net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86 #include <linux/random.h>
  87 #include <linux/uaccess.h>
  88 #include <asm/unaligned.h>
  89
  90 #include <linux/proc_fs.h>
  91 #include <linux/seq_file.h>
  92 #include <linux/export.h>
  93
  94 /* Set to 3 to get tracing... */
  95 #define ACONF_DEBUG 2
  96
  97 #if ACONF_DEBUG >= 3
  98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
  99 #else
 100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 101 #endif
 102
 103 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 104
 105 static inline u32 cstamp_delta(unsigned long cstamp)
 106 {
 107         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 108 }
 109
 110 #ifdef CONFIG_SYSCTL
 111 static void addrconf_sysctl_register(struct inet6_dev *idev);
 112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 113 #else
 114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 115 {
 116 }
 117
 118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 119 {
 120 }
 121 #endif
 122
 123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 125 static void ipv6_regen_rndid(unsigned long data);
 126
 127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 128 static int ipv6_count_addresses(struct inet6_dev *idev);
 129
 130 /*
 131  *      Configured unicast address hash table
 132  */
 133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 134 static DEFINE_SPINLOCK(addrconf_hash_lock);
 135
 136 static void addrconf_verify(void);
 137 static void addrconf_verify_rtnl(void);
 138 static void addrconf_verify_work(struct work_struct *);
 139
 140 static struct workqueue_struct *addrconf_wq;
 141 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 142
 143 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 144 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 145
 146 static void addrconf_type_change(struct net_device *dev,
 147                                  unsigned long event);
 148 static int addrconf_ifdown(struct net_device *dev, int how);
 149
 150 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 151                                                   int plen,
 152                                                   const struct net_device *dev,
 153                                                   u32 flags, u32 noflags);
 154
 155 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 156 static void addrconf_dad_work(struct work_struct *w);
 157 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 158 static void addrconf_dad_run(struct inet6_dev *idev);
 159 static void addrconf_rs_timer(unsigned long data);
 160 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 161 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 162
 163 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 164                                 struct prefix_info *pinfo);
 165 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 166                                struct net_device *dev);
 167
 168 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 169         .forwarding             = 0,
 170         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 171         .mtu6                   = IPV6_MIN_MTU,
 172         .accept_ra              = 1,
 173         .accept_redirects       = 1,
 174         .autoconf               = 1,
 175         .force_mld_version      = 0,
 176         .mldv1_unsolicited_report_interval = 10 * HZ,
 177         .mldv2_unsolicited_report_interval = HZ,
 178         .dad_transmits          = 1,
 179         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 180         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 181         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 182         .use_tempaddr           = 0,
 183         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 184         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 185         .regen_max_retry        = REGEN_MAX_RETRY,
 186         .max_desync_factor      = MAX_DESYNC_FACTOR,
 187         .max_addresses          = IPV6_MAX_ADDRESSES,
 188         .accept_ra_defrtr       = 1,
 189         .accept_ra_min_hop_limit= 1,
 190         .accept_ra_pinfo        = 1,
 191 #ifdef CONFIG_IPV6_ROUTER_PREF
 192         .accept_ra_rtr_pref     = 1,
 193         .rtr_probe_interval     = 60 * HZ,
 194 #ifdef CONFIG_IPV6_ROUTE_INFO
 195         .accept_ra_rt_info_max_plen = 0,
 196 #endif
 197 #endif
 198         .proxy_ndp              = 0,
 199         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 200         .disable_ipv6           = 0,
 201         .accept_dad             = 1,
 202         .suppress_frag_ndisc    = 1,
 203 };
 204
 205 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 206         .forwarding             = 0,
 207         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 208         .mtu6                   = IPV6_MIN_MTU,
 209         .accept_ra              = 1,
 210         .accept_redirects       = 1,
 211         .autoconf               = 1,
 212         .force_mld_version      = 0,
 213         .mldv1_unsolicited_report_interval = 10 * HZ,
 214         .mldv2_unsolicited_report_interval = HZ,
 215         .dad_transmits          = 1,
 216         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 217         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 218         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 219         .use_tempaddr           = 0,
 220         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 221         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 222         .regen_max_retry        = REGEN_MAX_RETRY,
 223         .max_desync_factor      = MAX_DESYNC_FACTOR,
 224         .max_addresses          = IPV6_MAX_ADDRESSES,
 225         .accept_ra_defrtr       = 1,
 226         .accept_ra_min_hop_limit= 1,
 227         .accept_ra_pinfo        = 1,
 228 #ifdef CONFIG_IPV6_ROUTER_PREF
 229         .accept_ra_rtr_pref     = 1,
 230         .rtr_probe_interval     = 60 * HZ,
 231 #ifdef CONFIG_IPV6_ROUTE_INFO
 232         .accept_ra_rt_info_max_plen = 0,
 233 #endif
 234 #endif
 235         .proxy_ndp              = 0,
 236         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 237         .disable_ipv6           = 0,
 238         .accept_dad             = 1,
 239         .suppress_frag_ndisc    = 1,
 240 };
 241
 242 /* Check if a valid qdisc is available */
 243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 244 {
 245         return !qdisc_tx_is_noop(dev);
 246 }
 247
 248 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 249 {
 250         if (del_timer(&idev->rs_timer))
 251                 __in6_dev_put(idev);
 252 }
 253
 254 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 255 {
 256         if (cancel_delayed_work(&ifp->dad_work))
 257                 __in6_ifa_put(ifp);
 258 }
 259
 260 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 261                                   unsigned long when)
 262 {
 263         if (!timer_pending(&idev->rs_timer))
 264                 in6_dev_hold(idev);
 265         mod_timer(&idev->rs_timer, jiffies + when);
 266 }
 267
 268 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 269                                    unsigned long delay)
 270 {
 271         if (!delayed_work_pending(&ifp->dad_work))
 272                 in6_ifa_hold(ifp);
 273         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 274 }
 275
 276 static int snmp6_alloc_dev(struct inet6_dev *idev)
 277 {
 278         int i;
 279
 280         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 281         if (!idev->stats.ipv6)
 282                 goto err_ip;
 283
 284         for_each_possible_cpu(i) {
 285                 struct ipstats_mib *addrconf_stats;
 286                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 287                 u64_stats_init(&addrconf_stats->syncp);
 288         }
 289
 290
 291         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 292                                         GFP_KERNEL);
 293         if (!idev->stats.icmpv6dev)
 294                 goto err_icmp;
 295         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 296                                            GFP_KERNEL);
 297         if (!idev->stats.icmpv6msgdev)
 298                 goto err_icmpmsg;
 299
 300         return 0;
 301
 302 err_icmpmsg:
 303         kfree(idev->stats.icmpv6dev);
 304 err_icmp:
 305         free_percpu(idev->stats.ipv6);
 306 err_ip:
 307         return -ENOMEM;
 308 }
 309
 310 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 311 {
 312         struct inet6_dev *ndev;
 313
 314         ASSERT_RTNL();
 315
 316         if (dev->mtu < IPV6_MIN_MTU)
 317                 return NULL;
 318
 319         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 320
 321         if (ndev == NULL)
 322                 return NULL;
 323
 324         rwlock_init(&ndev->lock);
 325         ndev->dev = dev;
 326         INIT_LIST_HEAD(&ndev->addr_list);
 327         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 328                     (unsigned long)ndev);
 329         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 330         ndev->cnf.mtu6 = dev->mtu;
 331         ndev->cnf.sysctl = NULL;
 332         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 333         if (ndev->nd_parms == NULL) {
 334                 kfree(ndev);
 335                 return NULL;
 336         }
 337         if (ndev->cnf.forwarding)
 338                 dev_disable_lro(dev);
 339         /* We refer to the device */
 340         dev_hold(dev);
 341
 342         if (snmp6_alloc_dev(ndev) < 0) {
 343                 ADBG(KERN_WARNING
 344                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 345                         __func__, dev->name);
 346                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 347                 dev_put(dev);
 348                 kfree(ndev);
 349                 return NULL;
 350         }
 351
 352         if (snmp6_register_dev(ndev) < 0) {
 353                 ADBG(KERN_WARNING
 354                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 355                         __func__, dev->name);
 356                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 357                 ndev->dead = 1;
 358                 in6_dev_finish_destroy(ndev);
 359                 return NULL;
 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         addrconf_sysctl_register(ndev);
 398         /* protected by rtnl_lock */
 399         rcu_assign_pointer(dev->ip6_ptr, ndev);
 400
 401         /* Join interface-local all-node multicast group */
 402         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 403
 404         /* Join all-node multicast group */
 405         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 406
 407         /* Join all-router multicast group if forwarding is set */
 408         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 409                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 410
 411         return ndev;
 412 }
 413
 414 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 415 {
 416         struct inet6_dev *idev;
 417
 418         ASSERT_RTNL();
 419
 420         idev = __in6_dev_get(dev);
 421         if (!idev) {
 422                 idev = ipv6_add_dev(dev);
 423                 if (!idev)
 424                         return NULL;
 425         }
 426
 427         if (dev->flags&IFF_UP)
 428                 ipv6_mc_up(idev);
 429         return idev;
 430 }
 431
 432 static int inet6_netconf_msgsize_devconf(int type)
 433 {
 434         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 435                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 436
 437         /* type -1 is used for ALL */
 438         if (type == -1 || type == NETCONFA_FORWARDING)
 439                 size += nla_total_size(4);
 440 #ifdef CONFIG_IPV6_MROUTE
 441         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 442                 size += nla_total_size(4);
 443 #endif
 444         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 445                 size += nla_total_size(4);
 446
 447         return size;
 448 }
 449
 450 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 451                                       struct ipv6_devconf *devconf, u32 portid,
 452                                       u32 seq, int event, unsigned int flags,
 453                                       int type)
 454 {
 455         struct nlmsghdr  *nlh;
 456         struct netconfmsg *ncm;
 457
 458         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 459                         flags);
 460         if (nlh == NULL)
 461                 return -EMSGSIZE;
 462
 463         ncm = nlmsg_data(nlh);
 464         ncm->ncm_family = AF_INET6;
 465
 466         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 467                 goto nla_put_failure;
 468
 469         /* type -1 is used for ALL */
 470         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 471             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 472                 goto nla_put_failure;
 473 #ifdef CONFIG_IPV6_MROUTE
 474         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 475             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 476                         devconf->mc_forwarding) < 0)
 477                 goto nla_put_failure;
 478 #endif
 479         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 480             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 481                 goto nla_put_failure;
 482
 483         return nlmsg_end(skb, nlh);
 484
 485 nla_put_failure:
 486         nlmsg_cancel(skb, nlh);
 487         return -EMSGSIZE;
 488 }
 489
 490 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 491                                   struct ipv6_devconf *devconf)
 492 {
 493         struct sk_buff *skb;
 494         int err = -ENOBUFS;
 495
 496         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 497         if (skb == NULL)
 498                 goto errout;
 499
 500         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 501                                          RTM_NEWNETCONF, 0, type);
 502         if (err < 0) {
 503                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 504                 WARN_ON(err == -EMSGSIZE);
 505                 kfree_skb(skb);
 506                 goto errout;
 507         }
 508         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 509         return;
 510 errout:
 511         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 512 }
 513
 514 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 515         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 516         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 517         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 518 };
 519
 520 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 521                                      struct nlmsghdr *nlh)
 522 {
 523         struct net *net = sock_net(in_skb->sk);
 524         struct nlattr *tb[NETCONFA_MAX+1];
 525         struct netconfmsg *ncm;
 526         struct sk_buff *skb;
 527         struct ipv6_devconf *devconf;
 528         struct inet6_dev *in6_dev;
 529         struct net_device *dev;
 530         int ifindex;
 531         int err;
 532
 533         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 534                           devconf_ipv6_policy);
 535         if (err < 0)
 536                 goto errout;
 537
 538         err = -EINVAL;
 539         if (!tb[NETCONFA_IFINDEX])
 540                 goto errout;
 541
 542         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 543         switch (ifindex) {
 544         case NETCONFA_IFINDEX_ALL:
 545                 devconf = net->ipv6.devconf_all;
 546                 break;
 547         case NETCONFA_IFINDEX_DEFAULT:
 548                 devconf = net->ipv6.devconf_dflt;
 549                 break;
 550         default:
 551                 dev = __dev_get_by_index(net, ifindex);
 552                 if (dev == NULL)
 553                         goto errout;
 554                 in6_dev = __in6_dev_get(dev);
 555                 if (in6_dev == NULL)
 556                         goto errout;
 557                 devconf = &in6_dev->cnf;
 558                 break;
 559         }
 560
 561         err = -ENOBUFS;
 562         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 563         if (skb == NULL)
 564                 goto errout;
 565
 566         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 567                                          NETLINK_CB(in_skb).portid,
 568                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 569                                          -1);
 570         if (err < 0) {
 571                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 572                 WARN_ON(err == -EMSGSIZE);
 573                 kfree_skb(skb);
 574                 goto errout;
 575         }
 576         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 577 errout:
 578         return err;
 579 }
 580
 581 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 582                                       struct netlink_callback *cb)
 583 {
 584         struct net *net = sock_net(skb->sk);
 585         int h, s_h;
 586         int idx, s_idx;
 587         struct net_device *dev;
 588         struct inet6_dev *idev;
 589         struct hlist_head *head;
 590
 591         s_h = cb->args[0];
 592         s_idx = idx = cb->args[1];
 593
 594         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 595                 idx = 0;
 596                 head = &net->dev_index_head[h];
 597                 rcu_read_lock();
 598                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 599                           net->dev_base_seq;
 600                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 601                         if (idx < s_idx)
 602                                 goto cont;
 603                         idev = __in6_dev_get(dev);
 604                         if (!idev)
 605                                 goto cont;
 606
 607                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 608                                                        &idev->cnf,
 609                                                        NETLINK_CB(cb->skb).portid,
 610                                                        cb->nlh->nlmsg_seq,
 611                                                        RTM_NEWNETCONF,
 612                                                        NLM_F_MULTI,
 613                                                        -1) <= 0) {
 614                                 rcu_read_unlock();
 615                                 goto done;
 616                         }
 617                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 618 cont:
 619                         idx++;
 620                 }
 621                 rcu_read_unlock();
 622         }
 623         if (h == NETDEV_HASHENTRIES) {
 624                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 625                                                net->ipv6.devconf_all,
 626                                                NETLINK_CB(cb->skb).portid,
 627                                                cb->nlh->nlmsg_seq,
 628                                                RTM_NEWNETCONF, NLM_F_MULTI,
 629                                                -1) <= 0)
 630                         goto done;
 631                 else
 632                         h++;
 633         }
 634         if (h == NETDEV_HASHENTRIES + 1) {
 635                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 636                                                net->ipv6.devconf_dflt,
 637                                                NETLINK_CB(cb->skb).portid,
 638                                                cb->nlh->nlmsg_seq,
 639                                                RTM_NEWNETCONF, NLM_F_MULTI,
 640                                                -1) <= 0)
 641                         goto done;
 642                 else
 643                         h++;
 644         }
 645 done:
 646         cb->args[0] = h;
 647         cb->args[1] = idx;
 648
 649         return skb->len;
 650 }
 651
 652 #ifdef CONFIG_SYSCTL
 653 static void dev_forward_change(struct inet6_dev *idev)
 654 {
 655         struct net_device *dev;
 656         struct inet6_ifaddr *ifa;
 657
 658         if (!idev)
 659                 return;
 660         dev = idev->dev;
 661         if (idev->cnf.forwarding)
 662                 dev_disable_lro(dev);
 663         if (dev->flags & IFF_MULTICAST) {
 664                 if (idev->cnf.forwarding) {
 665                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 666                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 667                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 668                 } else {
 669                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 670                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 671                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 672                 }
 673         }
 674
 675         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 676                 if (ifa->flags&IFA_F_TENTATIVE)
 677                         continue;
 678                 if (idev->cnf.forwarding)
 679                         addrconf_join_anycast(ifa);
 680                 else
 681                         addrconf_leave_anycast(ifa);
 682         }
 683         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 684                                      dev->ifindex, &idev->cnf);
 685 }
 686
 687
 688 static void addrconf_forward_change(struct net *net, __s32 newf)
 689 {
 690         struct net_device *dev;
 691         struct inet6_dev *idev;
 692
 693         for_each_netdev(net, dev) {
 694                 idev = __in6_dev_get(dev);
 695                 if (idev) {
 696                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 697                         idev->cnf.forwarding = newf;
 698                         if (changed)
 699                                 dev_forward_change(idev);
 700                 }
 701         }
 702 }
 703
 704 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 705 {
 706         struct net *net;
 707         int old;
 708
 709         if (!rtnl_trylock())
 710                 return restart_syscall();
 711
 712         net = (struct net *)table->extra2;
 713         old = *p;
 714         *p = newf;
 715
 716         if (p == &net->ipv6.devconf_dflt->forwarding) {
 717                 if ((!newf) ^ (!old))
 718                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 719                                                      NETCONFA_IFINDEX_DEFAULT,
 720                                                      net->ipv6.devconf_dflt);
 721                 rtnl_unlock();
 722                 return 0;
 723         }
 724
 725         if (p == &net->ipv6.devconf_all->forwarding) {
 726                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 727
 728                 net->ipv6.devconf_dflt->forwarding = newf;
 729                 if ((!newf) ^ (!old_dflt))
 730                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 731                                                      NETCONFA_IFINDEX_DEFAULT,
 732                                                      net->ipv6.devconf_dflt);
 733
 734                 addrconf_forward_change(net, newf);
 735                 if ((!newf) ^ (!old))
 736                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 737                                                      NETCONFA_IFINDEX_ALL,
 738                                                      net->ipv6.devconf_all);
 739         } else if ((!newf) ^ (!old))
 740                 dev_forward_change((struct inet6_dev *)table->extra1);
 741         rtnl_unlock();
 742
 743         if (newf)
 744                 rt6_purge_dflt_routers(net);
 745         return 1;
 746 }
 747 #endif
 748
 749 /* Nobody refers to this ifaddr, destroy it */
 750 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 751 {
 752         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 753
 754 #ifdef NET_REFCNT_DEBUG
 755         pr_debug("%s\n", __func__);
 756 #endif
 757
 758         in6_dev_put(ifp->idev);
 759
 760         if (cancel_delayed_work(&ifp->dad_work))
 761                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 762                           ifp);
 763
 764         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 765                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 766                 return;
 767         }
 768         ip6_rt_put(ifp->rt);
 769
 770         kfree_rcu(ifp, rcu);
 771 }
 772
 773 static void
 774 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 775 {
 776         struct list_head *p;
 777         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 778
 779         /*
 780          * Each device address list is sorted in order of scope -
 781          * global before linklocal.
