linux/net/ipv6/addrconf.c
<<
>>
Prefs
   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
  87#ifdef CONFIG_IPV6_PRIVACY
  88#include <linux/random.h>
  89#endif
  90
  91#include <linux/uaccess.h>
  92#include <asm/unaligned.h>
  93
  94#include <linux/proc_fs.h>
  95#include <linux/seq_file.h>
  96#include <linux/export.h>
  97
  98/* Set to 3 to get tracing... */
  99#define ACONF_DEBUG 2
 100
 101#if ACONF_DEBUG >= 3
 102#define ADBG(x) printk x
 103#else
 104#define ADBG(x)
 105#endif
 106
 107#define INFINITY_LIFE_TIME      0xFFFFFFFF
 108
 109static inline u32 cstamp_delta(unsigned long cstamp)
 110{
 111        return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112}
 113
 114#ifdef CONFIG_SYSCTL
 115static void addrconf_sysctl_register(struct inet6_dev *idev);
 116static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117#else
 118static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 119{
 120}
 121
 122static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 123{
 124}
 125#endif
 126
 127#ifdef CONFIG_IPV6_PRIVACY
 128static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130static void ipv6_regen_rndid(unsigned long data);
 131#endif
 132
 133static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134static int ipv6_count_addresses(struct inet6_dev *idev);
 135
 136/*
 137 *      Configured unicast address hash table
 138 */
 139static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 140static DEFINE_SPINLOCK(addrconf_hash_lock);
 141
 142static void addrconf_verify(unsigned long);
 143
 144static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 145static DEFINE_SPINLOCK(addrconf_verify_lock);
 146
 147static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 148static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 149
 150static void addrconf_type_change(struct net_device *dev,
 151                                 unsigned long event);
 152static int addrconf_ifdown(struct net_device *dev, int how);
 153
 154static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 155                                                  int plen,
 156                                                  const struct net_device *dev,
 157                                                  u32 flags, u32 noflags);
 158
 159static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 160static void addrconf_dad_timer(unsigned long data);
 161static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 162static void addrconf_dad_run(struct inet6_dev *idev);
 163static void addrconf_rs_timer(unsigned long data);
 164static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 166
 167static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 168                                struct prefix_info *pinfo);
 169static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 170                               struct net_device *dev);
 171
 172static struct ipv6_devconf ipv6_devconf __read_mostly = {
 173        .forwarding             = 0,
 174        .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 175        .mtu6                   = IPV6_MIN_MTU,
 176        .accept_ra              = 1,
 177        .accept_redirects       = 1,
 178        .autoconf               = 1,
 179        .force_mld_version      = 0,
 180        .dad_transmits          = 1,
 181        .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 182        .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 183        .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 184#ifdef CONFIG_IPV6_PRIVACY
 185        .use_tempaddr           = 0,
 186        .temp_valid_lft         = TEMP_VALID_LIFETIME,
 187        .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 188        .regen_max_retry        = REGEN_MAX_RETRY,
 189        .max_desync_factor      = MAX_DESYNC_FACTOR,
 190#endif
 191        .max_addresses          = IPV6_MAX_ADDRESSES,
 192        .accept_ra_defrtr       = 1,
 193        .accept_ra_pinfo        = 1,
 194#ifdef CONFIG_IPV6_ROUTER_PREF
 195        .accept_ra_rtr_pref     = 1,
 196        .rtr_probe_interval     = 60 * HZ,
 197#ifdef CONFIG_IPV6_ROUTE_INFO
 198        .accept_ra_rt_info_max_plen = 0,
 199#endif
 200#endif
 201        .proxy_ndp              = 0,
 202        .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 203        .disable_ipv6           = 0,
 204        .accept_dad             = 1,
 205};
 206
 207static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 208        .forwarding             = 0,
 209        .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 210        .mtu6                   = IPV6_MIN_MTU,
 211        .accept_ra              = 1,
 212        .accept_redirects       = 1,
 213        .autoconf               = 1,
 214        .dad_transmits          = 1,
 215        .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 216        .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 217        .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 218#ifdef CONFIG_IPV6_PRIVACY
 219        .use_tempaddr           = 0,
 220        .temp_valid_lft         = TEMP_VALID_LIFETIME,
 221        .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 222        .regen_max_retry        = REGEN_MAX_RETRY,
 223        .max_desync_factor      = MAX_DESYNC_FACTOR,
 224#endif
 225        .max_addresses          = IPV6_MAX_ADDRESSES,
 226        .accept_ra_defrtr       = 1,
 227        .accept_ra_pinfo        = 1,
 228#ifdef CONFIG_IPV6_ROUTER_PREF
 229        .accept_ra_rtr_pref     = 1,
 230        .rtr_probe_interval     = 60 * HZ,
 231#ifdef CONFIG_IPV6_ROUTE_INFO
 232        .accept_ra_rt_info_max_plen = 0,
