Pileus Git - ~andy/linux/blob - net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86 #include <linux/random.h>
  87 #include <linux/uaccess.h>
  88 #include <asm/unaligned.h>
  89
  90 #include <linux/proc_fs.h>
  91 #include <linux/seq_file.h>
  92 #include <linux/export.h>
  93
  94 /* Set to 3 to get tracing... */
  95 #define ACONF_DEBUG 2
  96
  97 #if ACONF_DEBUG >= 3
  98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
  99 #else
 100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 101 #endif
 102
 103 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 104
 105 static inline u32 cstamp_delta(unsigned long cstamp)
 106 {
 107         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 108 }
 109
 110 #ifdef CONFIG_SYSCTL
 111 static void addrconf_sysctl_register(struct inet6_dev *idev);
 112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 113 #else
 114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 115 {
 116 }
 117
 118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 119 {
 120 }
 121 #endif
 122
 123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 125 static void ipv6_regen_rndid(unsigned long data);
 126
 127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 128 static int ipv6_count_addresses(struct inet6_dev *idev);
 129
 130 /*
 131  *      Configured unicast address hash table
 132  */
 133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 134 static DEFINE_SPINLOCK(addrconf_hash_lock);
 135
 136 static void addrconf_verify(unsigned long);
 137
 138 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 139 static DEFINE_SPINLOCK(addrconf_verify_lock);
 140
 141 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 142 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 143
 144 static void addrconf_type_change(struct net_device *dev,
 145                                  unsigned long event);
 146 static int addrconf_ifdown(struct net_device *dev, int how);
 147
 148 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 149                                                   int plen,
 150                                                   const struct net_device *dev,
 151                                                   u32 flags, u32 noflags);
 152
 153 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_timer(unsigned long data);
 155 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 156 static void addrconf_dad_run(struct inet6_dev *idev);
 157 static void addrconf_rs_timer(unsigned long data);
 158 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 160
 161 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 162                                 struct prefix_info *pinfo);
 163 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 164                                struct net_device *dev);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .mldv1_unsolicited_report_interval = 10 * HZ,
 175         .mldv2_unsolicited_report_interval = HZ,
 176         .dad_transmits          = 1,
 177         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 178         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 179         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199         .suppress_frag_ndisc    = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .force_mld_version      = 0,
 210         .mldv1_unsolicited_report_interval = 10 * HZ,
 211         .mldv2_unsolicited_report_interval = HZ,
 212         .dad_transmits          = 1,
 213         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 214         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 215         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 216         .use_tempaddr           = 0,
 217         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 218         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 219         .regen_max_retry        = REGEN_MAX_RETRY,
 220         .max_desync_factor      = MAX_DESYNC_FACTOR,
 221         .max_addresses          = IPV6_MAX_ADDRESSES,
 222         .accept_ra_defrtr       = 1,
 223         .accept_ra_pinfo        = 1,
 224 #ifdef CONFIG_IPV6_ROUTER_PREF
 225         .accept_ra_rtr_pref     = 1,
 226         .rtr_probe_interval     = 60 * HZ,
 227 #ifdef CONFIG_IPV6_ROUTE_INFO
 228         .accept_ra_rt_info_max_plen = 0,
 229 #endif
 230 #endif
 231         .proxy_ndp              = 0,
 232         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 233         .disable_ipv6           = 0,
 234         .accept_dad             = 1,
 235         .suppress_frag_ndisc    = 1,
 236 };
 237
 238 /* Check if a valid qdisc is available */
 239 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 240 {
 241         return !qdisc_tx_is_noop(dev);
 242 }
 243
 244 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 245 {
 246         if (del_timer(&idev->rs_timer))
 247                 __in6_dev_put(idev);
 248 }
 249
 250 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 251 {
 252         if (del_timer(&ifp->dad_timer))
 253                 __in6_ifa_put(ifp);
 254 }
 255
 256 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 257                                   unsigned long when)
 258 {
 259         if (!timer_pending(&idev->rs_timer))
 260                 in6_dev_hold(idev);
 261         mod_timer(&idev->rs_timer, jiffies + when);
 262 }
 263
 264 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 265                                    unsigned long when)
 266 {
 267         if (!timer_pending(&ifp->dad_timer))
 268                 in6_ifa_hold(ifp);
 269         mod_timer(&ifp->dad_timer, jiffies + when);
 270 }
 271
 272 static int snmp6_alloc_dev(struct inet6_dev *idev)
 273 {
 274         int i;
 275
 276         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 277                           sizeof(struct ipstats_mib),
 278                           __alignof__(struct ipstats_mib)) < 0)
 279                 goto err_ip;
 280
 281         for_each_possible_cpu(i) {
 282                 struct ipstats_mib *addrconf_stats;
 283                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i);
 284                 u64_stats_init(&addrconf_stats->syncp);
 285 #if SNMP_ARRAY_SZ == 2
 286                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i);
 287                 u64_stats_init(&addrconf_stats->syncp);
 288 #endif
 289         }
 290
 291
 292         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 293                                         GFP_KERNEL);
 294         if (!idev->stats.icmpv6dev)
 295                 goto err_icmp;
 296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 297                                            GFP_KERNEL);
 298         if (!idev->stats.icmpv6msgdev)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         kfree(idev->stats.icmpv6dev);
 305 err_icmp:
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 312 {
 313         struct inet6_dev *ndev;
 314
 315         ASSERT_RTNL();
 316
 317         if (dev->mtu < IPV6_MIN_MTU)
 318                 return NULL;
 319
 320         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 321
 322         if (ndev == NULL)
 323                 return NULL;
 324
 325         rwlock_init(&ndev->lock);
 326         ndev->dev = dev;
 327         INIT_LIST_HEAD(&ndev->addr_list);
 328         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 329                     (unsigned long)ndev);
 330         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 331         ndev->cnf.mtu6 = dev->mtu;
 332         ndev->cnf.sysctl = NULL;
 333         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 334         if (ndev->nd_parms == NULL) {
 335                 kfree(ndev);
 336                 return NULL;
 337         }
 338         if (ndev->cnf.forwarding)
 339                 dev_disable_lro(dev);
 340         /* We refer to the device */
 341         dev_hold(dev);
 342
 343         if (snmp6_alloc_dev(ndev) < 0) {
 344                 ADBG(KERN_WARNING
 345                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 346                         __func__, dev->name);
 347                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 348                 dev_put(dev);
 349                 kfree(ndev);
 350                 return NULL;
 351         }
 352
 353         if (snmp6_register_dev(ndev) < 0) {
 354                 ADBG(KERN_WARNING
 355                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 356                         __func__, dev->name);
 357                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 358                 ndev->dead = 1;
 359                 in6_dev_finish_destroy(ndev);
 360                 return NULL;
 361         }
 362
 363         /* One reference from device.  We must do this before
 364          * we invoke __ipv6_regen_rndid().
 365          */
 366         in6_dev_hold(ndev);
 367
 368         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 369                 ndev->cnf.accept_dad = -1;
 370
 371 #if IS_ENABLED(CONFIG_IPV6_SIT)
 372         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 373                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 374                 ndev->cnf.rtr_solicits = 0;
 375         }
 376 #endif
 377
 378         INIT_LIST_HEAD(&ndev->tempaddr_list);
 379         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 380         if ((dev->flags&IFF_LOOPBACK) ||
 381             dev->type == ARPHRD_TUNNEL ||
 382             dev->type == ARPHRD_TUNNEL6 ||
 383             dev->type == ARPHRD_SIT ||
 384             dev->type == ARPHRD_NONE) {
 385                 ndev->cnf.use_tempaddr = -1;
 386         } else {
 387                 in6_dev_hold(ndev);
 388                 ipv6_regen_rndid((unsigned long) ndev);
 389         }
 390
 391         ndev->token = in6addr_any;
 392
 393         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 394                 ndev->if_flags |= IF_READY;
 395
 396         ipv6_mc_init_dev(ndev);
 397         ndev->tstamp = jiffies;
 398         addrconf_sysctl_register(ndev);
 399         /* protected by rtnl_lock */
 400         rcu_assign_pointer(dev->ip6_ptr, ndev);
 401
 402         /* Join interface-local all-node multicast group */
 403         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 404
 405         /* Join all-node multicast group */
 406         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 407
 408         /* Join all-router multicast group if forwarding is set */
 409         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 410                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 411
 412         return ndev;
 413 }
 414
 415 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 416 {
 417         struct inet6_dev *idev;
 418
 419         ASSERT_RTNL();
 420
 421         idev = __in6_dev_get(dev);
 422         if (!idev) {
 423                 idev = ipv6_add_dev(dev);
 424                 if (!idev)
 425                         return NULL;
 426         }
 427
 428         if (dev->flags&IFF_UP)
 429                 ipv6_mc_up(idev);
 430         return idev;
 431 }
 432
 433 static int inet6_netconf_msgsize_devconf(int type)
 434 {
 435         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 436                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 437
 438         /* type -1 is used for ALL */
 439         if (type == -1 || type == NETCONFA_FORWARDING)
 440                 size += nla_total_size(4);
 441 #ifdef CONFIG_IPV6_MROUTE
 442         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 443                 size += nla_total_size(4);
 444 #endif
 445         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 446                 size += nla_total_size(4);
 447
 448         return size;
 449 }
 450
 451 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 452                                       struct ipv6_devconf *devconf, u32 portid,
 453                                       u32 seq, int event, unsigned int flags,
 454                                       int type)
 455 {
 456         struct nlmsghdr  *nlh;
 457         struct netconfmsg *ncm;
 458
 459         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 460                         flags);
 461         if (nlh == NULL)
 462                 return -EMSGSIZE;
 463
 464         ncm = nlmsg_data(nlh);
 465         ncm->ncm_family = AF_INET6;
 466
 467         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 468                 goto nla_put_failure;
 469
 470         /* type -1 is used for ALL */
 471         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 472             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 473                 goto nla_put_failure;
 474 #ifdef CONFIG_IPV6_MROUTE
 475         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 476             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 477                         devconf->mc_forwarding) < 0)
 478                 goto nla_put_failure;
 479 #endif
 480         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 481             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 482                 goto nla_put_failure;
 483
 484         return nlmsg_end(skb, nlh);
 485
 486 nla_put_failure:
 487         nlmsg_cancel(skb, nlh);
 488         return -EMSGSIZE;
 489 }
 490
 491 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 492                                   struct ipv6_devconf *devconf)
 493 {
 494         struct sk_buff *skb;
 495         int err = -ENOBUFS;
 496
 497         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 498         if (skb == NULL)
 499                 goto errout;
 500
 501         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 502                                          RTM_NEWNETCONF, 0, type);
 503         if (err < 0) {
 504                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 505                 WARN_ON(err == -EMSGSIZE);
 506                 kfree_skb(skb);
 507                 goto errout;
 508         }
 509         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 510         return;
 511 errout:
 512         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 513 }
 514
 515 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 516         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 517         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 518         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 519 };
 520
 521 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 522                                      struct nlmsghdr *nlh)
 523 {
 524         struct net *net = sock_net(in_skb->sk);
 525         struct nlattr *tb[NETCONFA_MAX+1];
 526         struct netconfmsg *ncm;
 527         struct sk_buff *skb;
 528         struct ipv6_devconf *devconf;
 529         struct inet6_dev *in6_dev;
 530         struct net_device *dev;
 531         int ifindex;
 532         int err;
 533
 534         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 535                           devconf_ipv6_policy);
 536         if (err < 0)
 537                 goto errout;
 538
 539         err = EINVAL;
 540         if (!tb[NETCONFA_IFINDEX])
 541                 goto errout;
 542
 543         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 544         switch (ifindex) {
 545         case NETCONFA_IFINDEX_ALL:
 546                 devconf = net->ipv6.devconf_all;
 547                 break;
 548         case NETCONFA_IFINDEX_DEFAULT:
 549                 devconf = net->ipv6.devconf_dflt;
 550                 break;
 551         default:
 552                 dev = __dev_get_by_index(net, ifindex);
 553                 if (dev == NULL)
 554                         goto errout;
 555                 in6_dev = __in6_dev_get(dev);
 556                 if (in6_dev == NULL)
 557                         goto errout;
 558                 devconf = &in6_dev->cnf;
 559                 break;
 560         }
 561
 562         err = -ENOBUFS;
 563         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 564         if (skb == NULL)
 565                 goto errout;
 566
 567         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 568                                          NETLINK_CB(in_skb).portid,
 569                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 570                                          -1);
 571         if (err < 0) {
 572                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 573                 WARN_ON(err == -EMSGSIZE);
 574                 kfree_skb(skb);
 575                 goto errout;
 576         }
 577         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 578 errout:
 579         return err;
 580 }
 581
 582 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 583                                       struct netlink_callback *cb)
 584 {
 585         struct net *net = sock_net(skb->sk);
 586         int h, s_h;
 587         int idx, s_idx;
 588         struct net_device *dev;
 589         struct inet6_dev *idev;
 590         struct hlist_head *head;
 591
 592         s_h = cb->args[0];
 593         s_idx = idx = cb->args[1];
 594
 595         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 596                 idx = 0;
 597                 head = &net->dev_index_head[h];
 598                 rcu_read_lock();
 599                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 600                           net->dev_base_seq;
 601                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 602                         if (idx < s_idx)
 603                                 goto cont;
 604                         idev = __in6_dev_get(dev);
 605                         if (!idev)
 606                                 goto cont;
 607
 608                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 609                                                        &idev->cnf,
 610                                                        NETLINK_CB(cb->skb).portid,
 611                                                        cb->nlh->nlmsg_seq,
 612                                                        RTM_NEWNETCONF,
 613                                                        NLM_F_MULTI,
 614                                                        -1) <= 0) {
 615                                 rcu_read_unlock();
 616                                 goto done;
 617                         }
 618                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 619 cont:
 620                         idx++;
 621                 }
 622                 rcu_read_unlock();
 623         }
 624         if (h == NETDEV_HASHENTRIES) {
 625                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 626                                                net->ipv6.devconf_all,
 627                                                NETLINK_CB(cb->skb).portid,
 628                                                cb->nlh->nlmsg_seq,
 629                                                RTM_NEWNETCONF, NLM_F_MULTI,
 630                                                -1) <= 0)
 631                         goto done;
 632                 else
 633                         h++;
 634         }
 635         if (h == NETDEV_HASHENTRIES + 1) {
 636                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 637                                                net->ipv6.devconf_dflt,
 638                                                NETLINK_CB(cb->skb).portid,
 639                                                cb->nlh->nlmsg_seq,
 640                                                RTM_NEWNETCONF, NLM_F_MULTI,
 641                                                -1) <= 0)
 642                         goto done;
 643                 else
 644                         h++;
 645         }
 646 done:
 647         cb->args[0] = h;
 648         cb->args[1] = idx;
 649
 650         return skb->len;
 651 }
 652
 653 #ifdef CONFIG_SYSCTL
 654 static void dev_forward_change(struct inet6_dev *idev)
 655 {
 656         struct net_device *dev;
 657         struct inet6_ifaddr *ifa;
 658
 659         if (!idev)
 660                 return;
 661         dev = idev->dev;
 662         if (idev->cnf.forwarding)
 663                 dev_disable_lro(dev);
 664         if (dev->flags & IFF_MULTICAST) {
 665                 if (idev->cnf.forwarding) {
 666                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 667                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 668                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 669                 } else {
 670                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 671                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 672                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 673                 }
 674         }
 675
 676         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 677                 if (ifa->flags&IFA_F_TENTATIVE)
 678                         continue;
 679                 if (idev->cnf.forwarding)
 680                         addrconf_join_anycast(ifa);
 681                 else
 682                         addrconf_leave_anycast(ifa);
 683         }
 684         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 685                                      dev->ifindex, &idev->cnf);
 686 }
 687
 688
 689 static void addrconf_forward_change(struct net *net, __s32 newf)
 690 {
 691         struct net_device *dev;
 692         struct inet6_dev *idev;
 693
 694         for_each_netdev(net, dev) {
 695                 idev = __in6_dev_get(dev);
 696                 if (idev) {
 697                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 698                         idev->cnf.forwarding = newf;
 699                         if (changed)
 700                                 dev_forward_change(idev);
 701                 }
 702         }
 703 }
 704
 705 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 706 {
 707         struct net *net;
 708         int old;
 709
 710         if (!rtnl_trylock())
 711                 return restart_syscall();
 712
 713         net = (struct net *)table->extra2;
 714         old = *p;
 715         *p = newf;
 716
 717         if (p == &net->ipv6.devconf_dflt->forwarding) {
 718                 if ((!newf) ^ (!old))
 719                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 720                                                      NETCONFA_IFINDEX_DEFAULT,
 721                                                      net->ipv6.devconf_dflt);
 722                 rtnl_unlock();
 723                 return 0;
 724         }
 725
 726         if (p == &net->ipv6.devconf_all->forwarding) {
 727                 net->ipv6.devconf_dflt->forwarding = newf;
 728                 addrconf_forward_change(net, newf);
 729                 if ((!newf) ^ (!old))
 730                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 731                                                      NETCONFA_IFINDEX_ALL,
 732                                                      net->ipv6.devconf_all);
 733         } else if ((!newf) ^ (!old))
 734                 dev_forward_change((struct inet6_dev *)table->extra1);
 735         rtnl_unlock();
 736
 737         if (newf)
 738                 rt6_purge_dflt_routers(net);
 739         return 1;
 740 }
 741 #endif
 742
 743 /* Nobody refers to this ifaddr, destroy it */
 744 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 745 {
 746         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 747
 748 #ifdef NET_REFCNT_DEBUG
 749         pr_debug("%s\n", __func__);
 750 #endif
 751
 752         in6_dev_put(ifp->idev);
 753
 754         if (del_timer(&ifp->dad_timer))
 755                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 756
 757         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 758                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 759                 return;
 760         }
 761         ip6_rt_put(ifp->rt);
 762
 763         kfree_rcu(ifp, rcu);
 764 }
 765
 766 static void
 767 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 768 {
 769         struct list_head *p;
 770         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 771
 772         /*
 773          * Each device address list is sorted in order of scope -
 774          * global before linklocal.