 782          */
 783         list_for_each(p, &idev->addr_list) {
 784                 struct inet6_ifaddr *ifa
 785                         = list_entry(p, struct inet6_ifaddr, if_list);
 786                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 787                         break;
 788         }
 789
 790         list_add_tail(&ifp->if_list, p);
 791 }
 792
 793 static u32 inet6_addr_hash(const struct in6_addr *addr)
 794 {
 795         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 796 }
 797
 798 /* On success it returns ifp with increased reference count */
 799
 800 static struct inet6_ifaddr *
 801 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 802               const struct in6_addr *peer_addr, int pfxlen,
 803               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 804 {
 805         struct inet6_ifaddr *ifa = NULL;
 806         struct rt6_info *rt;
 807         unsigned int hash;
 808         int err = 0;
 809         int addr_type = ipv6_addr_type(addr);
 810
 811         if (addr_type == IPV6_ADDR_ANY ||
 812             addr_type & IPV6_ADDR_MULTICAST ||
 813             (!(idev->dev->flags & IFF_LOOPBACK) &&
 814              addr_type & IPV6_ADDR_LOOPBACK))
 815                 return ERR_PTR(-EADDRNOTAVAIL);
 816
 817         rcu_read_lock_bh();
 818         if (idev->dead) {
 819                 err = -ENODEV;                  /*XXX*/
 820                 goto out2;
 821         }
 822
 823         if (idev->cnf.disable_ipv6) {
 824                 err = -EACCES;
 825                 goto out2;
 826         }
 827
 828         spin_lock(&addrconf_hash_lock);
 829
 830         /* Ignore adding duplicate addresses on an interface */
 831         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 832                 ADBG("ipv6_add_addr: already assigned\n");
 833                 err = -EEXIST;
 834                 goto out;
 835         }
 836
 837         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 838
 839         if (ifa == NULL) {
 840                 ADBG("ipv6_add_addr: malloc failed\n");
 841                 err = -ENOBUFS;
 842                 goto out;
 843         }
 844
 845         rt = addrconf_dst_alloc(idev, addr, false);
 846         if (IS_ERR(rt)) {
 847                 err = PTR_ERR(rt);
 848                 goto out;
 849         }
 850
 851         neigh_parms_data_state_setall(idev->nd_parms);
 852
 853         ifa->addr = *addr;
 854         if (peer_addr)
 855                 ifa->peer_addr = *peer_addr;
 856
 857         spin_lock_init(&ifa->lock);
 858         spin_lock_init(&ifa->state_lock);
 859         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 860         INIT_HLIST_NODE(&ifa->addr_lst);
 861         ifa->scope = scope;
 862         ifa->prefix_len = pfxlen;
 863         ifa->flags = flags | IFA_F_TENTATIVE;
 864         ifa->valid_lft = valid_lft;
 865         ifa->prefered_lft = prefered_lft;
 866         ifa->cstamp = ifa->tstamp = jiffies;
 867         ifa->tokenized = false;
 868
 869         ifa->rt = rt;
 870
 871         ifa->idev = idev;
 872         in6_dev_hold(idev);
 873         /* For caller */
 874         in6_ifa_hold(ifa);
 875
 876         /* Add to big hash table */
 877         hash = inet6_addr_hash(addr);
 878
 879         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 880         spin_unlock(&addrconf_hash_lock);
 881
 882         write_lock(&idev->lock);
 883         /* Add to inet6_dev unicast addr list. */
 884         ipv6_link_dev_addr(idev, ifa);
 885
 886         if (ifa->flags&IFA_F_TEMPORARY) {
 887                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 888                 in6_ifa_hold(ifa);
 889         }
 890
 891         in6_ifa_hold(ifa);
 892         write_unlock(&idev->lock);
 893 out2:
 894         rcu_read_unlock_bh();
 895
 896         if (likely(err == 0))
 897                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 898         else {
 899                 kfree(ifa);
 900                 ifa = ERR_PTR(err);
 901         }
 902
 903         return ifa;
 904 out:
 905         spin_unlock(&addrconf_hash_lock);
 906         goto out2;
 907 }
 908
 909 enum cleanup_prefix_rt_t {
 910         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
 911         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
 912         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
 913 };
 914
 915 /*
 916  * Check, whether the prefix for ifp would still need a prefix route
 917  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
 918  * constants.
 919  *
 920  * 1) we don't purge prefix if address was not permanent.
 921  *    prefix is managed by its own lifetime.
 922  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
 923  * 3) if there are no addresses, delete prefix.
 924  * 4) if there are still other permanent address(es),
 925  *    corresponding prefix is still permanent.
 926  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
 927  *    don't purge the prefix, assume user space is managing it.
 928  * 6) otherwise, update prefix lifetime to the
 929  *    longest valid lifetime among the corresponding
 930  *    addresses on the device.
 931  *    Note: subsequent RA will update lifetime.
 932  **/
 933 static enum cleanup_prefix_rt_t
 934 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
 935 {
 936         struct inet6_ifaddr *ifa;
 937         struct inet6_dev *idev = ifp->idev;
 938         unsigned long lifetime;
 939         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
 940
 941         *expires = jiffies;
 942
 943         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 944                 if (ifa == ifp)
 945                         continue;
 946                 if (ifa->prefix_len != ifp->prefix_len ||
 947                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 948                                        ifp->prefix_len))
 949                         continue;
 950                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
 951                         return CLEANUP_PREFIX_RT_NOP;
 952
 953                 action = CLEANUP_PREFIX_RT_EXPIRE;
 954
 955                 spin_lock(&ifa->lock);
 956
 957                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 958                 /*
 959                  * Note: Because this address is
 960                  * not permanent, lifetime <
 961                  * LONG_MAX / HZ here.
 962                  */
 963                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
 964                         *expires = ifa->tstamp + lifetime * HZ;
 965                 spin_unlock(&ifa->lock);
 966         }
 967
 968         return action;
 969 }
 970
 971 static void
 972 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
 973 {
 974         struct rt6_info *rt;
 975
 976         rt = addrconf_get_prefix_route(&ifp->addr,
 977                                        ifp->prefix_len,
 978                                        ifp->idev->dev,
 979                                        0, RTF_GATEWAY | RTF_DEFAULT);
 980         if (rt) {
 981                 if (del_rt)
 982                         ip6_del_rt(rt);
 983                 else {
 984                         if (!(rt->rt6i_flags & RTF_EXPIRES))
 985                                 rt6_set_expires(rt, expires);
 986                         ip6_rt_put(rt);
 987                 }
 988         }
 989 }
 990
 991
 992 /* This function wants to get referenced ifp and releases it before return */
 993
 994 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 995 {
 996         int state;
 997         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
 998         unsigned long expires;
 999
1000         ASSERT_RTNL();
1001
1002         spin_lock_bh(&ifp->state_lock);
1003         state = ifp->state;
1004         ifp->state = INET6_IFADDR_STATE_DEAD;
1005         spin_unlock_bh(&ifp->state_lock);
1006
1007         if (state == INET6_IFADDR_STATE_DEAD)
1008                 goto out;
1009
1010         spin_lock_bh(&addrconf_hash_lock);
1011         hlist_del_init_rcu(&ifp->addr_lst);
1012         spin_unlock_bh(&addrconf_hash_lock);
1013
1014         write_lock_bh(&ifp->idev->lock);
1015
1016         if (ifp->flags&IFA_F_TEMPORARY) {
1017                 list_del(&ifp->tmp_list);
1018                 if (ifp->ifpub) {
1019                         in6_ifa_put(ifp->ifpub);
1020                         ifp->ifpub = NULL;
1021                 }
1022                 __in6_ifa_put(ifp);
1023         }
1024
1025         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1026                 action = check_cleanup_prefix_route(ifp, &expires);
1027
1028         list_del_init(&ifp->if_list);
1029         __in6_ifa_put(ifp);
1030
1031         write_unlock_bh(&ifp->idev->lock);
1032
1033         addrconf_del_dad_work(ifp);
1034
1035         ipv6_ifa_notify(RTM_DELADDR, ifp);
1036
1037         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1038
1039         if (action != CLEANUP_PREFIX_RT_NOP) {
1040                 cleanup_prefix_route(ifp, expires,
1041                         action == CLEANUP_PREFIX_RT_DEL);
1042         }
1043
1044         /* clean up prefsrc entries */
1045         rt6_remove_prefsrc(ifp);
1046 out:
1047         in6_ifa_put(ifp);
1048 }
1049
1050 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1051 {
1052         struct inet6_dev *idev = ifp->idev;
1053         struct in6_addr addr, *tmpaddr;
1054         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1055         unsigned long regen_advance;
1056         int tmp_plen;
1057         int ret = 0;
1058         u32 addr_flags;
1059         unsigned long now = jiffies;
1060
1061         write_lock_bh(&idev->lock);
1062         if (ift) {
1063                 spin_lock_bh(&ift->lock);
1064                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1065                 spin_unlock_bh(&ift->lock);
1066                 tmpaddr = &addr;
1067         } else {
1068                 tmpaddr = NULL;
1069         }
1070 retry:
1071         in6_dev_hold(idev);
1072         if (idev->cnf.use_tempaddr <= 0) {
1073                 write_unlock_bh(&idev->lock);
1074                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1075                 in6_dev_put(idev);
1076                 ret = -1;
1077                 goto out;
1078         }
1079         spin_lock_bh(&ifp->lock);
1080         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1081                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1082                 spin_unlock_bh(&ifp->lock);
1083                 write_unlock_bh(&idev->lock);
1084                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1085                         __func__);
1086                 in6_dev_put(idev);
1087                 ret = -1;
1088                 goto out;
1089         }
1090         in6_ifa_hold(ifp);
1091         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1092         __ipv6_try_regen_rndid(idev, tmpaddr);
1093         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1094         age = (now - ifp->tstamp) / HZ;
1095         tmp_valid_lft = min_t(__u32,
1096                               ifp->valid_lft,
1097                               idev->cnf.temp_valid_lft + age);
1098         tmp_prefered_lft = min_t(__u32,
1099                                  ifp->prefered_lft,
1100                                  idev->cnf.temp_prefered_lft + age -
1101                                  idev->cnf.max_desync_factor);
1102         tmp_plen = ifp->prefix_len;
1103         tmp_tstamp = ifp->tstamp;
1104         spin_unlock_bh(&ifp->lock);
1105
1106         regen_advance = idev->cnf.regen_max_retry *
1107                         idev->cnf.dad_transmits *
1108                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1109         write_unlock_bh(&idev->lock);
1110
1111         /* A temporary address is created only if this calculated Preferred
1112          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1113          * an implementation must not create a temporary address with a zero
1114          * Preferred Lifetime.
1115          * Use age calculation as in addrconf_verify to avoid unnecessary
1116          * temporary addresses being generated.
1117          */
1118         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1119         if (tmp_prefered_lft <= regen_advance + age) {
1120                 in6_ifa_put(ifp);
1121                 in6_dev_put(idev);
1122                 ret = -1;
1123                 goto out;
1124         }
1125
1126         addr_flags = IFA_F_TEMPORARY;
1127         /* set in addrconf_prefix_rcv() */
1128         if (ifp->flags & IFA_F_OPTIMISTIC)
1129                 addr_flags |= IFA_F_OPTIMISTIC;
1130
1131         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1132                             ipv6_addr_scope(&addr), addr_flags,
1133                             tmp_valid_lft, tmp_prefered_lft);
1134         if (IS_ERR(ift)) {
1135                 in6_ifa_put(ifp);
1136                 in6_dev_put(idev);
1137                 pr_info("%s: retry temporary address regeneration\n", __func__);
1138                 tmpaddr = &addr;
1139                 write_lock_bh(&idev->lock);
1140                 goto retry;
1141         }
1142
1143         spin_lock_bh(&ift->lock);
1144         ift->ifpub = ifp;
1145         ift->cstamp = now;
1146         ift->tstamp = tmp_tstamp;
1147         spin_unlock_bh(&ift->lock);
1148
1149         addrconf_dad_start(ift);
1150         in6_ifa_put(ift);
1151         in6_dev_put(idev);
1152 out:
1153         return ret;
1154 }
1155
1156 /*
1157  *      Choose an appropriate source address (RFC3484)
1158  */
1159 enum {
1160         IPV6_SADDR_RULE_INIT = 0,
1161         IPV6_SADDR_RULE_LOCAL,
1162         IPV6_SADDR_RULE_SCOPE,
1163         IPV6_SADDR_RULE_PREFERRED,
1164 #ifdef CONFIG_IPV6_MIP6
1165         IPV6_SADDR_RULE_HOA,
1166 #endif
1167         IPV6_SADDR_RULE_OIF,
1168         IPV6_SADDR_RULE_LABEL,
1169         IPV6_SADDR_RULE_PRIVACY,
1170         IPV6_SADDR_RULE_ORCHID,
1171         IPV6_SADDR_RULE_PREFIX,
1172         IPV6_SADDR_RULE_MAX
1173 };
1174
1175 struct ipv6_saddr_score {
1176         int                     rule;
1177         int                     addr_type;
1178         struct inet6_ifaddr     *ifa;
1179         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1180         int                     scopedist;
1181         int                     matchlen;
1182 };
1183
1184 struct ipv6_saddr_dst {
1185         const struct in6_addr *addr;
1186         int ifindex;
1187         int scope;
1188         int label;
1189         unsigned int prefs;
1190 };
1191
1192 static inline int ipv6_saddr_preferred(int type)
1193 {
1194         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1195                 return 1;
1196         return 0;
1197 }
1198
1199 static int ipv6_get_saddr_eval(struct net *net,
1200                                struct ipv6_saddr_score *score,
1201                                struct ipv6_saddr_dst *dst,
1202                                int i)
1203 {
1204         int ret;
1205
1206         if (i <= score->rule) {
1207                 switch (i) {
1208                 case IPV6_SADDR_RULE_SCOPE:
1209                         ret = score->scopedist;
1210                         break;
1211                 case IPV6_SADDR_RULE_PREFIX:
1212                         ret = score->matchlen;
1213                         break;
1214                 default:
1215                         ret = !!test_bit(i, score->scorebits);
1216                 }
1217                 goto out;
1218         }
1219
1220         switch (i) {
1221         case IPV6_SADDR_RULE_INIT:
1222                 /* Rule 0: remember if hiscore is not ready yet */
1223                 ret = !!score->ifa;
1224                 break;
1225         case IPV6_SADDR_RULE_LOCAL:
1226                 /* Rule 1: Prefer same address */
1227                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1228                 break;
1229         case IPV6_SADDR_RULE_SCOPE:
1230                 /* Rule 2: Prefer appropriate scope
1231                  *
1232                  *      ret
1233                  *       ^
1234                  *    -1 |  d 15
1235                  *    ---+--+-+---> scope
1236                  *       |
1237                  *       |             d is scope of the destination.
1238                  *  B-d  |  \
1239                  *       |   \      <- smaller scope is better if
1240                  *  B-15 |    \        if scope is enough for destination.
1241                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1242                  * d-C-1 | /
1243                  *       |/         <- greater is better
1244                  *   -C  /             if scope is not enough for destination.
1245                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1246                  *
1247                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1248                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1249                  * Assume B = 0 and we get C > 29.
1250                  */
1251                 ret = __ipv6_addr_src_scope(score->addr_type);
1252                 if (ret >= dst->scope)
1253                         ret = -ret;
1254                 else
1255                         ret -= 128;     /* 30 is enough */
1256                 score->scopedist = ret;
1257                 break;
1258         case IPV6_SADDR_RULE_PREFERRED:
1259                 /* Rule 3: Avoid deprecated and optimistic addresses */
1260                 ret = ipv6_saddr_preferred(score->addr_type) ||
1261                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1262                 break;
1263 #ifdef CONFIG_IPV6_MIP6
1264         case IPV6_SADDR_RULE_HOA:
1265             {
1266                 /* Rule 4: Prefer home address */
1267                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1268                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1269                 break;
1270             }
1271 #endif
1272         case IPV6_SADDR_RULE_OIF:
1273                 /* Rule 5: Prefer outgoing interface */
1274                 ret = (!dst->ifindex ||
1275                        dst->ifindex == score->ifa->idev->dev->ifindex);
1276                 break;
1277         case IPV6_SADDR_RULE_LABEL:
1278                 /* Rule 6: Prefer matching label */
1279                 ret = ipv6_addr_label(net,
1280                                       &score->ifa->addr, score->addr_type,
1281                                       score->ifa->idev->dev->ifindex) == dst->label;
1282                 break;
1283         case IPV6_SADDR_RULE_PRIVACY:
1284             {
1285                 /* Rule 7: Prefer public address
1286                  * Note: prefer temporary address if use_tempaddr >= 2
1287                  */
1288                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1289                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1290                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1291                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1292                 break;
1293             }
1294         case IPV6_SADDR_RULE_ORCHID:
1295                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1296                  *          non-ORCHID vs non-ORCHID
1297                  */
1298                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1299                         ipv6_addr_orchid(dst->addr));
1300                 break;
1301         case IPV6_SADDR_RULE_PREFIX:
1302                 /* Rule 8: Use longest matching prefix */
1303                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1304                 if (ret > score->ifa->prefix_len)
1305                         ret = score->ifa->prefix_len;
1306                 score->matchlen = ret;
1307                 break;
1308         default:
1309                 ret = 0;
1310         }
1311
1312         if (ret)
1313                 __set_bit(i, score->scorebits);
1314         score->rule = i;
1315 out:
1316         return ret;
1317 }
1318
1319 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1320                        const struct in6_addr *daddr, unsigned int prefs,
1321                        struct in6_addr *saddr)
1322 {
1323         struct ipv6_saddr_score scores[2],
1324                                 *score = &scores[0], *hiscore = &scores[1];
1325         struct ipv6_saddr_dst dst;
1326         struct net_device *dev;
1327         int dst_type;
1328
1329         dst_type = __ipv6_addr_type(daddr);
1330         dst.addr = daddr;
1331         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1332         dst.scope = __ipv6_addr_src_scope(dst_type);
1333         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1334         dst.prefs = prefs;
1335
1336         hiscore->rule = -1;
1337         hiscore->ifa = NULL;
1338
1339         rcu_read_lock();
1340
1341         for_each_netdev_rcu(net, dev) {
1342                 struct inet6_dev *idev;
1343
1344                 /* Candidate Source Address (section 4)
1345                  *  - multicast and link-local destination address,
1346                  *    the set of candidate source address MUST only
1347                  *    include addresses assigned to interfaces
1348                  *    belonging to the same link as the outgoing
1349                  *    interface.
1350                  * (- For site-local destination addresses, the
1351                  *    set of candidate source addresses MUST only
1352                  *    include addresses assigned to interfaces
1353                  *    belonging to the same site as the outgoing
1354                  *    interface.)
1355                  */
1356                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1357                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1358                     dst.ifindex && dev->ifindex != dst.ifindex)
1359                         continue;
1360
1361                 idev = __in6_dev_get(dev);
1362                 if (!idev)
1363                         continue;
1364
1365                 read_lock_bh(&idev->lock);
1366                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1367                         int i;
1368
1369                         /*
1370                          * - Tentative Address (RFC2462 section 5.4)
1371                          *  - A tentative address is not considered
1372                          *    "assigned to an interface" in the traditional
1373                          *    sense, unless it is also flagged as optimistic.
1374                          * - Candidate Source Address (section 4)
1375                          *  - In any case, anycast addresses, multicast
1376                          *    addresses, and the unspecified address MUST
1377                          *    NOT be included in a candidate set.
1378                          */
1379                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1380                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1381                                 continue;
1382
1383                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1384
1385                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1386                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1387                                 LIMIT_NETDEBUG(KERN_DEBUG
1388                                                "ADDRCONF: unspecified / multicast address "
1389                                                "assigned as unicast address on %s",
1390                                                dev->name);
1391                                 continue;
1392                         }
1393
1394                         score->rule = -1;
1395                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1396
1397                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1398                                 int minihiscore, miniscore;
1399
1400                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1401                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1402
1403                                 if (minihiscore > miniscore) {
1404                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1405                                             score->scopedist > 0) {
1406                                                 /*
1407                                                  * special case:
1408                                                  * each remaining entry
1409                                                  * has too small (not enough)
1410                                                  * scope, because ifa entries
1411                                                  * are sorted by their scope
1412                                                  * values.