 233#endif
 234#endif
 235        .proxy_ndp              = 0,
 236        .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 237        .disable_ipv6           = 0,
 238        .accept_dad             = 1,
 239};
 240
 241/* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 242const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 243const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 244const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 245const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 246const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 247const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 248const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 249
 250/* Check if a valid qdisc is available */
 251static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 252{
 253        return !qdisc_tx_is_noop(dev);
 254}
 255
 256static void addrconf_del_rs_timer(struct inet6_dev *idev)
 257{
 258        if (del_timer(&idev->rs_timer))
 259                __in6_dev_put(idev);
 260}
 261
 262static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 263{
 264        if (del_timer(&ifp->dad_timer))
 265                __in6_ifa_put(ifp);
 266}
 267
 268static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 269                                  unsigned long when)
 270{
 271        if (!timer_pending(&idev->rs_timer))
 272                in6_dev_hold(idev);
 273        mod_timer(&idev->rs_timer, jiffies + when);
 274}
 275
 276static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 277                                   unsigned long when)
 278{
 279        if (!timer_pending(&ifp->dad_timer))
 280                in6_ifa_hold(ifp);
 281        mod_timer(&ifp->dad_timer, jiffies + when);
 282}
 283
 284static int snmp6_alloc_dev(struct inet6_dev *idev)
 285{
 286        if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 287                          sizeof(struct ipstats_mib),
 288                          __alignof__(struct ipstats_mib)) < 0)
 289                goto err_ip;
 290        idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 291                                        GFP_KERNEL);
 292        if (!idev->stats.icmpv6dev)
 293                goto err_icmp;
 294        idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 295                                           GFP_KERNEL);
 296        if (!idev->stats.icmpv6msgdev)
 297                goto err_icmpmsg;
 298
 299        return 0;
 300
 301err_icmpmsg:
 302        kfree(idev->stats.icmpv6dev);
 303err_icmp:
 304        snmp_mib_free((void __percpu **)idev->stats.ipv6);
 305err_ip:
 306        return -ENOMEM;
 307}
 308
 309static void snmp6_free_dev(struct inet6_dev *idev)
 310{
 311        kfree(idev->stats.icmpv6msgdev);
 312        kfree(idev->stats.icmpv6dev);
 313        snmp_mib_free((void __percpu **)idev->stats.ipv6);
 314}
 315
 316/* Nobody refers to this device, we may destroy it. */
 317
 318void in6_dev_finish_destroy(struct inet6_dev *idev)
 319{
 320        struct net_device *dev = idev->dev;
 321
 322        WARN_ON(!list_empty(&idev->addr_list));
 323        WARN_ON(idev->mc_list != NULL);
 324        WARN_ON(timer_pending(&idev->rs_timer));
 325
 326#ifdef NET_REFCNT_DEBUG
 327        pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 328#endif
 329        dev_put(dev);
 330        if (!idev->dead) {
 331                pr_warn("Freeing alive inet6 device %p\n", idev);
 332                return;
 333        }
 334        snmp6_free_dev(idev);
 335        kfree_rcu(idev, rcu);
 336}
 337EXPORT_SYMBOL(in6_dev_finish_destroy);
 338
 339static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 340{
 341        struct inet6_dev *ndev;
 342
 343        ASSERT_RTNL();
 344
 345        if (dev->mtu < IPV6_MIN_MTU)
 346                return NULL;
 347
 348        ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 349
 350        if (ndev == NULL)
 351                return NULL;
 352
 353        rwlock_init(&ndev->lock);
 354        ndev->dev = dev;
 355        INIT_LIST_HEAD(&ndev->addr_list);
 356        setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 357                    (unsigned long)ndev);
 358        memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 359        ndev->cnf.mtu6 = dev->mtu;
 360        ndev->cnf.sysctl = NULL;
 361        ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 362        if (ndev->nd_parms == NULL) {
 363                kfree(ndev);
 364                return NULL;
 365        }
 366        if (ndev->cnf.forwarding)
 367                dev_disable_lro(dev);
 368        /* We refer to the device */
 369        dev_hold(dev);
 370
 371        if (snmp6_alloc_dev(ndev) < 0) {
 372                ADBG((KERN_WARNING
 373                        "%s: cannot allocate memory for statistics; dev=%s.\n",
 374                        __func__, dev->name));
 375                neigh_parms_release(&nd_tbl, ndev->nd_parms);
 376                dev_put(dev);
 377                kfree(ndev);
 378                return NULL;
 379        }
 380
 381        if (snmp6_register_dev(ndev) < 0) {
 382                ADBG((KERN_WARNING
 383                        "%s: cannot create /proc/net/dev_snmp6/%s\n",
 384                        __func__, dev->name));
 385                neigh_parms_release(&nd_tbl, ndev->nd_parms);
 386                ndev->dead = 1;
 387                in6_dev_finish_destroy(ndev);
 388                return NULL;
 389        }
 390
 391        /* One reference from device.  We must do this before
 392         * we invoke __ipv6_regen_rndid().
