releases(rcu_bh)
3584 {
3585         rcu_read_unlock_bh();
3586 }
3587
3588 static int if6_seq_show(struct seq_file *seq, void *v)
3589 {
3590         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3591         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3592                    &ifp->addr,
3593                    ifp->idev->dev->ifindex,
3594                    ifp->prefix_len,
3595                    ifp->scope,
3596                    (u8) ifp->flags,
3597                    ifp->idev->dev->name);
3598         return 0;
3599 }
3600
3601 static const struct seq_operations if6_seq_ops = {
3602         .start  = if6_seq_start,
3603         .next   = if6_seq_next,
3604         .show   = if6_seq_show,
3605         .stop   = if6_seq_stop,
3606 };
3607
3608 static int if6_seq_open(struct inode *inode, struct file *file)
3609 {
3610         return seq_open_net(inode, file, &if6_seq_ops,
3611                             sizeof(struct if6_iter_state));
3612 }
3613
3614 static const struct file_operations if6_fops = {
3615         .owner          = THIS_MODULE,
3616         .open           = if6_seq_open,
3617         .read           = seq_read,
3618         .llseek         = seq_lseek,
3619         .release        = seq_release_net,
3620 };
3621
3622 static int __net_init if6_proc_net_init(struct net *net)
3623 {
3624         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3625                 return -ENOMEM;
3626         return 0;
3627 }
3628
3629 static void __net_exit if6_proc_net_exit(struct net *net)
3630 {
3631         remove_proc_entry("if_inet6", net->proc_net);
3632 }
3633
3634 static struct pernet_operations if6_proc_net_ops = {
3635         .init = if6_proc_net_init,
3636         .exit = if6_proc_net_exit,
3637 };
3638
3639 int __init if6_proc_init(void)
3640 {
3641         return register_pernet_subsys(&if6_proc_net_ops);
3642 }
3643
3644 void if6_proc_exit(void)
3645 {
3646         unregister_pernet_subsys(&if6_proc_net_ops);
3647 }
3648 #endif  /* CONFIG_PROC_FS */
3649
3650 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3651 /* Check if address is a home address configured on any interface. */
3652 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3653 {
3654         int ret = 0;
3655         struct inet6_ifaddr *ifp = NULL;
3656         unsigned int hash = inet6_addr_hash(addr);
3657
3658         rcu_read_lock_bh();
3659         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3660                 if (!net_eq(dev_net(ifp->idev->dev), net))
3661                         continue;
3662                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3663                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3664                         ret = 1;
3665                         break;
3666                 }
3667         }
3668         rcu_read_unlock_bh();
3669         return ret;
3670 }
3671 #endif
3672
3673 /*
3674  *      Periodic address status verification
3675  */
3676
3677 static void addrconf_verify_rtnl(void)
3678 {
3679         unsigned long now, next, next_sec, next_sched;
3680         struct inet6_ifaddr *ifp;
3681         int i;
3682
3683         ASSERT_RTNL();
3684
3685         rcu_read_lock_bh();
3686         now = jiffies;
3687         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3688
3689         cancel_delayed_work(&addr_chk_work);
3690
3691         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3692 restart:
3693                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3694                         unsigned long age;
3695
3696                         /* When setting preferred_lft to a value not zero or
3697                          * infinity, while valid_lft is infinity
3698                          * IFA_F_PERMANENT has a non-infinity life time.
3699                          */
3700                         if ((ifp->flags & IFA_F_PERMANENT) &&
3701                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3702                                 continue;
3703
3704                         spin_lock(&ifp->lock);
3705                         /* We try to batch several events at once. */
3706                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3707
3708                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3709                             age >= ifp->valid_lft) {
3710                                 spin_unlock(&ifp->lock);
3711                                 in6_ifa_hold(ifp);
3712                                 ipv6_del_addr(ifp);
3713                                 goto restart;
3714                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3715                                 spin_unlock(&ifp->lock);
3716                                 continue;
3717                         } else if (age >= ifp->prefered_lft) {
3718                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3719                                 int deprecate = 0;
3720
3721                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3722                                         deprecate = 1;
3723                                         ifp->flags |= IFA_F_DEPRECATED;
3724                                 }
3725
3726                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3727                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3728                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3729
3730                                 spin_unlock(&ifp->lock);
3731
3732                                 if (deprecate) {
3733                                         in6_ifa_hold(ifp);
3734
3735                                         ipv6_ifa_notify(0, ifp);
3736                                         in6_ifa_put(ifp);
3737                                         goto restart;
3738                                 }
3739                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3740                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3741                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3742                                         ifp->idev->cnf.dad_transmits *
3743                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3744
3745                                 if (age >= ifp->prefered_lft - regen_advance) {
3746                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3747                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3748                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3749                                         if (!ifp->regen_count && ifpub) {
3750                                                 ifp->regen_count++;
3751                                                 in6_ifa_hold(ifp);
3752                                                 in6_ifa_hold(ifpub);
3753                                                 spin_unlock(&ifp->lock);
3754
3755                                                 spin_lock(&ifpub->lock);
3756                                                 ifpub->regen_count = 0;
3757                                                 spin_unlock(&ifpub->lock);
3758                                                 ipv6_create_tempaddr(ifpub, ifp);
3759                                                 in6_ifa_put(ifpub);
3760                                                 in6_ifa_put(ifp);
3761                                                 goto restart;
3762                                         }
3763                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3764                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3765                                 spin_unlock(&ifp->lock);
3766                         } else {
3767                                 /* ifp->prefered_lft <= ifp->valid_lft */
3768                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3769                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3770                                 spin_unlock(&ifp->lock);
3771                         }
3772                 }
3773         }
3774
3775         next_sec = round_jiffies_up(next);
3776         next_sched = next;
3777
3778         /* If rounded timeout is accurate enough, accept it. */
3779         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3780                 next_sched = next_sec;
3781
3782         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3783         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3784                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3785
3786         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3787               now, next, next_sec, next_sched);
3788         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3789         rcu_read_unlock_bh();
3790 }
3791
3792 static void addrconf_verify_work(struct work_struct *w)
3793 {
3794         rtnl_lock();
3795         addrconf_verify_rtnl();
3796         rtnl_unlock();
3797 }
3798
3799 static void addrconf_verify(void)
3800 {
3801         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3802 }
3803
3804 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3805                                      struct in6_addr **peer_pfx)
3806 {
3807         struct in6_addr *pfx = NULL;
3808
3809         *peer_pfx = NULL;
3810
3811         if (addr)
3812                 pfx = nla_data(addr);
3813
3814         if (local) {
3815                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3816                         *peer_pfx = pfx;
3817                 pfx = nla_data(local);
3818         }
3819
3820         return pfx;
3821 }
3822
3823 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3824         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3825         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3826         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3827         [IFA_FLAGS]             = { .len = sizeof(u32) },