 775          */
 776         list_for_each(p, &idev->addr_list) {
 777                 struct inet6_ifaddr *ifa
 778                         = list_entry(p, struct inet6_ifaddr, if_list);
 779                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 780                         break;
 781         }
 782
 783         list_add_tail(&ifp->if_list, p);
 784 }
 785
 786 static u32 inet6_addr_hash(const struct in6_addr *addr)
 787 {
 788         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 789 }
 790
 791 /* On success it returns ifp with increased reference count */
 792
 793 static struct inet6_ifaddr *
 794 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 795               const struct in6_addr *peer_addr, int pfxlen,
 796               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 797 {
 798         struct inet6_ifaddr *ifa = NULL;
 799         struct rt6_info *rt;
 800         unsigned int hash;
 801         int err = 0;
 802         int addr_type = ipv6_addr_type(addr);
 803
 804         if (addr_type == IPV6_ADDR_ANY ||
 805             addr_type & IPV6_ADDR_MULTICAST ||
 806             (!(idev->dev->flags & IFF_LOOPBACK) &&
 807              addr_type & IPV6_ADDR_LOOPBACK))
 808                 return ERR_PTR(-EADDRNOTAVAIL);
 809
 810         rcu_read_lock_bh();
 811         if (idev->dead) {
 812                 err = -ENODEV;                  /*XXX*/
 813                 goto out2;
 814         }
 815
 816         if (idev->cnf.disable_ipv6) {
 817                 err = -EACCES;
 818                 goto out2;
 819         }
 820
 821         spin_lock(&addrconf_hash_lock);
 822
 823         /* Ignore adding duplicate addresses on an interface */
 824         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 825                 ADBG("ipv6_add_addr: already assigned\n");
 826                 err = -EEXIST;
 827                 goto out;
 828         }
 829
 830         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 831
 832         if (ifa == NULL) {
 833                 ADBG("ipv6_add_addr: malloc failed\n");
 834                 err = -ENOBUFS;
 835                 goto out;
 836         }
 837
 838         rt = addrconf_dst_alloc(idev, addr, false);
 839         if (IS_ERR(rt)) {
 840                 err = PTR_ERR(rt);
 841                 goto out;
 842         }
 843
 844         neigh_parms_data_state_setall(idev->nd_parms);
 845
 846         ifa->addr = *addr;
 847         if (peer_addr)
 848                 ifa->peer_addr = *peer_addr;
 849
 850         spin_lock_init(&ifa->lock);
 851         spin_lock_init(&ifa->state_lock);
 852         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 853                     (unsigned long)ifa);
 854         INIT_HLIST_NODE(&ifa->addr_lst);
 855         ifa->scope = scope;
 856         ifa->prefix_len = pfxlen;
 857         ifa->flags = flags | IFA_F_TENTATIVE;
 858         ifa->valid_lft = valid_lft;
 859         ifa->prefered_lft = prefered_lft;
 860         ifa->cstamp = ifa->tstamp = jiffies;
 861         ifa->tokenized = false;
 862
 863         ifa->rt = rt;
 864
 865         ifa->idev = idev;
 866         in6_dev_hold(idev);
 867         /* For caller */
 868         in6_ifa_hold(ifa);
 869
 870         /* Add to big hash table */
 871         hash = inet6_addr_hash(addr);
 872
 873         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 874         spin_unlock(&addrconf_hash_lock);
 875
 876         write_lock(&idev->lock);
 877         /* Add to inet6_dev unicast addr list. */
 878         ipv6_link_dev_addr(idev, ifa);
 879
 880         if (ifa->flags&IFA_F_TEMPORARY) {
 881                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 882                 in6_ifa_hold(ifa);
 883         }
 884
 885         in6_ifa_hold(ifa);
 886         write_unlock(&idev->lock);
 887 out2:
 888         rcu_read_unlock_bh();
 889
 890         if (likely(err == 0))
 891                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 892         else {
 893                 kfree(ifa);
 894                 ifa = ERR_PTR(err);
 895         }
 896
 897         return ifa;
 898 out:
 899         spin_unlock(&addrconf_hash_lock);
 900         goto out2;
 901 }
 902
 903 enum cleanup_prefix_rt_t {
 904         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
 905         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
 906         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
 907 };
 908
 909 /*
 910  * Check, whether the prefix for ifp would still need a prefix route
 911  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
 912  * constants.
 913  *
 914  * 1) we don't purge prefix if address was not permanent.
 915  *    prefix is managed by its own lifetime.
 916  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
 917  * 3) if there are no addresses, delete prefix.
 918  * 4) if there are still other permanent address(es),
 919  *    corresponding prefix is still permanent.
 920  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
 921  *    don't purge the prefix, assume user space is managing it.
 922  * 6) otherwise, update prefix lifetime to the
 923  *    longest valid lifetime among the corresponding
 924  *    addresses on the device.
 925  *    Note: subsequent RA will update lifetime.
 926  **/
 927 static enum cleanup_prefix_rt_t
 928 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
 929 {
 930         struct inet6_ifaddr *ifa;
 931         struct inet6_dev *idev = ifp->idev;
 932         unsigned long lifetime;
 933         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
 934
 935         *expires = jiffies;
 936
 937         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 938                 if (ifa == ifp)
 939                         continue;
 940                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 941                                        ifp->prefix_len))
 942                         continue;
 943                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
 944                         return CLEANUP_PREFIX_RT_NOP;
 945
 946                 action = CLEANUP_PREFIX_RT_EXPIRE;
 947
 948                 spin_lock(&ifa->lock);
 949
 950                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 951                 /*
 952                  * Note: Because this address is
 953                  * not permanent, lifetime <
 954                  * LONG_MAX / HZ here.
 955                  */
 956                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
 957                         *expires = ifa->tstamp + lifetime * HZ;
 958                 spin_unlock(&ifa->lock);
 959         }
 960
 961         return action;
 962 }
 963
 964 static void
 965 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
 966 {
 967         struct rt6_info *rt;
 968
 969         rt = addrconf_get_prefix_route(&ifp->addr,
 970                                        ifp->prefix_len,
 971                                        ifp->idev->dev,
 972                                        0, RTF_GATEWAY | RTF_DEFAULT);
 973         if (rt) {
 974                 if (del_rt)
 975                         ip6_del_rt(rt);
 976                 else {
 977                         if (!(rt->rt6i_flags & RTF_EXPIRES))
 978                                 rt6_set_expires(rt, expires);
 979                         ip6_rt_put(rt);
 980                 }
 981         }
 982 }
 983
 984
 985 /* This function wants to get referenced ifp and releases it before return */
 986
 987 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 988 {
 989         int state;
 990         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
 991         unsigned long expires;
 992
 993         spin_lock_bh(&ifp->state_lock);
 994         state = ifp->state;
 995         ifp->state = INET6_IFADDR_STATE_DEAD;
 996         spin_unlock_bh(&ifp->state_lock);
 997
 998         if (state == INET6_IFADDR_STATE_DEAD)
 999                 goto out;
1000
1001         spin_lock_bh(&addrconf_hash_lock);
1002         hlist_del_init_rcu(&ifp->addr_lst);
1003         spin_unlock_bh(&addrconf_hash_lock);
1004
1005         write_lock_bh(&ifp->idev->lock);
1006
1007         if (ifp->flags&IFA_F_TEMPORARY) {
1008                 list_del(&ifp->tmp_list);
1009                 if (ifp->ifpub) {
1010                         in6_ifa_put(ifp->ifpub);
1011                         ifp->ifpub = NULL;
1012                 }
1013                 __in6_ifa_put(ifp);
1014         }
1015
1016         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1017                 action = check_cleanup_prefix_route(ifp, &expires);
1018
1019         list_del_init(&ifp->if_list);
1020         __in6_ifa_put(ifp);
1021
1022         write_unlock_bh(&ifp->idev->lock);
1023
1024         addrconf_del_dad_timer(ifp);
1025
1026         ipv6_ifa_notify(RTM_DELADDR, ifp);
1027
1028         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1029
1030         if (action != CLEANUP_PREFIX_RT_NOP) {
1031                 cleanup_prefix_route(ifp, expires,
1032                         action == CLEANUP_PREFIX_RT_DEL);
1033         }
1034
1035         /* clean up prefsrc entries */
1036         rt6_remove_prefsrc(ifp);
1037 out:
1038         in6_ifa_put(ifp);
1039 }
1040
1041 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1042 {
1043         struct inet6_dev *idev = ifp->idev;
1044         struct in6_addr addr, *tmpaddr;
1045         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1046         unsigned long regen_advance;
1047         int tmp_plen;
1048         int ret = 0;
1049         u32 addr_flags;
1050         unsigned long now = jiffies;
1051
1052         write_lock_bh(&idev->lock);
1053         if (ift) {
1054                 spin_lock_bh(&ift->lock);
1055                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1056                 spin_unlock_bh(&ift->lock);
1057                 tmpaddr = &addr;
1058         } else {
1059                 tmpaddr = NULL;
1060         }
1061 retry:
1062         in6_dev_hold(idev);
1063         if (idev->cnf.use_tempaddr <= 0) {
1064                 write_unlock_bh(&idev->lock);
1065                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1066                 in6_dev_put(idev);
1067                 ret = -1;
1068                 goto out;
1069         }
1070         spin_lock_bh(&ifp->lock);
1071         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1072                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1073                 spin_unlock_bh(&ifp->lock);
1074                 write_unlock_bh(&idev->lock);
1075                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1076                         __func__);
1077                 in6_dev_put(idev);
1078                 ret = -1;
1079                 goto out;
1080         }
1081         in6_ifa_hold(ifp);
1082         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1083         __ipv6_try_regen_rndid(idev, tmpaddr);
1084         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1085         age = (now - ifp->tstamp) / HZ;
1086         tmp_valid_lft = min_t(__u32,
1087                               ifp->valid_lft,
1088                               idev->cnf.temp_valid_lft + age);
1089         tmp_prefered_lft = min_t(__u32,
1090                                  ifp->prefered_lft,
1091                                  idev->cnf.temp_prefered_lft + age -
1092                                  idev->cnf.max_desync_factor);
1093         tmp_plen = ifp->prefix_len;
1094         tmp_tstamp = ifp->tstamp;
1095         spin_unlock_bh(&ifp->lock);
1096
1097         regen_advance = idev->cnf.regen_max_retry *
1098                         idev->cnf.dad_transmits *
1099                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1100         write_unlock_bh(&idev->lock);
1101
1102         /* A temporary address is created only if this calculated Preferred
1103          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1104          * an implementation must not create a temporary address with a zero
1105          * Preferred Lifetime.
1106          * Use age calculation as in addrconf_verify to avoid unnecessary
1107          * temporary addresses being generated.
1108          */
1109         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1110         if (tmp_prefered_lft <= regen_advance + age) {
1111                 in6_ifa_put(ifp);
1112                 in6_dev_put(idev);
1113                 ret = -1;
1114                 goto out;
1115         }
1116
1117         addr_flags = IFA_F_TEMPORARY;
1118         /* set in addrconf_prefix_rcv() */
1119         if (ifp->flags & IFA_F_OPTIMISTIC)
1120                 addr_flags |= IFA_F_OPTIMISTIC;
1121
1122         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1123                             ipv6_addr_scope(&addr), addr_flags,
1124                             tmp_valid_lft, tmp_prefered_lft);
1125         if (IS_ERR(ift)) {
1126                 in6_ifa_put(ifp);
1127                 in6_dev_put(idev);
1128                 pr_info("%s: retry temporary address regeneration\n", __func__);
1129                 tmpaddr = &addr;
1130                 write_lock_bh(&idev->lock);
1131                 goto retry;
1132         }
1133
1134         spin_lock_bh(&ift->lock);
1135         ift->ifpub = ifp;
1136         ift->cstamp = now;
1137         ift->tstamp = tmp_tstamp;
1138         spin_unlock_bh(&ift->lock);
1139
1140         addrconf_dad_start(ift);
1141         in6_ifa_put(ift);
1142         in6_dev_put(idev);
1143 out:
1144         return ret;
1145 }
1146
1147 /*
1148  *      Choose an appropriate source address (RFC3484)
1149  */
1150 enum {
1151         IPV6_SADDR_RULE_INIT = 0,
1152         IPV6_SADDR_RULE_LOCAL,
1153         IPV6_SADDR_RULE_SCOPE,
1154         IPV6_SADDR_RULE_PREFERRED,
1155 #ifdef CONFIG_IPV6_MIP6
1156         IPV6_SADDR_RULE_HOA,
1157 #endif
1158         IPV6_SADDR_RULE_OIF,
1159         IPV6_SADDR_RULE_LABEL,
1160         IPV6_SADDR_RULE_PRIVACY,
1161         IPV6_SADDR_RULE_ORCHID,
1162         IPV6_SADDR_RULE_PREFIX,
1163         IPV6_SADDR_RULE_MAX
1164 };
1165
1166 struct ipv6_saddr_score {
1167         int                     rule;
1168         int                     addr_type;
1169         struct inet6_ifaddr     *ifa;
1170         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1171         int                     scopedist;
1172         int                     matchlen;
1173 };
1174
1175 struct ipv6_saddr_dst {
1176         const struct in6_addr *addr;
1177         int ifindex;
1178         int scope;
1179         int label;
1180         unsigned int prefs;
1181 };
1182
1183 static inline int ipv6_saddr_preferred(int type)
1184 {
1185         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1186                 return 1;
1187         return 0;
1188 }
1189
1190 static int ipv6_get_saddr_eval(struct net *net,
1191                                struct ipv6_saddr_score *score,
1192                                struct ipv6_saddr_dst *dst,
1193                                int i)
1194 {
1195         int ret;
1196
1197         if (i <= score->rule) {
1198                 switch (i) {
1199                 case IPV6_SADDR_RULE_SCOPE:
1200                         ret = score->scopedist;
1201                         break;
1202                 case IPV6_SADDR_RULE_PREFIX:
1203                         ret = score->matchlen;
1204                         break;
1205                 default:
1206                         ret = !!test_bit(i, score->scorebits);
1207                 }
1208                 goto out;
1209         }
1210
1211         switch (i) {
1212         case IPV6_SADDR_RULE_INIT:
1213                 /* Rule 0: remember if hiscore is not ready yet */
1214                 ret = !!score->ifa;
1215                 break;
1216         case IPV6_SADDR_RULE_LOCAL:
1217                 /* Rule 1: Prefer same address */
1218                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1219                 break;
1220         case IPV6_SADDR_RULE_SCOPE:
1221                 /* Rule 2: Prefer appropriate scope
1222                  *
1223                  *      ret
1224                  *       ^
1225                  *    -1 |  d 15
1226                  *    ---+--+-+---> scope
1227                  *       |
1228                  *       |             d is scope of the destination.
1229                  *  B-d  |  \
1230                  *       |   \      <- smaller scope is better if
1231                  *  B-15 |    \        if scope is enough for destination.
1232                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1233                  * d-C-1 | /
1234                  *       |/         <- greater is better
1235                  *   -C  /             if scope is not enough for destination.
1236                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1237                  *
1238                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1239                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1240                  * Assume B = 0 and we get C > 29.
1241                  */
1242                 ret = __ipv6_addr_src_scope(score->addr_type);
1243                 if (ret >= dst->scope)
1244                         ret = -ret;
1245                 else
1246                         ret -= 128;     /* 30 is enough */
1247                 score->scopedist = ret;
1248                 break;
1249         case IPV6_SADDR_RULE_PREFERRED:
1250                 /* Rule 3: Avoid deprecated and optimistic addresses */
1251                 ret = ipv6_saddr_preferred(score->addr_type) ||
1252                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1253                 break;
1254 #ifdef CONFIG_IPV6_MIP6
1255         case IPV6_SADDR_RULE_HOA:
1256             {
1257                 /* Rule 4: Prefer home address */
1258                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1259                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1260                 break;
1261             }
1262 #endif
1263         case IPV6_SADDR_RULE_OIF:
1264                 /* Rule 5: Prefer outgoing interface */
1265                 ret = (!dst->ifindex ||
1266                        dst->ifindex == score->ifa->idev->dev->ifindex);
1267                 break;
1268         case IPV6_SADDR_RULE_LABEL:
1269                 /* Rule 6: Prefer matching label */
1270                 ret = ipv6_addr_label(net,
1271                                       &score->ifa->addr, score->addr_type,
1272                                       score->ifa->idev->dev->ifindex) == dst->label;
1273                 break;
1274         case IPV6_SADDR_RULE_PRIVACY:
1275             {
1276                 /* Rule 7: Prefer public address
1277                  * Note: prefer temporary address if use_tempaddr >= 2
1278                  */
1279                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1280                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1281                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1282                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1283                 break;
1284             }
1285         case IPV6_SADDR_RULE_ORCHID:
1286                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1287                  *          non-ORCHID vs non-ORCHID
1288                  */
1289                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1290                         ipv6_addr_orchid(dst->addr));
1291                 break;
1292         case IPV6_SADDR_RULE_PREFIX:
1293                 /* Rule 8: Use longest matching prefix */
1294                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1295                 if (ret > score->ifa->prefix_len)
1296                         ret = score->ifa->prefix_len;
1297                 score->matchlen = ret;
1298                 break;
1299         default:
1300                 ret = 0;
1301         }
1302
1303         if (ret)
1304                 __set_bit(i, score->scorebits);
1305         score->rule = i;
1306 out:
1307         return ret;
1308 }
1309
1310 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1311                        const struct in6_addr *daddr, unsigned int prefs,
1312                        struct in6_addr *saddr)
1313 {
1314         struct ipv6_saddr_score scores[2],
1315                                 *score = &scores[0], *hiscore = &scores[1];
1316         struct ipv6_saddr_dst dst;
1317         struct net_device *dev;
1318         int dst_type;
1319
1320         dst_type = __ipv6_addr_type(daddr);
1321         dst.addr = daddr;
1322         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1323         dst.scope = __ipv6_addr_src_scope(dst_type);
1324         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1325         dst.prefs = prefs;
1326
1327         hiscore->rule = -1;
1328         hiscore->ifa = NULL;
1329
1330         rcu_read_lock();
1331
1332         for_each_netdev_rcu(net, dev) {
1333                 struct inet6_dev *idev;
1334
1335                 /* Candidate Source Address (section 4)
1336                  *  - multicast and link-local destination address,
1337                  *    the set of candidate source address MUST only
1338                  *    include addresses assigned to interfaces
1339                  *    belonging to the same link as the outgoing
1340                  *    interface.
1341                  * (- For site-local destination addresses, the
1342                  *    set of candidate source addresses MUST only
1343                  *    include addresses assigned to interfaces
1344                  *    belonging to the same site as the outgoing
1345                  *    interface.)
1346                  */
1347                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1348                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1349                     dst.ifindex && dev->ifindex != dst.ifindex)
1350                         continue;
1351
1352                 idev = __in6_dev_get(dev);
1353                 if (!idev)
1354                         continue;
1355
1356                 read_lock_bh(&idev->lock);
1357                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1358                         int i;
1359
1360                         /*
1361                          * - Tentative Address (RFC2462 section 5.4)
1362                          *  - A tentative address is not considered
1363                          *    "assigned to an interface" in the traditional
1364                          *    sense, unless it is also flagged as optimistic.
1365                          * - Candidate Source Address (section 4)
1366                          *  - In any case, anycast addresses, multicast
1367                          *    addresses, and the unspecified address MUST
1368                          *    NOT be included in a candidate set.
1369                          */
1370                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1371                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1372                                 continue;
1373
1374                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1375
1376                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1377                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1378                                 LIMIT_NETDEBUG(KERN_DEBUG
1379                                                "ADDRCONF: unspecified / multicast address "
1380                                                "assigned as unicast address on %s",
1381                                                dev->name);
1382                                 continue;
1383                         }
1384
1385                         score->rule = -1;
1386                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1387
1388                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1389                                 int minihiscore, miniscore;
1390
1391                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1392                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1393
1394                                 if (minihiscore > miniscore) {
1395                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1396                                             score->scopedist > 0) {
1397                                                 /*
1398                                                  * special case:
1399                                                  * each remaining entry
1400                                                  * has too small (not enough)
1401                                                  * scope, because ifa entries
1402                                                  * are sorted by their scope
1403                                                  * values.