 393         */
 394        in6_dev_hold(ndev);
 395
 396        if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 397                ndev->cnf.accept_dad = -1;
 398
 399#if IS_ENABLED(CONFIG_IPV6_SIT)
 400        if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 401                pr_info("%s: Disabled Multicast RS\n", dev->name);
 402                ndev->cnf.rtr_solicits = 0;
 403        }
 404#endif
 405
 406#ifdef CONFIG_IPV6_PRIVACY
 407        INIT_LIST_HEAD(&ndev->tempaddr_list);
 408        setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 409        if ((dev->flags&IFF_LOOPBACK) ||
 410            dev->type == ARPHRD_TUNNEL ||
 411            dev->type == ARPHRD_TUNNEL6 ||
 412            dev->type == ARPHRD_SIT ||
 413            dev->type == ARPHRD_NONE) {
 414                ndev->cnf.use_tempaddr = -1;
 415        } else {
 416                in6_dev_hold(ndev);
 417                ipv6_regen_rndid((unsigned long) ndev);
 418        }
 419#endif
 420        ndev->token = in6addr_any;
 421
 422        if (netif_running(dev) && addrconf_qdisc_ok(dev))
 423                ndev->if_flags |= IF_READY;
 424
 425        ipv6_mc_init_dev(ndev);
 426        ndev->tstamp = jiffies;
 427        addrconf_sysctl_register(ndev);
 428        /* protected by rtnl_lock */
 429        rcu_assign_pointer(dev->ip6_ptr, ndev);
 430
 431        /* Join interface-local all-node multicast group */
 432        ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 433
 434        /* Join all-node multicast group */
 435        ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437        /* Join all-router multicast group if forwarding is set */
 438        if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 439                ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 440
 441        return ndev;
 442}
 443
 444static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 445{
 446        struct inet6_dev *idev;
 447
 448        ASSERT_RTNL();
 449
 450        idev = __in6_dev_get(dev);
 451        if (!idev) {
 452                idev = ipv6_add_dev(dev);
 453                if (!idev)
 454                        return NULL;
 455        }
 456
 457        if (dev->flags&IFF_UP)
 458                ipv6_mc_up(idev);
 459        return idev;
 460}
 461
 462static int inet6_netconf_msgsize_devconf(int type)
 463{
 464        int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 465                    + nla_total_size(4);        /* NETCONFA_IFINDEX */
 466
 467        /* type -1 is used for ALL */
 468        if (type == -1 || type == NETCONFA_FORWARDING)
 469                size += nla_total_size(4);
 470#ifdef CONFIG_IPV6_MROUTE
 471        if (type == -1 || type == NETCONFA_MC_FORWARDING)
 472                size += nla_total_size(4);
 473#endif
 474
 475        return size;
 476}
 477
 478static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 479                                      struct ipv6_devconf *devconf, u32 portid,
 480                                      u32 seq, int event, unsigned int flags,
 481                                      int type)
 482{
 483        struct nlmsghdr  *nlh;
 484        struct netconfmsg *ncm;
 485
 486        nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 487                        flags);
 488        if (nlh == NULL)
 489                return -EMSGSIZE;
 490
 491        ncm = nlmsg_data(nlh);
 492        ncm->ncm_family = AF_INET6;
 493
 494        if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 495                goto nla_put_failure;
 496
 497        /* type -1 is used for ALL */
 498        if ((type == -1 || type == NETCONFA_FORWARDING) &&
 499            nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 500                goto nla_put_failure;
 501#ifdef CONFIG_IPV6_MROUTE
 502        if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 503            nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 504                        devconf->mc_forwarding) < 0)
 505                goto nla_put_failure;
 506#endif
 507        return nlmsg_end(skb, nlh);
 508
 509nla_put_failure:
 510        nlmsg_cancel(skb, nlh);
 511        return -EMSGSIZE;
 512}
 513
 514void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 515                                  struct ipv6_devconf *devconf)
 516{
 517        struct sk_buff *skb;
 518        int err = -ENOBUFS;
 519
 520        skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 521        if (skb == NULL)
 522                goto errout;
 523
 524        err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 525                                         RTM_NEWNETCONF, 0, type);
 526        if (err < 0) {
 527                /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 528                WARN_ON(err == -EMSGSIZE);
 529                kfree_skb(skb);
 530                goto errout;
 531        }
 532        rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 533        return;
 534errout:
 535        rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 536}
 537
 538static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 539        [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 540        [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 541};
 542
 543static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 544                                     struct nlmsghdr *nlh)