3828 };
3829
3830 static int
3831 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3832 {
3833         struct net *net = sock_net(skb->sk);
3834         struct ifaddrmsg *ifm;
3835         struct nlattr *tb[IFA_MAX+1];
3836         struct in6_addr *pfx, *peer_pfx;
3837         u32 ifa_flags;
3838         int err;
3839
3840         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3841         if (err < 0)
3842                 return err;
3843
3844         ifm = nlmsg_data(nlh);
3845         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3846         if (pfx == NULL)
3847                 return -EINVAL;
3848
3849         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3850
3851         /* We ignore other flags so far. */
3852         ifa_flags &= IFA_F_MANAGETEMPADDR;
3853
3854         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3855                               ifm->ifa_prefixlen);
3856 }
3857
3858 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3859                              u32 prefered_lft, u32 valid_lft)
3860 {
3861         u32 flags;
3862         clock_t expires;
3863         unsigned long timeout;
3864         bool was_managetempaddr;
3865         bool had_prefixroute;
3866
3867         ASSERT_RTNL();
3868
3869         if (!valid_lft || (prefered_lft > valid_lft))
3870                 return -EINVAL;
3871
3872         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3873             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3874                 return -EINVAL;
3875
3876         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3877         if (addrconf_finite_timeout(timeout)) {
3878                 expires = jiffies_to_clock_t(timeout * HZ);
3879                 valid_lft = timeout;
3880                 flags = RTF_EXPIRES;
3881         } else {
3882                 expires = 0;
3883                 flags = 0;
3884                 ifa_flags |= IFA_F_PERMANENT;
3885         }
3886
3887         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3888         if (addrconf_finite_timeout(timeout)) {
3889                 if (timeout == 0)
3890                         ifa_flags |= IFA_F_DEPRECATED;
3891                 prefered_lft = timeout;
3892         }
3893
3894         spin_lock_bh(&ifp->lock);
3895         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3896         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3897                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
3898         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3899                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3900                         IFA_F_NOPREFIXROUTE);
3901         ifp->flags |= ifa_flags;
3902         ifp->tstamp = jiffies;
3903         ifp->valid_lft = valid_lft;
3904         ifp->prefered_lft = prefered_lft;
3905
3906         spin_unlock_bh(&ifp->lock);
3907         if (!(ifp->flags&IFA_F_TENTATIVE))
3908                 ipv6_ifa_notify(0, ifp);
3909
3910         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3911                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3912                                       expires, flags);
3913         } else if (had_prefixroute) {
3914                 enum cleanup_prefix_rt_t action;
3915                 unsigned long rt_expires;
3916
3917                 write_lock_bh(&ifp->idev->lock);
3918                 action = check_cleanup_prefix_route(ifp, &rt_expires);
3919                 write_unlock_bh(&ifp->idev->lock);
3920
3921                 if (action != CLEANUP_PREFIX_RT_NOP) {
3922                         cleanup_prefix_route(ifp, rt_expires,
3923                                 action == CLEANUP_PREFIX_RT_DEL);
3924                 }
3925         }
3926
3927         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3928                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3929                         valid_lft = prefered_lft = 0;
3930                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3931                                  !was_managetempaddr, jiffies);
3932         }
3933
3934         addrconf_verify_rtnl();
3935
3936         return 0;
3937 }
3938
3939 static int
3940 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3941 {
3942         struct net *net = sock_net(skb->sk);
3943         struct ifaddrmsg *ifm;
3944         struct nlattr *tb[IFA_MAX+1];
3945         struct in6_addr *pfx, *peer_pfx;
3946         struct inet6_ifaddr *ifa;
3947         struct net_device *dev;
3948         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3949         u32 ifa_flags;
3950         int err;
3951
3952         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3953         if (err < 0)
3954                 return err;
3955
3956         ifm = nlmsg_data(nlh);
3957         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3958         if (pfx == NULL)
3959                 return -EINVAL;
3960
3961         if (tb[IFA_CACHEINFO]) {
3962                 struct ifa_cacheinfo *ci;
3963
3964                 ci = nla_data(tb[IFA_CACHEINFO]);
3965                 valid_lft = ci->ifa_valid;
3966                 preferred_lft = ci->ifa_prefered;
3967         } else {
3968                 preferred_lft = INFINITY_LIFE_TIME;
3969                 valid_lft = INFINITY_LIFE_TIME;
3970         }
3971
3972         dev =  __dev_get_by_index(net, ifm->ifa_index);
3973         if (dev == NULL)
3974                 return -ENODEV;
3975
3976         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3977
3978         /* We ignore other flags so far. */
3979         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3980                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
3981
3982         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3983         if (ifa == NULL) {
3984                 /*
3985                  * It would be best to check for !NLM_F_CREATE here but
3986                  * userspace already relies on not having to provide this.
3987                  */
3988                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3989                                       ifm->ifa_prefixlen, ifa_flags,
3990                                       preferred_lft, valid_lft);
3991         }
3992
3993         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3994             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3995                 err = -EEXIST;
3996         else
3997                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3998
3999         in6_ifa_put(ifa);
4000
4001         return err;
4002 }
4003
4004 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4005                           u8 scope, int ifindex)
4006 {
4007         struct ifaddrmsg *ifm;
4008
4009         ifm = nlmsg_data(nlh);
4010         ifm->ifa_family = AF_INET6;
4011         ifm->ifa_prefixlen = prefixlen;
4012         ifm->ifa_flags = flags;
4013         ifm->ifa_scope = scope;
4014         ifm->ifa_index = ifindex;
4015 }
4016
4017 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4018                          unsigned long tstamp, u32 preferred, u32 valid)
4019 {
4020         struct ifa_cacheinfo ci;
4021
4022         ci.cstamp = cstamp_delta(cstamp);
4023         ci.tstamp = cstamp_delta(tstamp);
4024         ci.ifa_prefered = preferred;
4025         ci.ifa_valid = valid;
4026
4027         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4028 }
4029
4030 static inline int rt_scope(int ifa_scope)
4031 {
4032         if (ifa_scope & IFA_HOST)
4033                 return RT_SCOPE_HOST;
4034         else if (ifa_scope & IFA_LINK)
4035                 return RT_SCOPE_LINK;
4036         else if (ifa_scope & IFA_SITE)
4037                 return RT_SCOPE_SITE;
4038         else
4039                 return RT_SCOPE_UNIVERSE;
4040 }
4041
4042 static inline int inet6_ifaddr_msgsize(void)
4043 {
4044         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4045                + nla_total_size(16) /* IFA_LOCAL */
4046                + nla_total_size(16) /* IFA_ADDRESS */
4047                + nla_total_size(sizeof(struct ifa_cacheinfo))
4048                + nla_total_size(4)  /* IFA_FLAGS */;
4049 }
4050
4051 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4052                              u32 portid, u32 seq, int event, unsigned int flags)
4053 {
4054         struct nlmsghdr  *nlh;
4055         u32 preferred, valid;
4056
4057         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4058         if (nlh == NULL)
4059                 return -EMSGSIZE;
4060
4061         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4062                       ifa->idev->dev->ifindex);
4063
4064         if (!((ifa->flags&IFA_F_PERMANENT) &&
4065               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4066                 preferred = ifa->prefered_lft;
4067                 valid = ifa->valid_lft;
4068                 if (preferred != INFINITY_LIFE_TIME) {
4069                         long tval = (jiffies - ifa->tstamp)/HZ;
4070                         if (preferred > tval)
4071                                 preferred -= tval;
4072                         else
4073                                 preferred = 0;
4074                         if (valid != INFINITY_LIFE_TIME) {
4075                                 if (valid > tval)
4076                                         valid -= tval;
4077                                 else
4078                                         valid = 0;
4079                         }
4080                 }
4081         } else {
4082                 preferred = INFINITY_LIFE_TIME;
4083                 valid = INFINITY_LIFE_TIME;
4084         }
4085
4086         if (!ipv6_addr_any(&ifa->peer_addr)) {
4087                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