1404                                                  */
1405                                                 goto try_nextdev;
1406                                         }
1407                                         break;
1408                                 } else if (minihiscore < miniscore) {
1409                                         if (hiscore->ifa)
1410                                                 in6_ifa_put(hiscore->ifa);
1411
1412                                         in6_ifa_hold(score->ifa);
1413
1414                                         swap(hiscore, score);
1415
1416                                         /* restore our iterator */
1417                                         score->ifa = hiscore->ifa;
1418
1419                                         break;
1420                                 }
1421                         }
1422                 }
1423 try_nextdev:
1424                 read_unlock_bh(&idev->lock);
1425         }
1426         rcu_read_unlock();
1427
1428         if (!hiscore->ifa)
1429                 return -EADDRNOTAVAIL;
1430
1431         *saddr = hiscore->ifa->addr;
1432         in6_ifa_put(hiscore->ifa);
1433         return 0;
1434 }
1435 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1436
1437 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1438                       u32 banned_flags)
1439 {
1440         struct inet6_ifaddr *ifp;
1441         int err = -EADDRNOTAVAIL;
1442
1443         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1444                 if (ifp->scope > IFA_LINK)
1445                         break;
1446                 if (ifp->scope == IFA_LINK &&
1447                     !(ifp->flags & banned_flags)) {
1448                         *addr = ifp->addr;
1449                         err = 0;
1450                         break;
1451                 }
1452         }
1453         return err;
1454 }
1455
1456 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1457                     u32 banned_flags)
1458 {
1459         struct inet6_dev *idev;
1460         int err = -EADDRNOTAVAIL;
1461
1462         rcu_read_lock();
1463         idev = __in6_dev_get(dev);
1464         if (idev) {
1465                 read_lock_bh(&idev->lock);
1466                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1467                 read_unlock_bh(&idev->lock);
1468         }
1469         rcu_read_unlock();
1470         return err;
1471 }
1472
1473 static int ipv6_count_addresses(struct inet6_dev *idev)
1474 {
1475         int cnt = 0;
1476         struct inet6_ifaddr *ifp;
1477
1478         read_lock_bh(&idev->lock);
1479         list_for_each_entry(ifp, &idev->addr_list, if_list)
1480                 cnt++;
1481         read_unlock_bh(&idev->lock);
1482         return cnt;
1483 }
1484
1485 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1486                   const struct net_device *dev, int strict)
1487 {
1488         struct inet6_ifaddr *ifp;
1489         unsigned int hash = inet6_addr_hash(addr);
1490
1491         rcu_read_lock_bh();
1492         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1493                 if (!net_eq(dev_net(ifp->idev->dev), net))
1494                         continue;
1495                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1496                     !(ifp->flags&IFA_F_TENTATIVE) &&
1497                     (dev == NULL || ifp->idev->dev == dev ||
1498                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1499                         rcu_read_unlock_bh();
1500                         return 1;
1501                 }
1502         }
1503
1504         rcu_read_unlock_bh();
1505         return 0;
1506 }
1507 EXPORT_SYMBOL(ipv6_chk_addr);
1508
1509 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1510                                struct net_device *dev)
1511 {
1512         unsigned int hash = inet6_addr_hash(addr);
1513         struct inet6_ifaddr *ifp;
1514
1515         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1516                 if (!net_eq(dev_net(ifp->idev->dev), net))
1517                         continue;
1518                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1519                         if (dev == NULL || ifp->idev->dev == dev)
1520                                 return true;
1521                 }
1522         }
1523         return false;
1524 }
1525
1526 /* Compares an address/prefix_len with addresses on device @dev.
1527  * If one is found it returns true.
1528  */
1529 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1530         const unsigned int prefix_len, struct net_device *dev)
1531 {
1532         struct inet6_dev *idev;
1533         struct inet6_ifaddr *ifa;
1534         bool ret = false;
1535
1536         rcu_read_lock();
1537         idev = __in6_dev_get(dev);
1538         if (idev) {
1539                 read_lock_bh(&idev->lock);
1540                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1541                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1542                         if (ret)
1543                                 break;
1544                 }
1545                 read_unlock_bh(&idev->lock);
1546         }
1547         rcu_read_unlock();
1548
1549         return ret;
1550 }
1551 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1552
1553 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1554 {
1555         struct inet6_dev *idev;
1556         struct inet6_ifaddr *ifa;
1557         int     onlink;
1558
1559         onlink = 0;
1560         rcu_read_lock();
1561         idev = __in6_dev_get(dev);
1562         if (idev) {
1563                 read_lock_bh(&idev->lock);
1564                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1565                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1566                                                    ifa->prefix_len);
1567                         if (onlink)
1568                                 break;
1569                 }
1570                 read_unlock_bh(&idev->lock);
1571         }
1572         rcu_read_unlock();
1573         return onlink;
1574 }
1575 EXPORT_SYMBOL(ipv6_chk_prefix);
1576
1577 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1578                                      struct net_device *dev, int strict)
1579 {
1580         struct inet6_ifaddr *ifp, *result = NULL;
1581         unsigned int hash = inet6_addr_hash(addr);
1582
1583         rcu_read_lock_bh();
1584         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1585                 if (!net_eq(dev_net(ifp->idev->dev), net))
1586                         continue;
1587                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1588                         if (dev == NULL || ifp->idev->dev == dev ||
1589                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1590                                 result = ifp;
1591                                 in6_ifa_hold(ifp);
1592                                 break;
1593                         }
1594                 }
1595         }
1596         rcu_read_unlock_bh();
1597
1598         return result;
1599 }
1600
1601 /* Gets referenced address, destroys ifaddr */
1602
1603 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1604 {
1605         if (ifp->flags&IFA_F_PERMANENT) {
1606                 spin_lock_bh(&ifp->lock);
1607                 addrconf_del_dad_timer(ifp);
1608                 ifp->flags |= IFA_F_TENTATIVE;
1609                 if (dad_failed)
1610                         ifp->flags |= IFA_F_DADFAILED;
1611                 spin_unlock_bh(&ifp->lock);
1612                 if (dad_failed)
1613                         ipv6_ifa_notify(0, ifp);
1614                 in6_ifa_put(ifp);
1615         } else if (ifp->flags&IFA_F_TEMPORARY) {
1616                 struct inet6_ifaddr *ifpub;
1617                 spin_lock_bh(&ifp->lock);
1618                 ifpub = ifp->ifpub;
1619                 if (ifpub) {
1620                         in6_ifa_hold(ifpub);
1621                         spin_unlock_bh(&ifp->lock);
1622                         ipv6_create_tempaddr(ifpub, ifp);
1623                         in6_ifa_put(ifpub);
1624                 } else {
1625                         spin_unlock_bh(&ifp->lock);
1626                 }
1627                 ipv6_del_addr(ifp);
1628         } else
1629                 ipv6_del_addr(ifp);
1630 }
1631
1632 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1633 {
1634         int err = -ENOENT;
1635
1636         spin_lock(&ifp->state_lock);
1637         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1638                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1639                 err = 0;
1640         }
1641         spin_unlock(&ifp->state_lock);
1642
1643         return err;
1644 }
1645
1646 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1647 {
1648         struct inet6_dev *idev = ifp->idev;
1649
1650         if (addrconf_dad_end(ifp)) {
1651                 in6_ifa_put(ifp);
1652                 return;
1653         }
1654
1655         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1656                              ifp->idev->dev->name, &ifp->addr);
1657
1658         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1659                 struct in6_addr addr;
1660
1661                 addr.s6_addr32[0] = htonl(0xfe800000);
1662                 addr.s6_addr32[1] = 0;
1663
1664                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1665                     ipv6_addr_equal(&ifp->addr, &addr)) {
1666                         /* DAD failed for link-local based on MAC address */
1667                         idev->cnf.disable_ipv6 = 1;
1668
1669                         pr_info("%s: IPv6 being disabled!\n",
1670                                 ifp->idev->dev->name);
1671                 }
1672         }
1673
1674         addrconf_dad_stop(ifp, 1);
1675 }
1676
1677 /* Join to solicited addr multicast group. */
1678
1679 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1680 {
1681         struct in6_addr maddr;
1682
1683         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1684                 return;
1685
1686         addrconf_addr_solict_mult(addr, &maddr);
1687         ipv6_dev_mc_inc(dev, &maddr);
1688 }
1689
1690 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1691 {
1692         struct in6_addr maddr;
1693
1694         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1695                 return;
1696
1697         addrconf_addr_solict_mult(addr, &maddr);
1698         __ipv6_dev_mc_dec(idev, &maddr);
1699 }
1700
1701 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1702 {
1703         struct in6_addr addr;
1704         if (ifp->prefix_len >= 127) /* RFC 6164 */
1705                 return;
1706         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1707         if (ipv6_addr_any(&addr))
1708                 return;
1709         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1710 }
1711
1712 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1713 {
1714         struct in6_addr addr;
1715         if (ifp->prefix_len >= 127) /* RFC 6164 */
1716                 return;
1717         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1718         if (ipv6_addr_any(&addr))
1719                 return;
1720         __ipv6_dev_ac_dec(ifp->idev, &addr);
1721 }
1722
1723 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1724 {
1725         if (dev->addr_len != ETH_ALEN)
1726                 return -1;
1727         memcpy(eui, dev->dev_addr, 3);
1728         memcpy(eui + 5, dev->dev_addr + 3, 3);
1729
1730         /*
1731          * The zSeries OSA network cards can be shared among various
1732          * OS instances, but the OSA cards have only one MAC address.
1733          * This leads to duplicate address conflicts in conjunction
1734          * with IPv6 if more than one instance uses the same card.
1735          *
1736          * The driver for these cards can deliver a unique 16-bit
1737          * identifier for each instance sharing the same card.  It is
1738          * placed instead of 0xFFFE in the interface identifier.  The
1739          * "u" bit of the interface identifier is not inverted in this
1740          * case.  Hence the resulting interface identifier has local
1741          * scope according to RFC2373.
1742          */
1743         if (dev->dev_id) {
1744                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1745                 eui[4] = dev->dev_id & 0xFF;
1746         } else {
1747                 eui[3] = 0xFF;
1748                 eui[4] = 0xFE;
1749                 eui[0] ^= 2;
1750         }
1751         return 0;
1752 }
1753
1754 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1755 {
1756         if (dev->addr_len != IEEE802154_ADDR_LEN)
1757                 return -1;
1758         memcpy(eui, dev->dev_addr, 8);
1759         eui[0] ^= 2;
1760         return 0;
1761 }
1762
1763 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1764 {
1765         union fwnet_hwaddr *ha;
1766
1767         if (dev->addr_len != FWNET_ALEN)
1768                 return -1;
1769
1770         ha = (union fwnet_hwaddr *)dev->dev_addr;
1771
1772         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1773         eui[0] ^= 2;
1774         return 0;
1775 }
1776
1777 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1778 {
1779         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1780         if (dev->addr_len != ARCNET_ALEN)
1781                 return -1;
1782         memset(eui, 0, 7);
1783         eui[7] = *(u8 *)dev->dev_addr;
1784         return 0;
1785 }
1786
1787 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1788 {
1789         if (dev->addr_len != INFINIBAND_ALEN)
1790                 return -1;
1791         memcpy(eui, dev->dev_addr + 12, 8);
1792         eui[0] |= 2;
1793         return 0;
1794 }
1795
1796 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1797 {
1798         if (addr == 0)
1799                 return -1;
1800         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1801                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1802                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1803                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1804                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1805                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1806         eui[1] = 0;
1807         eui[2] = 0x5E;
1808         eui[3] = 0xFE;
1809         memcpy(eui + 4, &addr, 4);
1810         return 0;
1811 }
1812
1813 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1814 {
1815         if (dev->priv_flags & IFF_ISATAP)
1816                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1817         return -1;
1818 }
1819
1820 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1821 {
1822         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1823 }
1824
1825 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1826 {
1827         memcpy(eui, dev->perm_addr, 3);
1828         memcpy(eui + 5, dev->perm_addr + 3, 3);
1829         eui[3] = 0xFF;
1830         eui[4] = 0xFE;
1831         eui[0] ^= 2;
1832         return 0;
1833 }
1834
1835 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1836 {
1837         switch (dev->type) {
1838         case ARPHRD_ETHER:
1839         case ARPHRD_FDDI:
1840                 return addrconf_ifid_eui48(eui, dev);
1841         case ARPHRD_ARCNET:
1842                 return addrconf_ifid_arcnet(eui, dev);
1843         case ARPHRD_INFINIBAND:
1844                 return addrconf_ifid_infiniband(eui, dev);
1845         case ARPHRD_SIT:
1846                 return addrconf_ifid_sit(eui, dev);
1847         case ARPHRD_IPGRE:
1848                 return addrconf_ifid_gre(eui, dev);
1849         case ARPHRD_6LOWPAN:
1850         case ARPHRD_IEEE802154:
1851                 return addrconf_ifid_eui64(eui, dev);
1852         case ARPHRD_IEEE1394:
1853                 return addrconf_ifid_ieee1394(eui, dev);
1854         case ARPHRD_TUNNEL6:
1855                 return addrconf_ifid_ip6tnl(eui, dev);
1856         }
1857         return -1;
1858 }
1859
1860 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1861 {
1862         int err = -1;
1863         struct inet6_ifaddr *ifp;
1864
1865         read_lock_bh(&idev->lock);
1866         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1867                 if (ifp->scope > IFA_LINK)
1868                         break;
1869                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1870                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1871                         err = 0;
1872                         break;
1873                 }
1874         }
1875         read_unlock_bh(&idev->lock);
1876         return err;
1877 }
1878
1879 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1880 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1881 {
1882 regen:
1883         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1884         idev->rndid[0] &= ~0x02;
1885
1886         /*
1887          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1888          * check if generated address is not inappropriate
1889          *
1890          *  - Reserved subnet anycast (RFC 2526)
1891          *      11111101 11....11 1xxxxxxx
1892          *  - ISATAP (RFC4214) 6.1
1893          *      00-00-5E-FE-xx-xx-xx-xx
1894          *  - value 0
1895          *  - XXX: already assigned to an address on the device
1896          */
1897         if (idev->rndid[0] == 0xfd &&
1898             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1899             (idev->rndid[7]&0x80))
1900                 goto regen;
1901         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1902                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1903                         goto regen;
1904                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1905                         goto regen;
1906         }
1907 }
1908
1909 static void ipv6_regen_rndid(unsigned long data)
1910 {
1911         struct inet6_dev *idev = (struct inet6_dev *) data;
1912         unsigned long expires;
1913
1914         rcu_read_lock_bh();
1915         write_lock_bh(&idev->lock);
1916
1917         if (idev->dead)
1918                 goto out;
1919
1920         __ipv6_regen_rndid(idev);
1921
1922         expires = jiffies +
1923                 idev->cnf.temp_prefered_lft * HZ -
1924                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1925                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1926                 idev->cnf.max_desync_factor * HZ;
1927         if (time_before(expires, jiffies)) {
1928                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1929                         __func__, idev->dev->name);
1930                 goto out;
1931         }
1932
1933         if (!mod_timer(&idev->regen_timer, expires))
1934                 in6_dev_hold(idev);
1935
1936 out:
1937         write_unlock_bh(&idev->lock);
1938         rcu_read_unlock_bh();
1939         in6_dev_put(idev);
1940 }
1941
1942 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1943 {
1944         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1945                 __ipv6_regen_rndid(idev);
1946 }
1947
1948 /*
1949  *      Add prefix route.
1950  */
1951
1952 static void
1953 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1954                       unsigned long expires, u32 flags)
1955 {
1956         struct fib6_config cfg = {
1957                 .fc_table = RT6_TABLE_PREFIX,
1958                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1959                 .fc_ifindex = dev->ifindex,
1960                 .fc_expires = expires,
1961                 .fc_dst_len = plen,
1962                 .fc_flags = RTF_UP | flags,
1963                 .fc_nlinfo.nl_net = dev_net(dev),
1964                 .fc_protocol = RTPROT_KERNEL,
1965         };
1966
1967         cfg.fc_dst = *pfx;
1968
1969         /* Prevent useless cloning on PtP SIT.
1970            This thing is done here expecting that the whole
1971            class of non-broadcast devices need not cloning.
1972          */
1973 #if IS_ENABLED(CONFIG_IPV6_SIT)
1974         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1975                 cfg.fc_flags |= RTF_NONEXTHOP;
1976 #endif
1977
1978         ip6_route_add(&cfg);
1979 }
1980
1981
1982 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1983                                                   int plen,
1984                                                   const struct net_device *dev,
1985                                                   u32 flags, u32 noflags)
1986 {
1987         struct fib6_node *fn;
1988         struct rt6_info *rt = NULL;
1989         struct fib6_table *table;
1990
1991         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1992         if (table == NULL)
1993                 return NULL;
1994
1995         read_lock_bh(&table->tb6_lock);
1996         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1997         if (!fn)
1998                 goto out;
1999         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2000                 if (rt->dst.dev->ifindex != dev->ifindex)
2001                         continue;
2002                 if ((rt->rt6i_flags & flags) != flags)
2003                         continue;
2004                 if ((rt->rt6i_flags & noflags) != 0)
2005                         continue;
2006                 dst_hold(&rt->dst);
2007                 break;
2008         }
2009 out:
2010         read_unlock_bh(&table->tb6_lock);
2011         return rt;
2012 }
2013
2014
2015 /* Create "default" multicast route to the interface */
2016
2017 static void addrconf_add_mroute(struct net_device *dev)
2018 {
2019         struct fib6_config cfg = {
2020                 .fc_table = RT6_TABLE_LOCAL,
2021                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2022                 .fc_ifindex = dev->ifindex,
2023                 .fc_dst_len = 8,
2024                 .fc_flags = RTF_UP,
2025                 .fc_nlinfo.nl_net = dev_net(dev),
2026         };
2027
2028         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2029
2030         ip6_route_add(&cfg);
2031 }
2032
2033 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2034 {
2035         struct inet6_dev *idev;
2036
2037         ASSERT_RTNL();
2038
2039         idev = ipv6_find_idev(dev);
2040         if (!idev)
2041                 return ERR_PTR(-ENOBUFS);
2042
2043         if (idev->cnf.disable_ipv6)
2044                 return ERR_PTR(-EACCES);
2045
2046         /* Add default multicast route */
2047         if (!(dev->flags & IFF_LOOPBACK))
2048                 addrconf_add_mroute(dev);
2049
2050         return idev;
2051 }
2052
2053 static void manage_tempaddrs(struct inet6_dev *idev,
2054                              struct inet6_ifaddr *ifp,
2055                              __u32 valid_lft, __u32 prefered_lft,
2056                              bool create, unsigned long now)
2057 {
2058         u32 flags;
2059         struct inet6_ifaddr *ift;
2060
2061         read_lock_bh(&idev->lock);
2062         /* update all temporary addresses in the list */
2063         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2064                 int age, max_valid, max_prefered;
2065
2066                 if (ifp != ift->ifpub)
2067                         continue;
2068
2069                 /* RFC 4941 section 3.3:
2070                  * If a received option will extend the lifetime of a public
2071                  * address, the lifetimes of temporary addresses should
2072                  * be extended, subject to the overall constraint that no
2073                  * temporary addresses should ever remain "valid" or "preferred"
2074                  * for a time longer than (TEMP_VALID_LIFETIME) or
2075                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2076                  */
2077                 age = (now - ift->cstamp) / HZ;
2078                 max_valid = idev->cnf.temp_valid_lft - age;
2079                 if (max_valid < 0)
2080                         max_valid = 0;
2081
2082                 max_prefered = idev->cnf.temp_prefered_lft -
2083                                idev->cnf.max_desync_factor - age;
2084                 if (max_prefered < 0)
2085                         max_prefered = 0;
2086
2087                 if (valid_lft > max_valid)
2088                         valid_lft = max_valid;
2089
2090                 if (prefered_lft > max_prefered)
2091                         prefered_lft = max_prefered;
2092
2093                 spin_lock(&ift->lock);
2094                 flags = ift->flags;
2095                 ift->valid_lft = valid_lft;
2096                 ift->prefered_lft = prefered_lft;
2097                 ift->tstamp = now;
2098                 if (prefered_lft > 0)
2099                         ift->flags &= ~IFA_F_DEPRECATED;
2100
2101                 spin_unlock(&ift->lock);
2102                 if (!(flags&IFA_F_TENTATIVE))
2103                         ipv6_ifa_notify(0, ift);
2104         }
2105
2106         if ((create || list_empty(&idev->tempaddr_list)) &&
2107             idev->cnf.use_tempaddr > 0) {
2108                 /* When a new public address is created as described
2109                  * in [ADDRCONF], also create a new temporary address.