 545{
 546        struct net *net = sock_net(in_skb->sk);
 547        struct nlattr *tb[NETCONFA_MAX+1];
 548        struct netconfmsg *ncm;
 549        struct sk_buff *skb;
 550        struct ipv6_devconf *devconf;
 551        struct inet6_dev *in6_dev;
 552        struct net_device *dev;
 553        int ifindex;
 554        int err;
 555
 556        err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 557                          devconf_ipv6_policy);
 558        if (err < 0)
 559                goto errout;
 560
 561        err = EINVAL;
 562        if (!tb[NETCONFA_IFINDEX])
 563                goto errout;
 564
 565        ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 566        switch (ifindex) {
 567        case NETCONFA_IFINDEX_ALL:
 568                devconf = net->ipv6.devconf_all;
 569                break;
 570        case NETCONFA_IFINDEX_DEFAULT:
 571                devconf = net->ipv6.devconf_dflt;
 572                break;
 573        default:
 574                dev = __dev_get_by_index(net, ifindex);
 575                if (dev == NULL)
 576                        goto errout;
 577                in6_dev = __in6_dev_get(dev);
 578                if (in6_dev == NULL)
 579                        goto errout;
 580                devconf = &in6_dev->cnf;
 581                break;
 582        }
 583
 584        err = -ENOBUFS;
 585        skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 586        if (skb == NULL)
 587                goto errout;
 588
 589        err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 590                                         NETLINK_CB(in_skb).portid,
 591                                         nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 592                                         -1);
 593        if (err < 0) {
 594                /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 595                WARN_ON(err == -EMSGSIZE);
 596                kfree_skb(skb);
 597                goto errout;
 598        }
 599        err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 600errout:
 601        return err;
 602}
 603
 604static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 605                                      struct netlink_callback *cb)
 606{
 607        struct net *net = sock_net(skb->sk);
 608        int h, s_h;
 609        int idx, s_idx;
 610        struct net_device *dev;
 611        struct inet6_dev *idev;
 612        struct hlist_head *head;
 613
 614        s_h = cb->args[0];
 615        s_idx = idx = cb->args[1];
 616
 617        for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 618                idx = 0;
 619                head = &net->dev_index_head[h];
 620                rcu_read_lock();
 621                cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 622                          net->dev_base_seq;
 623                hlist_for_each_entry_rcu(dev, head, index_hlist) {
 624                        if (idx < s_idx)
 625                                goto cont;
 626                        idev = __in6_dev_get(dev);
 627                        if (!idev)
 628                                goto cont;
 629
 630                        if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 631                                                       &idev->cnf,
 632                                                       NETLINK_CB(cb->skb).portid,
 633                                                       cb->nlh->nlmsg_seq,
 634                                                       RTM_NEWNETCONF,
 635                                                       NLM_F_MULTI,
 636                                                       -1) <= 0) {
 637                                rcu_read_unlock();
 638                                goto done;
 639                        }
 640                        nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 641cont:
 642                        idx++;
 643                }
 644                rcu_read_unlock();
 645        }
 646        if (h == NETDEV_HASHENTRIES) {
 647                if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 648                                               net->ipv6.devconf_all,
 649                                               NETLINK_CB(cb->skb).portid,
 650                                               cb->nlh->nlmsg_seq,
 651                                               RTM_NEWNETCONF, NLM_F_MULTI,
 652                                               -1) <= 0)
 653                        goto done;
 654                else
 655                        h++;
 656        }
 657        if (h == NETDEV_HASHENTRIES + 1) {
 658                if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 659                                               net->ipv6.devconf_dflt,
 660                                               NETLINK_CB(cb->skb).portid,
 661                                               cb->nlh->nlmsg_seq,
 662                                               RTM_NEWNETCONF, NLM_F_MULTI,
 663                                               -1) <= 0)
 664                        goto done;
 665                else
 666                        h++;
 667        }
 668done:
 669        cb->args[0] = h;
 670        cb->args[1] = idx;
 671
 672        return skb->len;
 673}
 674
 675#ifdef CONFIG_SYSCTL