1413                                                  */
1414                                                 goto try_nextdev;
1415                                         }
1416                                         break;
1417                                 } else if (minihiscore < miniscore) {
1418                                         if (hiscore->ifa)
1419                                                 in6_ifa_put(hiscore->ifa);
1420
1421                                         in6_ifa_hold(score->ifa);
1422
1423                                         swap(hiscore, score);
1424
1425                                         /* restore our iterator */
1426                                         score->ifa = hiscore->ifa;
1427
1428                                         break;
1429                                 }
1430                         }
1431                 }
1432 try_nextdev:
1433                 read_unlock_bh(&idev->lock);
1434         }
1435         rcu_read_unlock();
1436
1437         if (!hiscore->ifa)
1438                 return -EADDRNOTAVAIL;
1439
1440         *saddr = hiscore->ifa->addr;
1441         in6_ifa_put(hiscore->ifa);
1442         return 0;
1443 }
1444 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1445
1446 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1447                       u32 banned_flags)
1448 {
1449         struct inet6_ifaddr *ifp;
1450         int err = -EADDRNOTAVAIL;
1451
1452         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1453                 if (ifp->scope > IFA_LINK)
1454                         break;
1455                 if (ifp->scope == IFA_LINK &&
1456                     !(ifp->flags & banned_flags)) {
1457                         *addr = ifp->addr;
1458                         err = 0;
1459                         break;
1460                 }
1461         }
1462         return err;
1463 }
1464
1465 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1466                     u32 banned_flags)
1467 {
1468         struct inet6_dev *idev;
1469         int err = -EADDRNOTAVAIL;
1470
1471         rcu_read_lock();
1472         idev = __in6_dev_get(dev);
1473         if (idev) {
1474                 read_lock_bh(&idev->lock);
1475                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1476                 read_unlock_bh(&idev->lock);
1477         }
1478         rcu_read_unlock();
1479         return err;
1480 }
1481
1482 static int ipv6_count_addresses(struct inet6_dev *idev)
1483 {
1484         int cnt = 0;
1485         struct inet6_ifaddr *ifp;
1486
1487         read_lock_bh(&idev->lock);
1488         list_for_each_entry(ifp, &idev->addr_list, if_list)
1489                 cnt++;
1490         read_unlock_bh(&idev->lock);
1491         return cnt;
1492 }
1493
1494 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1495                   const struct net_device *dev, int strict)
1496 {
1497         struct inet6_ifaddr *ifp;
1498         unsigned int hash = inet6_addr_hash(addr);
1499
1500         rcu_read_lock_bh();
1501         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1502                 if (!net_eq(dev_net(ifp->idev->dev), net))
1503                         continue;
1504                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1505                     !(ifp->flags&IFA_F_TENTATIVE) &&
1506                     (dev == NULL || ifp->idev->dev == dev ||
1507                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1508                         rcu_read_unlock_bh();
1509                         return 1;
1510                 }
1511         }
1512
1513         rcu_read_unlock_bh();
1514         return 0;
1515 }
1516 EXPORT_SYMBOL(ipv6_chk_addr);
1517
1518 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1519                                struct net_device *dev)
1520 {
1521         unsigned int hash = inet6_addr_hash(addr);
1522         struct inet6_ifaddr *ifp;
1523
1524         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1525                 if (!net_eq(dev_net(ifp->idev->dev), net))
1526                         continue;
1527                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1528                         if (dev == NULL || ifp->idev->dev == dev)
1529                                 return true;
1530                 }
1531         }
1532         return false;
1533 }
1534
1535 /* Compares an address/prefix_len with addresses on device @dev.
1536  * If one is found it returns true.
1537  */
1538 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1539         const unsigned int prefix_len, struct net_device *dev)
1540 {
1541         struct inet6_dev *idev;
1542         struct inet6_ifaddr *ifa;
1543         bool ret = false;
1544
1545         rcu_read_lock();
1546         idev = __in6_dev_get(dev);
1547         if (idev) {
1548                 read_lock_bh(&idev->lock);
1549                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1550                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1551                         if (ret)
1552                                 break;
1553                 }
1554                 read_unlock_bh(&idev->lock);
1555         }
1556         rcu_read_unlock();
1557
1558         return ret;
1559 }
1560 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1561
1562 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1563 {
1564         struct inet6_dev *idev;
1565         struct inet6_ifaddr *ifa;
1566         int     onlink;
1567
1568         onlink = 0;
1569         rcu_read_lock();
1570         idev = __in6_dev_get(dev);
1571         if (idev) {
1572                 read_lock_bh(&idev->lock);
1573                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1574                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1575                                                    ifa->prefix_len);
1576                         if (onlink)
1577                                 break;
1578                 }
1579                 read_unlock_bh(&idev->lock);
1580         }
1581         rcu_read_unlock();
1582         return onlink;
1583 }
1584 EXPORT_SYMBOL(ipv6_chk_prefix);
1585
1586 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1587                                      struct net_device *dev, int strict)
1588 {
1589         struct inet6_ifaddr *ifp, *result = NULL;
1590         unsigned int hash = inet6_addr_hash(addr);
1591
1592         rcu_read_lock_bh();
1593         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1594                 if (!net_eq(dev_net(ifp->idev->dev), net))
1595                         continue;
1596                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1597                         if (dev == NULL || ifp->idev->dev == dev ||
1598                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1599                                 result = ifp;
1600                                 in6_ifa_hold(ifp);
1601                                 break;
1602                         }
1603                 }
1604         }
1605         rcu_read_unlock_bh();
1606
1607         return result;
1608 }
1609
1610 /* Gets referenced address, destroys ifaddr */
1611
1612 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1613 {
1614         if (dad_failed)
1615                 ifp->flags |= IFA_F_DADFAILED;
1616
1617         if (ifp->flags&IFA_F_TEMPORARY) {
1618                 struct inet6_ifaddr *ifpub;
1619                 spin_lock_bh(&ifp->lock);
1620                 ifpub = ifp->ifpub;
1621                 if (ifpub) {
1622                         in6_ifa_hold(ifpub);
1623                         spin_unlock_bh(&ifp->lock);
1624                         ipv6_create_tempaddr(ifpub, ifp);
1625                         in6_ifa_put(ifpub);
1626                 } else {
1627                         spin_unlock_bh(&ifp->lock);
1628                 }
1629                 ipv6_del_addr(ifp);
1630         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1631                 spin_lock_bh(&ifp->lock);
1632                 addrconf_del_dad_work(ifp);
1633                 ifp->flags |= IFA_F_TENTATIVE;
1634                 spin_unlock_bh(&ifp->lock);
1635                 if (dad_failed)
1636                         ipv6_ifa_notify(0, ifp);
1637                 in6_ifa_put(ifp);
1638         } else {
1639                 ipv6_del_addr(ifp);
1640         }
1641 }
1642
1643 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1644 {
1645         int err = -ENOENT;
1646
1647         spin_lock_bh(&ifp->state_lock);
1648         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1649                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1650                 err = 0;
1651         }
1652         spin_unlock_bh(&ifp->state_lock);
1653
1654         return err;
1655 }
1656
1657 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1658 {
1659         struct inet6_dev *idev = ifp->idev;
1660
1661         if (addrconf_dad_end(ifp)) {
1662                 in6_ifa_put(ifp);
1663                 return;
1664         }
1665
1666         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1667                              ifp->idev->dev->name, &ifp->addr);
1668
1669         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1670                 struct in6_addr addr;
1671
1672                 addr.s6_addr32[0] = htonl(0xfe800000);
1673                 addr.s6_addr32[1] = 0;
1674
1675                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1676                     ipv6_addr_equal(&ifp->addr, &addr)) {
1677                         /* DAD failed for link-local based on MAC address */
1678                         idev->cnf.disable_ipv6 = 1;
1679
1680                         pr_info("%s: IPv6 being disabled!\n",
1681                                 ifp->idev->dev->name);
1682                 }
1683         }
1684
1685         spin_lock_bh(&ifp->state_lock);
1686         /* transition from _POSTDAD to _ERRDAD */
1687         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1688         spin_unlock_bh(&ifp->state_lock);
1689
1690         addrconf_mod_dad_work(ifp, 0);
1691         in6_ifa_put(ifp);
1692 }
1693
1694 /* Join to solicited addr multicast group.
1695  * caller must hold RTNL */
1696 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1697 {
1698         struct in6_addr maddr;
1699
1700         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1701                 return;
1702
1703         addrconf_addr_solict_mult(addr, &maddr);
1704         ipv6_dev_mc_inc(dev, &maddr);
1705 }
1706
1707 /* caller must hold RTNL */
1708 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1709 {
1710         struct in6_addr maddr;
1711
1712         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1713                 return;
1714
1715         addrconf_addr_solict_mult(addr, &maddr);
1716         __ipv6_dev_mc_dec(idev, &maddr);
1717 }
1718
1719 /* caller must hold RTNL */
1720 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1721 {
1722         struct in6_addr addr;
1723
1724         if (ifp->prefix_len >= 127) /* RFC 6164 */
1725                 return;
1726         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1727         if (ipv6_addr_any(&addr))
1728                 return;
1729         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1730 }
1731
1732 /* caller must hold RTNL */
1733 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1734 {
1735         struct in6_addr addr;
1736
1737         if (ifp->prefix_len >= 127) /* RFC 6164 */
1738                 return;
1739         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1740         if (ipv6_addr_any(&addr))
1741                 return;
1742         __ipv6_dev_ac_dec(ifp->idev, &addr);
1743 }
1744
1745 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1746 {
1747         if (dev->addr_len != ETH_ALEN)
1748                 return -1;
1749         memcpy(eui, dev->dev_addr, 3);
1750         memcpy(eui + 5, dev->dev_addr + 3, 3);
1751
1752         /*
1753          * The zSeries OSA network cards can be shared among various
1754          * OS instances, but the OSA cards have only one MAC address.
1755          * This leads to duplicate address conflicts in conjunction
1756          * with IPv6 if more than one instance uses the same card.
1757          *
1758          * The driver for these cards can deliver a unique 16-bit
1759          * identifier for each instance sharing the same card.  It is
1760          * placed instead of 0xFFFE in the interface identifier.  The
1761          * "u" bit of the interface identifier is not inverted in this
1762          * case.  Hence the resulting interface identifier has local
1763          * scope according to RFC2373.
1764          */
1765         if (dev->dev_id) {
1766                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1767                 eui[4] = dev->dev_id & 0xFF;
1768         } else {
1769                 eui[3] = 0xFF;
1770                 eui[4] = 0xFE;
1771                 eui[0] ^= 2;
1772         }
1773         return 0;
1774 }
1775
1776 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1777 {
1778         if (dev->addr_len != IEEE802154_ADDR_LEN)
1779                 return -1;
1780         memcpy(eui, dev->dev_addr, 8);
1781         eui[0] ^= 2;
1782         return 0;
1783 }
1784
1785 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1786 {
1787         union fwnet_hwaddr *ha;
1788
1789         if (dev->addr_len != FWNET_ALEN)
1790                 return -1;
1791
1792         ha = (union fwnet_hwaddr *)dev->dev_addr;
1793
1794         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1795         eui[0] ^= 2;
1796         return 0;
1797 }
1798
1799 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1800 {
1801         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1802         if (dev->addr_len != ARCNET_ALEN)
1803                 return -1;
1804         memset(eui, 0, 7);
1805         eui[7] = *(u8 *)dev->dev_addr;
1806         return 0;
1807 }
1808
1809 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1810 {
1811         if (dev->addr_len != INFINIBAND_ALEN)
1812                 return -1;
1813         memcpy(eui, dev->dev_addr + 12, 8);
1814         eui[0] |= 2;
1815         return 0;
1816 }
1817
1818 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1819 {
1820         if (addr == 0)
1821                 return -1;
1822         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1823                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1824                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1825                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1826                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1827                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1828         eui[1] = 0;
1829         eui[2] = 0x5E;
1830         eui[3] = 0xFE;
1831         memcpy(eui + 4, &addr, 4);
1832         return 0;
1833 }
1834
1835 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1836 {
1837         if (dev->priv_flags & IFF_ISATAP)
1838                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1839         return -1;
1840 }
1841
1842 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1843 {
1844         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1845 }
1846
1847 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1848 {
1849         memcpy(eui, dev->perm_addr, 3);
1850         memcpy(eui + 5, dev->perm_addr + 3, 3);
1851         eui[3] = 0xFF;
1852         eui[4] = 0xFE;
1853         eui[0] ^= 2;
1854         return 0;
1855 }
1856
1857 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1858 {
1859         switch (dev->type) {
1860         case ARPHRD_ETHER:
1861         case ARPHRD_FDDI:
1862                 return addrconf_ifid_eui48(eui, dev);
1863         case ARPHRD_ARCNET:
1864                 return addrconf_ifid_arcnet(eui, dev);
1865         case ARPHRD_INFINIBAND:
1866                 return addrconf_ifid_infiniband(eui, dev);
1867         case ARPHRD_SIT:
1868                 return addrconf_ifid_sit(eui, dev);
1869         case ARPHRD_IPGRE:
1870                 return addrconf_ifid_gre(eui, dev);
1871         case ARPHRD_6LOWPAN:
1872         case ARPHRD_IEEE802154:
1873                 return addrconf_ifid_eui64(eui, dev);
1874         case ARPHRD_IEEE1394:
1875                 return addrconf_ifid_ieee1394(eui, dev);
1876         case ARPHRD_TUNNEL6:
1877                 return addrconf_ifid_ip6tnl(eui, dev);
1878         }
1879         return -1;
1880 }
1881
1882 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1883 {
1884         int err = -1;
1885         struct inet6_ifaddr *ifp;
1886
1887         read_lock_bh(&idev->lock);
1888         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1889                 if (ifp->scope > IFA_LINK)
1890                         break;
1891                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1892                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1893                         err = 0;
1894                         break;
1895                 }
1896         }
1897         read_unlock_bh(&idev->lock);
1898         return err;
1899 }
1900
1901 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1902 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1903 {
1904 regen:
1905         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1906         idev->rndid[0] &= ~0x02;
1907
1908         /*
1909          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1910          * check if generated address is not inappropriate
1911          *
1912          *  - Reserved subnet anycast (RFC 2526)
1913          *      11111101 11....11 1xxxxxxx
1914          *  - ISATAP (RFC4214) 6.1
1915          *      00-00-5E-FE-xx-xx-xx-xx
1916          *  - value 0
1917          *  - XXX: already assigned to an address on the device
1918          */
1919         if (idev->rndid[0] == 0xfd &&
1920             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1921             (idev->rndid[7]&0x80))
1922                 goto regen;
1923         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1924                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1925                         goto regen;
1926                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1927                         goto regen;
1928         }
1929 }
1930
1931 static void ipv6_regen_rndid(unsigned long data)
1932 {
1933         struct inet6_dev *idev = (struct inet6_dev *) data;
1934         unsigned long expires;
1935
1936         rcu_read_lock_bh();
1937         write_lock_bh(&idev->lock);
1938
1939         if (idev->dead)
1940                 goto out;
1941
1942         __ipv6_regen_rndid(idev);
1943
1944         expires = jiffies +
1945                 idev->cnf.temp_prefered_lft * HZ -
1946                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1947                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1948                 idev->cnf.max_desync_factor * HZ;
1949         if (time_before(expires, jiffies)) {
1950                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1951                         __func__, idev->dev->name);
1952                 goto out;
1953         }
1954
1955         if (!mod_timer(&idev->regen_timer, expires))
1956                 in6_dev_hold(idev);
1957
1958 out:
1959         write_unlock_bh(&idev->lock);
1960         rcu_read_unlock_bh();
1961         in6_dev_put(idev);
1962 }
1963
1964 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1965 {
1966         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1967                 __ipv6_regen_rndid(idev);
1968 }
1969
1970 /*
1971  *      Add prefix route.
1972  */
1973
1974 static void
1975 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1976                       unsigned long expires, u32 flags)
1977 {
1978         struct fib6_config cfg = {
1979                 .fc_table = RT6_TABLE_PREFIX,
1980                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1981                 .fc_ifindex = dev->ifindex,
1982                 .fc_expires = expires,
1983                 .fc_dst_len = plen,
1984                 .fc_flags = RTF_UP | flags,
1985                 .fc_nlinfo.nl_net = dev_net(dev),
1986                 .fc_protocol = RTPROT_KERNEL,
1987         };
1988
1989         cfg.fc_dst = *pfx;
1990
1991         /* Prevent useless cloning on PtP SIT.
1992            This thing is done here expecting that the whole
1993            class of non-broadcast devices need not cloning.
1994          */
1995 #if IS_ENABLED(CONFIG_IPV6_SIT)
1996         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1997                 cfg.fc_flags |= RTF_NONEXTHOP;
1998 #endif
1999
2000         ip6_route_add(&cfg);
2001 }
2002
2003
2004 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2005                                                   int plen,
2006                                                   const struct net_device *dev,
2007                                                   u32 flags, u32 noflags)
2008 {
2009         struct fib6_node *fn;
2010         struct rt6_info *rt = NULL;
2011         struct fib6_table *table;
2012
2013         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2014         if (table == NULL)
2015                 return NULL;
2016
2017         read_lock_bh(&table->tb6_lock);
2018         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2019         if (!fn)
2020                 goto out;
2021         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2022                 if (rt->dst.dev->ifindex != dev->ifindex)
2023                         continue;
2024                 if ((rt->rt6i_flags & flags) != flags)
2025                         continue;
2026                 if ((rt->rt6i_flags & noflags) != 0)
2027                         continue;
2028                 dst_hold(&rt->dst);
2029                 break;
2030         }
2031 out:
2032         read_unlock_bh(&table->tb6_lock);
2033         return rt;
2034 }
2035
2036
2037 /* Create "default" multicast route to the interface */
2038
2039 static void addrconf_add_mroute(struct net_device *dev)
2040 {
2041         struct fib6_config cfg = {
2042                 .fc_table = RT6_TABLE_LOCAL,
2043                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2044                 .fc_ifindex = dev->ifindex,
2045                 .fc_dst_len = 8,
2046                 .fc_flags = RTF_UP,
2047                 .fc_nlinfo.nl_net = dev_net(dev),
2048         };
2049
2050         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2051
2052         ip6_route_add(&cfg);
2053 }
2054
2055 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2056 {
2057         struct inet6_dev *idev;
2058
2059         ASSERT_RTNL();
2060
2061         idev = ipv6_find_idev(dev);
2062         if (!idev)
2063                 return ERR_PTR(-ENOBUFS);
2064
2065         if (idev->cnf.disable_ipv6)
2066                 return ERR_PTR(-EACCES);
2067
2068         /* Add default multicast route */
2069         if (!(dev->flags & IFF_LOOPBACK))
2070                 addrconf_add_mroute(dev);
2071
2072         return idev;
2073 }
2074
2075 static void manage_tempaddrs(struct inet6_dev *idev,
2076                              struct inet6_ifaddr *ifp,
2077                              __u32 valid_lft, __u32 prefered_lft,
2078                              bool create, unsigned long now)
2079 {
2080         u32 flags;
2081         struct inet6_ifaddr *ift;
2082
2083         read_lock_bh(&idev->lock);
2084         /* update all temporary addresses in the list */
2085         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2086                 int age, max_valid, max_prefered;
2087
2088                 if (ifp != ift->ifpub)
2089                         continue;
2090
2091                 /* RFC 4941 section 3.3:
2092                  * If a received option will extend the lifetime of a public
2093                  * address, the lifetimes of temporary addresses should
2094                  * be extended, subject to the overall constraint that no
2095                  * temporary addresses should ever remain "valid" or "preferred"
2096                  * for a time longer than (TEMP_VALID_LIFETIME) or
2097                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2098                  */
2099                 age = (now - ift->cstamp) / HZ;
2100                 max_valid = idev->cnf.temp_valid_lft - age;
2101                 if (max_valid < 0)
2102                         max_valid = 0;
2103
2104                 max_prefered = idev->cnf.temp_prefered_lft -
2105                                idev->cnf.max_desync_factor - age;
2106                 if (max_prefered < 0)
2107                         max_prefered = 0;
2108
2109                 if (valid_lft > max_valid)
2110                         valid_lft = max_valid;
2111
2112                 if (prefered_lft > max_prefered)
2113                         prefered_lft = max_prefered;
2114
2115                 spin_lock(&ift->lock);
2116                 flags = ift->flags;
2117                 ift->valid_lft = valid_lft;
2118                 ift->prefered_lft = prefered_lft;
2119                 ift->tstamp = now;
2120                 if (prefered_lft > 0)
2121                         ift->flags &= ~IFA_F_DEPRECATED;
2122
2123                 spin_unlock(&ift->lock);
2124                 if (!(flags&IFA_F_TENTATIVE))
2125                         ipv6_ifa_notify(0, ift);
2126         }
2127
2128         if ((create || list_empty(&idev->tempaddr_list)) &&
2129             idev->cnf.use_tempaddr > 0) {
2130                 /* When a new public address is created as described
2131                  * in [ADDRCONF], also create a new temporary address.