2110                  * Also create a temporary address if it's enabled but
2111                  * no temporary address currently exists.
2112                  */
2113                 read_unlock_bh(&idev->lock);
2114                 ipv6_create_tempaddr(ifp, NULL);
2115         } else {
2116                 read_unlock_bh(&idev->lock);
2117         }
2118 }
2119
2120 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2121 {
2122         struct prefix_info *pinfo;
2123         __u32 valid_lft;
2124         __u32 prefered_lft;
2125         int addr_type;
2126         struct inet6_dev *in6_dev;
2127         struct net *net = dev_net(dev);
2128
2129         pinfo = (struct prefix_info *) opt;
2130
2131         if (len < sizeof(struct prefix_info)) {
2132                 ADBG("addrconf: prefix option too short\n");
2133                 return;
2134         }
2135
2136         /*
2137          *      Validation checks ([ADDRCONF], page 19)
2138          */
2139
2140         addr_type = ipv6_addr_type(&pinfo->prefix);
2141
2142         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2143                 return;
2144
2145         valid_lft = ntohl(pinfo->valid);
2146         prefered_lft = ntohl(pinfo->prefered);
2147
2148         if (prefered_lft > valid_lft) {
2149                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2150                 return;
2151         }
2152
2153         in6_dev = in6_dev_get(dev);
2154
2155         if (in6_dev == NULL) {
2156                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2157                                     dev->name);
2158                 return;
2159         }
2160
2161         /*
2162          *      Two things going on here:
2163          *      1) Add routes for on-link prefixes
2164          *      2) Configure prefixes with the auto flag set
2165          */
2166
2167         if (pinfo->onlink) {
2168                 struct rt6_info *rt;
2169                 unsigned long rt_expires;
2170
2171                 /* Avoid arithmetic overflow. Really, we could
2172                  * save rt_expires in seconds, likely valid_lft,
2173                  * but it would require division in fib gc, that it
2174                  * not good.
2175                  */
2176                 if (HZ > USER_HZ)
2177                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2178                 else
2179                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2180
2181                 if (addrconf_finite_timeout(rt_expires))
2182                         rt_expires *= HZ;
2183
2184                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2185                                                pinfo->prefix_len,
2186                                                dev,
2187                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2188                                                RTF_GATEWAY | RTF_DEFAULT);
2189
2190                 if (rt) {
2191                         /* Autoconf prefix route */
2192                         if (valid_lft == 0) {
2193                                 ip6_del_rt(rt);
2194                                 rt = NULL;
2195                         } else if (addrconf_finite_timeout(rt_expires)) {
2196                                 /* not infinity */
2197                                 rt6_set_expires(rt, jiffies + rt_expires);
2198                         } else {
2199                                 rt6_clean_expires(rt);
2200                         }
2201                 } else if (valid_lft) {
2202                         clock_t expires = 0;
2203                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2204                         if (addrconf_finite_timeout(rt_expires)) {
2205                                 /* not infinity */
2206                                 flags |= RTF_EXPIRES;
2207                                 expires = jiffies_to_clock_t(rt_expires);
2208                         }
2209                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2210                                               dev, expires, flags);
2211                 }
2212                 ip6_rt_put(rt);
2213         }
2214
2215         /* Try to figure out our local address for this prefix */
2216
2217         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2218                 struct inet6_ifaddr *ifp;
2219                 struct in6_addr addr;
2220                 int create = 0, update_lft = 0;
2221                 bool tokenized = false;
2222
2223                 if (pinfo->prefix_len == 64) {
2224                         memcpy(&addr, &pinfo->prefix, 8);
2225
2226                         if (!ipv6_addr_any(&in6_dev->token)) {
2227                                 read_lock_bh(&in6_dev->lock);
2228                                 memcpy(addr.s6_addr + 8,
2229                                        in6_dev->token.s6_addr + 8, 8);
2230                                 read_unlock_bh(&in6_dev->lock);
2231                                 tokenized = true;
2232                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2233                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2234                                 in6_dev_put(in6_dev);
2235                                 return;
2236                         }
2237                         goto ok;
2238                 }
2239                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2240                                     pinfo->prefix_len);
2241                 in6_dev_put(in6_dev);
2242                 return;
2243
2244 ok:
2245
2246                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2247
2248                 if (ifp == NULL && valid_lft) {
2249                         int max_addresses = in6_dev->cnf.max_addresses;
2250                         u32 addr_flags = 0;
2251
2252 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2253                         if (in6_dev->cnf.optimistic_dad &&
2254                             !net->ipv6.devconf_all->forwarding && sllao)
2255                                 addr_flags = IFA_F_OPTIMISTIC;
2256 #endif
2257
2258                         /* Do not allow to create too much of autoconfigured
2259                          * addresses; this would be too easy way to crash kernel.
2260                          */
2261                         if (!max_addresses ||
2262                             ipv6_count_addresses(in6_dev) < max_addresses)
2263                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2264                                                     pinfo->prefix_len,
2265                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2266                                                     addr_flags, valid_lft,
2267                                                     prefered_lft);
2268
2269                         if (IS_ERR_OR_NULL(ifp)) {
2270                                 in6_dev_put(in6_dev);
2271                                 return;
2272                         }
2273
2274                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2275                         update_lft = 0;
2276                         create = 1;
2277                         ifp->cstamp = jiffies;
2278                         ifp->tokenized = tokenized;
2279                         addrconf_dad_start(ifp);
2280                 }
2281
2282                 if (ifp) {
2283                         u32 flags;
2284                         unsigned long now;
2285                         u32 stored_lft;
2286
2287                         /* update lifetime (RFC2462 5.5.3 e) */
2288                         spin_lock(&ifp->lock);
2289                         now = jiffies;
2290                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2291                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2292                         else
2293                                 stored_lft = 0;
2294                         if (!update_lft && !create && stored_lft) {
2295                                 const u32 minimum_lft = min(
2296                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2297                                 valid_lft = max(valid_lft, minimum_lft);
2298
2299                                 /* RFC4862 Section 5.5.3e:
2300                                  * "Note that the preferred lifetime of the
2301                                  *  corresponding address is always reset to
2302                                  *  the Preferred Lifetime in the received
2303                                  *  Prefix Information option, regardless of
2304                                  *  whether the valid lifetime is also reset or
2305                                  *  ignored."
2306                                  *
2307                                  * So we should always update prefered_lft here.
2308                                  */
2309                                 update_lft = 1;
2310                         }
2311
2312                         if (update_lft) {
2313                                 ifp->valid_lft = valid_lft;
2314                                 ifp->prefered_lft = prefered_lft;
2315                                 ifp->tstamp = now;
2316                                 flags = ifp->flags;
2317                                 ifp->flags &= ~IFA_F_DEPRECATED;
2318                                 spin_unlock(&ifp->lock);
2319
2320                                 if (!(flags&IFA_F_TENTATIVE))
2321                                         ipv6_ifa_notify(0, ifp);
2322                         } else
2323                                 spin_unlock(&ifp->lock);
2324
2325                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2326                                          create, now);
2327
2328                         in6_ifa_put(ifp);
2329                         addrconf_verify(0);
2330                 }
2331         }
2332         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2333         in6_dev_put(in6_dev);
2334 }
2335
2336 /*
2337  *      Set destination address.
2338  *      Special case for SIT interfaces where we create a new "virtual"
2339  *      device.
2340  */
2341 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2342 {
2343         struct in6_ifreq ireq;
2344         struct net_device *dev;
2345         int err = -EINVAL;
2346
2347         rtnl_lock();
2348
2349         err = -EFAULT;
2350         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2351                 goto err_exit;
2352
2353         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2354
2355         err = -ENODEV;
2356         if (dev == NULL)
2357                 goto err_exit;
2358
2359 #if IS_ENABLED(CONFIG_IPV6_SIT)
2360         if (dev->type == ARPHRD_SIT) {
2361                 const struct net_device_ops *ops = dev->netdev_ops;
2362                 struct ifreq ifr;
2363                 struct ip_tunnel_parm p;
2364
2365                 err = -EADDRNOTAVAIL;
2366                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2367                         goto err_exit;
2368
2369                 memset(&p, 0, sizeof(p));
2370                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2371                 p.iph.saddr = 0;
2372                 p.iph.version = 4;
2373                 p.iph.ihl = 5;
2374                 p.iph.protocol = IPPROTO_IPV6;
2375                 p.iph.ttl = 64;
2376                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2377
2378                 if (ops->ndo_do_ioctl) {
2379                         mm_segment_t oldfs = get_fs();
2380
2381                         set_fs(KERNEL_DS);
2382                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2383                         set_fs(oldfs);
2384                 } else
2385                         err = -EOPNOTSUPP;
2386
2387                 if (err == 0) {
2388                         err = -ENOBUFS;
2389                         dev = __dev_get_by_name(net, p.name);
2390                         if (!dev)
2391                                 goto err_exit;
2392                         err = dev_open(dev);
2393                 }
2394         }
2395 #endif
2396
2397 err_exit:
2398         rtnl_unlock();
2399         return err;
2400 }
2401
2402 /*
2403  *      Manual configuration of address on an interface
2404  */
2405 static int inet6_addr_add(struct net *net, int ifindex,
2406                           const struct in6_addr *pfx,
2407                           const struct in6_addr *peer_pfx,
2408                           unsigned int plen, __u32 ifa_flags,
2409                           __u32 prefered_lft, __u32 valid_lft)
2410 {
2411         struct inet6_ifaddr *ifp;
2412         struct inet6_dev *idev;
2413         struct net_device *dev;
2414         int scope;
2415         u32 flags;
2416         clock_t expires;
2417         unsigned long timeout;
2418
2419         ASSERT_RTNL();
2420
2421         if (plen > 128)
2422                 return -EINVAL;
2423
2424         /* check the lifetime */
2425         if (!valid_lft || prefered_lft > valid_lft)
2426                 return -EINVAL;
2427
2428         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2429                 return -EINVAL;
2430
2431         dev = __dev_get_by_index(net, ifindex);
2432         if (!dev)
2433                 return -ENODEV;
2434
2435         idev = addrconf_add_dev(dev);
2436         if (IS_ERR(idev))
2437                 return PTR_ERR(idev);
2438
2439         scope = ipv6_addr_scope(pfx);
2440
2441         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2442         if (addrconf_finite_timeout(timeout)) {
2443                 expires = jiffies_to_clock_t(timeout * HZ);
2444                 valid_lft = timeout;
2445                 flags = RTF_EXPIRES;
2446         } else {
2447                 expires = 0;
2448                 flags = 0;
2449                 ifa_flags |= IFA_F_PERMANENT;
2450         }
2451
2452         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2453         if (addrconf_finite_timeout(timeout)) {
2454                 if (timeout == 0)
2455                         ifa_flags |= IFA_F_DEPRECATED;
2456                 prefered_lft = timeout;
2457         }
2458
2459         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2460                             valid_lft, prefered_lft);
2461
2462         if (!IS_ERR(ifp)) {
2463                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2464                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2465                                               expires, flags);
2466                 }
2467
2468                 /*
2469                  * Note that section 3.1 of RFC 4429 indicates
2470                  * that the Optimistic flag should not be set for
2471                  * manually configured addresses
2472                  */
2473                 addrconf_dad_start(ifp);
2474                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2475                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2476                                          true, jiffies);
2477                 in6_ifa_put(ifp);
2478                 addrconf_verify(0);
2479                 return 0;
2480         }
2481
2482         return PTR_ERR(ifp);
2483 }
2484
2485 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2486                           unsigned int plen)
2487 {
2488         struct inet6_ifaddr *ifp;
2489         struct inet6_dev *idev;
2490         struct net_device *dev;
2491
2492         if (plen > 128)
2493                 return -EINVAL;
2494
2495         dev = __dev_get_by_index(net, ifindex);
2496         if (!dev)
2497                 return -ENODEV;
2498
2499         if ((idev = __in6_dev_get(dev)) == NULL)
2500                 return -ENXIO;
2501
2502         read_lock_bh(&idev->lock);
2503         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2504                 if (ifp->prefix_len == plen &&
2505                     ipv6_addr_equal(pfx, &ifp->addr)) {
2506                         in6_ifa_hold(ifp);
2507                         read_unlock_bh(&idev->lock);
2508
2509                         ipv6_del_addr(ifp);
2510                         return 0;
2511                 }
2512         }
2513         read_unlock_bh(&idev->lock);
2514         return -EADDRNOTAVAIL;
2515 }
2516
2517
2518 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2519 {
2520         struct in6_ifreq ireq;
2521         int err;
2522
2523         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2524                 return -EPERM;
2525
2526         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2527                 return -EFAULT;
2528
2529         rtnl_lock();
2530         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2531                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2532                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2533         rtnl_unlock();
2534         return err;
2535 }
2536
2537 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2538 {
2539         struct in6_ifreq ireq;
2540         int err;
2541
2542         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2543                 return -EPERM;
2544
2545         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2546                 return -EFAULT;
2547
2548         rtnl_lock();
2549         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2550                              ireq.ifr6_prefixlen);
2551         rtnl_unlock();
2552         return err;
2553 }
2554
2555 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2556                      int plen, int scope)
2557 {
2558         struct inet6_ifaddr *ifp;
2559
2560         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2561                             scope, IFA_F_PERMANENT,
2562                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2563         if (!IS_ERR(ifp)) {
2564                 spin_lock_bh(&ifp->lock);
2565                 ifp->flags &= ~IFA_F_TENTATIVE;
2566                 spin_unlock_bh(&ifp->lock);
2567                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2568                 in6_ifa_put(ifp);
2569         }
2570 }
2571
2572 #if IS_ENABLED(CONFIG_IPV6_SIT)
2573 static void sit_add_v4_addrs(struct inet6_dev *idev)
2574 {
2575         struct in6_addr addr;
2576         struct net_device *dev;
2577         struct net *net = dev_net(idev->dev);
2578         int scope, plen;
2579         u32 pflags = 0;
2580
2581         ASSERT_RTNL();
2582
2583         memset(&addr, 0, sizeof(struct in6_addr));
2584         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2585
2586         if (idev->dev->flags&IFF_POINTOPOINT) {
2587                 addr.s6_addr32[0] = htonl(0xfe800000);
2588                 scope = IFA_LINK;
2589                 plen = 64;
2590         } else {
2591                 scope = IPV6_ADDR_COMPATv4;
2592                 plen = 96;
2593                 pflags |= RTF_NONEXTHOP;
2594         }
2595
2596         if (addr.s6_addr32[3]) {
2597                 add_addr(idev, &addr, plen, scope);
2598                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2599                 return;
2600         }
2601
2602         for_each_netdev(net, dev) {
2603                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2604                 if (in_dev && (dev->flags & IFF_UP)) {
2605                         struct in_ifaddr *ifa;
2606
2607                         int flag = scope;
2608
2609                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2610
2611                                 addr.s6_addr32[3] = ifa->ifa_local;
2612
2613                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2614                                         continue;
2615                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2616                                         if (idev->dev->flags&IFF_POINTOPOINT)
2617                                                 continue;
2618                                         flag |= IFA_HOST;
2619                                 }
2620
2621                                 add_addr(idev, &addr, plen, flag);
2622                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2623                                                       pflags);
2624                         }
2625                 }
2626         }
2627 }
2628 #endif
2629
2630 static void init_loopback(struct net_device *dev)
2631 {
2632         struct inet6_dev  *idev;
2633         struct net_device *sp_dev;
2634         struct inet6_ifaddr *sp_ifa;
2635         struct rt6_info *sp_rt;
2636
2637         /* ::1 */
2638
2639         ASSERT_RTNL();
2640
2641         if ((idev = ipv6_find_idev(dev)) == NULL) {
2642                 pr_debug("%s: add_dev failed\n", __func__);
2643                 return;
2644         }
2645
2646         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2647
2648         /* Add routes to other interface's IPv6 addresses */
2649         for_each_netdev(dev_net(dev), sp_dev) {
2650                 if (!strcmp(sp_dev->name, dev->name))
2651                         continue;
2652
2653                 idev = __in6_dev_get(sp_dev);
2654                 if (!idev)
2655                         continue;
2656
2657                 read_lock_bh(&idev->lock);
2658                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2659
2660                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2661                                 continue;
2662
2663                         if (sp_ifa->rt) {
2664                                 /* This dst has been added to garbage list when
2665                                  * lo device down, release this obsolete dst and
2666                                  * reallocate a new router for ifa.