4088                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
4089                         goto error;
4090         } else
4091                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4092                         goto error;
4093
4094         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4095                 goto error;
4096
4097         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4098                 goto error;
4099
4100         nlmsg_end(skb, nlh);
4101         return 0;
4102
4103 error:
4104         nlmsg_cancel(skb, nlh);
4105         return -EMSGSIZE;
4106 }
4107
4108 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4109                                 u32 portid, u32 seq, int event, u16 flags)
4110 {
4111         struct nlmsghdr  *nlh;
4112         u8 scope = RT_SCOPE_UNIVERSE;
4113         int ifindex = ifmca->idev->dev->ifindex;
4114
4115         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4116                 scope = RT_SCOPE_SITE;
4117
4118         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4119         if (nlh == NULL)
4120                 return -EMSGSIZE;
4121
4122         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4123         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4124             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4125                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4126                 nlmsg_cancel(skb, nlh);
4127                 return -EMSGSIZE;
4128         }
4129
4130         nlmsg_end(skb, nlh);
4131         return 0;
4132 }
4133
4134 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4135                                 u32 portid, u32 seq, int event, unsigned int flags)
4136 {
4137         struct nlmsghdr  *nlh;
4138         u8 scope = RT_SCOPE_UNIVERSE;
4139         int ifindex = ifaca->aca_idev->dev->ifindex;
4140
4141         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4142                 scope = RT_SCOPE_SITE;
4143
4144         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4145         if (nlh == NULL)
4146                 return -EMSGSIZE;
4147
4148         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4149         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4150             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4151                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4152                 nlmsg_cancel(skb, nlh);
4153                 return -EMSGSIZE;
4154         }
4155
4156         nlmsg_end(skb, nlh);
4157         return 0;
4158 }
4159
4160 enum addr_type_t {
4161         UNICAST_ADDR,
4162         MULTICAST_ADDR,
4163         ANYCAST_ADDR,
4164 };
4165
4166 /* called with rcu_read_lock() */
4167 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4168                           struct netlink_callback *cb, enum addr_type_t type,
4169                           int s_ip_idx, int *p_ip_idx)
4170 {
4171         struct ifmcaddr6 *ifmca;
4172         struct ifacaddr6 *ifaca;
4173         int err = 1;
4174         int ip_idx = *p_ip_idx;
4175
4176         read_lock_bh(&idev->lock);
4177         switch (type) {
4178         case UNICAST_ADDR: {
4179                 struct inet6_ifaddr *ifa;
4180
4181                 /* unicast address incl. temp addr */
4182                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4183                         if (++ip_idx < s_ip_idx)
4184                                 continue;
4185                         err = inet6_fill_ifaddr(skb, ifa,
4186                                                 NETLINK_CB(cb->skb).portid,
4187                                                 cb->nlh->nlmsg_seq,
4188                                                 RTM_NEWADDR,
4189                                                 NLM_F_MULTI);
4190                         if (err < 0)
4191                                 break;
4192                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4193                 }
4194                 break;
4195         }
4196         case MULTICAST_ADDR:
4197                 /* multicast address */
4198                 for (ifmca = idev->mc_list; ifmca;
4199                      ifmca = ifmca->next, ip_idx++) {
4200                         if (ip_idx < s_ip_idx)
4201                                 continue;
4202                         err = inet6_fill_ifmcaddr(skb, ifmca,
4203                                                   NETLINK_CB(cb->skb).portid,
4204                                                   cb->nlh->nlmsg_seq,
4205                                                   RTM_GETMULTICAST,
4206                                                   NLM_F_MULTI);
4207                         if (err < 0)
4208                                 break;
4209                 }
4210                 break;
4211         case ANYCAST_ADDR:
4212                 /* anycast address */
4213                 for (ifaca = idev->ac_list; ifaca;
4214                      ifaca = ifaca->aca_next, ip_idx++) {
4215                         if (ip_idx < s_ip_idx)
4216                                 continue;
4217                         err = inet6_fill_ifacaddr(skb, ifaca,
4218                                                   NETLINK_CB(cb->skb).portid,
4219                                                   cb->nlh->nlmsg_seq,
4220                                                   RTM_GETANYCAST,
4221                                                   NLM_F_MULTI);
4222                         if (err < 0)
4223                                 break;
4224                 }
4225                 break;
4226         default:
4227                 break;
4228         }
4229         read_unlock_bh(&idev->lock);
4230         *p_ip_idx = ip_idx;
4231         return err;
4232 }
4233
4234 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4235                            enum addr_type_t type)
4236 {
4237         struct net *net = sock_net(skb->sk);
4238         int h, s_h;
4239         int idx, ip_idx;
4240         int s_idx, s_ip_idx;
4241         struct net_device *dev;
4242         struct inet6_dev *idev;
4243         struct hlist_head *head;
4244
4245         s_h = cb->args[0];
4246         s_idx = idx = cb->args[1];
4247         s_ip_idx = ip_idx = cb->args[2];
4248
4249         rcu_read_lock();
4250         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4251         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4252                 idx = 0;
4253                 head = &net->dev_index_head[h];
4254                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4255                         if (idx < s_idx)
4256                                 goto cont;
4257                         if (h > s_h || idx > s_idx)
4258                                 s_ip_idx = 0;
4259                         ip_idx = 0;
4260                         idev = __in6_dev_get(dev);
4261                         if (!idev)
4262                                 goto cont;
4263
4264                         if (in6_dump_addrs(idev, skb, cb, type,
4265                                            s_ip_idx, &ip_idx) < 0)
4266                                 goto done;
4267 cont:
4268                         idx++;
4269                 }
4270         }
4271 done:
4272         rcu_read_unlock();
4273         cb->args[0] = h;
4274         cb->args[1] = idx;
4275         cb->args[2] = ip_idx;
4276
4277         return skb->len;
4278 }
4279
4280 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4281 {
4282         enum addr_type_t type = UNICAST_ADDR;
4283
4284         return inet6_dump_addr(skb, cb, type);
4285 }
4286
4287 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4288 {
4289         enum addr_type_t type = MULTICAST_ADDR;
4290
4291         return inet6_dump_addr(skb, cb, type);
4292 }
4293
4294
4295 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4296 {
4297         enum addr_type_t type = ANYCAST_ADDR;
4298
4299         return inet6_dump_addr(skb, cb, type);
4300 }
4301
4302 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4303 {
4304         struct net *net = sock_net(in_skb->sk);
4305         struct ifaddrmsg *ifm;
4306         struct nlattr *tb[IFA_MAX+1];
4307         struct in6_addr *addr = NULL, *peer;
4308         struct net_device *dev = NULL;
4309         struct inet6_ifaddr *ifa;
4310         struct sk_buff *skb;
4311         int err;
4312
4313         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4314         if (err < 0)
4315                 goto errout;
4316
4317         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4318         if (addr == NULL) {
4319                 err = -EINVAL;
4320                 goto errout;
4321         }
4322
4323         ifm = nlmsg_data(nlh);
4324         if (ifm->ifa_index)
4325                 dev = __dev_get_by_index(net, ifm->ifa_index);
4326
4327         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4328         if (!ifa) {
4329                 err = -EADDRNOTAVAIL;
4330                 goto errout;
4331         }
4332
4333         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4334         if (!skb) {
4335                 err = -ENOBUFS;
4336                 goto errout_ifa;
4337         }
4338
4339         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4340                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4341         if (err < 0) {
4342                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4343                 WARN_ON(err == -EMSGSIZE);
4344                 kfree_skb(skb);
4345                 goto errout_ifa;
4346         }
4347         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4348 errout_ifa:
4349         in6_ifa_put(ifa);
4350 errout:
4351         return err;
4352 }
4353
4354 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4355 {
4356         struct sk_buff *skb;
4357         struct net *net = dev_net(ifa->idev->dev);