2132                  * Also create a temporary address if it's enabled but
2133                  * no temporary address currently exists.
2134                  */
2135                 read_unlock_bh(&idev->lock);
2136                 ipv6_create_tempaddr(ifp, NULL);
2137         } else {
2138                 read_unlock_bh(&idev->lock);
2139         }
2140 }
2141
2142 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2143 {
2144         struct prefix_info *pinfo;
2145         __u32 valid_lft;
2146         __u32 prefered_lft;
2147         int addr_type;
2148         struct inet6_dev *in6_dev;
2149         struct net *net = dev_net(dev);
2150
2151         pinfo = (struct prefix_info *) opt;
2152
2153         if (len < sizeof(struct prefix_info)) {
2154                 ADBG("addrconf: prefix option too short\n");
2155                 return;
2156         }
2157
2158         /*
2159          *      Validation checks ([ADDRCONF], page 19)
2160          */
2161
2162         addr_type = ipv6_addr_type(&pinfo->prefix);
2163
2164         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2165                 return;
2166
2167         valid_lft = ntohl(pinfo->valid);
2168         prefered_lft = ntohl(pinfo->prefered);
2169
2170         if (prefered_lft > valid_lft) {
2171                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2172                 return;
2173         }
2174
2175         in6_dev = in6_dev_get(dev);
2176
2177         if (in6_dev == NULL) {
2178                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2179                                     dev->name);
2180                 return;
2181         }
2182
2183         /*
2184          *      Two things going on here:
2185          *      1) Add routes for on-link prefixes
2186          *      2) Configure prefixes with the auto flag set
2187          */
2188
2189         if (pinfo->onlink) {
2190                 struct rt6_info *rt;
2191                 unsigned long rt_expires;
2192
2193                 /* Avoid arithmetic overflow. Really, we could
2194                  * save rt_expires in seconds, likely valid_lft,
2195                  * but it would require division in fib gc, that it
2196                  * not good.
2197                  */
2198                 if (HZ > USER_HZ)
2199                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2200                 else
2201                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2202
2203                 if (addrconf_finite_timeout(rt_expires))
2204                         rt_expires *= HZ;
2205
2206                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2207                                                pinfo->prefix_len,
2208                                                dev,
2209                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2210                                                RTF_GATEWAY | RTF_DEFAULT);
2211
2212                 if (rt) {
2213                         /* Autoconf prefix route */
2214                         if (valid_lft == 0) {
2215                                 ip6_del_rt(rt);
2216                                 rt = NULL;
2217                         } else if (addrconf_finite_timeout(rt_expires)) {
2218                                 /* not infinity */
2219                                 rt6_set_expires(rt, jiffies + rt_expires);
2220                         } else {
2221                                 rt6_clean_expires(rt);
2222                         }
2223                 } else if (valid_lft) {
2224                         clock_t expires = 0;
2225                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2226                         if (addrconf_finite_timeout(rt_expires)) {
2227                                 /* not infinity */
2228                                 flags |= RTF_EXPIRES;
2229                                 expires = jiffies_to_clock_t(rt_expires);
2230                         }
2231                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2232                                               dev, expires, flags);
2233                 }
2234                 ip6_rt_put(rt);
2235         }
2236
2237         /* Try to figure out our local address for this prefix */
2238
2239         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2240                 struct inet6_ifaddr *ifp;
2241                 struct in6_addr addr;
2242                 int create = 0, update_lft = 0;
2243                 bool tokenized = false;
2244
2245                 if (pinfo->prefix_len == 64) {
2246                         memcpy(&addr, &pinfo->prefix, 8);
2247
2248                         if (!ipv6_addr_any(&in6_dev->token)) {
2249                                 read_lock_bh(&in6_dev->lock);
2250                                 memcpy(addr.s6_addr + 8,
2251                                        in6_dev->token.s6_addr + 8, 8);
2252                                 read_unlock_bh(&in6_dev->lock);
2253                                 tokenized = true;
2254                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2255                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2256                                 in6_dev_put(in6_dev);
2257                                 return;
2258                         }
2259                         goto ok;
2260                 }
2261                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2262                                     pinfo->prefix_len);
2263                 in6_dev_put(in6_dev);
2264                 return;
2265
2266 ok:
2267
2268                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2269
2270                 if (ifp == NULL && valid_lft) {
2271                         int max_addresses = in6_dev->cnf.max_addresses;
2272                         u32 addr_flags = 0;
2273
2274 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2275                         if (in6_dev->cnf.optimistic_dad &&
2276                             !net->ipv6.devconf_all->forwarding && sllao)
2277                                 addr_flags = IFA_F_OPTIMISTIC;
2278 #endif
2279
2280                         /* Do not allow to create too much of autoconfigured
2281                          * addresses; this would be too easy way to crash kernel.
2282                          */
2283                         if (!max_addresses ||
2284                             ipv6_count_addresses(in6_dev) < max_addresses)
2285                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2286                                                     pinfo->prefix_len,
2287                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2288                                                     addr_flags, valid_lft,
2289                                                     prefered_lft);
2290
2291                         if (IS_ERR_OR_NULL(ifp)) {
2292                                 in6_dev_put(in6_dev);
2293                                 return;
2294                         }
2295
2296                         update_lft = 0;
2297                         create = 1;
2298                         spin_lock_bh(&ifp->lock);
2299                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2300                         ifp->cstamp = jiffies;
2301                         ifp->tokenized = tokenized;
2302                         spin_unlock_bh(&ifp->lock);
2303                         addrconf_dad_start(ifp);
2304                 }
2305
2306                 if (ifp) {
2307                         u32 flags;
2308                         unsigned long now;
2309                         u32 stored_lft;
2310
2311                         /* update lifetime (RFC2462 5.5.3 e) */
2312                         spin_lock(&ifp->lock);
2313                         now = jiffies;
2314                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2315                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2316                         else
2317                                 stored_lft = 0;
2318                         if (!update_lft && !create && stored_lft) {
2319                                 const u32 minimum_lft = min(
2320                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2321                                 valid_lft = max(valid_lft, minimum_lft);
2322
2323                                 /* RFC4862 Section 5.5.3e:
2324                                  * "Note that the preferred lifetime of the
2325                                  *  corresponding address is always reset to
2326                                  *  the Preferred Lifetime in the received
2327                                  *  Prefix Information option, regardless of
2328                                  *  whether the valid lifetime is also reset or
2329                                  *  ignored."
2330                                  *
2331                                  * So we should always update prefered_lft here.
2332                                  */
2333                                 update_lft = 1;
2334                         }
2335
2336                         if (update_lft) {
2337                                 ifp->valid_lft = valid_lft;
2338                                 ifp->prefered_lft = prefered_lft;
2339                                 ifp->tstamp = now;
2340                                 flags = ifp->flags;
2341                                 ifp->flags &= ~IFA_F_DEPRECATED;
2342                                 spin_unlock(&ifp->lock);
2343
2344                                 if (!(flags&IFA_F_TENTATIVE))
2345                                         ipv6_ifa_notify(0, ifp);
2346                         } else
2347                                 spin_unlock(&ifp->lock);
2348
2349                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2350                                          create, now);
2351
2352                         in6_ifa_put(ifp);
2353                         addrconf_verify();
2354                 }
2355         }
2356         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2357         in6_dev_put(in6_dev);
2358 }
2359
2360 /*
2361  *      Set destination address.
2362  *      Special case for SIT interfaces where we create a new "virtual"
2363  *      device.
2364  */
2365 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2366 {
2367         struct in6_ifreq ireq;
2368         struct net_device *dev;
2369         int err = -EINVAL;
2370
2371         rtnl_lock();
2372
2373         err = -EFAULT;
2374         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2375                 goto err_exit;
2376
2377         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2378
2379         err = -ENODEV;
2380         if (dev == NULL)
2381                 goto err_exit;
2382
2383 #if IS_ENABLED(CONFIG_IPV6_SIT)
2384         if (dev->type == ARPHRD_SIT) {
2385                 const struct net_device_ops *ops = dev->netdev_ops;
2386                 struct ifreq ifr;
2387                 struct ip_tunnel_parm p;
2388
2389                 err = -EADDRNOTAVAIL;
2390                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2391                         goto err_exit;
2392
2393                 memset(&p, 0, sizeof(p));
2394                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2395                 p.iph.saddr = 0;
2396                 p.iph.version = 4;
2397                 p.iph.ihl = 5;
2398                 p.iph.protocol = IPPROTO_IPV6;
2399                 p.iph.ttl = 64;
2400                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2401
2402                 if (ops->ndo_do_ioctl) {
2403                         mm_segment_t oldfs = get_fs();
2404
2405                         set_fs(KERNEL_DS);
2406                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2407                         set_fs(oldfs);
2408                 } else
2409                         err = -EOPNOTSUPP;
2410
2411                 if (err == 0) {
2412                         err = -ENOBUFS;
2413                         dev = __dev_get_by_name(net, p.name);
2414                         if (!dev)
2415                                 goto err_exit;
2416                         err = dev_open(dev);
2417                 }
2418         }
2419 #endif
2420
2421 err_exit:
2422         rtnl_unlock();
2423         return err;
2424 }
2425
2426 /*
2427  *      Manual configuration of address on an interface
2428  */
2429 static int inet6_addr_add(struct net *net, int ifindex,
2430                           const struct in6_addr *pfx,
2431                           const struct in6_addr *peer_pfx,
2432                           unsigned int plen, __u32 ifa_flags,
2433                           __u32 prefered_lft, __u32 valid_lft)
2434 {
2435         struct inet6_ifaddr *ifp;
2436         struct inet6_dev *idev;
2437         struct net_device *dev;
2438         int scope;
2439         u32 flags;
2440         clock_t expires;
2441         unsigned long timeout;
2442
2443         ASSERT_RTNL();
2444
2445         if (plen > 128)
2446                 return -EINVAL;
2447
2448         /* check the lifetime */
2449         if (!valid_lft || prefered_lft > valid_lft)
2450                 return -EINVAL;
2451
2452         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2453                 return -EINVAL;
2454
2455         dev = __dev_get_by_index(net, ifindex);
2456         if (!dev)
2457                 return -ENODEV;
2458
2459         idev = addrconf_add_dev(dev);
2460         if (IS_ERR(idev))
2461                 return PTR_ERR(idev);
2462
2463         scope = ipv6_addr_scope(pfx);
2464
2465         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2466         if (addrconf_finite_timeout(timeout)) {
2467                 expires = jiffies_to_clock_t(timeout * HZ);
2468                 valid_lft = timeout;
2469                 flags = RTF_EXPIRES;
2470         } else {
2471                 expires = 0;
2472                 flags = 0;
2473                 ifa_flags |= IFA_F_PERMANENT;
2474         }
2475
2476         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2477         if (addrconf_finite_timeout(timeout)) {
2478                 if (timeout == 0)
2479                         ifa_flags |= IFA_F_DEPRECATED;
2480                 prefered_lft = timeout;
2481         }
2482
2483         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2484                             valid_lft, prefered_lft);
2485
2486         if (!IS_ERR(ifp)) {
2487                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2488                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2489                                               expires, flags);
2490                 }
2491
2492                 /*
2493                  * Note that section 3.1 of RFC 4429 indicates
2494                  * that the Optimistic flag should not be set for
2495                  * manually configured addresses
2496                  */
2497                 addrconf_dad_start(ifp);
2498                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2499                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2500                                          true, jiffies);
2501                 in6_ifa_put(ifp);
2502                 addrconf_verify_rtnl();
2503                 return 0;
2504         }
2505
2506         return PTR_ERR(ifp);
2507 }
2508
2509 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2510                           const struct in6_addr *pfx, unsigned int plen)
2511 {
2512         struct inet6_ifaddr *ifp;
2513         struct inet6_dev *idev;
2514         struct net_device *dev;
2515
2516         if (plen > 128)
2517                 return -EINVAL;
2518
2519         dev = __dev_get_by_index(net, ifindex);
2520         if (!dev)
2521                 return -ENODEV;
2522
2523         if ((idev = __in6_dev_get(dev)) == NULL)
2524                 return -ENXIO;
2525
2526         read_lock_bh(&idev->lock);
2527         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2528                 if (ifp->prefix_len == plen &&
2529                     ipv6_addr_equal(pfx, &ifp->addr)) {
2530                         in6_ifa_hold(ifp);
2531                         read_unlock_bh(&idev->lock);
2532
2533                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2534                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2535                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2536                                                  jiffies);
2537                         ipv6_del_addr(ifp);
2538                         addrconf_verify_rtnl();
2539                         return 0;
2540                 }
2541         }
2542         read_unlock_bh(&idev->lock);
2543         return -EADDRNOTAVAIL;
2544 }
2545
2546
2547 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2548 {
2549         struct in6_ifreq ireq;
2550         int err;
2551
2552         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2553                 return -EPERM;
2554
2555         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2556                 return -EFAULT;
2557
2558         rtnl_lock();
2559         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2560                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2561                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2562         rtnl_unlock();
2563         return err;
2564 }
2565
2566 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2567 {
2568         struct in6_ifreq ireq;
2569         int err;
2570
2571         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2572                 return -EPERM;
2573
2574         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2575                 return -EFAULT;
2576
2577         rtnl_lock();
2578         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2579                              ireq.ifr6_prefixlen);
2580         rtnl_unlock();
2581         return err;
2582 }
2583
2584 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2585                      int plen, int scope)
2586 {
2587         struct inet6_ifaddr *ifp;
2588
2589         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2590                             scope, IFA_F_PERMANENT,
2591                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2592         if (!IS_ERR(ifp)) {
2593                 spin_lock_bh(&ifp->lock);
2594                 ifp->flags &= ~IFA_F_TENTATIVE;
2595                 spin_unlock_bh(&ifp->lock);
2596                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2597                 in6_ifa_put(ifp);
2598         }
2599 }
2600
2601 #if IS_ENABLED(CONFIG_IPV6_SIT)
2602 static void sit_add_v4_addrs(struct inet6_dev *idev)
2603 {
2604         struct in6_addr addr;
2605         struct net_device *dev;
2606         struct net *net = dev_net(idev->dev);
2607         int scope, plen;
2608         u32 pflags = 0;
2609
2610         ASSERT_RTNL();
2611
2612         memset(&addr, 0, sizeof(struct in6_addr));
2613         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2614
2615         if (idev->dev->flags&IFF_POINTOPOINT) {
2616                 addr.s6_addr32[0] = htonl(0xfe800000);
2617                 scope = IFA_LINK;
2618                 plen = 64;
2619         } else {
2620                 scope = IPV6_ADDR_COMPATv4;
2621                 plen = 96;
2622                 pflags |= RTF_NONEXTHOP;
2623         }
2624
2625         if (addr.s6_addr32[3]) {
2626                 add_addr(idev, &addr, plen, scope);
2627                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2628                 return;
2629         }
2630
2631         for_each_netdev(net, dev) {
2632                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2633                 if (in_dev && (dev->flags & IFF_UP)) {
2634                         struct in_ifaddr *ifa;
2635
2636                         int flag = scope;
2637
2638                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2639
2640                                 addr.s6_addr32[3] = ifa->ifa_local;
2641
2642                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2643                                         continue;
2644                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2645                                         if (idev->dev->flags&IFF_POINTOPOINT)
2646                                                 continue;
2647                                         flag |= IFA_HOST;
2648                                 }
2649
2650                                 add_addr(idev, &addr, plen, flag);
2651                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2652                                                       pflags);
2653                         }
2654                 }
2655         }
2656 }
2657 #endif
2658
2659 static void init_loopback(struct net_device *dev)
2660 {
2661         struct inet6_dev  *idev;
2662         struct net_device *sp_dev;
2663         struct inet6_ifaddr *sp_ifa;
2664         struct rt6_info *sp_rt;
2665
2666         /* ::1 */
2667
2668         ASSERT_RTNL();
2669
2670         if ((idev = ipv6_find_idev(dev)) == NULL) {
2671                 pr_debug("%s: add_dev failed\n", __func__);
2672                 return;
2673         }
2674
2675         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2676
2677         /* Add routes to other interface's IPv6 addresses */
2678         for_each_netdev(dev_net(dev), sp_dev) {
2679                 if (!strcmp(sp_dev->name, dev->name))
2680                         continue;
2681
2682                 idev = __in6_dev_get(sp_dev);
2683                 if (!idev)
2684                         continue;
2685
2686                 read_lock_bh(&idev->lock);
2687                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2688
2689                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2690                                 continue;
2691
2692                         if (sp_ifa->rt) {
2693                                 /* This dst has been added to garbage list when
2694                                  * lo device down, release this obsolete dst and
2695                                  * reallocate a new router for ifa.