2667                                  */
2668                                 if (sp_ifa->rt->dst.obsolete > 0) {
2669                                         ip6_rt_put(sp_ifa->rt);
2670                                         sp_ifa->rt = NULL;
2671                                 } else {
2672                                         continue;
2673                                 }
2674                         }
2675
2676                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2677
2678                         /* Failure cases are ignored */
2679                         if (!IS_ERR(sp_rt)) {
2680                                 sp_ifa->rt = sp_rt;
2681                                 ip6_ins_rt(sp_rt);
2682                         }
2683                 }
2684                 read_unlock_bh(&idev->lock);
2685         }
2686 }
2687
2688 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2689 {
2690         struct inet6_ifaddr *ifp;
2691         u32 addr_flags = IFA_F_PERMANENT;
2692
2693 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2694         if (idev->cnf.optimistic_dad &&
2695             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2696                 addr_flags |= IFA_F_OPTIMISTIC;
2697 #endif
2698
2699
2700         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2701                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2702         if (!IS_ERR(ifp)) {
2703                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2704                 addrconf_dad_start(ifp);
2705                 in6_ifa_put(ifp);
2706         }
2707 }
2708
2709 static void addrconf_dev_config(struct net_device *dev)
2710 {
2711         struct in6_addr addr;
2712         struct inet6_dev *idev;
2713
2714         ASSERT_RTNL();
2715
2716         if ((dev->type != ARPHRD_ETHER) &&
2717             (dev->type != ARPHRD_FDDI) &&
2718             (dev->type != ARPHRD_ARCNET) &&
2719             (dev->type != ARPHRD_INFINIBAND) &&
2720             (dev->type != ARPHRD_IEEE802154) &&
2721             (dev->type != ARPHRD_IEEE1394) &&
2722             (dev->type != ARPHRD_TUNNEL6) &&
2723             (dev->type != ARPHRD_6LOWPAN)) {
2724                 /* Alas, we support only Ethernet autoconfiguration. */
2725                 return;
2726         }
2727
2728         idev = addrconf_add_dev(dev);
2729         if (IS_ERR(idev))
2730                 return;
2731
2732         memset(&addr, 0, sizeof(struct in6_addr));
2733         addr.s6_addr32[0] = htonl(0xFE800000);
2734
2735         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2736                 addrconf_add_linklocal(idev, &addr);
2737 }
2738
2739 #if IS_ENABLED(CONFIG_IPV6_SIT)
2740 static void addrconf_sit_config(struct net_device *dev)
2741 {
2742         struct inet6_dev *idev;
2743
2744         ASSERT_RTNL();
2745
2746         /*
2747          * Configure the tunnel with one of our IPv4
2748          * addresses... we should configure all of
2749          * our v4 addrs in the tunnel
2750          */
2751
2752         if ((idev = ipv6_find_idev(dev)) == NULL) {
2753                 pr_debug("%s: add_dev failed\n", __func__);
2754                 return;
2755         }
2756
2757         if (dev->priv_flags & IFF_ISATAP) {
2758                 struct in6_addr addr;
2759
2760                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2761                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2762                         addrconf_add_linklocal(idev, &addr);
2763                 return;
2764         }
2765
2766         sit_add_v4_addrs(idev);
2767
2768         if (dev->flags&IFF_POINTOPOINT)
2769                 addrconf_add_mroute(dev);
2770 }
2771 #endif
2772
2773 #if IS_ENABLED(CONFIG_NET_IPGRE)
2774 static void addrconf_gre_config(struct net_device *dev)
2775 {
2776         struct inet6_dev *idev;
2777         struct in6_addr addr;
2778
2779         ASSERT_RTNL();
2780
2781         if ((idev = ipv6_find_idev(dev)) == NULL) {
2782                 pr_debug("%s: add_dev failed\n", __func__);
2783                 return;
2784         }
2785
2786         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2787         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2788                 addrconf_add_linklocal(idev, &addr);
2789         else
2790                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2791 }
2792 #endif
2793
2794 static inline int
2795 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2796 {
2797         struct in6_addr lladdr;
2798
2799         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2800                 addrconf_add_linklocal(idev, &lladdr);
2801                 return 0;
2802         }
2803         return -1;
2804 }
2805
2806 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2807                            void *ptr)
2808 {
2809         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2810         struct inet6_dev *idev = __in6_dev_get(dev);
2811         int run_pending = 0;
2812         int err;
2813
2814         switch (event) {
2815         case NETDEV_REGISTER:
2816                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2817                         idev = ipv6_add_dev(dev);
2818                         if (!idev)
2819                                 return notifier_from_errno(-ENOMEM);
2820                 }
2821                 break;
2822
2823         case NETDEV_UP:
2824         case NETDEV_CHANGE:
2825                 if (dev->flags & IFF_SLAVE)
2826                         break;
2827
2828                 if (event == NETDEV_UP) {
2829                         if (!addrconf_qdisc_ok(dev)) {
2830                                 /* device is not ready yet. */
2831                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2832                                         dev->name);
2833                                 break;
2834                         }
2835
2836                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2837                                 idev = ipv6_add_dev(dev);
2838
2839                         if (idev) {
2840                                 idev->if_flags |= IF_READY;
2841                                 run_pending = 1;
2842                         }
2843                 } else {
2844                         if (!addrconf_qdisc_ok(dev)) {
2845                                 /* device is still not ready. */
2846                                 break;
2847                         }
2848
2849                         if (idev) {
2850                                 if (idev->if_flags & IF_READY)
2851                                         /* device is already configured. */
2852                                         break;
2853                                 idev->if_flags |= IF_READY;
2854                         }
2855
2856                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2857                                 dev->name);
2858
2859                         run_pending = 1;
2860                 }
2861
2862                 switch (dev->type) {
2863 #if IS_ENABLED(CONFIG_IPV6_SIT)
2864                 case ARPHRD_SIT:
2865                         addrconf_sit_config(dev);
2866                         break;
2867 #endif
2868 #if IS_ENABLED(CONFIG_NET_IPGRE)
2869                 case ARPHRD_IPGRE:
2870                         addrconf_gre_config(dev);
2871                         break;
2872 #endif
2873                 case ARPHRD_LOOPBACK:
2874                         init_loopback(dev);
2875                         break;
2876
2877                 default:
2878                         addrconf_dev_config(dev);
2879                         break;
2880                 }
2881
2882                 if (idev) {
2883                         if (run_pending)
2884                                 addrconf_dad_run(idev);
2885
2886                         /*
2887                          * If the MTU changed during the interface down,
2888                          * when the interface up, the changed MTU must be
2889                          * reflected in the idev as well as routers.
2890                          */
2891                         if (idev->cnf.mtu6 != dev->mtu &&
2892                             dev->mtu >= IPV6_MIN_MTU) {
2893                                 rt6_mtu_change(dev, dev->mtu);
2894                                 idev->cnf.mtu6 = dev->mtu;
2895                         }
2896                         idev->tstamp = jiffies;
2897                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2898
2899                         /*
2900                          * If the changed mtu during down is lower than
2901                          * IPV6_MIN_MTU stop IPv6 on this interface.
2902                          */
2903                         if (dev->mtu < IPV6_MIN_MTU)
2904                                 addrconf_ifdown(dev, 1);
2905                 }
2906                 break;
2907
2908         case NETDEV_CHANGEMTU:
2909                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2910                         rt6_mtu_change(dev, dev->mtu);
2911                         idev->cnf.mtu6 = dev->mtu;
2912                         break;
2913                 }
2914
2915                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2916                         idev = ipv6_add_dev(dev);
2917                         if (idev)
2918                                 break;
2919                 }
2920
2921                 /*
2922                  * if MTU under IPV6_MIN_MTU.
2923                  * Stop IPv6 on this interface.
2924                  */
2925
2926         case NETDEV_DOWN:
2927         case NETDEV_UNREGISTER:
2928                 /*
2929                  *      Remove all addresses from this interface.
2930                  */
2931                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2932                 break;
2933
2934         case NETDEV_CHANGENAME:
2935                 if (idev) {
2936                         snmp6_unregister_dev(idev);
2937                         addrconf_sysctl_unregister(idev);
2938                         addrconf_sysctl_register(idev);
2939                         err = snmp6_register_dev(idev);
2940                         if (err)
2941                                 return notifier_from_errno(err);
2942                 }
2943                 break;
2944
2945         case NETDEV_PRE_TYPE_CHANGE:
2946         case NETDEV_POST_TYPE_CHANGE:
2947                 addrconf_type_change(dev, event);
2948                 break;
2949         }
2950
2951         return NOTIFY_OK;
2952 }
2953
2954 /*
2955  *      addrconf module should be notified of a device going up
2956  */
2957 static struct notifier_block ipv6_dev_notf = {
2958         .notifier_call = addrconf_notify,
2959 };
2960
2961 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2962 {
2963         struct inet6_dev *idev;
2964         ASSERT_RTNL();
2965
2966         idev = __in6_dev_get(dev);
2967
2968         if (event == NETDEV_POST_TYPE_CHANGE)
2969                 ipv6_mc_remap(idev);
2970         else if (event == NETDEV_PRE_TYPE_CHANGE)
2971                 ipv6_mc_unmap(idev);
2972 }
2973
2974 static int addrconf_ifdown(struct net_device *dev, int how)
2975 {
2976         struct net *net = dev_net(dev);
2977         struct inet6_dev *idev;
2978         struct inet6_ifaddr *ifa;
2979         int state, i;
2980
2981         ASSERT_RTNL();
2982
2983         rt6_ifdown(net, dev);
2984         neigh_ifdown(&nd_tbl, dev);
2985
2986         idev = __in6_dev_get(dev);
2987         if (idev == NULL)
2988                 return -ENODEV;
2989
2990         /*
2991          * Step 1: remove reference to ipv6 device from parent device.
2992          *         Do not dev_put!
2993          */
2994         if (how) {
2995                 idev->dead = 1;
2996
2997                 /* protected by rtnl_lock */
2998                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2999
3000                 /* Step 1.5: remove snmp6 entry */
3001                 snmp6_unregister_dev(idev);
3002
3003         }
3004
3005         /* Step 2: clear hash table */
3006         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3007                 struct hlist_head *h = &inet6_addr_lst[i];
3008
3009                 spin_lock_bh(&addrconf_hash_lock);
3010         restart:
3011                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3012                         if (ifa->idev == idev) {
3013                                 hlist_del_init_rcu(&ifa->addr_lst);
3014                                 addrconf_del_dad_timer(ifa);
3015                                 goto restart;
3016                         }
3017                 }
3018                 spin_unlock_bh(&addrconf_hash_lock);
3019         }
3020
3021         write_lock_bh(&idev->lock);
3022
3023         addrconf_del_rs_timer(idev);
3024
3025         /* Step 2: clear flags for stateless addrconf */
3026         if (!how)
3027                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3028
3029         if (how && del_timer(&idev->regen_timer))
3030                 in6_dev_put(idev);
3031
3032         /* Step 3: clear tempaddr list */
3033         while (!list_empty(&idev->tempaddr_list)) {
3034                 ifa = list_first_entry(&idev->tempaddr_list,
3035                                        struct inet6_ifaddr, tmp_list);
3036                 list_del(&ifa->tmp_list);
3037                 write_unlock_bh(&idev->lock);
3038                 spin_lock_bh(&ifa->lock);
3039
3040                 if (ifa->ifpub) {
3041                         in6_ifa_put(ifa->ifpub);
3042                         ifa->ifpub = NULL;
3043                 }
3044                 spin_unlock_bh(&ifa->lock);
3045                 in6_ifa_put(ifa);
3046                 write_lock_bh(&idev->lock);
3047         }
3048
3049         while (!list_empty(&idev->addr_list)) {
3050                 ifa = list_first_entry(&idev->addr_list,
3051                                        struct inet6_ifaddr, if_list);
3052                 addrconf_del_dad_timer(ifa);
3053
3054                 list_del(&ifa->if_list);
3055
3056                 write_unlock_bh(&idev->lock);
3057
3058                 spin_lock_bh(&ifa->state_lock);
3059                 state = ifa->state;
3060                 ifa->state = INET6_IFADDR_STATE_DEAD;
3061                 spin_unlock_bh(&ifa->state_lock);
3062
3063                 if (state != INET6_IFADDR_STATE_DEAD) {
3064                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3065                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3066                 }
3067                 in6_ifa_put(ifa);
3068
3069                 write_lock_bh(&idev->lock);
3070         }
3071
3072         write_unlock_bh(&idev->lock);
3073
3074         /* Step 5: Discard multicast list */
3075         if (how)
3076                 ipv6_mc_destroy_dev(idev);
3077         else
3078                 ipv6_mc_down(idev);
3079
3080         idev->tstamp = jiffies;
3081
3082         /* Last: Shot the device (if unregistered) */
3083         if (how) {
3084                 addrconf_sysctl_unregister(idev);
3085                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3086                 neigh_ifdown(&nd_tbl, dev);
3087                 in6_dev_put(idev);
3088         }
3089         return 0;
3090 }
3091
3092 static void addrconf_rs_timer(unsigned long data)
3093 {
3094         struct inet6_dev *idev = (struct inet6_dev *)data;
3095         struct net_device *dev = idev->dev;
3096         struct in6_addr lladdr;
3097
3098         write_lock(&idev->lock);
3099         if (idev->dead || !(idev->if_flags & IF_READY))
3100                 goto out;
3101
3102         if (!ipv6_accept_ra(idev))
3103                 goto out;
3104
3105         /* Announcement received after solicitation was sent */
3106         if (idev->if_flags & IF_RA_RCVD)
3107                 goto out;
3108
3109         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3110                 write_unlock(&idev->lock);
3111                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3112                         ndisc_send_rs(dev, &lladdr,
3113                                       &in6addr_linklocal_allrouters);
3114                 else
3115                         goto put;
3116
3117                 write_lock(&idev->lock);
3118                 /* The wait after the last probe can be shorter */
3119                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3120                                              idev->cnf.rtr_solicits) ?
3121                                       idev->cnf.rtr_solicit_delay :
3122                                       idev->cnf.rtr_solicit_interval);
3123         } else {
3124                 /*
3125                  * Note: we do not support deprecated "all on-link"
3126                  * assumption any longer.
3127                  */
3128                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3129         }
3130
3131 out:
3132         write_unlock(&idev->lock);
3133 put:
3134         in6_dev_put(idev);
3135 }
3136
3137 /*
3138  *      Duplicate Address Detection
3139  */
3140 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3141 {
3142         unsigned long rand_num;
3143         struct inet6_dev *idev = ifp->idev;
3144
3145         if (ifp->flags & IFA_F_OPTIMISTIC)
3146                 rand_num = 0;
3147         else
3148                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3149
3150         ifp->dad_probes = idev->cnf.dad_transmits;
3151         addrconf_mod_dad_timer(ifp, rand_num);
3152 }
3153
3154 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3155 {
3156         struct inet6_dev *idev = ifp->idev;
3157         struct net_device *dev = idev->dev;
3158
3159         addrconf_join_solict(dev, &ifp->addr);
3160
3161         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3162
3163         read_lock_bh(&idev->lock);
3164         spin_lock(&ifp->lock);
3165         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3166                 goto out;
3167
3168         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3169             idev->cnf.accept_dad < 1 ||
3170             !(ifp->flags&IFA_F_TENTATIVE) ||
3171             ifp->flags & IFA_F_NODAD) {
3172                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3173                 spin_unlock(&ifp->lock);
3174                 read_unlock_bh(&idev->lock);
3175
3176                 addrconf_dad_completed(ifp);
3177                 return;
3178         }
3179
3180         if (!(idev->if_flags & IF_READY)) {
3181                 spin_unlock(&ifp->lock);
3182                 read_unlock_bh(&idev->lock);
3183                 /*
3184                  * If the device is not ready:
3185                  * - keep it tentative if it is a permanent address.
3186                  * - otherwise, kill it.
3187                  */
3188                 in6_ifa_hold(ifp);
3189                 addrconf_dad_stop(ifp, 0);
3190                 return;
3191         }
3192
3193         /*
3194          * Optimistic nodes can start receiving
3195          * Frames right away
3196          */
3197         if (ifp->flags & IFA_F_OPTIMISTIC)
3198                 ip6_ins_rt(ifp->rt);
3199
3200         addrconf_dad_kick(ifp);
3201 out:
3202         spin_unlock(&ifp->lock);
3203         read_unlock_bh(&idev->lock);
3204 }
3205
3206 static void addrconf_dad_timer(unsigned long data)
3207 {
3208         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3209         struct inet6_dev *idev = ifp->idev;
3210         struct in6_addr mcaddr;
3211
3212         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3213                 goto out;
3214
3215         write_lock(&idev->lock);
3216         if (idev->dead || !(idev->if_flags & IF_READY)) {
3217                 write_unlock(&idev->lock);
3218                 goto out;
3219         }
3220
3221         spin_lock(&ifp->lock);
3222         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3223                 spin_unlock(&ifp->lock);
3224                 write_unlock(&idev->lock);
3225                 goto out;
3226         }
3227
3228         if (ifp->dad_probes == 0) {
3229                 /*
3230                  * DAD was successful
3231                  */
3232
3233                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3234                 spin_unlock(&ifp->lock);
3235                 write_unlock(&idev->lock);
3236
3237                 addrconf_dad_completed(ifp);
3238
3239                 goto out;
3240         }
3241
3242         ifp->dad_probes--;
3243         addrconf_mod_dad_timer(ifp,
3244                                NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3245         spin_unlock(&ifp->lock);
3246         write_unlock(&idev->lock);
3247
3248         /* send a neighbour solicitation for our addr */
3249         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3250         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3251 out:
3252         in6_ifa_put(ifp);
3253 }
3254
3255 /* ifp->idev must be at least read locked */
3256 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3257 {
3258         struct inet6_ifaddr *ifpiter;
3259         struct inet6_dev *idev = ifp->idev;
3260
3261         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3262                 if (ifpiter->scope > IFA_LINK)
3263                         break;
3264                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3265                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3266                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3267                     IFA_F_PERMANENT)
3268                         return false;
3269         }
3270         return true;
3271 }
3272
3273 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3274 {
3275         struct net_device *dev = ifp->idev->dev;
3276         struct in6_addr lladdr;
3277         bool send_rs, send_mld;
3278
3279         addrconf_del_dad_timer(ifp);
3280
3281         /*
3282          *      Configure the address for reception. Now it is valid.
3283          */
3284
3285         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3286
3287         /* If added prefix is link local and we are prepared to process
3288            router advertisements, start sending router solicitations.
3289          */
3290
3291         read_lock_bh(&ifp->idev->lock);
3292         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3293         send_rs = send_mld &&
3294                   ipv6_accept_ra(ifp->idev) &&
3295                   ifp->idev->cnf.rtr_solicits > 0 &&
3296                   (dev->flags&IFF_LOOPBACK) == 0;
3297         read_unlock_bh(&ifp->idev->lock);
3298
3299         /* While dad is in progress mld report's source address is in6_addrany.
3300          * Resend with proper ll now.