4358         int err = -ENOBUFS;
4359
4360         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4361         if (skb == NULL)
4362                 goto errout;
4363
4364         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4365         if (err < 0) {
4366                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4367                 WARN_ON(err == -EMSGSIZE);
4368                 kfree_skb(skb);
4369                 goto errout;
4370         }
4371         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4372         return;
4373 errout:
4374         if (err < 0)
4375                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4376 }
4377
4378 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4379                                 __s32 *array, int bytes)
4380 {
4381         BUG_ON(bytes < (DEVCONF_MAX * 4));
4382
4383         memset(array, 0, bytes);
4384         array[DEVCONF_FORWARDING] = cnf->forwarding;
4385         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4386         array[DEVCONF_MTU6] = cnf->mtu6;
4387         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4388         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4389         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4390         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4391         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4392         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4393                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4394         array[DEVCONF_RTR_SOLICIT_DELAY] =
4395                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4396         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4397         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4398                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4399         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4400                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4401         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4402         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4403         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4404         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4405         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4406         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4407         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4408         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4409 #ifdef CONFIG_IPV6_ROUTER_PREF
4410         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4411         array[DEVCONF_RTR_PROBE_INTERVAL] =
4412                 jiffies_to_msecs(cnf->rtr_probe_interval);
4413 #ifdef CONFIG_IPV6_ROUTE_INFO
4414         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4415 #endif
4416 #endif
4417         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4418         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4419 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4420         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4421         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4422 #endif
4423 #ifdef CONFIG_IPV6_MROUTE
4424         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4425 #endif
4426         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4427         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4428         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4429         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4430         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4431         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4432         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4433 }
4434
4435 static inline size_t inet6_ifla6_size(void)
4436 {
4437         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4438              + nla_total_size(sizeof(struct ifla_cacheinfo))
4439              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4440              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4441              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4442              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4443 }
4444
4445 static inline size_t inet6_if_nlmsg_size(void)
4446 {
4447         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4448                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4449                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4450                + nla_total_size(4) /* IFLA_MTU */
4451                + nla_total_size(4) /* IFLA_LINK */
4452                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4453 }
4454
4455 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4456                                       int items, int bytes)
4457 {
4458         int i;
4459         int pad = bytes - sizeof(u64) * items;
4460         BUG_ON(pad < 0);
4461
4462         /* Use put_unaligned() because stats may not be aligned for u64. */
4463         put_unaligned(items, &stats[0]);
4464         for (i = 1; i < items; i++)
4465                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4466
4467         memset(&stats[items], 0, pad);
4468 }
4469
4470 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4471                                       int items, int bytes, size_t syncpoff)
4472 {
4473         int i;
4474         int pad = bytes - sizeof(u64) * items;
4475         BUG_ON(pad < 0);
4476
4477         /* Use put_unaligned() because stats may not be aligned for u64. */
4478         put_unaligned(items, &stats[0]);
4479         for (i = 1; i < items; i++)
4480                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4481
4482         memset(&stats[items], 0, pad);
4483 }
4484
4485 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4486                              int bytes)
4487 {
4488         switch (attrtype) {
4489         case IFLA_INET6_STATS:
4490                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4491                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4492                 break;
4493         case IFLA_INET6_ICMP6STATS:
4494                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4495                 break;
4496         }
4497 }
4498
4499 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4500 {
4501         struct nlattr *nla;
4502         struct ifla_cacheinfo ci;
4503
4504         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4505                 goto nla_put_failure;
4506         ci.max_reasm_len = IPV6_MAXPLEN;
4507         ci.tstamp = cstamp_delta(idev->tstamp);
4508         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4509         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4510         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4511                 goto nla_put_failure;
4512         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4513         if (nla == NULL)
4514                 goto nla_put_failure;
4515         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4516
4517         /* XXX - MC not implemented */
4518
4519         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4520         if (nla == NULL)
4521                 goto nla_put_failure;
4522         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4523
4524         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4525         if (nla == NULL)
4526                 goto nla_put_failure;
4527         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4528
4529         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4530         if (nla == NULL)
4531                 goto nla_put_failure;
4532
4533         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4534                 goto nla_put_failure;
4535
4536         read_lock_bh(&idev->lock);
4537         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4538         read_unlock_bh(&idev->lock);
4539
4540         return 0;
4541
4542 nla_put_failure:
4543         return -EMSGSIZE;
4544 }
4545
4546 static size_t inet6_get_link_af_size(const struct net_device *dev)
4547 {
4548         if (!__in6_dev_get(dev))
4549                 return 0;
4550
4551         return inet6_ifla6_size();
4552 }
4553
4554 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4555 {
4556         struct inet6_dev *idev = __in6_dev_get(dev);
4557
4558         if (!idev)
4559                 return -ENODATA;
4560
4561         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4562                 return -EMSGSIZE;
4563
4564         return 0;
4565 }
4566
4567 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4568 {
4569         struct inet6_ifaddr *ifp;
4570         struct net_device *dev = idev->dev;
4571         bool update_rs = false;
4572         struct in6_addr ll_addr;
4573
4574         ASSERT_RTNL();
4575
4576         if (token == NULL)
4577                 return -EINVAL;
4578         if (ipv6_addr_any(token))
4579                 return -EINVAL;
4580         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4581                 return -EINVAL;
4582         if (!ipv6_accept_ra(idev))
4583                 return -EINVAL;
4584         if (idev->cnf.rtr_solicits <= 0)
4585                 return -EINVAL;
4586
4587         write_lock_bh(&idev->lock);
4588
4589         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4590         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4591
4592         write_unlock_bh(&idev->lock);
4593
4594         if (!idev->dead && (idev->if_flags & IF_READY) &&
4595             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4596                              IFA_F_OPTIMISTIC)) {
4597
4598                 /* If we're not ready, then normal ifup will take care
4599                  * of this. Otherwise, we need to request our rs here.