2696                                  */
2697                                 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
2698                                         ip6_rt_put(sp_ifa->rt);
2699                                         sp_ifa->rt = NULL;
2700                                 } else {
2701                                         continue;
2702                                 }
2703                         }
2704
2705                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2706
2707                         /* Failure cases are ignored */
2708                         if (!IS_ERR(sp_rt)) {
2709                                 sp_ifa->rt = sp_rt;
2710                                 ip6_ins_rt(sp_rt);
2711                         }
2712                 }
2713                 read_unlock_bh(&idev->lock);
2714         }
2715 }
2716
2717 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2718 {
2719         struct inet6_ifaddr *ifp;
2720         u32 addr_flags = IFA_F_PERMANENT;
2721
2722 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2723         if (idev->cnf.optimistic_dad &&
2724             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2725                 addr_flags |= IFA_F_OPTIMISTIC;
2726 #endif
2727
2728
2729         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2730                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2731         if (!IS_ERR(ifp)) {
2732                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2733                 addrconf_dad_start(ifp);
2734                 in6_ifa_put(ifp);
2735         }
2736 }
2737
2738 static void addrconf_dev_config(struct net_device *dev)
2739 {
2740         struct in6_addr addr;
2741         struct inet6_dev *idev;
2742
2743         ASSERT_RTNL();
2744
2745         if ((dev->type != ARPHRD_ETHER) &&
2746             (dev->type != ARPHRD_FDDI) &&
2747             (dev->type != ARPHRD_ARCNET) &&
2748             (dev->type != ARPHRD_INFINIBAND) &&
2749             (dev->type != ARPHRD_IEEE802154) &&
2750             (dev->type != ARPHRD_IEEE1394) &&
2751             (dev->type != ARPHRD_TUNNEL6) &&
2752             (dev->type != ARPHRD_6LOWPAN)) {
2753                 /* Alas, we support only Ethernet autoconfiguration. */
2754                 return;
2755         }
2756
2757         idev = addrconf_add_dev(dev);
2758         if (IS_ERR(idev))
2759                 return;
2760
2761         memset(&addr, 0, sizeof(struct in6_addr));
2762         addr.s6_addr32[0] = htonl(0xFE800000);
2763
2764         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2765                 addrconf_add_linklocal(idev, &addr);
2766 }
2767
2768 #if IS_ENABLED(CONFIG_IPV6_SIT)
2769 static void addrconf_sit_config(struct net_device *dev)
2770 {
2771         struct inet6_dev *idev;
2772
2773         ASSERT_RTNL();
2774
2775         /*
2776          * Configure the tunnel with one of our IPv4
2777          * addresses... we should configure all of
2778          * our v4 addrs in the tunnel
2779          */
2780
2781         if ((idev = ipv6_find_idev(dev)) == NULL) {
2782                 pr_debug("%s: add_dev failed\n", __func__);
2783                 return;
2784         }
2785
2786         if (dev->priv_flags & IFF_ISATAP) {
2787                 struct in6_addr addr;
2788
2789                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2790                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2791                         addrconf_add_linklocal(idev, &addr);
2792                 return;
2793         }
2794
2795         sit_add_v4_addrs(idev);
2796
2797         if (dev->flags&IFF_POINTOPOINT)
2798                 addrconf_add_mroute(dev);
2799 }
2800 #endif
2801
2802 #if IS_ENABLED(CONFIG_NET_IPGRE)
2803 static void addrconf_gre_config(struct net_device *dev)
2804 {
2805         struct inet6_dev *idev;
2806         struct in6_addr addr;
2807
2808         ASSERT_RTNL();
2809
2810         if ((idev = ipv6_find_idev(dev)) == NULL) {
2811                 pr_debug("%s: add_dev failed\n", __func__);
2812                 return;
2813         }
2814
2815         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2816         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2817                 addrconf_add_linklocal(idev, &addr);
2818         else
2819                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2820 }
2821 #endif
2822
2823 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2824                            void *ptr)
2825 {
2826         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2827         struct inet6_dev *idev = __in6_dev_get(dev);
2828         struct net *net = dev_net(dev);
2829         int run_pending = 0;
2830         int err;
2831
2832         switch (event) {
2833         case NETDEV_REGISTER:
2834                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2835                         idev = ipv6_add_dev(dev);
2836                         if (!idev)
2837                                 return notifier_from_errno(-ENOMEM);
2838                 }
2839                 break;
2840
2841         case NETDEV_UP:
2842         case NETDEV_CHANGE:
2843                 if (dev->flags & IFF_SLAVE)
2844                         break;
2845
2846                 if (event == NETDEV_UP) {
2847                         if (!addrconf_qdisc_ok(dev)) {
2848                                 /* device is not ready yet. */
2849                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2850                                         dev->name);
2851                                 break;
2852                         }
2853
2854                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2855                                 idev = ipv6_add_dev(dev);
2856
2857                         if (idev) {
2858                                 idev->if_flags |= IF_READY;
2859                                 run_pending = 1;
2860                         }
2861                 } else {
2862                         if (!addrconf_qdisc_ok(dev)) {
2863                                 /* device is still not ready. */
2864                                 break;
2865                         }
2866
2867                         if (idev) {
2868                                 if (idev->if_flags & IF_READY)
2869                                         /* device is already configured. */
2870                                         break;
2871                                 idev->if_flags |= IF_READY;
2872                         }
2873
2874                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2875                                 dev->name);
2876
2877                         run_pending = 1;
2878                 }
2879
2880                 switch (dev->type) {
2881 #if IS_ENABLED(CONFIG_IPV6_SIT)
2882                 case ARPHRD_SIT:
2883                         addrconf_sit_config(dev);
2884                         break;
2885 #endif
2886 #if IS_ENABLED(CONFIG_NET_IPGRE)
2887                 case ARPHRD_IPGRE:
2888                         addrconf_gre_config(dev);
2889                         break;
2890 #endif
2891                 case ARPHRD_LOOPBACK:
2892                         init_loopback(dev);
2893                         break;
2894
2895                 default:
2896                         addrconf_dev_config(dev);
2897                         break;
2898                 }
2899
2900                 if (idev) {
2901                         if (run_pending)
2902                                 addrconf_dad_run(idev);
2903
2904                         /*
2905                          * If the MTU changed during the interface down,
2906                          * when the interface up, the changed MTU must be
2907                          * reflected in the idev as well as routers.
2908                          */
2909                         if (idev->cnf.mtu6 != dev->mtu &&
2910                             dev->mtu >= IPV6_MIN_MTU) {
2911                                 rt6_mtu_change(dev, dev->mtu);
2912                                 idev->cnf.mtu6 = dev->mtu;
2913                         }
2914                         idev->tstamp = jiffies;
2915                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2916
2917                         /*
2918                          * If the changed mtu during down is lower than
2919                          * IPV6_MIN_MTU stop IPv6 on this interface.
2920                          */
2921                         if (dev->mtu < IPV6_MIN_MTU)
2922                                 addrconf_ifdown(dev, dev != net->loopback_dev);
2923                 }
2924                 break;
2925
2926         case NETDEV_CHANGEMTU:
2927                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2928                         rt6_mtu_change(dev, dev->mtu);
2929                         idev->cnf.mtu6 = dev->mtu;
2930                         break;
2931                 }
2932
2933                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2934                         idev = ipv6_add_dev(dev);
2935                         if (idev)
2936                                 break;
2937                 }
2938
2939                 /*
2940                  * if MTU under IPV6_MIN_MTU.
2941                  * Stop IPv6 on this interface.
2942                  */
2943                 addrconf_ifdown(dev, dev != net->loopback_dev);
2944                 break;
2945
2946         case NETDEV_DOWN:
2947         case NETDEV_UNREGISTER:
2948                 /*
2949                  *      Remove all addresses from this interface.
2950                  */
2951                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2952                 break;
2953
2954         case NETDEV_CHANGENAME:
2955                 if (idev) {
2956                         snmp6_unregister_dev(idev);
2957                         addrconf_sysctl_unregister(idev);
2958                         addrconf_sysctl_register(idev);
2959                         err = snmp6_register_dev(idev);
2960                         if (err)
2961                                 return notifier_from_errno(err);
2962                 }
2963                 break;
2964
2965         case NETDEV_PRE_TYPE_CHANGE:
2966         case NETDEV_POST_TYPE_CHANGE:
2967                 addrconf_type_change(dev, event);
2968                 break;
2969         }
2970
2971         return NOTIFY_OK;
2972 }
2973
2974 /*
2975  *      addrconf module should be notified of a device going up
2976  */
2977 static struct notifier_block ipv6_dev_notf = {
2978         .notifier_call = addrconf_notify,
2979         .priority = ADDRCONF_NOTIFY_PRIORITY,
2980 };
2981
2982 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2983 {
2984         struct inet6_dev *idev;
2985         ASSERT_RTNL();
2986
2987         idev = __in6_dev_get(dev);
2988
2989         if (event == NETDEV_POST_TYPE_CHANGE)
2990                 ipv6_mc_remap(idev);
2991         else if (event == NETDEV_PRE_TYPE_CHANGE)
2992                 ipv6_mc_unmap(idev);
2993 }
2994
2995 static int addrconf_ifdown(struct net_device *dev, int how)
2996 {
2997         struct net *net = dev_net(dev);
2998         struct inet6_dev *idev;
2999         struct inet6_ifaddr *ifa;
3000         int state, i;
3001
3002         ASSERT_RTNL();
3003
3004         rt6_ifdown(net, dev);
3005         neigh_ifdown(&nd_tbl, dev);
3006
3007         idev = __in6_dev_get(dev);
3008         if (idev == NULL)
3009                 return -ENODEV;
3010
3011         /*
3012          * Step 1: remove reference to ipv6 device from parent device.
3013          *         Do not dev_put!
3014          */
3015         if (how) {
3016                 idev->dead = 1;
3017
3018                 /* protected by rtnl_lock */
3019                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3020
3021                 /* Step 1.5: remove snmp6 entry */
3022                 snmp6_unregister_dev(idev);
3023
3024         }
3025
3026         /* Step 2: clear hash table */
3027         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3028                 struct hlist_head *h = &inet6_addr_lst[i];
3029
3030                 spin_lock_bh(&addrconf_hash_lock);
3031         restart:
3032                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3033                         if (ifa->idev == idev) {
3034                                 hlist_del_init_rcu(&ifa->addr_lst);
3035                                 addrconf_del_dad_work(ifa);
3036                                 goto restart;
3037                         }
3038                 }
3039                 spin_unlock_bh(&addrconf_hash_lock);
3040         }
3041
3042         write_lock_bh(&idev->lock);
3043
3044         addrconf_del_rs_timer(idev);
3045
3046         /* Step 2: clear flags for stateless addrconf */
3047         if (!how)
3048                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3049
3050         if (how && del_timer(&idev->regen_timer))
3051                 in6_dev_put(idev);
3052
3053         /* Step 3: clear tempaddr list */
3054         while (!list_empty(&idev->tempaddr_list)) {
3055                 ifa = list_first_entry(&idev->tempaddr_list,
3056                                        struct inet6_ifaddr, tmp_list);
3057                 list_del(&ifa->tmp_list);
3058                 write_unlock_bh(&idev->lock);
3059                 spin_lock_bh(&ifa->lock);
3060
3061                 if (ifa->ifpub) {
3062                         in6_ifa_put(ifa->ifpub);
3063                         ifa->ifpub = NULL;
3064                 }
3065                 spin_unlock_bh(&ifa->lock);
3066                 in6_ifa_put(ifa);
3067                 write_lock_bh(&idev->lock);
3068         }
3069
3070         while (!list_empty(&idev->addr_list)) {
3071                 ifa = list_first_entry(&idev->addr_list,
3072                                        struct inet6_ifaddr, if_list);
3073                 addrconf_del_dad_work(ifa);
3074
3075                 list_del(&ifa->if_list);
3076
3077                 write_unlock_bh(&idev->lock);
3078
3079                 spin_lock_bh(&ifa->state_lock);
3080                 state = ifa->state;
3081                 ifa->state = INET6_IFADDR_STATE_DEAD;
3082                 spin_unlock_bh(&ifa->state_lock);
3083
3084                 if (state != INET6_IFADDR_STATE_DEAD) {
3085                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3086                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3087                 }
3088                 in6_ifa_put(ifa);
3089
3090                 write_lock_bh(&idev->lock);
3091         }
3092
3093         write_unlock_bh(&idev->lock);
3094
3095         /* Step 5: Discard multicast list */
3096         if (how)
3097                 ipv6_mc_destroy_dev(idev);
3098         else
3099                 ipv6_mc_down(idev);
3100
3101         idev->tstamp = jiffies;
3102
3103         /* Last: Shot the device (if unregistered) */
3104         if (how) {
3105                 addrconf_sysctl_unregister(idev);
3106                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3107                 neigh_ifdown(&nd_tbl, dev);
3108                 in6_dev_put(idev);
3109         }
3110         return 0;
3111 }
3112
3113 static void addrconf_rs_timer(unsigned long data)
3114 {
3115         struct inet6_dev *idev = (struct inet6_dev *)data;
3116         struct net_device *dev = idev->dev;
3117         struct in6_addr lladdr;
3118
3119         write_lock(&idev->lock);
3120         if (idev->dead || !(idev->if_flags & IF_READY))
3121                 goto out;
3122
3123         if (!ipv6_accept_ra(idev))
3124                 goto out;
3125
3126         /* Announcement received after solicitation was sent */
3127         if (idev->if_flags & IF_RA_RCVD)
3128                 goto out;
3129
3130         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3131                 write_unlock(&idev->lock);
3132                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3133                         ndisc_send_rs(dev, &lladdr,
3134                                       &in6addr_linklocal_allrouters);
3135                 else
3136                         goto put;
3137
3138                 write_lock(&idev->lock);
3139                 /* The wait after the last probe can be shorter */
3140                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3141                                              idev->cnf.rtr_solicits) ?
3142                                       idev->cnf.rtr_solicit_delay :
3143                                       idev->cnf.rtr_solicit_interval);
3144         } else {
3145                 /*
3146                  * Note: we do not support deprecated "all on-link"
3147                  * assumption any longer.
3148                  */
3149                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3150         }
3151
3152 out:
3153         write_unlock(&idev->lock);
3154 put:
3155         in6_dev_put(idev);
3156 }
3157
3158 /*
3159  *      Duplicate Address Detection
3160  */
3161 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3162 {
3163         unsigned long rand_num;
3164         struct inet6_dev *idev = ifp->idev;
3165
3166         if (ifp->flags & IFA_F_OPTIMISTIC)
3167                 rand_num = 0;
3168         else
3169                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3170
3171         ifp->dad_probes = idev->cnf.dad_transmits;
3172         addrconf_mod_dad_work(ifp, rand_num);
3173 }
3174
3175 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3176 {
3177         struct inet6_dev *idev = ifp->idev;
3178         struct net_device *dev = idev->dev;
3179
3180         addrconf_join_solict(dev, &ifp->addr);
3181
3182         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3183
3184         read_lock_bh(&idev->lock);
3185         spin_lock(&ifp->lock);
3186         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3187                 goto out;
3188
3189         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3190             idev->cnf.accept_dad < 1 ||
3191             !(ifp->flags&IFA_F_TENTATIVE) ||
3192             ifp->flags & IFA_F_NODAD) {
3193                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3194                 spin_unlock(&ifp->lock);
3195                 read_unlock_bh(&idev->lock);
3196
3197                 addrconf_dad_completed(ifp);
3198                 return;
3199         }
3200
3201         if (!(idev->if_flags & IF_READY)) {
3202                 spin_unlock(&ifp->lock);
3203                 read_unlock_bh(&idev->lock);
3204                 /*
3205                  * If the device is not ready:
3206                  * - keep it tentative if it is a permanent address.
3207                  * - otherwise, kill it.
3208                  */
3209                 in6_ifa_hold(ifp);
3210                 addrconf_dad_stop(ifp, 0);
3211                 return;
3212         }
3213
3214         /*
3215          * Optimistic nodes can start receiving
3216          * Frames right away
3217          */
3218         if (ifp->flags & IFA_F_OPTIMISTIC)
3219                 ip6_ins_rt(ifp->rt);
3220
3221         addrconf_dad_kick(ifp);
3222 out:
3223         spin_unlock(&ifp->lock);
3224         read_unlock_bh(&idev->lock);
3225 }
3226
3227 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3228 {
3229         bool begin_dad = false;
3230
3231         spin_lock_bh(&ifp->state_lock);
3232         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3233                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3234                 begin_dad = true;
3235         }
3236         spin_unlock_bh(&ifp->state_lock);
3237
3238         if (begin_dad)
3239                 addrconf_mod_dad_work(ifp, 0);
3240 }
3241
3242 static void addrconf_dad_work(struct work_struct *w)
3243 {
3244         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3245                                                 struct inet6_ifaddr,
3246                                                 dad_work);
3247         struct inet6_dev *idev = ifp->idev;
3248         struct in6_addr mcaddr;
3249
3250         enum {
3251                 DAD_PROCESS,
3252                 DAD_BEGIN,
3253                 DAD_ABORT,
3254         } action = DAD_PROCESS;
3255
3256         rtnl_lock();
3257
3258         spin_lock_bh(&ifp->state_lock);
3259         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3260                 action = DAD_BEGIN;
3261                 ifp->state = INET6_IFADDR_STATE_DAD;
3262         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3263                 action = DAD_ABORT;
3264                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3265         }
3266         spin_unlock_bh(&ifp->state_lock);
3267
3268         if (action == DAD_BEGIN) {
3269                 addrconf_dad_begin(ifp);
3270                 goto out;
3271         } else if (action == DAD_ABORT) {
3272                 in6_ifa_hold(ifp);
3273                 addrconf_dad_stop(ifp, 1);
3274                 goto out;
3275         }
3276
3277         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3278                 goto out;
3279
3280         write_lock_bh(&idev->lock);
3281         if (idev->dead || !(idev->if_flags & IF_READY)) {
3282                 write_unlock_bh(&idev->lock);
3283                 goto out;
3284         }
3285
3286         spin_lock(&ifp->lock);
3287         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3288                 spin_unlock(&ifp->lock);
3289                 write_unlock_bh(&idev->lock);
3290                 goto out;
3291         }
3292
3293         if (ifp->dad_probes == 0) {
3294                 /*
3295                  * DAD was successful
3296                  */
3297
3298                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3299                 spin_unlock(&ifp->lock);
3300                 write_unlock_bh(&idev->lock);
3301
3302                 addrconf_dad_completed(ifp);
3303
3304                 goto out;
3305         }
3306
3307         ifp->dad_probes--;
3308         addrconf_mod_dad_work(ifp,
3309                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3310         spin_unlock(&ifp->lock);
3311         write_unlock_bh(&idev->lock);
3312
3313         /* send a neighbour solicitation for our addr */
3314         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3315         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3316 out:
3317         in6_ifa_put(ifp);
3318         rtnl_unlock();
3319 }
3320
3321 /* ifp->idev must be at least read locked */
3322 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3323 {
3324         struct inet6_ifaddr *ifpiter;
3325         struct inet6_dev *idev = ifp->idev;
3326
3327         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3328                 if (ifpiter->scope > IFA_LINK)
3329                         break;
3330                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3331                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3332                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3333                     IFA_F_PERMANENT)
3334                         return false;
3335         }
3336         return true;
3337 }
3338
3339 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3340 {
3341         struct net_device *dev = ifp->idev->dev;
3342         struct in6_addr lladdr;
3343         bool send_rs, send_mld;
3344
3345         addrconf_del_dad_work(ifp);
3346
3347         /*
3348          *      Configure the address for reception. Now it is valid.
3349          */
3350
3351         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3352
3353         /* If added prefix is link local and we are prepared to process
3354            router advertisements, start sending router solicitations.
3355          */
3356
3357         read_lock_bh(&ifp->idev->lock);
3358         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3359         send_rs = send_mld &&
3360                   ipv6_accept_ra(ifp->idev) &&
3361                   ifp->idev->cnf.rtr_solicits > 0 &&
3362                   (dev->flags&IFF_LOOPBACK) == 0;
3363         read_unlock_bh(&ifp->idev->lock);
3364
3365         /* While dad is in progress mld report's source address is in6_addrany.
3366          * Resend with proper ll now.