3301          */
3302         if (send_mld)
3303                 ipv6_mc_dad_complete(ifp->idev);
3304
3305         if (send_rs) {
3306                 /*
3307                  *      If a host as already performed a random delay
3308                  *      [...] as part of DAD [...] there is no need
3309                  *      to delay again before sending the first RS
3310                  */
3311                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3312                         return;
3313                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3314
3315                 write_lock_bh(&ifp->idev->lock);
3316                 spin_lock(&ifp->lock);
3317                 ifp->idev->rs_probes = 1;
3318                 ifp->idev->if_flags |= IF_RS_SENT;
3319                 addrconf_mod_rs_timer(ifp->idev,
3320                                       ifp->idev->cnf.rtr_solicit_interval);
3321                 spin_unlock(&ifp->lock);
3322                 write_unlock_bh(&ifp->idev->lock);
3323         }
3324 }
3325
3326 static void addrconf_dad_run(struct inet6_dev *idev)
3327 {
3328         struct inet6_ifaddr *ifp;
3329
3330         read_lock_bh(&idev->lock);
3331         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3332                 spin_lock(&ifp->lock);
3333                 if (ifp->flags & IFA_F_TENTATIVE &&
3334                     ifp->state == INET6_IFADDR_STATE_DAD)
3335                         addrconf_dad_kick(ifp);
3336                 spin_unlock(&ifp->lock);
3337         }
3338         read_unlock_bh(&idev->lock);
3339 }
3340
3341 #ifdef CONFIG_PROC_FS
3342 struct if6_iter_state {
3343         struct seq_net_private p;
3344         int bucket;
3345         int offset;
3346 };
3347
3348 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3349 {
3350         struct inet6_ifaddr *ifa = NULL;
3351         struct if6_iter_state *state = seq->private;
3352         struct net *net = seq_file_net(seq);
3353         int p = 0;
3354
3355         /* initial bucket if pos is 0 */
3356         if (pos == 0) {
3357                 state->bucket = 0;
3358                 state->offset = 0;
3359         }
3360
3361         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3362                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3363                                          addr_lst) {
3364                         if (!net_eq(dev_net(ifa->idev->dev), net))
3365                                 continue;
3366                         /* sync with offset */
3367                         if (p < state->offset) {
3368                                 p++;
3369                                 continue;
3370                         }
3371                         state->offset++;
3372                         return ifa;
3373                 }
3374
3375                 /* prepare for next bucket */
3376                 state->offset = 0;
3377                 p = 0;
3378         }
3379         return NULL;
3380 }
3381
3382 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3383                                          struct inet6_ifaddr *ifa)
3384 {
3385         struct if6_iter_state *state = seq->private;
3386         struct net *net = seq_file_net(seq);
3387
3388         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3389                 if (!net_eq(dev_net(ifa->idev->dev), net))
3390                         continue;
3391                 state->offset++;
3392                 return ifa;
3393         }
3394
3395         while (++state->bucket < IN6_ADDR_HSIZE) {
3396                 state->offset = 0;
3397                 hlist_for_each_entry_rcu_bh(ifa,
3398                                      &inet6_addr_lst[state->bucket], addr_lst) {
3399                         if (!net_eq(dev_net(ifa->idev->dev), net))
3400                                 continue;
3401                         state->offset++;
3402                         return ifa;
3403                 }
3404         }
3405
3406         return NULL;
3407 }
3408
3409 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3410         __acquires(rcu_bh)
3411 {
3412         rcu_read_lock_bh();
3413         return if6_get_first(seq, *pos);
3414 }
3415
3416 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3417 {
3418         struct inet6_ifaddr *ifa;
3419
3420         ifa = if6_get_next(seq, v);
3421         ++*pos;
3422         return ifa;
3423 }
3424
3425 static void if6_seq_stop(struct seq_file *seq, void *v)
3426         __releases(rcu_bh)
3427 {
3428         rcu_read_unlock_bh();
3429 }
3430
3431 static int if6_seq_show(struct seq_file *seq, void *v)
3432 {
3433         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3434         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3435                    &ifp->addr,
3436                    ifp->idev->dev->ifindex,
3437                    ifp->prefix_len,
3438                    ifp->scope,
3439                    (u8) ifp->flags,
3440                    ifp->idev->dev->name);
3441         return 0;
3442 }
3443
3444 static const struct seq_operations if6_seq_ops = {
3445         .start  = if6_seq_start,
3446         .next   = if6_seq_next,
3447         .show   = if6_seq_show,
3448         .stop   = if6_seq_stop,
3449 };
3450
3451 static int if6_seq_open(struct inode *inode, struct file *file)
3452 {
3453         return seq_open_net(inode, file, &if6_seq_ops,
3454                             sizeof(struct if6_iter_state));
3455 }
3456
3457 static const struct file_operations if6_fops = {
3458         .owner          = THIS_MODULE,
3459         .open           = if6_seq_open,
3460         .read           = seq_read,
3461         .llseek         = seq_lseek,
3462         .release        = seq_release_net,
3463 };
3464
3465 static int __net_init if6_proc_net_init(struct net *net)
3466 {
3467         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3468                 return -ENOMEM;
3469         return 0;
3470 }
3471
3472 static void __net_exit if6_proc_net_exit(struct net *net)
3473 {
3474         remove_proc_entry("if_inet6", net->proc_net);
3475 }
3476
3477 static struct pernet_operations if6_proc_net_ops = {
3478        .init = if6_proc_net_init,
3479        .exit = if6_proc_net_exit,
3480 };
3481
3482 int __init if6_proc_init(void)
3483 {
3484         return register_pernet_subsys(&if6_proc_net_ops);
3485 }
3486
3487 void if6_proc_exit(void)
3488 {
3489         unregister_pernet_subsys(&if6_proc_net_ops);
3490 }
3491 #endif  /* CONFIG_PROC_FS */
3492
3493 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3494 /* Check if address is a home address configured on any interface. */
3495 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3496 {
3497         int ret = 0;
3498         struct inet6_ifaddr *ifp = NULL;
3499         unsigned int hash = inet6_addr_hash(addr);
3500
3501         rcu_read_lock_bh();
3502         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3503                 if (!net_eq(dev_net(ifp->idev->dev), net))
3504                         continue;
3505                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3506                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3507                         ret = 1;
3508                         break;
3509                 }
3510         }
3511         rcu_read_unlock_bh();
3512         return ret;
3513 }
3514 #endif
3515
3516 /*
3517  *      Periodic address status verification
3518  */
3519
3520 static void addrconf_verify(unsigned long foo)
3521 {
3522         unsigned long now, next, next_sec, next_sched;
3523         struct inet6_ifaddr *ifp;
3524         int i;
3525
3526         rcu_read_lock_bh();
3527         spin_lock(&addrconf_verify_lock);
3528         now = jiffies;
3529         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3530
3531         del_timer(&addr_chk_timer);
3532
3533         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3534 restart:
3535                 hlist_for_each_entry_rcu_bh(ifp,
3536                                          &inet6_addr_lst[i], addr_lst) {
3537                         unsigned long age;
3538
3539                         /* When setting preferred_lft to a value not zero or
3540                          * infinity, while valid_lft is infinity
3541                          * IFA_F_PERMANENT has a non-infinity life time.
3542                          */
3543                         if ((ifp->flags & IFA_F_PERMANENT) &&
3544                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3545                                 continue;
3546
3547                         spin_lock(&ifp->lock);
3548                         /* We try to batch several events at once. */
3549                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3550
3551                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3552                             age >= ifp->valid_lft) {
3553                                 spin_unlock(&ifp->lock);
3554                                 in6_ifa_hold(ifp);
3555                                 ipv6_del_addr(ifp);
3556                                 goto restart;
3557                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3558                                 spin_unlock(&ifp->lock);
3559                                 continue;
3560                         } else if (age >= ifp->prefered_lft) {
3561                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3562                                 int deprecate = 0;
3563
3564                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3565                                         deprecate = 1;
3566                                         ifp->flags |= IFA_F_DEPRECATED;
3567                                 }
3568
3569                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3570                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3571                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3572
3573                                 spin_unlock(&ifp->lock);
3574
3575                                 if (deprecate) {
3576                                         in6_ifa_hold(ifp);
3577
3578                                         ipv6_ifa_notify(0, ifp);
3579                                         in6_ifa_put(ifp);
3580                                         goto restart;
3581                                 }
3582                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3583                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3584                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3585                                         ifp->idev->cnf.dad_transmits *
3586                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3587
3588                                 if (age >= ifp->prefered_lft - regen_advance) {
3589                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3590                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3591                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3592                                         if (!ifp->regen_count && ifpub) {
3593                                                 ifp->regen_count++;
3594                                                 in6_ifa_hold(ifp);
3595                                                 in6_ifa_hold(ifpub);
3596                                                 spin_unlock(&ifp->lock);
3597
3598                                                 spin_lock(&ifpub->lock);
3599                                                 ifpub->regen_count = 0;
3600                                                 spin_unlock(&ifpub->lock);
3601                                                 ipv6_create_tempaddr(ifpub, ifp);
3602                                                 in6_ifa_put(ifpub);
3603                                                 in6_ifa_put(ifp);
3604                                                 goto restart;
3605                                         }
3606                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3607                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3608                                 spin_unlock(&ifp->lock);
3609                         } else {
3610                                 /* ifp->prefered_lft <= ifp->valid_lft */
3611                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3612                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3613                                 spin_unlock(&ifp->lock);
3614                         }
3615                 }
3616         }
3617
3618         next_sec = round_jiffies_up(next);
3619         next_sched = next;
3620
3621         /* If rounded timeout is accurate enough, accept it. */
3622         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3623                 next_sched = next_sec;
3624
3625         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3626         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3627                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3628
3629         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3630               now, next, next_sec, next_sched);
3631
3632         addr_chk_timer.expires = next_sched;
3633         add_timer(&addr_chk_timer);
3634         spin_unlock(&addrconf_verify_lock);
3635         rcu_read_unlock_bh();
3636 }
3637
3638 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3639                                      struct in6_addr **peer_pfx)
3640 {
3641         struct in6_addr *pfx = NULL;
3642
3643         *peer_pfx = NULL;
3644
3645         if (addr)
3646                 pfx = nla_data(addr);
3647
3648         if (local) {
3649                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3650                         *peer_pfx = pfx;
3651                 pfx = nla_data(local);
3652         }
3653
3654         return pfx;
3655 }
3656
3657 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3658         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3659         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3660         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3661         [IFA_FLAGS]             = { .len = sizeof(u32) },
3662 };
3663
3664 static int
3665 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3666 {
3667         struct net *net = sock_net(skb->sk);
3668         struct ifaddrmsg *ifm;
3669         struct nlattr *tb[IFA_MAX+1];
3670         struct in6_addr *pfx, *peer_pfx;
3671         int err;
3672
3673         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3674         if (err < 0)
3675                 return err;
3676
3677         ifm = nlmsg_data(nlh);
3678         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3679         if (pfx == NULL)
3680                 return -EINVAL;
3681
3682         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3683 }
3684
3685 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3686                              u32 prefered_lft, u32 valid_lft)
3687 {
3688         u32 flags;
3689         clock_t expires;
3690         unsigned long timeout;
3691         bool was_managetempaddr;
3692         bool had_prefixroute;
3693
3694         if (!valid_lft || (prefered_lft > valid_lft))
3695                 return -EINVAL;
3696
3697         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3698             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3699                 return -EINVAL;
3700
3701         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3702         if (addrconf_finite_timeout(timeout)) {
3703                 expires = jiffies_to_clock_t(timeout * HZ);
3704                 valid_lft = timeout;
3705                 flags = RTF_EXPIRES;
3706         } else {
3707                 expires = 0;
3708                 flags = 0;
3709                 ifa_flags |= IFA_F_PERMANENT;
3710         }
3711
3712         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3713         if (addrconf_finite_timeout(timeout)) {
3714                 if (timeout == 0)
3715                         ifa_flags |= IFA_F_DEPRECATED;
3716                 prefered_lft = timeout;
3717         }
3718
3719         spin_lock_bh(&ifp->lock);
3720         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3721         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3722                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
3723         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3724                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3725                         IFA_F_NOPREFIXROUTE);
3726         ifp->flags |= ifa_flags;
3727         ifp->tstamp = jiffies;
3728         ifp->valid_lft = valid_lft;
3729         ifp->prefered_lft = prefered_lft;
3730
3731         spin_unlock_bh(&ifp->lock);
3732         if (!(ifp->flags&IFA_F_TENTATIVE))
3733                 ipv6_ifa_notify(0, ifp);
3734
3735         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3736                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3737                                       expires, flags);
3738         } else if (had_prefixroute) {
3739                 enum cleanup_prefix_rt_t action;
3740                 unsigned long rt_expires;
3741
3742                 write_lock_bh(&ifp->idev->lock);
3743                 action = check_cleanup_prefix_route(ifp, &rt_expires);
3744                 write_unlock_bh(&ifp->idev->lock);
3745
3746                 if (action != CLEANUP_PREFIX_RT_NOP) {
3747                         cleanup_prefix_route(ifp, rt_expires,
3748                                 action == CLEANUP_PREFIX_RT_DEL);
3749                 }
3750         }
3751
3752         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3753                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3754                         valid_lft = prefered_lft = 0;
3755                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3756                                  !was_managetempaddr, jiffies);
3757         }
3758
3759         addrconf_verify(0);
3760
3761         return 0;
3762 }
3763
3764 static int
3765 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3766 {
3767         struct net *net = sock_net(skb->sk);
3768         struct ifaddrmsg *ifm;
3769         struct nlattr *tb[IFA_MAX+1];
3770         struct in6_addr *pfx, *peer_pfx;
3771         struct inet6_ifaddr *ifa;
3772         struct net_device *dev;
3773         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3774         u32 ifa_flags;
3775         int err;
3776
3777         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3778         if (err < 0)
3779                 return err;
3780
3781         ifm = nlmsg_data(nlh);
3782         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3783         if (pfx == NULL)
3784                 return -EINVAL;
3785
3786         if (tb[IFA_CACHEINFO]) {
3787                 struct ifa_cacheinfo *ci;
3788
3789                 ci = nla_data(tb[IFA_CACHEINFO]);
3790                 valid_lft = ci->ifa_valid;
3791                 preferred_lft = ci->ifa_prefered;
3792         } else {
3793                 preferred_lft = INFINITY_LIFE_TIME;
3794                 valid_lft = INFINITY_LIFE_TIME;
3795         }
3796
3797         dev =  __dev_get_by_index(net, ifm->ifa_index);
3798         if (dev == NULL)
3799                 return -ENODEV;
3800
3801         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3802
3803         /* We ignore other flags so far. */
3804         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3805                      IFA_F_NOPREFIXROUTE;
3806
3807         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3808         if (ifa == NULL) {
3809                 /*
3810                  * It would be best to check for !NLM_F_CREATE here but
3811                  * userspace already relies on not having to provide this.
3812                  */
3813                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3814                                       ifm->ifa_prefixlen, ifa_flags,
3815                                       preferred_lft, valid_lft);
3816         }
3817
3818         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3819             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3820                 err = -EEXIST;
3821         else
3822                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3823
3824         in6_ifa_put(ifa);
3825
3826         return err;
3827 }
3828
3829 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3830                           u8 scope, int ifindex)
3831 {
3832         struct ifaddrmsg *ifm;
3833
3834         ifm = nlmsg_data(nlh);
3835         ifm->ifa_family = AF_INET6;
3836         ifm->ifa_prefixlen = prefixlen;
3837         ifm->ifa_flags = flags;
3838         ifm->ifa_scope = scope;
3839         ifm->ifa_index = ifindex;
3840 }
3841
3842 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3843                          unsigned long tstamp, u32 preferred, u32 valid)
3844 {
3845         struct ifa_cacheinfo ci;
3846
3847         ci.cstamp = cstamp_delta(cstamp);
3848         ci.tstamp = cstamp_delta(tstamp);
3849         ci.ifa_prefered = preferred;
3850         ci.ifa_valid = valid;
3851
3852         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3853 }
3854
3855 static inline int rt_scope(int ifa_scope)
3856 {
3857         if (ifa_scope & IFA_HOST)
3858                 return RT_SCOPE_HOST;
3859         else if (ifa_scope & IFA_LINK)
3860                 return RT_SCOPE_LINK;
3861         else if (ifa_scope & IFA_SITE)
3862                 return RT_SCOPE_SITE;
3863         else
3864                 return RT_SCOPE_UNIVERSE;
3865 }
3866
3867 static inline int inet6_ifaddr_msgsize(void)
3868 {
3869         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3870                + nla_total_size(16) /* IFA_LOCAL */
3871                + nla_total_size(16) /* IFA_ADDRESS */
3872                + nla_total_size(sizeof(struct ifa_cacheinfo))
3873                + nla_total_size(4)  /* IFA_FLAGS */;
3874 }
3875
3876 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3877                              u32 portid, u32 seq, int event, unsigned int flags)
3878 {
3879         struct nlmsghdr  *nlh;
3880         u32 preferred, valid;
3881
3882         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3883         if (nlh == NULL)
3884                 return -EMSGSIZE;
3885
3886         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3887                       ifa->idev->dev->ifindex);
3888
3889         if (!((ifa->flags&IFA_F_PERMANENT) &&
3890               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3891                 preferred = ifa->prefered_lft;
3892                 valid = ifa->valid_lft;
3893                 if (preferred != INFINITY_LIFE_TIME) {
3894                         long tval = (jiffies - ifa->tstamp)/HZ;
3895                         if (preferred > tval)
3896                                 preferred -= tval;
3897                         else
3898                                 preferred = 0;
3899                         if (valid != INFINITY_LIFE_TIME) {
3900                                 if (valid > tval)
3901                                         valid -= tval;
3902                                 else
3903                                         valid = 0;
3904                         }
3905                 }
3906         } else {
3907                 preferred = INFINITY_LIFE_TIME;
3908                 valid = INFINITY_LIFE_TIME;
3909         }
3910
3911         if (!ipv6_addr_any(&ifa->peer_addr)) {
3912                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3913                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3914                         goto error;
3915         } else
3916                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3917                         goto error;
3918
3919         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3920                 goto error;
3921
3922         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
3923                 goto error;
3924
3925         return nlmsg_end(skb, nlh);
3926
3927 error:
3928         nlmsg_cancel(skb, nlh);
3929         return -EMSGSIZE;
3930 }
3931
3932 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3933                                 u32 portid, u32 seq, int event, u16 flags)
3934 {
3935         struct nlmsghdr  *nlh;
3936         u8 scope = RT_SCOPE_UNIVERSE;
3937         int ifindex = ifmca->idev->dev->ifindex;
3938
3939         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3940                 scope = RT_SCOPE_SITE;
3941
3942         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3943         if (nlh == NULL)
3944                 return -EMSGSIZE;
3945
3946         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3947         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3948             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3949                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3950                 nlmsg_cancel(skb, nlh);
3951                 return -EMSGSIZE;
3952         }
3953
3954         return nlmsg_end(skb, nlh);
3955 }
3956
3957 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3958                                 u32 portid, u32 seq, int event, unsigned int flags)
3959 {
3960         struct nlmsghdr  *nlh;
3961         u8 scope = RT_SCOPE_UNIVERSE;
3962         int ifindex = ifaca->aca_idev->dev->ifindex;
3963
3964         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3965                 scope = RT_SCOPE_SITE;
3966
3967         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3968         if (nlh == NULL)
3969                 return -EMSGSIZE;
3970
3971         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3972         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3973             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3974                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3975                 nlmsg_cancel(skb, nlh);
3976                 return -EMSGSIZE;
3977         }
3978
3979         return nlmsg_end(skb, nlh);
3980 }
3981
3982 enum addr_type_t {
3983         UNICAST_ADDR,
3984         MULTICAST_ADDR,
3985         ANYCAST_ADDR,
3986 };
3987
3988 /* called with rcu_read_lock() */
3989 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3990                           struct netlink_callback *cb, enum addr_type_t type,
3991                           int s_ip_idx, int *p_ip_idx)
3992 {
3993         struct ifmcaddr6 *ifmca;
3994         struct ifacaddr6 *ifaca;
3995         int err = 1;
3996         int ip_idx = *p_ip_idx;
3997
3998         read_lock_bh(&idev->lock);
3999         switch (type) {
4000         case UNICAST_ADDR: {
4001                 struct inet6_ifaddr *ifa;