4600                  */
4601                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4602                 update_rs = true;
4603         }
4604
4605         write_lock_bh(&idev->lock);
4606
4607         if (update_rs) {
4608                 idev->if_flags |= IF_RS_SENT;
4609                 idev->rs_probes = 1;
4610                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4611         }
4612
4613         /* Well, that's kinda nasty ... */
4614         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4615                 spin_lock(&ifp->lock);
4616                 if (ifp->tokenized) {
4617                         ifp->valid_lft = 0;
4618                         ifp->prefered_lft = 0;
4619                 }
4620                 spin_unlock(&ifp->lock);
4621         }
4622
4623         write_unlock_bh(&idev->lock);
4624         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4625         addrconf_verify_rtnl();
4626         return 0;
4627 }
4628
4629 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4630         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4631         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4632 };
4633
4634 static int inet6_validate_link_af(const struct net_device *dev,
4635                                   const struct nlattr *nla)
4636 {
4637         struct nlattr *tb[IFLA_INET6_MAX + 1];
4638
4639         if (dev && !__in6_dev_get(dev))
4640                 return -EAFNOSUPPORT;
4641
4642         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4643 }
4644
4645 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4646 {
4647         int err = -EINVAL;
4648         struct inet6_dev *idev = __in6_dev_get(dev);
4649         struct nlattr *tb[IFLA_INET6_MAX + 1];
4650
4651         if (!idev)
4652                 return -EAFNOSUPPORT;
4653
4654         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4655                 BUG();
4656
4657         if (tb[IFLA_INET6_TOKEN]) {
4658                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4659                 if (err)
4660                         return err;
4661         }
4662
4663         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4664                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4665
4666                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4667                     mode != IN6_ADDR_GEN_MODE_NONE)
4668                         return -EINVAL;
4669                 idev->addr_gen_mode = mode;
4670                 err = 0;
4671         }
4672
4673         return err;
4674 }
4675
4676 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4677                              u32 portid, u32 seq, int event, unsigned int flags)
4678 {
4679         struct net_device *dev = idev->dev;
4680         struct ifinfomsg *hdr;
4681         struct nlmsghdr *nlh;
4682         void *protoinfo;
4683
4684         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4685         if (nlh == NULL)
4686                 return -EMSGSIZE;
4687
4688         hdr = nlmsg_data(nlh);
4689         hdr->ifi_family = AF_INET6;
4690         hdr->__ifi_pad = 0;
4691         hdr->ifi_type = dev->type;
4692         hdr->ifi_index = dev->ifindex;
4693         hdr->ifi_flags = dev_get_flags(dev);
4694         hdr->ifi_change = 0;
4695
4696         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4697             (dev->addr_len &&
4698              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4699             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4700             (dev->ifindex != dev->iflink &&
4701              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4702                 goto nla_put_failure;
4703         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4704         if (protoinfo == NULL)
4705                 goto nla_put_failure;
4706
4707         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4708                 goto nla_put_failure;
4709
4710         nla_nest_end(skb, protoinfo);
4711         nlmsg_end(skb, nlh);
4712         return 0;
4713
4714 nla_put_failure:
4715         nlmsg_cancel(skb, nlh);
4716         return -EMSGSIZE;
4717 }
4718
4719 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4720 {
4721         struct net *net = sock_net(skb->sk);
4722         int h, s_h;
4723         int idx = 0, s_idx;
4724         struct net_device *dev;
4725         struct inet6_dev *idev;
4726         struct hlist_head *head;
4727
4728         s_h = cb->args[0];
4729         s_idx = cb->args[1];
4730
4731         rcu_read_lock();
4732         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4733                 idx = 0;
4734                 head = &net->dev_index_head[h];
4735                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4736                         if (idx < s_idx)
4737                                 goto cont;
4738                         idev = __in6_dev_get(dev);
4739                         if (!idev)
4740                                 goto cont;
4741                         if (inet6_fill_ifinfo(skb, idev,
4742                                               NETLINK_CB(cb->skb).portid,
4743                                               cb->nlh->nlmsg_seq,
4744                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
4745                                 goto out;
4746 cont:
4747                         idx++;
4748                 }
4749         }
4750 out:
4751         rcu_read_unlock();
4752         cb->args[1] = idx;
4753         cb->args[0] = h;
4754
4755         return skb->len;
4756 }
4757
4758 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4759 {
4760         struct sk_buff *skb;
4761         struct net *net = dev_net(idev->dev);
4762         int err = -ENOBUFS;
4763
4764         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4765         if (skb == NULL)
4766                 goto errout;
4767
4768         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4769         if (err < 0) {
4770                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4771                 WARN_ON(err == -EMSGSIZE);
4772                 kfree_skb(skb);
4773                 goto errout;
4774         }
4775         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4776         return;
4777 errout:
4778         if (err < 0)
4779                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4780 }
4781
4782 static inline size_t inet6_prefix_nlmsg_size(void)
4783 {
4784         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4785                + nla_total_size(sizeof(struct in6_addr))
4786                + nla_total_size(sizeof(struct prefix_cacheinfo));
4787 }
4788
4789 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4790                              struct prefix_info *pinfo, u32 portid, u32 seq,
4791                              int event, unsigned int flags)
4792 {
4793         struct prefixmsg *pmsg;
4794         struct nlmsghdr *nlh;
4795         struct prefix_cacheinfo ci;
4796
4797         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4798         if (nlh == NULL)
4799                 return -EMSGSIZE;
4800
4801         pmsg = nlmsg_data(nlh);
4802         pmsg->prefix_family = AF_INET6;
4803         pmsg->prefix_pad1 = 0;
4804         pmsg->prefix_pad2 = 0;
4805         pmsg->prefix_ifindex = idev->dev->ifindex;
4806         pmsg->prefix_len = pinfo->prefix_len;
4807         pmsg->prefix_type = pinfo->type;
4808         pmsg->prefix_pad3 = 0;
4809         pmsg->prefix_flags = 0;
4810         if (pinfo->onlink)
4811                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4812         if (pinfo->autoconf)
4813                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4814
4815         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4816                 goto nla_put_failure;
4817         ci.preferred_time = ntohl(pinfo->prefered);
4818         ci.valid_time = ntohl(pinfo->valid);
4819         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4820                 goto nla_put_failure;
4821         nlmsg_end(skb, nlh);
4822         return 0;
4823
4824 nla_put_failure:
4825         nlmsg_cancel(skb, nlh);
4826         return -EMSGSIZE;
4827 }
4828
4829 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4830                          struct prefix_info *pinfo)
4831 {
4832         struct sk_buff *skb;
4833         struct net *net = dev_net(idev->dev);
4834         int err = -ENOBUFS;
4835
4836         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4837         if (skb == NULL)
4838                 goto errout;
4839
4840         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4841         if (err < 0) {
4842                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4843                 WARN_ON(err == -EMSGSIZE);
4844                 kfree_skb(skb);
4845                 goto errout;
4846         }
4847         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4848         return;
4849 errout:
4850         if (err < 0)
4851                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4852 }
4853
4854 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4855 {
4856         struct net *net = dev_net(ifp->idev->dev);
4857
4858         if (event)
4859                 ASSERT_RTNL();
4860
4861         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4862
4863         switch (event) {
4864         case RTM_NEWADDR:
4865                 /*
4866                  * If the address was optimistic
4867                  * we inserted the route at the start of
4868                  * our DAD process, so we don't need
4869                  * to do it again
4870                  */
4871                 if (!(ifp->rt->rt6i_node))
4872                         ip6_ins_rt(ifp->rt);
4873                 if (ifp->idev->cnf.forwarding)
4874                         addrconf_join_anycast(ifp);
4875                 if (!ipv6_addr_any(&ifp->peer_addr))
4876                         addrconf_prefix_route(&ifp->peer_addr, 128,
4877                                               ifp->idev->dev, 0, 0);
4878                 break;
4879         case RTM_DELADDR:
4880                 if (ifp->idev->cnf.forwarding)
4881                         addrconf_leave_anycast(ifp);
4882                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4883                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4884                         struct rt6_info *rt;
4885
4886                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
4887                                                        ifp->idev->dev, 0, 0);
4888                         if (rt && ip6_del_rt(rt))
4889                                 dst_free(&rt->dst);
4890                 }
4891                 dst_hold(&ifp->rt->dst);
4892
4893                 if (ip6_del_rt(ifp->rt))
4894                         dst_free(&ifp->rt->dst);
4895
4896                 rt_genid_bump_ipv6(net);
4897                 break;
4898         }
4899         atomic_inc(&net->ipv6.dev_addr_genid);
4900 }
4901
4902 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4903 {
4904         rcu_read_lock_bh();
4905         if (likely(ifp->idev->dead == 0))
4906                 __ipv6_ifa_notify(event, ifp);
4907         rcu_read_unlock_bh();
4908 }
4909
4910 #ifdef CONFIG_SYSCTL
4911
4912 static
4913 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4914                            void __user *buffer, size_t *lenp, loff_t *ppos)
4915 {
4916         int *valp = ctl->data;
4917         int val = *valp;
4918         loff_t pos = *ppos;
4919         struct ctl_table lctl;
4920         int ret;
4921
4922         /*
4923          * ctl->data points to idev->cnf.forwarding, we should
4924          * not modify it until we get the rtnl lock.
