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