4002
4003                 /* unicast address incl. temp addr */
4004                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4005                         if (++ip_idx < s_ip_idx)
4006                                 continue;
4007                         err = inet6_fill_ifaddr(skb, ifa,
4008                                                 NETLINK_CB(cb->skb).portid,
4009                                                 cb->nlh->nlmsg_seq,
4010                                                 RTM_NEWADDR,
4011                                                 NLM_F_MULTI);
4012                         if (err <= 0)
4013                                 break;
4014                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4015                 }
4016                 break;
4017         }
4018         case MULTICAST_ADDR:
4019                 /* multicast address */
4020                 for (ifmca = idev->mc_list; ifmca;
4021                      ifmca = ifmca->next, ip_idx++) {
4022                         if (ip_idx < s_ip_idx)
4023                                 continue;
4024                         err = inet6_fill_ifmcaddr(skb, ifmca,
4025                                                   NETLINK_CB(cb->skb).portid,
4026                                                   cb->nlh->nlmsg_seq,
4027                                                   RTM_GETMULTICAST,
4028                                                   NLM_F_MULTI);
4029                         if (err <= 0)
4030                                 break;
4031                 }
4032                 break;
4033         case ANYCAST_ADDR:
4034                 /* anycast address */
4035                 for (ifaca = idev->ac_list; ifaca;
4036                      ifaca = ifaca->aca_next, ip_idx++) {
4037                         if (ip_idx < s_ip_idx)
4038                                 continue;
4039                         err = inet6_fill_ifacaddr(skb, ifaca,
4040                                                   NETLINK_CB(cb->skb).portid,
4041                                                   cb->nlh->nlmsg_seq,
4042                                                   RTM_GETANYCAST,
4043                                                   NLM_F_MULTI);
4044                         if (err <= 0)
4045                                 break;
4046                 }
4047                 break;
4048         default:
4049                 break;
4050         }
4051         read_unlock_bh(&idev->lock);
4052         *p_ip_idx = ip_idx;
4053         return err;
4054 }
4055
4056 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4057                            enum addr_type_t type)
4058 {
4059         struct net *net = sock_net(skb->sk);
4060         int h, s_h;
4061         int idx, ip_idx;
4062         int s_idx, s_ip_idx;
4063         struct net_device *dev;
4064         struct inet6_dev *idev;
4065         struct hlist_head *head;
4066
4067         s_h = cb->args[0];
4068         s_idx = idx = cb->args[1];
4069         s_ip_idx = ip_idx = cb->args[2];
4070
4071         rcu_read_lock();
4072         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4073         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4074                 idx = 0;
4075                 head = &net->dev_index_head[h];
4076                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4077                         if (idx < s_idx)
4078                                 goto cont;
4079                         if (h > s_h || idx > s_idx)
4080                                 s_ip_idx = 0;
4081                         ip_idx = 0;
4082                         idev = __in6_dev_get(dev);
4083                         if (!idev)
4084                                 goto cont;
4085
4086                         if (in6_dump_addrs(idev, skb, cb, type,
4087                                            s_ip_idx, &ip_idx) <= 0)
4088                                 goto done;
4089 cont:
4090                         idx++;
4091                 }
4092         }
4093 done:
4094         rcu_read_unlock();
4095         cb->args[0] = h;
4096         cb->args[1] = idx;
4097         cb->args[2] = ip_idx;
4098
4099         return skb->len;
4100 }
4101
4102 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4103 {
4104         enum addr_type_t type = UNICAST_ADDR;
4105
4106         return inet6_dump_addr(skb, cb, type);
4107 }
4108
4109 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4110 {
4111         enum addr_type_t type = MULTICAST_ADDR;
4112
4113         return inet6_dump_addr(skb, cb, type);
4114 }
4115
4116
4117 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4118 {
4119         enum addr_type_t type = ANYCAST_ADDR;
4120
4121         return inet6_dump_addr(skb, cb, type);
4122 }
4123
4124 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4125 {
4126         struct net *net = sock_net(in_skb->sk);
4127         struct ifaddrmsg *ifm;
4128         struct nlattr *tb[IFA_MAX+1];
4129         struct in6_addr *addr = NULL, *peer;
4130         struct net_device *dev = NULL;
4131         struct inet6_ifaddr *ifa;
4132         struct sk_buff *skb;
4133         int err;
4134
4135         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4136         if (err < 0)
4137                 goto errout;
4138
4139         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4140         if (addr == NULL) {
4141                 err = -EINVAL;
4142                 goto errout;
4143         }
4144
4145         ifm = nlmsg_data(nlh);
4146         if (ifm->ifa_index)
4147                 dev = __dev_get_by_index(net, ifm->ifa_index);
4148
4149         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4150         if (!ifa) {
4151                 err = -EADDRNOTAVAIL;
4152                 goto errout;
4153         }
4154
4155         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4156         if (!skb) {
4157                 err = -ENOBUFS;
4158                 goto errout_ifa;
4159         }
4160
4161         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4162                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4163         if (err < 0) {
4164                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4165                 WARN_ON(err == -EMSGSIZE);
4166                 kfree_skb(skb);
4167                 goto errout_ifa;
4168         }
4169         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4170 errout_ifa:
4171         in6_ifa_put(ifa);
4172 errout:
4173         return err;
4174 }
4175
4176 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4177 {
4178         struct sk_buff *skb;
4179         struct net *net = dev_net(ifa->idev->dev);
4180         int err = -ENOBUFS;
4181
4182         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4183         if (skb == NULL)
4184                 goto errout;
4185
4186         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4187         if (err < 0) {
4188                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4189                 WARN_ON(err == -EMSGSIZE);
4190                 kfree_skb(skb);
4191                 goto errout;
4192         }
4193         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4194         return;
4195 errout:
4196         if (err < 0)
4197                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4198 }
4199
4200 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4201                                 __s32 *array, int bytes)
4202 {
4203         BUG_ON(bytes < (DEVCONF_MAX * 4));
4204
4205         memset(array, 0, bytes);
4206         array[DEVCONF_FORWARDING] = cnf->forwarding;
4207         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4208         array[DEVCONF_MTU6] = cnf->mtu6;
4209         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4210         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4211         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4212         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4213         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4214         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4215                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4216         array[DEVCONF_RTR_SOLICIT_DELAY] =
4217                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4218         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4219         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4220                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4221         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4222                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4223         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4224         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4225         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4226         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4227         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4228         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4229         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4230         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4231 #ifdef CONFIG_IPV6_ROUTER_PREF
4232         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4233         array[DEVCONF_RTR_PROBE_INTERVAL] =
4234                 jiffies_to_msecs(cnf->rtr_probe_interval);
4235 #ifdef CONFIG_IPV6_ROUTE_INFO
4236         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4237 #endif
4238 #endif
4239         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4240         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4241 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4242         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4243 #endif
4244 #ifdef CONFIG_IPV6_MROUTE
4245         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4246 #endif
4247         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4248         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4249         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4250         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4251         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4252 }
4253
4254 static inline size_t inet6_ifla6_size(void)
4255 {
4256         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4257              + nla_total_size(sizeof(struct ifla_cacheinfo))
4258              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4259              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4260              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4261              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4262 }
4263
4264 static inline size_t inet6_if_nlmsg_size(void)
4265 {
4266         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4267                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4268                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4269                + nla_total_size(4) /* IFLA_MTU */
4270                + nla_total_size(4) /* IFLA_LINK */
4271                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4272 }
4273
4274 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4275                                       int items, int bytes)
4276 {
4277         int i;
4278         int pad = bytes - sizeof(u64) * items;
4279         BUG_ON(pad < 0);
4280
4281         /* Use put_unaligned() because stats may not be aligned for u64. */
4282         put_unaligned(items, &stats[0]);
4283         for (i = 1; i < items; i++)
4284                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4285
4286         memset(&stats[items], 0, pad);
4287 }
4288
4289 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4290                                       int items, int bytes, size_t syncpoff)
4291 {
4292         int i;
4293         int pad = bytes - sizeof(u64) * items;
4294         BUG_ON(pad < 0);
4295
4296         /* Use put_unaligned() because stats may not be aligned for u64. */
4297         put_unaligned(items, &stats[0]);
4298         for (i = 1; i < items; i++)
4299                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4300
4301         memset(&stats[items], 0, pad);
4302 }
4303
4304 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4305                              int bytes)
4306 {
4307         switch (attrtype) {
4308         case IFLA_INET6_STATS:
4309                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4310                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4311                 break;
4312         case IFLA_INET6_ICMP6STATS:
4313                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4314                 break;
4315         }
4316 }
4317
4318 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4319 {
4320         struct nlattr *nla;
4321         struct ifla_cacheinfo ci;
4322
4323         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4324                 goto nla_put_failure;
4325         ci.max_reasm_len = IPV6_MAXPLEN;
4326         ci.tstamp = cstamp_delta(idev->tstamp);
4327         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4328         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4329         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4330                 goto nla_put_failure;
4331         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4332         if (nla == NULL)
4333                 goto nla_put_failure;
4334         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4335
4336         /* XXX - MC not implemented */
4337
4338         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4339         if (nla == NULL)
4340                 goto nla_put_failure;
4341         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4342
4343         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4344         if (nla == NULL)
4345                 goto nla_put_failure;
4346         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4347
4348         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4349         if (nla == NULL)
4350                 goto nla_put_failure;
4351         read_lock_bh(&idev->lock);
4352         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4353         read_unlock_bh(&idev->lock);
4354
4355         return 0;
4356
4357 nla_put_failure:
4358         return -EMSGSIZE;
4359 }
4360
4361 static size_t inet6_get_link_af_size(const struct net_device *dev)
4362 {
4363         if (!__in6_dev_get(dev))
4364                 return 0;
4365
4366         return inet6_ifla6_size();
4367 }
4368
4369 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4370 {
4371         struct inet6_dev *idev = __in6_dev_get(dev);
4372
4373         if (!idev)
4374                 return -ENODATA;
4375
4376         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4377                 return -EMSGSIZE;
4378
4379         return 0;
4380 }
4381
4382 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4383 {
4384         struct inet6_ifaddr *ifp;
4385         struct net_device *dev = idev->dev;
4386         bool update_rs = false;
4387         struct in6_addr ll_addr;
4388
4389         if (token == NULL)
4390                 return -EINVAL;
4391         if (ipv6_addr_any(token))
4392                 return -EINVAL;
4393         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4394                 return -EINVAL;
4395         if (!ipv6_accept_ra(idev))
4396                 return -EINVAL;
4397         if (idev->cnf.rtr_solicits <= 0)
4398                 return -EINVAL;
4399
4400         write_lock_bh(&idev->lock);
4401
4402         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4403         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4404
4405         write_unlock_bh(&idev->lock);
4406
4407         if (!idev->dead && (idev->if_flags & IF_READY) &&
4408             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4409                              IFA_F_OPTIMISTIC)) {
4410
4411                 /* If we're not ready, then normal ifup will take care
4412                  * of this. Otherwise, we need to request our rs here.
4413                  */
4414                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4415                 update_rs = true;
4416         }
4417
4418         write_lock_bh(&idev->lock);
4419
4420         if (update_rs) {
4421                 idev->if_flags |= IF_RS_SENT;
4422                 idev->rs_probes = 1;
4423                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4424         }
4425
4426         /* Well, that's kinda nasty ... */
4427         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4428                 spin_lock(&ifp->lock);
4429                 if (ifp->tokenized) {
4430                         ifp->valid_lft = 0;
4431                         ifp->prefered_lft = 0;
4432                 }
4433                 spin_unlock(&ifp->lock);
4434         }
4435
4436         write_unlock_bh(&idev->lock);
4437         addrconf_verify(0);
4438         return 0;
4439 }
4440
4441 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4442 {
4443         int err = -EINVAL;
4444         struct inet6_dev *idev = __in6_dev_get(dev);
4445         struct nlattr *tb[IFLA_INET6_MAX + 1];
4446
4447         if (!idev)
4448                 return -EAFNOSUPPORT;
4449
4450         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4451                 BUG();
4452
4453         if (tb[IFLA_INET6_TOKEN])
4454                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4455
4456         return err;
4457 }
4458
4459 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4460                              u32 portid, u32 seq, int event, unsigned int flags)
4461 {
4462         struct net_device *dev = idev->dev;
4463         struct ifinfomsg *hdr;
4464         struct nlmsghdr *nlh;
4465         void *protoinfo;
4466
4467         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4468         if (nlh == NULL)
4469                 return -EMSGSIZE;
4470
4471         hdr = nlmsg_data(nlh);
4472         hdr->ifi_family = AF_INET6;
4473         hdr->__ifi_pad = 0;
4474         hdr->ifi_type = dev->type;
4475         hdr->ifi_index = dev->ifindex;
4476         hdr->ifi_flags = dev_get_flags(dev);
4477         hdr->ifi_change = 0;
4478
4479         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4480             (dev->addr_len &&
4481              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4482             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4483             (dev->ifindex != dev->iflink &&
4484              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4485                 goto nla_put_failure;
4486         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4487         if (protoinfo == NULL)
4488                 goto nla_put_failure;
4489
4490         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4491                 goto nla_put_failure;
4492
4493         nla_nest_end(skb, protoinfo);
4494         return nlmsg_end(skb, nlh);
4495
4496 nla_put_failure:
4497         nlmsg_cancel(skb, nlh);
4498         return -EMSGSIZE;
4499 }
4500
4501 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4502 {
4503         struct net *net = sock_net(skb->sk);
4504         int h, s_h;
4505         int idx = 0, s_idx;
4506         struct net_device *dev;
4507         struct inet6_dev *idev;
4508         struct hlist_head *head;
4509
4510         s_h = cb->args[0];
4511         s_idx = cb->args[1];
4512
4513         rcu_read_lock();
4514         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4515                 idx = 0;
4516                 head = &net->dev_index_head[h];
4517                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4518                         if (idx < s_idx)
4519                                 goto cont;
4520                         idev = __in6_dev_get(dev);
4521                         if (!idev)
4522                                 goto cont;
4523                         if (inet6_fill_ifinfo(skb, idev,
4524                                               NETLINK_CB(cb->skb).portid,
4525                                               cb->nlh->nlmsg_seq,
4526                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4527                                 goto out;
4528 cont:
4529                         idx++;
4530                 }
4531         }
4532 out:
4533         rcu_read_unlock();
4534         cb->args[1] = idx;
4535         cb->args[0] = h;
4536
4537         return skb->len;
4538 }
4539
4540 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4541 {
4542         struct sk_buff *skb;
4543         struct net *net = dev_net(idev->dev);
4544         int err = -ENOBUFS;
4545
4546         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4547         if (skb == NULL)
4548                 goto errout;
4549
4550         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4551         if (err < 0) {
4552                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4553                 WARN_ON(err == -EMSGSIZE);
4554                 kfree_skb(skb);
4555                 goto errout;
4556         }
4557         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4558         return;
4559 errout:
4560         if (err < 0)
4561                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4562 }
4563
4564 static inline size_t inet6_prefix_nlmsg_size(void)
4565 {
4566         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4567                + nla_total_size(sizeof(struct in6_addr))
4568                + nla_total_size(sizeof(struct prefix_cacheinfo));
4569 }
4570
4571 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4572                              struct prefix_info *pinfo, u32 portid, u32 seq,
4573                              int event, unsigned int flags)
4574 {
4575         struct prefixmsg *pmsg;
4576         struct nlmsghdr *nlh;
4577         struct prefix_cacheinfo ci;
4578
4579         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4580         if (nlh == NULL)
4581                 return -EMSGSIZE;
4582
4583         pmsg = nlmsg_data(nlh);
4584         pmsg->prefix_family = AF_INET6;
4585         pmsg->prefix_pad1 = 0;
4586         pmsg->prefix_pad2 = 0;
4587         pmsg->prefix_ifindex = idev->dev->ifindex;
4588         pmsg->prefix_len = pinfo->prefix_len;
4589         pmsg->prefix_type = pinfo->type;
4590         pmsg->prefix_pad3 = 0;
4591         pmsg->prefix_flags = 0;
4592         if (pinfo->onlink)
4593                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4594         if (pinfo->autoconf)
4595                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4596
4597         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4598                 goto nla_put_failure;
4599         ci.preferred_time = ntohl(pinfo->prefered);
4600         ci.valid_time = ntohl(pinfo->valid);
4601         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4602                 goto nla_put_failure;
4603         return nlmsg_end(skb, nlh);
4604
4605 nla_put_failure:
4606         nlmsg_cancel(skb, nlh);
4607         return -EMSGSIZE;
4608 }
4609
4610 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4611                          struct prefix_info *pinfo)
4612 {
4613         struct sk_buff *skb;
4614         struct net *net = dev_net(idev->dev);
4615         int err = -ENOBUFS;
4616
4617         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4618         if (skb == NULL)
4619                 goto errout;
4620
4621         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4622         if (err < 0) {
4623                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4624                 WARN_ON(err == -EMSGSIZE);
4625                 kfree_skb(skb);
4626                 goto errout;
4627         }
4628         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4629         return;
4630 errout:
4631         if (err < 0)
4632                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4633 }
4634
4635 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4636 {
4637         struct net *net = dev_net(ifp->idev->dev);
4638
4639         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4640
4641         switch (event) {
4642         case RTM_NEWADDR:
4643                 /*
4644                  * If the address was optimistic
4645                  * we inserted the route at the start of
4646                  * our DAD process, so we don't need
4647                  * to do it again
4648                  */
4649                 if (!(ifp->rt->rt6i_node))
4650                         ip6_ins_rt(ifp->rt);
4651                 if (ifp->idev->cnf.forwarding)
4652                         addrconf_join_anycast(ifp);
4653                 if (!ipv6_addr_any(&ifp->peer_addr))
4654                         addrconf_prefix_route(&ifp->peer_addr, 128,
4655                                               ifp->idev->dev, 0, 0);
4656                 break;
4657         case RTM_DELADDR:
4658                 if (ifp->idev->cnf.forwarding)
4659                         addrconf_leave_anycast(ifp);
4660                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4661                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4662                         struct rt6_info *rt;
4663                         struct net_device *dev = ifp->idev->dev;
4664
4665                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4666                                         dev->ifindex, 1);
4667                         if (rt) {
4668                                 dst_hold(&rt->dst);
4669                                 if (ip6_del_rt(rt))
4670                                         dst_free(&rt->dst);
4671                         }
4672                 }
4673                 dst_hold(&ifp->rt->dst);
4674
4675                 if (ip6_del_rt(ifp->rt))
4676                         dst_free(&ifp->rt->dst);
4677                 break;
4678         }
4679         atomic_inc(&net->ipv6.dev_addr_genid);
4680         rt_genid_bump_ipv6(net);
4681 }
4682
4683 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4684 {
4685         rcu_read_lock_bh();
4686         if (likely(ifp->idev->dead == 0))
4687                 __ipv6_ifa_notify(event, ifp);
4688         rcu_read_unlock_bh();
4689 }
4690
4691 #ifdef CONFIG_SYSCTL
4692
4693 static
4694 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4695                            void __user *buffer, size_t *lenp, loff_t *ppos)
4696 {
4697         int *valp = ctl->data;
4698         int val = *valp;
4699         loff_t pos = *ppos;
4700         struct ctl_table lctl;
4701         int ret;
4702
4703         /*
4704          * ctl->data points to idev->cnf.forwarding, we should
4705          * not modify it until we get the rtnl lock.