4925          */
4926         lctl = *ctl;
4927         lctl.data = &val;
4928
4929         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4930
4931         if (write)
4932                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4933         if (ret)
4934                 *ppos = pos;
4935         return ret;
4936 }
4937
4938 static
4939 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
4940                         void __user *buffer, size_t *lenp, loff_t *ppos)
4941 {
4942         struct inet6_dev *idev = ctl->extra1;
4943         int min_mtu = IPV6_MIN_MTU;
4944         struct ctl_table lctl;
4945
4946         lctl = *ctl;
4947         lctl.extra1 = &min_mtu;
4948         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
4949
4950         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
4951 }
4952
4953 static void dev_disable_change(struct inet6_dev *idev)
4954 {
4955         struct netdev_notifier_info info;
4956
4957         if (!idev || !idev->dev)
4958                 return;
4959
4960         netdev_notifier_info_init(&info, idev->dev);
4961         if (idev->cnf.disable_ipv6)
4962                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4963         else
4964                 addrconf_notify(NULL, NETDEV_UP, &info);
4965 }
4966
4967 static void addrconf_disable_change(struct net *net, __s32 newf)
4968 {
4969         struct net_device *dev;
4970         struct inet6_dev *idev;
4971
4972         rcu_read_lock();
4973         for_each_netdev_rcu(net, dev) {
4974                 idev = __in6_dev_get(dev);
4975                 if (idev) {
4976                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4977                         idev->cnf.disable_ipv6 = newf;
4978                         if (changed)
4979                                 dev_disable_change(idev);
4980                 }
4981         }
4982         rcu_read_unlock();
4983 }
4984
4985 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4986 {
4987         struct net *net;
4988         int old;
4989
4990         if (!rtnl_trylock())
4991                 return restart_syscall();
4992
4993         net = (struct net *)table->extra2;
4994         old = *p;
4995         *p = newf;
4996
4997         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4998                 rtnl_unlock();
4999                 return 0;
5000         }
5001
5002         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5003                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5004                 addrconf_disable_change(net, newf);
5005         } else if ((!newf) ^ (!old))
5006                 dev_disable_change((struct inet6_dev *)table->extra1);
5007
5008         rtnl_unlock();
5009         return 0;
5010 }
5011
5012 static
5013 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5014                             void __user *buffer, size_t *lenp, loff_t *ppos)
5015 {
5016         int *valp = ctl->data;
5017         int val = *valp;
5018         loff_t pos = *ppos;
5019         struct ctl_table lctl;
5020         int ret;
5021
5022         /*
5023          * ctl->data points to idev->cnf.disable_ipv6, we should
5024          * not modify it until we get the rtnl lock.
5025          */
5026         lctl = *ctl;
5027         lctl.data = &val;
5028
5029         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5030
5031         if (write)
5032                 ret = addrconf_disable_ipv6(ctl, valp, val);
5033         if (ret)
5034                 *ppos = pos;
5035         return ret;
5036 }
5037
5038 static
5039 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5040                               void __user *buffer, size_t *lenp, loff_t *ppos)
5041 {
5042         int *valp = ctl->data;
5043         int ret;
5044         int old, new;
5045
5046         old = *valp;
5047         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5048         new = *valp;
5049
5050         if (write && old != new) {
5051                 struct net *net = ctl->extra2;
5052
5053                 if (!rtnl_trylock())
5054                         return restart_syscall();
5055
5056                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5057                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5058                                                      NETCONFA_IFINDEX_DEFAULT,
5059                                                      net->ipv6.devconf_dflt);
5060                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5061                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5062                                                      NETCONFA_IFINDEX_ALL,
5063                                                      net->ipv6.devconf_all);
5064                 else {
5065                         struct inet6_dev *idev = ctl->extra1;
5066
5067                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5068                                                      idev->dev->ifindex,
5069                                                      &idev->cnf);
5070                 }
5071                 rtnl_unlock();
5072         }
5073
5074         return ret;
5075 }
5076
5077
5078 static struct addrconf_sysctl_table
5079 {
5080         struct ctl_table_header *sysctl_header;
5081         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5082 } addrconf_sysctl __read_mostly = {
5083         .sysctl_header = NULL,
5084         .addrconf_vars = {
5085                 {
5086                         .procname       = "forwarding",
5087                         .data           = &ipv6_devconf.forwarding,
5088                         .maxlen         = sizeof(int),
5089                         .mode           = 0644,
5090                         .proc_handler   = addrconf_sysctl_forward,
5091                 },
5092                 {
5093                         .procname       = "hop_limit",
5094                         .data           = &ipv6_devconf.hop_limit,
5095                         .maxlen         = sizeof(int),
5096                         .mode           = 0644,
5097                         .proc_handler   = proc_dointvec,
5098                 },
5099                 {
5100                         .procname       = "mtu",
5101                         .data           = &ipv6_devconf.mtu6,
5102                         .maxlen         = sizeof(int),
5103                         .mode           = 0644,
5104                         .proc_handler   = addrconf_sysctl_mtu,
5105                 },
5106                 {
5107                         .procname       = "accept_ra",
5108                         .data           = &ipv6_devconf.accept_ra,
5109                         .maxlen         = sizeof(int),
5110                         .mode           = 0644,
5111                         .proc_handler   = proc_dointvec,
5112                 },
5113                 {
5114                         .procname       = "accept_redirects",
5115                         .data           = &ipv6_devconf.accept_redirects,
5116                         .maxlen         = sizeof(int),
5117                         .mode           = 0644,
5118                         .proc_handler   = proc_dointvec,
5119                 },
5120                 {
5121                         .procname       = "autoconf",
5122                         .data           = &ipv6_devconf.autoconf,
5123                         .maxlen         = sizeof(int),
5124                         .mode           = 0644,
5125                         .proc_handler   = proc_dointvec,
5126                 },
5127                 {
5128                         .procname       = "dad_transmits",
5129                         .data           = &ipv6_devconf.dad_transmits,
5130                         .maxlen         = sizeof(int),
5131                         .mode           = 0644,
5132                         .proc_handler   = proc_dointvec,
5133                 },
5134                 {
5135                         .procname       = "router_solicitations",
5136                         .data           = &ipv6_devconf.rtr_solicits,
5137                         .maxlen         = sizeof(int),
5138                         .mode           = 0644,
5139                         .proc_handler   = proc_dointvec,
5140                 },
5141                 {
5142                         .procname       = "router_solicitation_interval",
5143                         .data           = &ipv6_devconf.rtr_solicit_interval,
5144                         .maxlen         = sizeof(int),
5145                         .mode           = 0644,
5146                         .proc_handler   = proc_dointvec_jiffies,
5147                 },
5148                 {
5149                         .procname       = "router_solicitation_delay",
5150                         .data           = &ipv6_devconf.rtr_solicit_delay,
5151                         .maxlen         = sizeof(int),
5152                         .mode           = 0644,
5153                         .proc_handler   = proc_dointvec_jiffies,
5154                 },
5155                 {
5156                         .procname       = "force_mld_version",
5157                         .data           = &ipv6_devconf.force_mld_version,
5158                         .maxlen         = sizeof(int),
5159                         .mode           = 0644,