3367          */
3368         if (send_mld)
3369                 ipv6_mc_dad_complete(ifp->idev);
3370
3371         if (send_rs) {
3372                 /*
3373                  *      If a host as already performed a random delay
3374                  *      [...] as part of DAD [...] there is no need
3375                  *      to delay again before sending the first RS
3376                  */
3377                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3378                         return;
3379                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3380
3381                 write_lock_bh(&ifp->idev->lock);
3382                 spin_lock(&ifp->lock);
3383                 ifp->idev->rs_probes = 1;
3384                 ifp->idev->if_flags |= IF_RS_SENT;
3385                 addrconf_mod_rs_timer(ifp->idev,
3386                                       ifp->idev->cnf.rtr_solicit_interval);
3387                 spin_unlock(&ifp->lock);
3388                 write_unlock_bh(&ifp->idev->lock);
3389         }
3390 }
3391
3392 static void addrconf_dad_run(struct inet6_dev *idev)
3393 {
3394         struct inet6_ifaddr *ifp;
3395
3396         read_lock_bh(&idev->lock);
3397         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3398                 spin_lock(&ifp->lock);
3399                 if (ifp->flags & IFA_F_TENTATIVE &&
3400                     ifp->state == INET6_IFADDR_STATE_DAD)
3401                         addrconf_dad_kick(ifp);
3402                 spin_unlock(&ifp->lock);
3403         }
3404         read_unlock_bh(&idev->lock);
3405 }
3406
3407 #ifdef CONFIG_PROC_FS
3408 struct if6_iter_state {
3409         struct seq_net_private p;
3410         int bucket;
3411         int offset;
3412 };
3413
3414 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3415 {
3416         struct inet6_ifaddr *ifa = NULL;
3417         struct if6_iter_state *state = seq->private;
3418         struct net *net = seq_file_net(seq);
3419         int p = 0;
3420
3421         /* initial bucket if pos is 0 */
3422         if (pos == 0) {
3423                 state->bucket = 0;
3424                 state->offset = 0;
3425         }
3426
3427         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3428                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3429                                          addr_lst) {
3430                         if (!net_eq(dev_net(ifa->idev->dev), net))
3431                                 continue;
3432                         /* sync with offset */
3433                         if (p < state->offset) {
3434                                 p++;
3435                                 continue;
3436                         }
3437                         state->offset++;
3438                         return ifa;
3439                 }
3440
3441                 /* prepare for next bucket */
3442                 state->offset = 0;
3443                 p = 0;
3444         }
3445         return NULL;
3446 }
3447
3448 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3449                                          struct inet6_ifaddr *ifa)
3450 {
3451         struct if6_iter_state *state = seq->private;
3452         struct net *net = seq_file_net(seq);
3453
3454         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3455                 if (!net_eq(dev_net(ifa->idev->dev), net))
3456                         continue;
3457                 state->offset++;
3458                 return ifa;
3459         }
3460
3461         while (++state->bucket < IN6_ADDR_HSIZE) {
3462                 state->offset = 0;
3463                 hlist_for_each_entry_rcu_bh(ifa,
3464                                      &inet6_addr_lst[state->bucket], addr_lst) {
3465                         if (!net_eq(dev_net(ifa->idev->dev), net))
3466                                 continue;
3467                         state->offset++;
3468                         return ifa;
3469                 }
3470         }
3471
3472         return NULL;
3473 }
3474
3475 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3476         __acquires(rcu_bh)
3477 {
3478         rcu_read_lock_bh();
3479         return if6_get_first(seq, *pos);
3480 }
3481
3482 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3483 {
3484         struct inet6_ifaddr *ifa;
3485
3486         ifa = if6_get_next(seq, v);
3487         ++*pos;
3488         return ifa;
3489 }
3490
3491 static void if6_seq_stop(struct seq_file *seq, void *v)
3492         __releases(rcu_bh)
3493 {
3494         rcu_read_unlock_bh();
3495 }
3496
3497 static int if6_seq_show(struct seq_file *seq, void *v)
3498 {
3499         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3500         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3501                    &ifp->addr,
3502                    ifp->idev->dev->ifindex,
3503                    ifp->prefix_len,
3504                    ifp->scope,
3505                    (u8) ifp->flags,
3506                    ifp->idev->dev->name);
3507         return 0;
3508 }
3509
3510 static const struct seq_operations if6_seq_ops = {
3511         .start  = if6_seq_start,
3512         .next   = if6_seq_next,
3513         .show   = if6_seq_show,
3514         .stop   = if6_seq_stop,
3515 };
3516
3517 static int if6_seq_open(struct inode *inode, struct file *file)
3518 {
3519         return seq_open_net(inode, file, &if6_seq_ops,
3520                             sizeof(struct if6_iter_state));
3521 }
3522
3523 static const struct file_operations if6_fops = {
3524         .owner          = THIS_MODULE,
3525         .open           = if6_seq_open,
3526         .read           = seq_read,
3527         .llseek         = seq_lseek,
3528         .release        = seq_release_net,
3529 };
3530
3531 static int __net_init if6_proc_net_init(struct net *net)
3532 {
3533         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3534                 return -ENOMEM;
3535         return 0;
3536 }
3537
3538 static void __net_exit if6_proc_net_exit(struct net *net)
3539 {
3540         remove_proc_entry("if_inet6", net->proc_net);
3541 }
3542
3543 static struct pernet_operations if6_proc_net_ops = {
3544        .init = if6_proc_net_init,
3545        .exit = if6_proc_net_exit,
3546 };
3547
3548 int __init if6_proc_init(void)
3549 {
3550         return register_pernet_subsys(&if6_proc_net_ops);
3551 }
3552
3553 void if6_proc_exit(void)
3554 {
3555         unregister_pernet_subsys(&if6_proc_net_ops);
3556 }
3557 #endif  /* CONFIG_PROC_FS */
3558
3559 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3560 /* Check if address is a home address configured on any interface. */
3561 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3562 {
3563         int ret = 0;
3564         struct inet6_ifaddr *ifp = NULL;
3565         unsigned int hash = inet6_addr_hash(addr);
3566
3567         rcu_read_lock_bh();
3568         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3569                 if (!net_eq(dev_net(ifp->idev->dev), net))
3570                         continue;
3571                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3572                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3573                         ret = 1;
3574                         break;
3575                 }
3576         }
3577         rcu_read_unlock_bh();
3578         return ret;
3579 }
3580 #endif
3581
3582 /*
3583  *      Periodic address status verification
3584  */
3585
3586 static void addrconf_verify_rtnl(void)
3587 {
3588         unsigned long now, next, next_sec, next_sched;
3589         struct inet6_ifaddr *ifp;
3590         int i;
3591
3592         ASSERT_RTNL();
3593
3594         rcu_read_lock_bh();
3595         now = jiffies;
3596         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3597
3598         cancel_delayed_work(&addr_chk_work);
3599
3600         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3601 restart:
3602                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3603                         unsigned long age;
3604
3605                         /* When setting preferred_lft to a value not zero or
3606                          * infinity, while valid_lft is infinity
3607                          * IFA_F_PERMANENT has a non-infinity life time.
3608                          */
3609                         if ((ifp->flags & IFA_F_PERMANENT) &&
3610                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3611                                 continue;
3612
3613                         spin_lock(&ifp->lock);
3614                         /* We try to batch several events at once. */
3615                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3616
3617                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3618                             age >= ifp->valid_lft) {
3619                                 spin_unlock(&ifp->lock);
3620                                 in6_ifa_hold(ifp);
3621                                 ipv6_del_addr(ifp);
3622                                 goto restart;
3623                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3624                                 spin_unlock(&ifp->lock);
3625                                 continue;
3626                         } else if (age >= ifp->prefered_lft) {
3627                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3628                                 int deprecate = 0;
3629
3630                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3631                                         deprecate = 1;
3632                                         ifp->flags |= IFA_F_DEPRECATED;
3633                                 }
3634
3635                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3636                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3637                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3638
3639                                 spin_unlock(&ifp->lock);
3640
3641                                 if (deprecate) {
3642                                         in6_ifa_hold(ifp);
3643
3644                                         ipv6_ifa_notify(0, ifp);
3645                                         in6_ifa_put(ifp);
3646                                         goto restart;
3647                                 }
3648                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3649                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3650                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3651                                         ifp->idev->cnf.dad_transmits *
3652                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3653
3654                                 if (age >= ifp->prefered_lft - regen_advance) {
3655                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3656                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3657                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3658                                         if (!ifp->regen_count && ifpub) {
3659                                                 ifp->regen_count++;
3660                                                 in6_ifa_hold(ifp);
3661                                                 in6_ifa_hold(ifpub);
3662                                                 spin_unlock(&ifp->lock);
3663
3664                                                 spin_lock(&ifpub->lock);
3665                                                 ifpub->regen_count = 0;
3666                                                 spin_unlock(&ifpub->lock);
3667                                                 ipv6_create_tempaddr(ifpub, ifp);
3668                                                 in6_ifa_put(ifpub);
3669                                                 in6_ifa_put(ifp);
3670                                                 goto restart;
3671                                         }
3672                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3673                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3674                                 spin_unlock(&ifp->lock);
3675                         } else {
3676                                 /* ifp->prefered_lft <= ifp->valid_lft */
3677                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3678                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3679                                 spin_unlock(&ifp->lock);
3680                         }
3681                 }
3682         }
3683
3684         next_sec = round_jiffies_up(next);
3685         next_sched = next;
3686
3687         /* If rounded timeout is accurate enough, accept it. */
3688         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3689                 next_sched = next_sec;
3690
3691         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3692         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3693                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3694
3695         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3696               now, next, next_sec, next_sched);
3697         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3698         rcu_read_unlock_bh();
3699 }
3700
3701 static void addrconf_verify_work(struct work_struct *w)
3702 {
3703         rtnl_lock();
3704         addrconf_verify_rtnl();
3705         rtnl_unlock();
3706 }
3707
3708 static void addrconf_verify(void)
3709 {
3710         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3711 }
3712
3713 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3714                                      struct in6_addr **peer_pfx)
3715 {
3716         struct in6_addr *pfx = NULL;
3717
3718         *peer_pfx = NULL;
3719
3720         if (addr)
3721                 pfx = nla_data(addr);
3722
3723         if (local) {
3724                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3725                         *peer_pfx = pfx;
3726                 pfx = nla_data(local);
3727         }
3728
3729         return pfx;
3730 }
3731
3732 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3733         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3734         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3735         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3736         [IFA_FLAGS]             = { .len = sizeof(u32) },
3737 };
3738
3739 static int
3740 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3741 {
3742         struct net *net = sock_net(skb->sk);
3743         struct ifaddrmsg *ifm;
3744         struct nlattr *tb[IFA_MAX+1];
3745         struct in6_addr *pfx, *peer_pfx;
3746         u32 ifa_flags;
3747         int err;
3748
3749         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3750         if (err < 0)
3751                 return err;
3752
3753         ifm = nlmsg_data(nlh);
3754         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3755         if (pfx == NULL)
3756                 return -EINVAL;
3757
3758         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3759
3760         /* We ignore other flags so far. */
3761         ifa_flags &= IFA_F_MANAGETEMPADDR;
3762
3763         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3764                               ifm->ifa_prefixlen);
3765 }
3766
3767 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3768                              u32 prefered_lft, u32 valid_lft)
3769 {
3770         u32 flags;
3771         clock_t expires;
3772         unsigned long timeout;
3773         bool was_managetempaddr;
3774         bool had_prefixroute;
3775
3776         ASSERT_RTNL();
3777
3778         if (!valid_lft || (prefered_lft > valid_lft))
3779                 return -EINVAL;
3780
3781         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3782             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3783                 return -EINVAL;
3784
3785         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3786         if (addrconf_finite_timeout(timeout)) {
3787                 expires = jiffies_to_clock_t(timeout * HZ);
3788                 valid_lft = timeout;
3789                 flags = RTF_EXPIRES;
3790         } else {
3791                 expires = 0;
3792                 flags = 0;
3793                 ifa_flags |= IFA_F_PERMANENT;
3794         }
3795
3796         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3797         if (addrconf_finite_timeout(timeout)) {
3798                 if (timeout == 0)
3799                         ifa_flags |= IFA_F_DEPRECATED;
3800                 prefered_lft = timeout;
3801         }
3802
3803         spin_lock_bh(&ifp->lock);
3804         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3805         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3806                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
3807         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3808                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3809                         IFA_F_NOPREFIXROUTE);
3810         ifp->flags |= ifa_flags;
3811         ifp->tstamp = jiffies;
3812         ifp->valid_lft = valid_lft;
3813         ifp->prefered_lft = prefered_lft;
3814
3815         spin_unlock_bh(&ifp->lock);
3816         if (!(ifp->flags&IFA_F_TENTATIVE))
3817                 ipv6_ifa_notify(0, ifp);
3818
3819         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3820                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3821                                       expires, flags);
3822         } else if (had_prefixroute) {
3823                 enum cleanup_prefix_rt_t action;
3824                 unsigned long rt_expires;
3825
3826                 write_lock_bh(&ifp->idev->lock);
3827                 action = check_cleanup_prefix_route(ifp, &rt_expires);
3828                 write_unlock_bh(&ifp->idev->lock);
3829
3830                 if (action != CLEANUP_PREFIX_RT_NOP) {
3831                         cleanup_prefix_route(ifp, rt_expires,
3832                                 action == CLEANUP_PREFIX_RT_DEL);
3833                 }
3834         }
3835
3836         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3837                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3838                         valid_lft = prefered_lft = 0;
3839                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3840                                  !was_managetempaddr, jiffies);
3841         }
3842
3843         addrconf_verify_rtnl();
3844
3845         return 0;
3846 }
3847
3848 static int
3849 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3850 {
3851         struct net *net = sock_net(skb->sk);
3852         struct ifaddrmsg *ifm;
3853         struct nlattr *tb[IFA_MAX+1];
3854         struct in6_addr *pfx, *peer_pfx;
3855         struct inet6_ifaddr *ifa;
3856         struct net_device *dev;
3857         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3858         u32 ifa_flags;
3859         int err;
3860
3861         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3862         if (err < 0)
3863                 return err;
3864
3865         ifm = nlmsg_data(nlh);
3866         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3867         if (pfx == NULL)
3868                 return -EINVAL;
3869
3870         if (tb[IFA_CACHEINFO]) {
3871                 struct ifa_cacheinfo *ci;
3872
3873                 ci = nla_data(tb[IFA_CACHEINFO]);
3874                 valid_lft = ci->ifa_valid;
3875                 preferred_lft = ci->ifa_prefered;
3876         } else {
3877                 preferred_lft = INFINITY_LIFE_TIME;
3878                 valid_lft = INFINITY_LIFE_TIME;
3879         }
3880
3881         dev =  __dev_get_by_index(net, ifm->ifa_index);
3882         if (dev == NULL)
3883                 return -ENODEV;
3884
3885         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3886
3887         /* We ignore other flags so far. */
3888         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3889                      IFA_F_NOPREFIXROUTE;
3890
3891         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3892         if (ifa == NULL) {
3893                 /*
3894                  * It would be best to check for !NLM_F_CREATE here but
3895                  * userspace already relies on not having to provide this.
3896                  */
3897                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3898                                       ifm->ifa_prefixlen, ifa_flags,
3899                                       preferred_lft, valid_lft);
3900         }
3901
3902         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3903             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3904                 err = -EEXIST;
3905         else
3906                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3907
3908         in6_ifa_put(ifa);
3909
3910         return err;
3911 }
3912
3913 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3914                           u8 scope, int ifindex)
3915 {
3916         struct ifaddrmsg *ifm;
3917
3918         ifm = nlmsg_data(nlh);
3919         ifm->ifa_family = AF_INET6;
3920         ifm->ifa_prefixlen = prefixlen;
3921         ifm->ifa_flags = flags;
3922         ifm->ifa_scope = scope;
3923         ifm->ifa_index = ifindex;
3924 }
3925
3926 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3927                          unsigned long tstamp, u32 preferred, u32 valid)
3928 {
3929         struct ifa_cacheinfo ci;
3930
3931         ci.cstamp = cstamp_delta(cstamp);
3932         ci.tstamp = cstamp_delta(tstamp);
3933         ci.ifa_prefered = preferred;
3934         ci.ifa_valid = valid;
3935
3936         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3937 }
3938
3939 static inline int rt_scope(int ifa_scope)
3940 {
3941         if (ifa_scope & IFA_HOST)
3942                 return RT_SCOPE_HOST;
3943         else if (ifa_scope & IFA_LINK)
3944                 return RT_SCOPE_LINK;
3945         else if (ifa_scope & IFA_SITE)
3946                 return RT_SCOPE_SITE;
3947         else
3948                 return RT_SCOPE_UNIVERSE;
3949 }
3950
3951 static inline int inet6_ifaddr_msgsize(void)
3952 {
3953         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3954                + nla_total_size(16) /* IFA_LOCAL */
3955                + nla_total_size(16) /* IFA_ADDRESS */
3956                + nla_total_size(sizeof(struct ifa_cacheinfo))
3957                + nla_total_size(4)  /* IFA_FLAGS */;
3958 }
3959
3960 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3961                              u32 portid, u32 seq, int event, unsigned int flags)
3962 {
3963         struct nlmsghdr  *nlh;
3964         u32 preferred, valid;
3965
3966         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3967         if (nlh == NULL)
3968                 return -EMSGSIZE;
3969
3970         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3971                       ifa->idev->dev->ifindex);
3972
3973         if (!((ifa->flags&IFA_F_PERMANENT) &&
3974               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3975                 preferred = ifa->prefered_lft;
3976                 valid = ifa->valid_lft;
3977                 if (preferred != INFINITY_LIFE_TIME) {
3978                         long tval = (jiffies - ifa->tstamp)/HZ;
3979                         if (preferred > tval)
3980                                 preferred -= tval;
3981                         else
3982                                 preferred = 0;
3983                         if (valid != INFINITY_LIFE_TIME) {
3984                                 if (valid > tval)
3985                                         valid -= tval;
3986                                 else
3987                                         valid = 0;
3988                         }
3989                 }
3990         } else {
3991                 preferred = INFINITY_LIFE_TIME;
3992                 valid = INFINITY_LIFE_TIME;
3993         }
3994
3995         if (!ipv6_addr_any(&ifa->peer_addr)) {
3996                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3997                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3998                         goto error;
3999         } else
4000                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4001                         goto error;
4002
4003         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4004                 goto error;
4005
4006         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4007                 goto error;
4008
4009         return nlmsg_end(skb, nlh);
4010
4011 error:
4012         nlmsg_cancel(skb, nlh);
4013         return -EMSGSIZE;
4014 }
4015
4016 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4017                                 u32 portid, u32 seq, int event, u16 flags)
4018 {
4019         struct nlmsghdr  *nlh;
4020         u8 scope = RT_SCOPE_UNIVERSE;
4021         int ifindex = ifmca->idev->dev->ifindex;
4022
4023         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4024                 scope = RT_SCOPE_SITE;
4025
4026         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4027         if (nlh == NULL)
4028                 return -EMSGSIZE;
4029
4030         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4031         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4032             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4033                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4034                 nlmsg_cancel(skb, nlh);
4035                 return -EMSGSIZE;
4036         }
4037
4038         return nlmsg_end(skb, nlh);
4039 }
4040
4041 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4042                                 u32 portid, u32 seq, int event, unsigned int flags)
4043 {
4044         struct nlmsghdr  *nlh;
4045         u8 scope = RT_SCOPE_UNIVERSE;
4046         int ifindex = ifaca->aca_idev->dev->ifindex;
4047
4048         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4049                 scope = RT_SCOPE_SITE;
4050
4051         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4052         if (nlh == NULL)
4053                 return -EMSGSIZE;
4054
4055         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4056         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4057             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4058                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4059                 nlmsg_cancel(skb, nlh);
4060                 return -EMSGSIZE;
4061         }
4062
4063         return nlmsg_end(skb, nlh);
4064 }
4065
4066 enum addr_type_t {
4067         UNICAST_ADDR,
4068         MULTICAST_ADDR,
4069         ANYCAST_ADDR,
4070 };
4071
4072 /* called with rcu_read_lock() */
4073 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4074                           struct netlink_callback *cb, enum addr_type_t type,
4075                           int s_ip_idx, int *p_ip_idx)
4076 {
4077         struct ifmcaddr6 *ifmca;
4078         struct ifacaddr6 *ifaca;
4079         int err = 1;
4080         int ip_idx = *p_ip_idx;
4081
4082         read_lock_bh(&idev->lock);
4083         switch (type) {
4084         case UNICAST_ADDR: {
4085                 struct inet6_ifaddr *ifa;
4086
4087                 /* unicast address incl. temp addr */
4088                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4089                         if (ip_idx < s_ip_idx)
4090                                 goto next;
4091                         err = inet6_fill_ifaddr(skb, ifa,
4092                                                 NETLINK_CB(cb->skb).portid,
4093                                                 cb->nlh->nlmsg_seq,
4094                                                 RTM_NEWADDR,
4095                                                 NLM_F_MULTI);
4096                         if (err <= 0)
4097                                 break;
4098                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4099 next:
4100                         ip_idx++;
4101                 }
4102                 break;
4103         }
4104         case MULTICAST_ADDR:
4105                 /* multicast address */
4106                 for (ifmca = idev->mc_list; ifmca;
4107                      ifmca = ifmca->next, ip_idx++) {
4108                         if (ip_idx < s_ip_idx)
4109                                 continue;
4110                         err = inet6_fill_ifmcaddr(skb, ifmca,
4111                                                   NETLINK_CB(cb->skb).portid,
4112                                                   cb->nlh->nlmsg_seq,
4113                                                   RTM_GETMULTICAST,
4114                                                   NLM_F_MULTI);