4706          */
4707         lctl = *ctl;
4708         lctl.data = &val;
4709
4710         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4711
4712         if (write)
4713                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4714         if (ret)
4715                 *ppos = pos;
4716         return ret;
4717 }
4718
4719 static void dev_disable_change(struct inet6_dev *idev)
4720 {
4721         struct netdev_notifier_info info;
4722
4723         if (!idev || !idev->dev)
4724                 return;
4725
4726         netdev_notifier_info_init(&info, idev->dev);
4727         if (idev->cnf.disable_ipv6)
4728                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4729         else
4730                 addrconf_notify(NULL, NETDEV_UP, &info);
4731 }
4732
4733 static void addrconf_disable_change(struct net *net, __s32 newf)
4734 {
4735         struct net_device *dev;
4736         struct inet6_dev *idev;
4737
4738         rcu_read_lock();
4739         for_each_netdev_rcu(net, dev) {
4740                 idev = __in6_dev_get(dev);
4741                 if (idev) {
4742                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4743                         idev->cnf.disable_ipv6 = newf;
4744                         if (changed)
4745                                 dev_disable_change(idev);
4746                 }
4747         }
4748         rcu_read_unlock();
4749 }
4750
4751 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4752 {
4753         struct net *net;
4754         int old;
4755
4756         if (!rtnl_trylock())
4757                 return restart_syscall();
4758
4759         net = (struct net *)table->extra2;
4760         old = *p;
4761         *p = newf;
4762
4763         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4764                 rtnl_unlock();
4765                 return 0;
4766         }
4767
4768         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4769                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4770                 addrconf_disable_change(net, newf);
4771         } else if ((!newf) ^ (!old))
4772                 dev_disable_change((struct inet6_dev *)table->extra1);
4773
4774         rtnl_unlock();
4775         return 0;
4776 }
4777
4778 static
4779 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4780                             void __user *buffer, size_t *lenp, loff_t *ppos)
4781 {
4782         int *valp = ctl->data;
4783         int val = *valp;
4784         loff_t pos = *ppos;
4785         struct ctl_table lctl;
4786         int ret;
4787
4788         /*
4789          * ctl->data points to idev->cnf.disable_ipv6, we should
4790          * not modify it until we get the rtnl lock.
4791          */
4792         lctl = *ctl;
4793         lctl.data = &val;
4794
4795         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4796
4797         if (write)
4798                 ret = addrconf_disable_ipv6(ctl, valp, val);
4799         if (ret)
4800                 *ppos = pos;
4801         return ret;
4802 }
4803
4804 static
4805 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4806                               void __user *buffer, size_t *lenp, loff_t *ppos)
4807 {
4808         int *valp = ctl->data;
4809         int ret;
4810         int old, new;
4811
4812         old = *valp;
4813         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4814         new = *valp;
4815
4816         if (write && old != new) {
4817                 struct net *net = ctl->extra2;
4818
4819                 if (!rtnl_trylock())
4820                         return restart_syscall();
4821
4822                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4823                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4824                                                      NETCONFA_IFINDEX_DEFAULT,
4825                                                      net->ipv6.devconf_dflt);
4826                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4827                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4828                                                      NETCONFA_IFINDEX_ALL,
4829                                                      net->ipv6.devconf_all);
4830                 else {
4831                         struct inet6_dev *idev = ctl->extra1;
4832
4833                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4834                                                      idev->dev->ifindex,
4835                                                      &idev->cnf);
4836                 }
4837                 rtnl_unlock();
4838         }
4839
4840         return ret;
4841 }
4842
4843
4844 static struct addrconf_sysctl_table
4845 {
4846         struct ctl_table_header *sysctl_header;
4847         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4848 } addrconf_sysctl __read_mostly = {
4849         .sysctl_header = NULL,
4850         .addrconf_vars = {
4851                 {
4852                         .procname       = "forwarding",
4853                         .data           = &ipv6_devconf.forwarding,
4854                         .maxlen         = sizeof(int),
4855                         .mode           = 0644,
4856                         .proc_handler   = addrconf_sysctl_forward,
4857                 },
4858                 {
4859                         .procname       = "hop_limit",
4860                         .data           = &ipv6_devconf.hop_limit,
4861                         .maxlen         = sizeof(int),
4862                         .mode           = 0644,
4863                         .proc_handler   = proc_dointvec,
4864                 },
4865                 {
4866                         .procname       = "mtu",
4867                         .data           = &ipv6_devconf.mtu6,
4868                         .maxlen         = sizeof(int),
4869                         .mode           = 0644,
4870                         .proc_handler   = proc_dointvec,
4871                 },
4872                 {
4873                         .procname       = "accept_ra",
4874                         .data           = &ipv6_devconf.accept_ra,
4875                         .maxlen         = sizeof(int),
4876                         .mode           = 0644,
4877                         .proc_handler   = proc_dointvec,
4878                 },
4879                 {
4880                         .procname       = "accept_redirects",
4881                         .data           = &ipv6_devconf.accept_redirects,
4882                         .maxlen         = sizeof(int),
4883                         .mode           = 0644,
4884                         .proc_handler   = proc_dointvec,
4885                 },
4886                 {
4887                         .procname       = "autoconf",
4888                         .data           = &ipv6_devconf.autoconf,
4889                         .maxlen         = sizeof(int),
4890                         .mode           = 0644,
4891                         .proc_handler   = proc_dointvec,
4892                 },
4893                 {
4894                         .procname       = "dad_transmits",
4895                         .data           = &ipv6_devconf.dad_transmits,
4896                         .maxlen         = sizeof(int),
4897                         .mode           = 0644,
4898                         .proc_handler   = proc_dointvec,
4899                 },
4900                 {
4901                         .procname       = "router_solicitations",
4902                         .data           = &ipv6_devconf.rtr_solicits,
4903                         .maxlen         = sizeof(int),
4904                         .mode           = 0644,
4905                         .proc_handler   = proc_dointvec,
4906                 },
4907                 {
4908                         .procname       = "router_solicitation_interval",
4909                         .data           = &ipv6_devconf.rtr_solicit_interval,
4910                         .maxlen         = sizeof(int),
4911                         .mode           = 0644,
4912                         .proc_handler   = proc_dointvec_jiffies,
4913                 },
4914                 {
4915                         .procname       = "router_solicitation_delay",
4916                         .data           = &ipv6_devconf.rtr_solicit_delay,
4917                         .maxlen         = sizeof(int),
4918                         .mode           = 0644,
4919                         .proc_handler   = proc_dointvec_jiffies,
4920                 },
4921                 {
4922                         .procname       = "force_mld_version",
4923                         .data           = &ipv6_devconf.force_mld_version,
4924                         .maxlen         = sizeof(int),
4925                         .mode           = 0644,
4926                         .proc_handler   = proc_dointvec,
4927                 },
4928                 {
4929                         .procname       = "mldv1_unsolicited_report_interval",
4930                         .data           =
4931                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4932                         .maxlen         = sizeof(int),
4933                         .mode           = 0644,
4934                         .proc_handler   = proc_dointvec_ms_jiffies,
4935                 },
4936                 {
4937                         .procname       = "mldv2_unsolicited_report_interval",
4938                         .data           =
4939                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4940                         .maxlen         = sizeof(int),
4941                         .mode           = 0644,
4942                         .proc_handler   = proc_dointvec_ms_jiffies,
4943                 },
4944                 {
4945                         .procname       = "use_tempaddr",
4946                         .data           = &ipv6_devconf.use_tempaddr,
4947                         .maxlen         = sizeof(int),
4948                         .mode           = 0644,
4949                         .proc_handler   = proc_dointvec,
4950                 },
4951                 {
4952                         .procname       = "temp_valid_lft",
4953                         .data           = &ipv6_devconf.temp_valid_lft,
4954                         .maxlen         = sizeof(int),
4955                         .mode           = 0644,
4956                         .proc_handler   = proc_dointvec,
4957                 },
4958                 {
4959                         .procname       = "temp_prefered_lft",
4960                         .data           = &ipv6_devconf.temp_prefered_lft,
4961                         .maxlen         = sizeof(int),
4962                         .mode           = 0644,
4963                         .proc_handler   = proc_dointvec,
4964                 },
4965                 {
4966                         .procname       = "regen_max_retry",
4967                         .data           = &ipv6_devconf.regen_max_retry,
4968                         .maxlen         = sizeof(int),
4969                         .mode           = 0644,
4970                         .proc_handler   = proc_dointvec,
4971                 },
4972                 {
4973                         .procname       = "max_desync_factor",
4974                         .data           = &ipv6_devconf.max_desync_factor,
4975                         .maxlen         = sizeof(int),
4976                         .mode           = 0644,
4977                         .proc_handler   = proc_dointvec,
4978                 },
4979                 {
4980                         .procname       = "max_addresses",
4981                         .data           = &ipv6_devconf.max_addresses,
4982                         .maxlen         = sizeof(int),
4983                         .mode           = 0644,
4984                         .proc_handler   = proc_dointvec,
4985                 },
4986                 {
4987                         .procname       = "accept_ra_defrtr",
4988                         .data           = &ipv6_devconf.accept_ra_defrtr,
4989                         .maxlen         = sizeof(int),
4990                         .mode           = 0644,
4991                         .proc_handler   = proc_dointvec,
4992                 },
4993                 {
4994                         .procname       = "accept_ra_pinfo",
4995                         .data           = &ipv6_devconf.accept_ra_pinfo,
4996                         .maxlen         = sizeof(int),
4997                         .mode           = 0644,
4998                         .proc_handler   = proc_dointvec,
4999                 },
5000 #ifdef CONFIG_IPV6_ROUTER_PREF
5001                 {
5002                         .procname       = "accept_ra_rtr_pref",
5003                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5004                         .maxlen         = sizeof(int),
5005                         .mode           = 0644,
5006                         .proc_handler   = proc_dointvec,
5007                 },
5008                 {
5009                         .procname       = "router_probe_interval",
5010                         .data           = &ipv6_devconf.rtr_probe_interval,
5011                         .maxlen         = sizeof(int),
5012                         .mode           = 0644,
5013                         .proc_handler   = proc_dointvec_jiffies,
5014                 },
5015 #ifdef CONFIG_IPV6_ROUTE_INFO
5016                 {
5017                         .procname       = "accept_ra_rt_info_max_plen",
5018                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5019                         .maxlen         = sizeof(int),
5020                         .mode           = 0644,
5021                         .proc_handler   = proc_dointvec,
5022                 },
5023 #endif
5024 #endif
5025                 {
5026                         .procname       = "proxy_ndp",
5027                         .data           = &ipv6_devconf.proxy_ndp,
5028                         .maxlen         = sizeof(int),
5029                         .mode           = 0644,
5030                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5031                 },
5032                 {
5033                         .procname       = "accept_source_route",
5034                         .data           = &ipv6_devconf.accept_source_route,
5035                         .maxlen         = sizeof(int),
5036                         .mode           = 0644,
5037                         .proc_handler   = proc_dointvec,
5038                 },
5039 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5040                 {
5041                         .procname       = "optimistic_dad",
5042                         .data           = &ipv6_devconf.optimistic_dad,
5043                         .maxlen         = sizeof(int),
5044                         .mode           = 0644,
5045                         .proc_handler   = proc_dointvec,
5046
5047                 },
5048 #endif
5049 #ifdef CONFIG_IPV6_MROUTE
5050                 {
5051                         .procname       = "mc_forwarding",
5052                         .data           = &ipv6_devconf.mc_forwarding,
5053                         .maxlen         = sizeof(int),
5054                         .mode           = 0444,
5055                         .proc_handler   = proc_dointvec,
5056                 },
5057 #endif
5058                 {
5059                         .procname       = "disable_ipv6",
5060                         .data           = &ipv6_devconf.disable_ipv6,
5061                         .maxlen         = sizeof(int),
5062                         .mode           = 0644,
5063                         .proc_handler   = addrconf_sysctl_disable,
5064                 },
5065                 {
5066                         .procname       = "accept_dad",
5067                         .data           = &ipv6_devconf.accept_dad,
5068                         .maxlen         = sizeof(int),
5069                         .mode           = 0644,
5070                         .proc_handler   = proc_dointvec,
5071                 },
5072                 {
5073                         .procname       = "force_tllao",
5074                         .data           = &ipv6_devconf.force_tllao,
5075                         .maxlen         = sizeof(int),
5076                         .mode           = 0644,
5077                         .proc_handler   = proc_dointvec
5078                 },
5079                 {
5080                         .procname       = "ndisc_notify",
5081                         .data           = &ipv6_devconf.ndisc_notify,
5082                         .maxlen         = sizeof(int),
5083                         .mode           = 0644,
5084                         .proc_handler   = proc_dointvec
5085                 },
5086                 {
5087                         .procname       = "suppress_frag_ndisc",
5088                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5089                         .maxlen         = sizeof(int),
5090                         .mode           = 0644,
5091                         .proc_handler   = proc_dointvec
5092                 },
5093                 {
5094                         /* sentinel */
5095                 }
5096         },
5097 };
5098
5099 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5100                 struct inet6_dev *idev, struct ipv6_devconf *p)
5101 {
5102         int i;
5103         struct addrconf_sysctl_table *t;
5104         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5105
5106         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5107         if (t == NULL)
5108                 goto out;
5109
5110         for (i = 0; t->addrconf_vars[i].data; i++) {
5111                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5112                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5113                 t->addrconf_vars[i].extra2 = net;
5114         }
5115
5116         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5117
5118         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5119         if (t->sysctl_header == NULL)
5120                 goto free;
5121
5122         p->sysctl = t;
5123         return 0;
5124
5125 free:
5126         kfree(t);
5127 out:
5128         return -ENOBUFS;
5129 }
5130
5131 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5132 {
5133         struct addrconf_sysctl_table *t;
5134
5135         if (p->sysctl == NULL)
5136                 return;
5137
5138         t = p->sysctl;
5139         p->sysctl = NULL;
5140         unregister_net_sysctl_table(t->sysctl_header);
5141         kfree(t);
5142 }
5143
5144 static void addrconf_sysctl_register(struct inet6_dev *idev)
5145 {
5146         neigh_sysctl_register(idev->dev, idev->nd_parms,
5147                               &ndisc_ifinfo_sysctl_change);
5148         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5149                                         idev, &idev->cnf);
5150 }
5151
5152 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5153 {
5154         __addrconf_sysctl_unregister(&idev->cnf);
5155         neigh_sysctl_unregister(idev->nd_parms);
5156 }
5157
5158
5159 #endif
5160
5161 static int __net_init addrconf_init_net(struct net *net)
5162 {
5163         int err = -ENOMEM;
5164         struct ipv6_devconf *all, *dflt;
5165
5166         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5167         if (all == NULL)
5168                 goto err_alloc_all;
5169
5170         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5171         if (dflt == NULL)
5172                 goto err_alloc_dflt;
5173
5174         /* these will be inherited by all namespaces */
5175         dflt->autoconf = ipv6_defaults.autoconf;
5176         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5177
5178         net->ipv6.devconf_all = all;
5179         net->ipv6.devconf_dflt = dflt;
5180
5181 #ifdef CONFIG_SYSCTL
5182         err = __addrconf_sysctl_register(net, "all", NULL, all);
5183         if (err < 0)
5184                 goto err_reg_all;
5185
5186         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5187         if (err < 0)
5188                 goto err_reg_dflt;
5189 #endif
5190         return 0;
5191
5192 #ifdef CONFIG_SYSCTL
5193 err_reg_dflt:
5194         __addrconf_sysctl_unregister(all);
5195 err_reg_all:
5196         kfree(dflt);
5197 #endif
5198 err_alloc_dflt:
5199         kfree(all);
5200 err_alloc_all:
5201         return err;
5202 }
5203
5204 static void __net_exit addrconf_exit_net(struct net *net)
5205 {
5206 #ifdef CONFIG_SYSCTL
5207         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5208         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5209 #endif
5210         if (!net_eq(net, &init_net)) {
5211                 kfree(net->ipv6.devconf_dflt);
5212                 kfree(net->ipv6.devconf_all);
5213         }
5214 }
5215
5216 static struct pernet_operations addrconf_ops = {
5217         .init = addrconf_init_net,
5218         .exit = addrconf_exit_net,
5219 };
5220
5221 static struct rtnl_af_ops inet6_ops = {
5222         .family           = AF_INET6,
5223         .fill_link_af     = inet6_fill_link_af,
5224         .get_link_af_size = inet6_get_link_af_size,
5225         .set_link_af      = inet6_set_link_af,
5226 };
5227
5228 /*
5229  *      Init / cleanup code
5230  */
5231
5232 int __init addrconf_init(void)
5233 {
5234         int i, err;
5235
5236         err = ipv6_addr_label_init();
5237         if (err < 0) {
5238                 pr_crit("%s: cannot initialize default policy table: %d\n",
5239                         __func__, err);
5240                 goto out;
5241         }
5242
5243         err = register_pernet_subsys(&addrconf_ops);
5244         if (err < 0)
5245                 goto out_addrlabel;
5246
5247         /* The addrconf netdev notifier requires that loopback_dev
5248          * has it's ipv6 private information allocated and setup
5249          * before it can bring up and give link-local addresses
5250          * to other devices which are up.
5251          *
5252          * Unfortunately, loopback_dev is not necessarily the first
5253          * entry in the global dev_base list of net devices.  In fact,
5254          * it is likely to be the very last entry on that list.
5255          * So this causes the notifier registry below to try and
5256          * give link-local addresses to all devices besides loopback_dev
5257          * first, then loopback_dev, which cases all the non-loopback_dev
5258          * devices to fail to get a link-local address.
5259          *
5260          * So, as a temporary fix, allocate the ipv6 structure for
5261          * loopback_dev first by hand.
5262          * Longer term, all of the dependencies ipv6 has upon the loopback
5263          * device and it being up should be removed.
5264          */
5265         rtnl_lock();
5266         if (!ipv6_add_dev(init_net.loopback_dev))
5267                 err = -ENOMEM;
5268         rtnl_unlock();
5269         if (err)
5270                 goto errlo;
5271
5272         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5273                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5274
5275         register_netdevice_notifier(&ipv6_dev_notf);
5276
5277         addrconf_verify(0);
5278
5279         rtnl_af_register(&inet6_ops);
5280
5281         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5282                               NULL);
5283         if (err < 0)
5284                 goto errout;
5285
5286         /* Only the first call to __rtnl_register can fail */
5287         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5288         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5289         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5290                         inet6_dump_ifaddr, NULL);
5291         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5292                         inet6_dump_ifmcaddr, NULL);
5293         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5294                         inet6_dump_ifacaddr, NULL);
5295         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5296                         inet6_netconf_dump_devconf, NULL);
5297
5298         ipv6_addr_label_rtnl_register();
5299
5300         return 0;
5301 errout:
5302         rtnl_af_unregister(&inet6_ops);
5303         unregister_netdevice_notifier(&ipv6_dev_notf);
5304 errlo:
5305         unregister_pernet_subsys(&addrconf_ops);
5306 out_addrlabel:
5307         ipv6_addr_label_cleanup();
5308 out:
5309         return err;
5310 }
5311
5312 void addrconf_cleanup(void)
5313 {
5314         struct net_device *dev;
5315         int i;
5316
5317         unregister_netdevice_notifier(&ipv6_dev_notf);
5318         unregister_pernet_subsys(&addrconf_ops);
5319         ipv6_addr_label_cleanup();
5320
5321         rtnl_lock();
5322
5323         __rtnl_af_unregister(&inet6_ops);
5324
5325         /* clean dev list */
5326         for_each_netdev(&init_net, dev) {
5327                 if (__in6_dev_get(dev) == NULL)
5328                         continue;
5329                 addrconf_ifdown(dev, 1);
5330         }
5331         addrconf_ifdown(init_net.loopback_dev, 2);
5332
5333         /*
5334          *      Check hash table.
5335          */
5336         spin_lock_bh(&addrconf_hash_lock);
5337         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5338                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5339         spin_unlock_bh(&addrconf_hash_lock);
5340
5341         del_timer(&addr_chk_timer);
5342         rtnl_unlock();
5343 }