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

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