4115                         if (err <= 0)
4116                                 break;
4117                 }
4118                 break;
4119         case ANYCAST_ADDR:
4120                 /* anycast address */
4121                 for (ifaca = idev->ac_list; ifaca;
4122                      ifaca = ifaca->aca_next, ip_idx++) {
4123                         if (ip_idx < s_ip_idx)
4124                                 continue;
4125                         err = inet6_fill_ifacaddr(skb, ifaca,
4126                                                   NETLINK_CB(cb->skb).portid,
4127                                                   cb->nlh->nlmsg_seq,
4128                                                   RTM_GETANYCAST,
4129                                                   NLM_F_MULTI);
4130                         if (err <= 0)
4131                                 break;
4132                 }
4133                 break;
4134         default:
4135                 break;
4136         }
4137         read_unlock_bh(&idev->lock);
4138         *p_ip_idx = ip_idx;
4139         return err;
4140 }
4141
4142 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4143                            enum addr_type_t type)
4144 {
4145         struct net *net = sock_net(skb->sk);
4146         int h, s_h;
4147         int idx, ip_idx;
4148         int s_idx, s_ip_idx;
4149         struct net_device *dev;
4150         struct inet6_dev *idev;
4151         struct hlist_head *head;
4152
4153         s_h = cb->args[0];
4154         s_idx = idx = cb->args[1];
4155         s_ip_idx = ip_idx = cb->args[2];
4156
4157         rcu_read_lock();
4158         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4159         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4160                 idx = 0;
4161                 head = &net->dev_index_head[h];
4162                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4163                         if (idx < s_idx)
4164                                 goto cont;
4165                         if (h > s_h || idx > s_idx)
4166                                 s_ip_idx = 0;
4167                         ip_idx = 0;
4168                         idev = __in6_dev_get(dev);
4169                         if (!idev)
4170                                 goto cont;
4171
4172                         if (in6_dump_addrs(idev, skb, cb, type,
4173                                            s_ip_idx, &ip_idx) <= 0)
4174                                 goto done;
4175 cont:
4176                         idx++;
4177                 }
4178         }
4179 done:
4180         rcu_read_unlock();
4181         cb->args[0] = h;
4182         cb->args[1] = idx;
4183         cb->args[2] = ip_idx;
4184
4185         return skb->len;
4186 }
4187
4188 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4189 {
4190         enum addr_type_t type = UNICAST_ADDR;
4191
4192         return inet6_dump_addr(skb, cb, type);
4193 }
4194
4195 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4196 {
4197         enum addr_type_t type = MULTICAST_ADDR;
4198
4199         return inet6_dump_addr(skb, cb, type);
4200 }
4201
4202
4203 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4204 {
4205         enum addr_type_t type = ANYCAST_ADDR;
4206
4207         return inet6_dump_addr(skb, cb, type);
4208 }
4209
4210 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4211 {
4212         struct net *net = sock_net(in_skb->sk);
4213         struct ifaddrmsg *ifm;
4214         struct nlattr *tb[IFA_MAX+1];
4215         struct in6_addr *addr = NULL, *peer;
4216         struct net_device *dev = NULL;
4217         struct inet6_ifaddr *ifa;
4218         struct sk_buff *skb;
4219         int err;
4220
4221         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4222         if (err < 0)
4223                 goto errout;
4224
4225         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4226         if (addr == NULL) {
4227                 err = -EINVAL;
4228                 goto errout;
4229         }
4230
4231         ifm = nlmsg_data(nlh);
4232         if (ifm->ifa_index)
4233                 dev = __dev_get_by_index(net, ifm->ifa_index);
4234
4235         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4236         if (!ifa) {
4237                 err = -EADDRNOTAVAIL;
4238                 goto errout;
4239         }
4240
4241         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4242         if (!skb) {
4243                 err = -ENOBUFS;
4244                 goto errout_ifa;
4245         }
4246
4247         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4248                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4249         if (err < 0) {
4250                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4251                 WARN_ON(err == -EMSGSIZE);
4252                 kfree_skb(skb);
4253                 goto errout_ifa;
4254         }
4255         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4256 errout_ifa:
4257         in6_ifa_put(ifa);
4258 errout:
4259         return err;
4260 }
4261
4262 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4263 {
4264         struct sk_buff *skb;
4265         struct net *net = dev_net(ifa->idev->dev);
4266         int err = -ENOBUFS;
4267
4268         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4269         if (skb == NULL)
4270                 goto errout;
4271
4272         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4273         if (err < 0) {
4274                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4275                 WARN_ON(err == -EMSGSIZE);
4276                 kfree_skb(skb);
4277                 goto errout;
4278         }
4279         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4280         return;
4281 errout:
4282         if (err < 0)
4283                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4284 }
4285
4286 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4287                                 __s32 *array, int bytes)
4288 {
4289         BUG_ON(bytes < (DEVCONF_MAX * 4));
4290
4291         memset(array, 0, bytes);
4292         array[DEVCONF_FORWARDING] = cnf->forwarding;
4293         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4294         array[DEVCONF_MTU6] = cnf->mtu6;
4295         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4296         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4297         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4298         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4299         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4300         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4301                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4302         array[DEVCONF_RTR_SOLICIT_DELAY] =
4303                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4304         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4305         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4306                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4307         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4308                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4309         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4310         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4311         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4312         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4313         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4314         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4315         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4316         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4317         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4318 #ifdef CONFIG_IPV6_ROUTER_PREF
4319         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4320         array[DEVCONF_RTR_PROBE_INTERVAL] =
4321                 jiffies_to_msecs(cnf->rtr_probe_interval);
4322 #ifdef CONFIG_IPV6_ROUTE_INFO
4323         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4324 #endif
4325 #endif
4326         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4327         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4328 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4329         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4330 #endif
4331 #ifdef CONFIG_IPV6_MROUTE
4332         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4333 #endif
4334         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4335         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4336         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4337         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4338         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4339 }
4340
4341 static inline size_t inet6_ifla6_size(void)
4342 {
4343         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4344              + nla_total_size(sizeof(struct ifla_cacheinfo))
4345              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4346              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4347              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4348              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4349 }
4350
4351 static inline size_t inet6_if_nlmsg_size(void)
4352 {
4353         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4354                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4355                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4356                + nla_total_size(4) /* IFLA_MTU */
4357                + nla_total_size(4) /* IFLA_LINK */
4358                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4359 }
4360
4361 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4362                                       int items, int bytes)
4363 {
4364         int i;
4365         int pad = bytes - sizeof(u64) * items;
4366         BUG_ON(pad < 0);
4367
4368         /* Use put_unaligned() because stats may not be aligned for u64. */
4369         put_unaligned(items, &stats[0]);
4370         for (i = 1; i < items; i++)
4371                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4372
4373         memset(&stats[items], 0, pad);
4374 }
4375
4376 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4377                                       int items, int bytes, size_t syncpoff)
4378 {
4379         int i;
4380         int pad = bytes - sizeof(u64) * items;
4381         BUG_ON(pad < 0);
4382
4383         /* Use put_unaligned() because stats may not be aligned for u64. */
4384         put_unaligned(items, &stats[0]);
4385         for (i = 1; i < items; i++)
4386                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4387
4388         memset(&stats[items], 0, pad);
4389 }
4390
4391 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4392                              int bytes)
4393 {
4394         switch (attrtype) {
4395         case IFLA_INET6_STATS:
4396                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4397                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4398                 break;
4399         case IFLA_INET6_ICMP6STATS:
4400                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4401                 break;
4402         }
4403 }
4404
4405 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4406 {
4407         struct nlattr *nla;
4408         struct ifla_cacheinfo ci;
4409
4410         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4411                 goto nla_put_failure;
4412         ci.max_reasm_len = IPV6_MAXPLEN;
4413         ci.tstamp = cstamp_delta(idev->tstamp);
4414         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4415         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4416         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4417                 goto nla_put_failure;
4418         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4419         if (nla == NULL)
4420                 goto nla_put_failure;
4421         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4422
4423         /* XXX - MC not implemented */
4424
4425         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4426         if (nla == NULL)
4427                 goto nla_put_failure;
4428         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4429
4430         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4431         if (nla == NULL)
4432                 goto nla_put_failure;
4433         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4434
4435         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4436         if (nla == NULL)
4437                 goto nla_put_failure;
4438         read_lock_bh(&idev->lock);
4439         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4440         read_unlock_bh(&idev->lock);
4441
4442         return 0;
4443
4444 nla_put_failure:
4445         return -EMSGSIZE;
4446 }
4447
4448 static size_t inet6_get_link_af_size(const struct net_device *dev)
4449 {
4450         if (!__in6_dev_get(dev))
4451                 return 0;
4452
4453         return inet6_ifla6_size();
4454 }
4455
4456 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4457 {
4458         struct inet6_dev *idev = __in6_dev_get(dev);
4459
4460         if (!idev)
4461                 return -ENODATA;
4462
4463         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4464                 return -EMSGSIZE;
4465
4466         return 0;
4467 }
4468
4469 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4470 {
4471         struct inet6_ifaddr *ifp;
4472         struct net_device *dev = idev->dev;
4473         bool clear_token, update_rs = false;
4474         struct in6_addr ll_addr;
4475
4476         ASSERT_RTNL();
4477
4478         if (token == NULL)
4479                 return -EINVAL;
4480         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4481                 return -EINVAL;
4482         if (!ipv6_accept_ra(idev))
4483                 return -EINVAL;
4484         if (idev->cnf.rtr_solicits <= 0)
4485                 return -EINVAL;
4486
4487         write_lock_bh(&idev->lock);
4488
4489         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4490         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4491
4492         write_unlock_bh(&idev->lock);
4493
4494         clear_token = ipv6_addr_any(token);
4495         if (clear_token)
4496                 goto update_lft;
4497
4498         if (!idev->dead && (idev->if_flags & IF_READY) &&
4499             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4500                              IFA_F_OPTIMISTIC)) {
4501                 /* If we're not ready, then normal ifup will take care
4502                  * of this. Otherwise, we need to request our rs here.
4503                  */
4504                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4505                 update_rs = true;
4506         }
4507
4508 update_lft:
4509         write_lock_bh(&idev->lock);
4510
4511         if (update_rs) {
4512                 idev->if_flags |= IF_RS_SENT;
4513                 idev->rs_probes = 1;
4514                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4515         }
4516
4517         /* Well, that's kinda nasty ... */
4518         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4519                 spin_lock(&ifp->lock);
4520                 if (ifp->tokenized) {
4521                         ifp->valid_lft = 0;
4522                         ifp->prefered_lft = 0;
4523                 }
4524                 spin_unlock(&ifp->lock);
4525         }
4526
4527         write_unlock_bh(&idev->lock);
4528         addrconf_verify_rtnl();
4529         return 0;
4530 }
4531
4532 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4533 {
4534         int err = -EINVAL;
4535         struct inet6_dev *idev = __in6_dev_get(dev);
4536         struct nlattr *tb[IFLA_INET6_MAX + 1];
4537
4538         if (!idev)
4539                 return -EAFNOSUPPORT;
4540
4541         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4542                 BUG();
4543
4544         if (tb[IFLA_INET6_TOKEN])
4545                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4546
4547         return err;
4548 }
4549
4550 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4551                              u32 portid, u32 seq, int event, unsigned int flags)
4552 {
4553         struct net_device *dev = idev->dev;
4554         struct ifinfomsg *hdr;
4555         struct nlmsghdr *nlh;
4556         void *protoinfo;
4557
4558         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4559         if (nlh == NULL)
4560                 return -EMSGSIZE;
4561
4562         hdr = nlmsg_data(nlh);
4563         hdr->ifi_family = AF_INET6;
4564         hdr->__ifi_pad = 0;
4565         hdr->ifi_type = dev->type;
4566         hdr->ifi_index = dev->ifindex;
4567         hdr->ifi_flags = dev_get_flags(dev);
4568         hdr->ifi_change = 0;
4569
4570         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4571             (dev->addr_len &&
4572              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4573             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4574             (dev->ifindex != dev->iflink &&
4575              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4576                 goto nla_put_failure;
4577         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4578         if (protoinfo == NULL)
4579                 goto nla_put_failure;
4580
4581         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4582                 goto nla_put_failure;
4583
4584         nla_nest_end(skb, protoinfo);
4585         return nlmsg_end(skb, nlh);
4586
4587 nla_put_failure:
4588         nlmsg_cancel(skb, nlh);
4589         return -EMSGSIZE;
4590 }
4591
4592 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4593 {
4594         struct net *net = sock_net(skb->sk);
4595         int h, s_h;
4596         int idx = 0, s_idx;
4597         struct net_device *dev;
4598         struct inet6_dev *idev;
4599         struct hlist_head *head;
4600
4601         s_h = cb->args[0];
4602         s_idx = cb->args[1];
4603
4604         rcu_read_lock();
4605         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4606                 idx = 0;
4607                 head = &net->dev_index_head[h];
4608                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4609                         if (idx < s_idx)
4610                                 goto cont;
4611                         idev = __in6_dev_get(dev);
4612                         if (!idev)
4613                                 goto cont;
4614                         if (inet6_fill_ifinfo(skb, idev,
4615                                               NETLINK_CB(cb->skb).portid,
4616                                               cb->nlh->nlmsg_seq,
4617                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4618                                 goto out;
4619 cont:
4620                         idx++;
4621                 }
4622         }
4623 out:
4624         rcu_read_unlock();
4625         cb->args[1] = idx;
4626         cb->args[0] = h;
4627
4628         return skb->len;
4629 }
4630
4631 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4632 {
4633         struct sk_buff *skb;
4634         struct net *net = dev_net(idev->dev);
4635         int err = -ENOBUFS;
4636
4637         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4638         if (skb == NULL)
4639                 goto errout;
4640
4641         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4642         if (err < 0) {
4643                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4644                 WARN_ON(err == -EMSGSIZE);
4645                 kfree_skb(skb);
4646                 goto errout;
4647         }
4648         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4649         return;
4650 errout:
4651         if (err < 0)
4652                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4653 }
4654
4655 static inline size_t inet6_prefix_nlmsg_size(void)
4656 {
4657         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4658                + nla_total_size(sizeof(struct in6_addr))
4659                + nla_total_size(sizeof(struct prefix_cacheinfo));
4660 }
4661
4662 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4663                              struct prefix_info *pinfo, u32 portid, u32 seq,
4664                              int event, unsigned int flags)
4665 {
4666         struct prefixmsg *pmsg;
4667         struct nlmsghdr *nlh;
4668         struct prefix_cacheinfo ci;
4669
4670         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4671         if (nlh == NULL)
4672                 return -EMSGSIZE;
4673
4674         pmsg = nlmsg_data(nlh);
4675         pmsg->prefix_family = AF_INET6;
4676         pmsg->prefix_pad1 = 0;
4677         pmsg->prefix_pad2 = 0;
4678         pmsg->prefix_ifindex = idev->dev->ifindex;
4679         pmsg->prefix_len = pinfo->prefix_len;
4680         pmsg->prefix_type = pinfo->type;
4681         pmsg->prefix_pad3 = 0;
4682         pmsg->prefix_flags = 0;
4683         if (pinfo->onlink)
4684                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4685         if (pinfo->autoconf)
4686                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4687
4688         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4689                 goto nla_put_failure;
4690         ci.preferred_time = ntohl(pinfo->prefered);
4691         ci.valid_time = ntohl(pinfo->valid);
4692         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4693                 goto nla_put_failure;
4694         return nlmsg_end(skb, nlh);
4695
4696 nla_put_failure:
4697         nlmsg_cancel(skb, nlh);
4698         return -EMSGSIZE;
4699 }
4700
4701 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4702                          struct prefix_info *pinfo)
4703 {
4704         struct sk_buff *skb;
4705         struct net *net = dev_net(idev->dev);
4706         int err = -ENOBUFS;
4707
4708         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4709         if (skb == NULL)
4710                 goto errout;
4711
4712         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4713         if (err < 0) {
4714                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4715                 WARN_ON(err == -EMSGSIZE);
4716                 kfree_skb(skb);
4717                 goto errout;
4718         }
4719         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4720         return;
4721 errout:
4722         if (err < 0)
4723                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4724 }
4725
4726 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4727 {
4728         struct net *net = dev_net(ifp->idev->dev);
4729
4730         if (event)
4731                 ASSERT_RTNL();
4732
4733         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4734
4735         switch (event) {
4736         case RTM_NEWADDR:
4737                 /*
4738                  * If the address was optimistic
4739                  * we inserted the route at the start of
4740                  * our DAD process, so we don't need
4741                  * to do it again
4742                  */
4743                 if (!rcu_access_pointer(ifp->rt->rt6i_node))
4744                         ip6_ins_rt(ifp->rt);
4745                 if (ifp->idev->cnf.forwarding)
4746                         addrconf_join_anycast(ifp);
4747                 if (!ipv6_addr_any(&ifp->peer_addr))
4748                         addrconf_prefix_route(&ifp->peer_addr, 128,
4749                                               ifp->idev->dev, 0, 0);
4750                 break;
4751         case RTM_DELADDR:
4752                 if (ifp->idev->cnf.forwarding)
4753                         addrconf_leave_anycast(ifp);
4754                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4755                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4756                         struct rt6_info *rt;
4757                         struct net_device *dev = ifp->idev->dev;
4758
4759                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4760                                         dev->ifindex, 1);
4761                         if (rt) {
4762                                 dst_hold(&rt->dst);
4763                                 if (ip6_del_rt(rt))
4764                                         dst_free(&rt->dst);
4765                         }
4766                 }
4767                 dst_hold(&ifp->rt->dst);
4768
4769                 if (ip6_del_rt(ifp->rt))
4770                         dst_free(&ifp->rt->dst);
4771
4772                 rt_genid_bump_ipv6(net);
4773                 break;
4774         }
4775         atomic_inc(&net->ipv6.dev_addr_genid);
4776 }
4777
4778 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4779 {
4780         rcu_read_lock_bh();
4781         if (likely(ifp->idev->dead == 0))
4782                 __ipv6_ifa_notify(event, ifp);
4783         rcu_read_unlock_bh();
4784 }
4785
4786 #ifdef CONFIG_SYSCTL
4787
4788 static
4789 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4790                            void __user *buffer, size_t *lenp, loff_t *ppos)
4791 {
4792         int *valp = ctl->data;
4793         int val = *valp;
4794         loff_t pos = *ppos;
4795         struct ctl_table lctl;
4796         int ret;
4797
4798         /*
4799          * ctl->data points to idev->cnf.forwarding, we should
4800          * not modify it until we get the rtnl lock.
4801          */
4802         lctl = *ctl;
4803         lctl.data = &val;
4804
4805         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4806
4807         if (write)
4808                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4809         if (ret)
4810                 *ppos = pos;
4811         return ret;
4812 }
4813
4814 static
4815 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
4816                         void __user *buffer, size_t *lenp, loff_t *ppos)
4817 {
4818         struct inet6_dev *idev = ctl->extra1;
4819         int min_mtu = IPV6_MIN_MTU;
4820         struct ctl_table lctl;
4821
4822         lctl = *ctl;
4823         lctl.extra1 = &min_mtu;
4824         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
4825
4826         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
4827 }
4828
4829 static void dev_disable_change(struct inet6_dev *idev)
4830 {
4831         struct netdev_notifier_info info;
4832
4833         if (!idev || !idev->dev)
4834                 return;
4835
4836         netdev_notifier_info_init(&info, idev->dev);
4837         if (idev->cnf.disable_ipv6)
4838                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4839         else
4840                 addrconf_notify(NULL, NETDEV_UP, &info);
4841 }
4842
4843 static void addrconf_disable_change(struct net *net, __s32 newf)
4844 {
4845         struct net_device *dev;
4846         struct inet6_dev *idev;
4847
4848         for_each_netdev(net, dev) {
4849                 idev = __in6_dev_get(dev);
4850                 if (idev) {
4851                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4852                         idev->cnf.disable_ipv6 = newf;
4853                         if (changed)
4854                                 dev_disable_change(idev);
4855                 }
4856         }
4857 }
4858
4859 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4860 {
4861         struct net *net;
4862         int old;
4863
4864         if (!rtnl_trylock())
4865                 return restart_syscall();
4866
4867         net = (struct net *)table->extra2;
4868         old = *p;
4869         *p = newf;
4870
4871         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4872                 rtnl_unlock();
4873                 return 0;
4874         }
4875
4876         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4877                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4878                 addrconf_disable_change(net, newf);
4879         } else if ((!newf) ^ (!old))
4880                 dev_disable_change((struct inet6_dev *)table->extra1);
4881
4882         rtnl_unlock();
4883         return 0;
4884 }
4885
4886 static
4887 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4888                             void __user *buffer, size_t *lenp, loff_t *ppos)
4889 {
4890         int *valp = ctl->data;
4891         int val = *valp;
4892         loff_t pos = *ppos;
4893         struct ctl_table lctl;
4894         int ret;
4895
4896         /*
4897          * ctl->data points to idev->cnf.disable_ipv6, we should
4898          * not modify it until we get the rtnl lock.