5160                         .proc_handler   = proc_dointvec,
5161                 },
5162                 {
5163                         .procname       = "mldv1_unsolicited_report_interval",
5164                         .data           =
5165                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5166                         .maxlen         = sizeof(int),
5167                         .mode           = 0644,
5168                         .proc_handler   = proc_dointvec_ms_jiffies,
5169                 },
5170                 {
5171                         .procname       = "mldv2_unsolicited_report_interval",
5172                         .data           =
5173                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5174                         .maxlen         = sizeof(int),
5175                         .mode           = 0644,
5176                         .proc_handler   = proc_dointvec_ms_jiffies,
5177                 },
5178                 {
5179                         .procname       = "use_tempaddr",
5180                         .data           = &ipv6_devconf.use_tempaddr,
5181                         .maxlen         = sizeof(int),
5182                         .mode           = 0644,
5183                         .proc_handler   = proc_dointvec,
5184                 },
5185                 {
5186                         .procname       = "temp_valid_lft",
5187                         .data           = &ipv6_devconf.temp_valid_lft,
5188                         .maxlen         = sizeof(int),
5189                         .mode           = 0644,
5190                         .proc_handler   = proc_dointvec,
5191                 },
5192                 {
5193                         .procname       = "temp_prefered_lft",
5194                         .data           = &ipv6_devconf.temp_prefered_lft,
5195                         .maxlen         = sizeof(int),
5196                         .mode           = 0644,
5197                         .proc_handler   = proc_dointvec,
5198                 },
5199                 {
5200                         .procname       = "regen_max_retry",
5201                         .data           = &ipv6_devconf.regen_max_retry,
5202                         .maxlen         = sizeof(int),
5203                         .mode           = 0644,
5204                         .proc_handler   = proc_dointvec,
5205                 },
5206                 {
5207                         .procname       = "max_desync_factor",
5208                         .data           = &ipv6_devconf.max_desync_factor,
5209                         .maxlen         = sizeof(int),
5210                         .mode           = 0644,
5211                         .proc_handler   = proc_dointvec,
5212                 },
5213                 {
5214                         .procname       = "max_addresses",
5215                         .data           = &ipv6_devconf.max_addresses,
5216                         .maxlen         = sizeof(int),
5217                         .mode           = 0644,
5218                         .proc_handler   = proc_dointvec,
5219                 },
5220                 {
5221                         .procname       = "accept_ra_defrtr",
5222                         .data           = &ipv6_devconf.accept_ra_defrtr,
5223                         .maxlen         = sizeof(int),
5224                         .mode           = 0644,
5225                         .proc_handler   = proc_dointvec,
5226                 },
5227                 {
5228                         .procname       = "accept_ra_pinfo",
5229                         .data           = &ipv6_devconf.accept_ra_pinfo,
5230                         .maxlen         = sizeof(int),
5231                         .mode           = 0644,
5232                         .proc_handler   = proc_dointvec,
5233                 },
5234 #ifdef CONFIG_IPV6_ROUTER_PREF
5235                 {
5236                         .procname       = "accept_ra_rtr_pref",
5237                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5238                         .maxlen         = sizeof(int),
5239                         .mode           = 0644,
5240                         .proc_handler   = proc_dointvec,
5241                 },
5242                 {
5243                         .procname       = "router_probe_interval",
5244                         .data           = &ipv6_devconf.rtr_probe_interval,
5245                         .maxlen         = sizeof(int),
5246                         .mode           = 0644,
5247                         .proc_handler   = proc_dointvec_jiffies,
5248                 },
5249 #ifdef CONFIG_IPV6_ROUTE_INFO
5250                 {
5251                         .procname       = "accept_ra_rt_info_max_plen",
5252                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5253                         .maxlen         = sizeof(int),
5254                         .mode           = 0644,
5255                         .proc_handler   = proc_dointvec,
5256                 },
5257 #endif
5258 #endif
5259                 {
5260                         .procname       = "proxy_ndp",
5261                         .data           = &ipv6_devconf.proxy_ndp,
5262                         .maxlen         = sizeof(int),
5263                         .mode           = 0644,
5264                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5265                 },
5266                 {
5267                         .procname       = "accept_source_route",
5268                         .data           = &ipv6_devconf.accept_source_route,
5269                         .maxlen         = sizeof(int),
5270                         .mode           = 0644,
5271                         .proc_handler   = proc_dointvec,
5272                 },
5273 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5274                 {
5275                         .procname       = "optimistic_dad",
5276                         .data           = &ipv6_devconf.optimistic_dad,
5277                         .maxlen         = sizeof(int),
5278                         .mode           = 0644,
5279                         .proc_handler   = proc_dointvec,
5280
5281                 },
5282                 {
5283                         .procname       = "use_optimistic",
5284                         .data           = &ipv6_devconf.use_optimistic,
5285                         .maxlen         = sizeof(int),
5286                         .mode           = 0644,
5287                         .proc_handler   = proc_dointvec,
5288
5289                 },
5290 #endif
5291 #ifdef CONFIG_IPV6_MROUTE
5292                 {
5293                         .procname       = "mc_forwarding",
5294                         .data           = &ipv6_devconf.mc_forwarding,
5295                         .maxlen         = sizeof(int),
5296                         .mode           = 0444,
5297                         .proc_handler   = proc_dointvec,
5298                 },
5299 #endif
5300                 {
5301                         .procname       = "disable_ipv6",
5302                         .data           = &ipv6_devconf.disable_ipv6,
5303                         .maxlen         = sizeof(int),
5304                         .mode           = 0644,
5305                         .proc_handler   = addrconf_sysctl_disable,
5306                 },
5307                 {
5308                         .procname       = "accept_dad",
5309                         .data           = &ipv6_devconf.accept_dad,
5310                         .maxlen         = sizeof(int),
5311                         .mode           = 0644,
5312                         .proc_handler   = proc_dointvec,
5313                 },
5314                 {
5315                         .procname       = "force_tllao",
5316                         .data           = &ipv6_devconf.force_tllao,
5317                         .maxlen         = sizeof(int),
5318                         .mode           = 0644,
5319                         .proc_handler   = proc_dointvec
5320                 },
5321                 {
5322                         .procname       = "ndisc_notify",
5323                         .data           = &ipv6_devconf.ndisc_notify,
5324                         .maxlen         = sizeof(int),
5325                         .mode           = 0644,
5326                         .proc_handler   = proc_dointvec
5327                 },
5328                 {
5329                         .procname       = "suppress_frag_ndisc",
5330                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5331                         .maxlen         = sizeof(int),
5332                         .mode           = 0644,
5333                         .proc_handler   = proc_dointvec
5334                 },
5335                 {
5336                         .procname       = "accept_ra_from_local",
5337                         .data           = &ipv6_devconf.accept_ra_from_local,
5338                         .maxlen         = sizeof(int),
5339                         .mode           = 0644,
5340                         .proc_handler   = proc_dointvec,
5341                 },
5342                 {
5343                         .procname       = "accept_ra_mtu",
5344                         .data           = &ipv6_devconf.accept_ra_mtu,