4899          */
4900         lctl = *ctl;
4901         lctl.data = &val;
4902
4903         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4904
4905         if (write)
4906                 ret = addrconf_disable_ipv6(ctl, valp, val);
4907         if (ret)
4908                 *ppos = pos;
4909         return ret;
4910 }
4911
4912 static
4913 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4914                               void __user *buffer, size_t *lenp, loff_t *ppos)
4915 {
4916         int *valp = ctl->data;
4917         int ret;
4918         int old, new;
4919
4920         old = *valp;
4921         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4922         new = *valp;
4923
4924         if (write && old != new) {
4925                 struct net *net = ctl->extra2;
4926
4927                 if (!rtnl_trylock())
4928                         return restart_syscall();
4929
4930                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4931                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4932                                                      NETCONFA_IFINDEX_DEFAULT,
4933                                                      net->ipv6.devconf_dflt);
4934                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4935                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4936                                                      NETCONFA_IFINDEX_ALL,
4937                                                      net->ipv6.devconf_all);
4938                 else {
4939                         struct inet6_dev *idev = ctl->extra1;
4940
4941                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4942                                                      idev->dev->ifindex,
4943                                                      &idev->cnf);
4944                 }
4945                 rtnl_unlock();
4946         }
4947
4948         return ret;
4949 }
4950
4951
4952 static struct addrconf_sysctl_table
4953 {
4954         struct ctl_table_header *sysctl_header;
4955         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4956 } addrconf_sysctl __read_mostly = {
4957         .sysctl_header = NULL,
4958         .addrconf_vars = {
4959                 {
4960                         .procname       = "forwarding",
4961                         .data           = &ipv6_devconf.forwarding,
4962                         .maxlen         = sizeof(int),
4963                         .mode           = 0644,
4964                         .proc_handler   = addrconf_sysctl_forward,
4965                 },
4966                 {
4967                         .procname       = "hop_limit",
4968                         .data           = &ipv6_devconf.hop_limit,
4969                         .maxlen         = sizeof(int),
4970                         .mode           = 0644,
4971                         .proc_handler   = proc_dointvec,
4972                 },
4973                 {
4974                         .procname       = "mtu",
4975                         .data           = &ipv6_devconf.mtu6,
4976                         .maxlen         = sizeof(int),
4977                         .mode           = 0644,
4978                         .proc_handler   = addrconf_sysctl_mtu,
4979                 },
4980                 {
4981                         .procname       = "accept_ra",
4982                         .data           = &ipv6_devconf.accept_ra,
4983                         .maxlen         = sizeof(int),
4984                         .mode           = 0644,
4985                         .proc_handler   = proc_dointvec,
4986                 },
4987                 {
4988                         .procname       = "accept_redirects",
4989                         .data           = &ipv6_devconf.accept_redirects,
4990                         .maxlen         = sizeof(int),
4991                         .mode           = 0644,
4992                         .proc_handler   = proc_dointvec,
4993                 },
4994                 {
4995                         .procname       = "autoconf",
4996                         .data           = &ipv6_devconf.autoconf,
4997                         .maxlen         = sizeof(int),
4998                         .mode           = 0644,
4999                         .proc_handler   = proc_dointvec,
5000                 },
5001                 {
5002                         .procname       = "dad_transmits",
5003                         .data           = &ipv6_devconf.dad_transmits,
5004                         .maxlen         = sizeof(int),
5005                         .mode           = 0644,
5006                         .proc_handler   = proc_dointvec,
5007                 },
5008                 {
5009                         .procname       = "router_solicitations",
5010                         .data           = &ipv6_devconf.rtr_solicits,
5011                         .maxlen         = sizeof(int),
5012                         .mode           = 0644,
5013                         .proc_handler   = proc_dointvec,
5014                 },
5015                 {
5016                         .procname       = "router_solicitation_interval",
5017                         .data           = &ipv6_devconf.rtr_solicit_interval,
5018                         .maxlen         = sizeof(int),
5019                         .mode           = 0644,
5020                         .proc_handler   = proc_dointvec_jiffies,
5021                 },
5022                 {
5023                         .procname       = "router_solicitation_delay",
5024                         .data           = &ipv6_devconf.rtr_solicit_delay,
5025                         .maxlen         = sizeof(int),
5026                         .mode           = 0644,
5027                         .proc_handler   = proc_dointvec_jiffies,
5028                 },
5029                 {
5030                         .procname       = "force_mld_version",
5031                         .data           = &ipv6_devconf.force_mld_version,
5032                         .maxlen         = sizeof(int),
5033                         .mode           = 0644,
5034                         .proc_handler   = proc_dointvec,
5035                 },
5036                 {
5037                         .procname       = "mldv1_unsolicited_report_interval",
5038                         .data           =
5039                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5040                         .maxlen         = sizeof(int),
5041                         .mode           = 0644,
5042                         .proc_handler   = proc_dointvec_ms_jiffies,
5043                 },
5044                 {
5045                         .procname       = "mldv2_unsolicited_report_interval",
5046                         .data           =
5047                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5048                         .maxlen         = sizeof(int),
5049                         .mode           = 0644,
5050                         .proc_handler   = proc_dointvec_ms_jiffies,
5051                 },
5052                 {
5053                         .procname       = "use_tempaddr",
5054                         .data           = &ipv6_devconf.use_tempaddr,
5055                         .maxlen         = sizeof(int),
5056                         .mode           = 0644,
5057                         .proc_handler   = proc_dointvec,
5058                 },
5059                 {
5060                         .procname       = "temp_valid_lft",
5061                         .data           = &ipv6_devconf.temp_valid_lft,
5062                         .maxlen         = sizeof(int),
5063                         .mode           = 0644,
5064                         .proc_handler   = proc_dointvec,
5065                 },
5066                 {
5067                         .procname       = "temp_prefered_lft",
5068                         .data           = &ipv6_devconf.temp_prefered_lft,
5069                         .maxlen         = sizeof(int),
5070                         .mode           = 0644,
5071                         .proc_handler   = proc_dointvec,
5072                 },
5073                 {
5074                         .procname       = "regen_max_retry",
5075                         .data           = &ipv6_devconf.regen_max_retry,
5076                         .maxlen         = sizeof(int),
5077                         .mode           = 0644,
5078                         .proc_handler   = proc_dointvec,
5079                 },
5080                 {
5081                         .procname       = "max_desync_factor",
5082                         .data           = &ipv6_devconf.max_desync_factor,
5083                         .maxlen         = sizeof(int),
5084                         .mode           = 0644,
5085                         .proc_handler   = proc_dointvec,
5086                 },
5087                 {
5088                         .procname       = "max_addresses",
5089                         .data           = &ipv6_devconf.max_addresses,
5090                         .maxlen         = sizeof(int),
5091                         .mode           = 0644,
5092                         .proc_handler   = proc_dointvec,
5093                 },
5094                 {
5095                         .procname       = "accept_ra_defrtr",
5096                         .data           = &ipv6_devconf.accept_ra_defrtr,
5097                         .maxlen         = sizeof(int),
5098                         .mode           = 0644,
5099                         .proc_handler   = proc_dointvec,
5100                 },
5101                 {
5102                         .procname       = "accept_ra_min_hop_limit",
5103                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5104                         .maxlen         = sizeof(int),
5105                         .mode           = 0644,
5106                         .proc_handler   = proc_dointvec,
5107                 },
5108                 {
5109                         .procname       = "accept_ra_pinfo",
5110                         .data           = &ipv6_devconf.accept_ra_pinfo,
5111                         .maxlen         = sizeof(int),
5112                         .mode           = 0644,
5113                         .proc_handler   = proc_dointvec,
5114                 },
5115 #ifdef CONFIG_IPV6_ROUTER_PREF
5116                 {
5117                         .procname       = "accept_ra_rtr_pref",
5118                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5119                         .maxlen         = sizeof(int),
5120                         .mode           = 0644,
5121                         .proc_handler   = proc_dointvec,
5122                 },
5123                 {
5124                         .procname       = "router_probe_interval",
5125                         .data           = &ipv6_devconf.rtr_probe_interval,
5126                         .maxlen         = sizeof(int),
5127                         .mode           = 0644,
5128                         .proc_handler   = proc_dointvec_jiffies,
5129                 },
5130 #ifdef CONFIG_IPV6_ROUTE_INFO
5131                 {
5132                         .procname       = "accept_ra_rt_info_max_plen",
5133                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5134                         .maxlen         = sizeof(int),
5135                         .mode           = 0644,
5136                         .proc_handler   = proc_dointvec,
5137                 },
5138 #endif
5139 #endif
5140                 {
5141                         .procname       = "proxy_ndp",
5142                         .data           = &ipv6_devconf.proxy_ndp,
5143                         .maxlen         = sizeof(int),
5144                         .mode           = 0644,
5145                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5146                 },
5147                 {
5148                         .procname       = "accept_source_route",
5149                         .data           = &ipv6_devconf.accept_source_route,
5150                         .maxlen         = sizeof(int),
5151                         .mode           = 0644,
5152                         .proc_handler   = proc_dointvec,
5153                 },
5154 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5155                 {
5156                         .procname       = "optimistic_dad",
5157                         .data           = &ipv6_devconf.optimistic_dad,
5158                         .maxlen         = sizeof(int),
5159                         .mode           = 0644,
5160                         .proc_handler   = proc_dointvec,
5161
5162                 },
5163 #endif
5164 #ifdef CONFIG_IPV6_MROUTE
5165                 {
5166                         .procname       = "mc_forwarding",
5167                         .data           = &ipv6_devconf.mc_forwarding,
5168                         .maxlen         = sizeof(int),
5169                         .mode           = 0444,
5170                         .proc_handler   = proc_dointvec,
5171                 },
5172 #endif
5173                 {
5174                         .procname       = "disable_ipv6",
5175                         .data           = &ipv6_devconf.disable_ipv6,
5176                         .maxlen         = sizeof(int),
5177                         .mode           = 0644,
5178                         .proc_handler   = addrconf_sysctl_disable,
5179                 },
5180                 {
5181                         .procname       = "accept_dad",
5182                         .data           = &ipv6_devconf.accept_dad,
5183                         .maxlen         = sizeof(int),
5184                         .mode           = 0644,
5185                         .proc_handler   = proc_dointvec,
5186                 },
5187                 {
5188                         .procname       = "force_tllao",
5189                         .data           = &ipv6_devconf.force_tllao,
5190                         .maxlen         = sizeof(int),
5191                         .mode           = 0644,
5192                         .proc_handler   = proc_dointvec
5193                 },
5194                 {
5195                         .procname       = "ndisc_notify",
5196                         .data           = &ipv6_devconf.ndisc_notify,
5197                         .maxlen         = sizeof(int),
5198                         .mode           = 0644,
5199                         .proc_handler   = proc_dointvec
5200                 },
5201                 {
5202                         .procname       = "suppress_frag_ndisc",
5203                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5204                         .maxlen         = sizeof(int),
5205                         .mode           = 0644,
5206                         .proc_handler   = proc_dointvec
5207                 },
5208                 {
5209                         /* sentinel */
5210                 }
5211         },
5212 };
5213
5214 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5215                 struct inet6_dev *idev, struct ipv6_devconf *p)
5216 {
5217         int i;
5218         struct addrconf_sysctl_table *t;
5219         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5220
5221         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5222         if (t == NULL)
5223                 goto out;
5224
5225         for (i = 0; t->addrconf_vars[i].data; i++) {
5226                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5227                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5228                 t->addrconf_vars[i].extra2 = net;
5229         }
5230
5231         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5232
5233         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5234         if (t->sysctl_header == NULL)
5235                 goto free;
5236
5237         p->sysctl = t;
5238         return 0;
5239
5240 free:
5241         kfree(t);
5242 out:
5243         return -ENOBUFS;
5244 }
5245
5246 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5247 {
5248         struct addrconf_sysctl_table *t;
5249
5250         if (p->sysctl == NULL)
5251                 return;
5252
5253         t = p->sysctl;
5254         p->sysctl = NULL;
5255         unregister_net_sysctl_table(t->sysctl_header);
5256         kfree(t);
5257 }
5258
5259 static void addrconf_sysctl_register(struct inet6_dev *idev)
5260 {
5261         neigh_sysctl_register(idev->dev, idev->nd_parms,
5262                               &ndisc_ifinfo_sysctl_change);
5263         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5264                                         idev, &idev->cnf);
5265 }
5266
5267 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5268 {
5269         __addrconf_sysctl_unregister(&idev->cnf);
5270         neigh_sysctl_unregister(idev->nd_parms);
5271 }
5272
5273
5274 #endif
5275
5276 static int __net_init addrconf_init_net(struct net *net)
5277 {
5278         int err = -ENOMEM;
5279         struct ipv6_devconf *all, *dflt;
5280
5281         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5282         if (all == NULL)
5283                 goto err_alloc_all;
5284
5285         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5286         if (dflt == NULL)
5287                 goto err_alloc_dflt;
5288
5289         /* these will be inherited by all namespaces */
5290         dflt->autoconf = ipv6_defaults.autoconf;
5291         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5292
5293         net->ipv6.devconf_all = all;
5294         net->ipv6.devconf_dflt = dflt;
5295
5296 #ifdef CONFIG_SYSCTL
5297         err = __addrconf_sysctl_register(net, "all", NULL, all);
5298         if (err < 0)
5299                 goto err_reg_all;
5300
5301         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5302         if (err < 0)
5303                 goto err_reg_dflt;
5304 #endif
5305         return 0;
5306
5307 #ifdef CONFIG_SYSCTL
5308 err_reg_dflt:
5309         __addrconf_sysctl_unregister(all);
5310 err_reg_all:
5311         kfree(dflt);
5312 #endif
5313 err_alloc_dflt:
5314         kfree(all);
5315 err_alloc_all:
5316         return err;
5317 }
5318
5319 static void __net_exit addrconf_exit_net(struct net *net)
5320 {
5321 #ifdef CONFIG_SYSCTL
5322         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5323         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5324 #endif
5325         if (!net_eq(net, &init_net)) {
5326                 kfree(net->ipv6.devconf_dflt);
5327                 kfree(net->ipv6.devconf_all);
5328         }
5329 }
5330
5331 static struct pernet_operations addrconf_ops = {
5332         .init = addrconf_init_net,
5333         .exit = addrconf_exit_net,
5334 };
5335
5336 static struct rtnl_af_ops inet6_ops = {
5337         .family           = AF_INET6,
5338         .fill_link_af     = inet6_fill_link_af,
5339         .get_link_af_size = inet6_get_link_af_size,
5340         .set_link_af      = inet6_set_link_af,
5341 };
5342
5343 /*
5344  *      Init / cleanup code
5345  */
5346
5347 int __init addrconf_init(void)
5348 {
5349         int i, err;
5350
5351         err = ipv6_addr_label_init();
5352         if (err < 0) {
5353                 pr_crit("%s: cannot initialize default policy table: %d\n",
5354                         __func__, err);
5355                 goto out;
5356         }
5357
5358         err = register_pernet_subsys(&addrconf_ops);
5359         if (err < 0)
5360                 goto out_addrlabel;
5361
5362         addrconf_wq = create_workqueue("ipv6_addrconf");
5363         if (!addrconf_wq) {
5364                 err = -ENOMEM;
5365                 goto out_nowq;
5366         }
5367
5368         /* The addrconf netdev notifier requires that loopback_dev
5369          * has it's ipv6 private information allocated and setup
5370          * before it can bring up and give link-local addresses
5371          * to other devices which are up.
5372          *
5373          * Unfortunately, loopback_dev is not necessarily the first
5374          * entry in the global dev_base list of net devices.  In fact,
5375          * it is likely to be the very last entry on that list.
5376          * So this causes the notifier registry below to try and
5377          * give link-local addresses to all devices besides loopback_dev
5378          * first, then loopback_dev, which cases all the non-loopback_dev
5379          * devices to fail to get a link-local address.
5380          *
5381          * So, as a temporary fix, allocate the ipv6 structure for
5382          * loopback_dev first by hand.
5383          * Longer term, all of the dependencies ipv6 has upon the loopback
5384          * device and it being up should be removed.
5385          */
5386         rtnl_lock();
5387         if (!ipv6_add_dev(init_net.loopback_dev))
5388                 err = -ENOMEM;
5389         rtnl_unlock();
5390         if (err)
5391                 goto errlo;
5392
5393         ip6_route_init_special_entries();
5394
5395         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5396                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5397
5398         register_netdevice_notifier(&ipv6_dev_notf);
5399
5400         addrconf_verify();
5401
5402         rtnl_af_register(&inet6_ops);
5403
5404         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5405                               NULL);
5406         if (err < 0)
5407                 goto errout;
5408
5409         /* Only the first call to __rtnl_register can fail */
5410         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5411         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5412         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5413                         inet6_dump_ifaddr, NULL);
5414         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5415                         inet6_dump_ifmcaddr, NULL);
5416         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5417                         inet6_dump_ifacaddr, NULL);
5418         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5419                         inet6_netconf_dump_devconf, NULL);
5420
5421         ipv6_addr_label_rtnl_register();
5422
5423         return 0;
5424 errout:
5425         rtnl_af_unregister(&inet6_ops);
5426         unregister_netdevice_notifier(&ipv6_dev_notf);
5427 errlo:
5428         destroy_workqueue(addrconf_wq);
5429 out_nowq:
5430         unregister_pernet_subsys(&addrconf_ops);
5431 out_addrlabel:
5432         ipv6_addr_label_cleanup();
5433 out:
5434         return err;
5435 }
5436
5437 void addrconf_cleanup(void)
5438 {
5439         struct net_device *dev;
5440         int i;
5441
5442         unregister_netdevice_notifier(&ipv6_dev_notf);
5443         unregister_pernet_subsys(&addrconf_ops);
5444         ipv6_addr_label_cleanup();
5445
5446         rtnl_lock();
5447
5448         __rtnl_af_unregister(&inet6_ops);
5449
5450         /* clean dev list */
5451         for_each_netdev(&init_net, dev) {
5452                 if (__in6_dev_get(dev) == NULL)
5453                         continue;
5454                 addrconf_ifdown(dev, 1);
5455         }
5456         addrconf_ifdown(init_net.loopback_dev, 2);
5457
5458         /*
5459          *      Check hash table.
5460          */
5461         spin_lock_bh(&addrconf_hash_lock);
5462         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5463                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5464         spin_unlock_bh(&addrconf_hash_lock);
5465         cancel_delayed_work(&addr_chk_work);
5466         rtnl_unlock();
5467
5468         destroy_workqueue(addrconf_wq);
5469 }