5345                         .maxlen         = sizeof(int),
5346                         .mode           = 0644,
5347                         .proc_handler   = proc_dointvec,
5348                 },
5349                 {
5350                         /* sentinel */
5351                 }
5352         },
5353 };
5354
5355 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5356                 struct inet6_dev *idev, struct ipv6_devconf *p)
5357 {
5358         int i;
5359         struct addrconf_sysctl_table *t;
5360         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5361
5362         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5363         if (t == NULL)
5364                 goto out;
5365
5366         for (i = 0; t->addrconf_vars[i].data; i++) {
5367                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5368                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5369                 t->addrconf_vars[i].extra2 = net;
5370         }
5371
5372         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5373
5374         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5375         if (t->sysctl_header == NULL)
5376                 goto free;
5377
5378         p->sysctl = t;
5379         return 0;
5380
5381 free:
5382         kfree(t);
5383 out:
5384         return -ENOBUFS;
5385 }
5386
5387 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5388 {
5389         struct addrconf_sysctl_table *t;
5390
5391         if (p->sysctl == NULL)
5392                 return;
5393
5394         t = p->sysctl;
5395         p->sysctl = NULL;
5396         unregister_net_sysctl_table(t->sysctl_header);
5397         kfree(t);
5398 }
5399
5400 static int addrconf_sysctl_register(struct inet6_dev *idev)
5401 {
5402         int err;
5403
5404         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5405                 return -EINVAL;
5406
5407         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5408                                     &ndisc_ifinfo_sysctl_change);
5409         if (err)
5410                 return err;
5411         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5412                                          idev, &idev->cnf);
5413         if (err)
5414                 neigh_sysctl_unregister(idev->nd_parms);
5415
5416         return err;
5417 }
5418
5419 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5420 {
5421         __addrconf_sysctl_unregister(&idev->cnf);
5422         neigh_sysctl_unregister(idev->nd_parms);
5423 }
5424
5425
5426 #endif
5427
5428 static int __net_init addrconf_init_net(struct net *net)
5429 {
5430         int err = -ENOMEM;
5431         struct ipv6_devconf *all, *dflt;
5432
5433         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5434         if (all == NULL)
5435                 goto err_alloc_all;
5436
5437         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5438         if (dflt == NULL)
5439                 goto err_alloc_dflt;
5440
5441         /* these will be inherited by all namespaces */
5442         dflt->autoconf = ipv6_defaults.autoconf;
5443         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5444
5445         net->ipv6.devconf_all = all;
5446         net->ipv6.devconf_dflt = dflt;
5447
5448 #ifdef CONFIG_SYSCTL
5449         err = __addrconf_sysctl_register(net, "all", NULL, all);
5450         if (err < 0)
5451                 goto err_reg_all;
5452
5453         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5454         if (err < 0)
5455                 goto err_reg_dflt;
5456 #endif
5457         return 0;
5458
5459 #ifdef CONFIG_SYSCTL
5460 err_reg_dflt:
5461         __addrconf_sysctl_unregister(all);
5462 err_reg_all:
5463         kfree(dflt);
5464 #endif
5465 err_alloc_dflt:
5466         kfree(all);
5467 err_alloc_all:
5468         return err;
5469 }
5470
5471 static void __net_exit addrconf_exit_net(struct net *net)
5472 {
5473 #ifdef CONFIG_SYSCTL
5474         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5475         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5476 #endif
5477         kfree(net->ipv6.devconf_dflt);
5478         kfree(net->ipv6.devconf_all);
5479 }
5480
5481 static struct pernet_operations addrconf_ops = {
5482         .init = addrconf_init_net,
5483         .exit = addrconf_exit_net,
5484 };
5485
5486 static struct rtnl_af_ops inet6_ops __read_mostly = {
5487         .family           = AF_INET6,
5488         .fill_link_af     = inet6_fill_link_af,
5489         .get_link_af_size = inet6_get_link_af_size,
5490         .validate_link_af = inet6_validate_link_af,
5491         .set_link_af      = inet6_set_link_af,
5492 };
5493
5494 /*
5495  *      Init / cleanup code
5496  */
5497
5498 int __init addrconf_init(void)
5499 {
5500         struct inet6_dev *idev;
5501         int i, err;
5502
5503         err = ipv6_addr_label_init();
5504         if (err < 0) {
5505                 pr_crit("%s: cannot initialize default policy table: %d\n",
5506                         __func__, err);
5507                 goto out;
5508         }
5509
5510         err = register_pernet_subsys(&addrconf_ops);
5511         if (err < 0)
5512                 goto out_addrlabel;
5513
5514         addrconf_wq = create_workqueue("ipv6_addrconf");
5515         if (!addrconf_wq) {
5516                 err = -ENOMEM;
5517                 goto out_nowq;
5518         }
5519
5520         /* The addrconf netdev notifier requires that loopback_dev
5521          * has it's ipv6 private information allocated and setup
5522          * before it can bring up and give link-local addresses
5523          * to other devices which are up.
5524          *
5525          * Unfortunately, loopback_dev is not necessarily the first
5526          * entry in the global dev_base list of net devices.  In fact,
5527          * it is likely to be the very last entry on that list.
5528          * So this causes the notifier registry below to try and
5529          * give link-local addresses to all devices besides loopback_dev
5530          * first, then loopback_dev, which cases all the non-loopback_dev
5531          * devices to fail to get a link-local address.
5532          *
5533          * So, as a temporary fix, allocate the ipv6 structure for
5534          * loopback_dev first by hand.
5535          * Longer term, all of the dependencies ipv6 has upon the loopback
5536          * device and it being up should be removed.
5537          */
5538         rtnl_lock();
5539         idev = ipv6_add_dev(init_net.loopback_dev);
5540         rtnl_unlock();
5541         if (IS_ERR(idev)) {
5542                 err = PTR_ERR(idev);
5543                 goto errlo;
5544         }
5545
5546         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5547                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5548
5549         register_netdevice_notifier(&ipv6_dev_notf);
5550
5551         addrconf_verify();
5552
5553         rtnl_af_register(&inet6_ops);
5554
5555         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5556                               NULL);
5557         if (err < 0)
5558                 goto errout;
5559
5560         /* Only the first call to __rtnl_register can fail */
5561         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5562         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5563         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5564                         inet6_dump_ifaddr, NULL);
5565         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5566                         inet6_dump_ifmcaddr, NULL);
5567         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5568                         inet6_dump_ifacaddr, NULL);
5569         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5570                         inet6_netconf_dump_devconf, NULL);
5571
5572         ipv6_addr_label_rtnl_register();
5573
5574         return 0;
5575 errout:
5576         rtnl_af_unregister(&inet6_ops);
5577         unregister_netdevice_notifier(&ipv6_dev_notf);
5578 errlo:
5579         destroy_workqueue(addrconf_wq);
5580 out_nowq:
5581         unregister_pernet_subsys(&addrconf_ops);
5582 out_addrlabel:
5583         ipv6_addr_label_cleanup();
5584 out:
5585         return err;
5586 }
5587
5588 void addrconf_cleanup(void)
5589 {
5590         struct net_device *dev;
5591         int i;
5592
5593         unregister_netdevice_notifier(&ipv6_dev_notf);
5594         unregister_pernet_subsys(&addrconf_ops);
5595         ipv6_addr_label_cleanup();
5596
5597         rtnl_lock();
5598
5599         __rtnl_af_unregister(&inet6_ops);
5600
5601         /* clean dev list */
5602         for_each_netdev(&init_net, dev) {
5603                 if (__in6_dev_get(dev) == NULL)
5604                         continue;
5605                 addrconf_ifdown(dev, 1);
5606         }
5607         addrconf_ifdown(init_net.loopback_dev, 2);
5608
5609         /*
5610          *      Check hash table.
5611          */
5612         spin_lock_bh(&addrconf_hash_lock);
5613         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5614                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5615         spin_unlock_bh(&addrconf_hash_lock);
5616         cancel_delayed_work(&addr_chk_work);
5617         rtnl_unlock();
5618
5619         destroy_workqueue(addrconf_wq);
5620 }