 676static void dev_forward_change(struct inet6_dev *idev)
 677{
 678        struct net_device *dev;
 679        struct inet6_ifaddr *ifa;
 680
 681        if (!idev)
 682                return;
 683        dev = idev->dev;
 684        if (idev->cnf.forwarding)
 685                dev_disable_lro(dev);
 686        if (dev->flags & IFF_MULTICAST) {
 687                if (idev->cnf.forwarding) {
 688                        ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 689                        ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 690                        ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 691                } else {
 692                        ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 693                        ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 694                        ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 695                }
 696        }
 697
 698        list_for_each_entry(ifa, &idev->addr_list, if_list) {
 699                if (ifa->flags&IFA_F_TENTATIVE)
 700                        continue;
 701                if (idev->cnf.forwarding)
 702                        addrconf_join_anycast(ifa);
 703                else
 704                        addrconf_leave_anycast(ifa);
 705        }
 706        inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 707                                     dev->ifindex, &idev->cnf);
 708}
 709
 710
 711static void addrconf_forward_change(struct net *net, __s32 newf)
 712{
 713        struct net_device *dev;
 714        struct inet6_dev *idev;
 715
 716        for_each_netdev(net, dev) {
 717                idev = __in6_dev_get(dev);
 718                if (idev) {
 719                        int changed = (!idev->cnf.forwarding) ^ (!newf);
 720                        idev->cnf.forwarding = newf;
 721                        if (changed)
 722                                dev_forward_change(idev);
 723                }
 724        }
 725}
 726
 727static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 728{
 729        struct net *net;
 730        int old;
 731
 732        if (!rtnl_trylock())
 733                return restart_syscall();
 734
 735        net = (struct net *)table->extra2;
 736        old = *p;
 737        *p = newf;
 738
 739        if (p == &net->ipv6.devconf_dflt->forwarding) {
 740                if ((!newf) ^ (!old))
 741                        inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 742                                                     NETCONFA_IFINDEX_DEFAULT,
 743                                                     net->ipv6.devconf_dflt);
 744                rtnl_unlock();
 745                return 0;
 746        }
 747
 748        if (p == &net->ipv6.devconf_all->forwarding) {
 749                net->ipv6.devconf_dflt->forwarding = newf;
 750                addrconf_forward_change(net, newf);
 751                if ((!newf) ^ (!old))
 752                        inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 753                                                     NETCONFA_IFINDEX_ALL,
 754                                                     net->ipv6.devconf_all);
 755        } else if ((!newf) ^ (!old))
 756                dev_forward_change((struct inet6_dev *)table->extra1);
 757        rtnl_unlock();
 758
 759        if (newf)
 760                rt6_purge_dflt_routers(net);
 761        return 1;
 762}
 763#endif
 764
 765/* Nobody refers to this ifaddr, destroy it */
 766void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 767{
 768        WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 769
 770#ifdef NET_REFCNT_DEBUG
 771        pr_debug("%s\n", __func__);
 772#endif
 773
 774        in6_dev_put(ifp->idev);
 775
 776        if (del_timer(&ifp->dad_timer))
 777                pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 778
 779        if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 780                pr_warn("Freeing alive inet6 address %p\n", ifp);
 781                return;
 782        }
 783        ip6_rt_put(ifp->rt);
 784
 785        kfree_rcu(ifp, rcu);
 786}
 787
 788static void
 789ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 790{
 791        struct list_head *p;
 792        int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 793
 794        /*
 795         * Each device address list is sorted in order of scope -
 796         * global before linklocal.
 797         */
 798        list_for_each(p, &idev->addr_list) {
 799                struct inet6_ifaddr *ifa
 800                        = list_entry(p, struct inet6_ifaddr, if_list);
 801                if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 802                        break;
 803        }
 804
 805        list_add_tail(&ifp->if_list, p);
 806}
 807
 808static u32 inet6_addr_hash(const struct in6_addr *addr)
 809{
 810        return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 811}
 812
 813/* On success it returns ifp with increased reference count */
 814
 815static struct inet6_ifaddr *
 816ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 817              const struct in6_addr *peer_addr, int pfxlen,
 818              int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 819{
 820        struct inet6_ifaddr *ifa = NULL;
 821        struct rt6_info *rt;
 822        unsigned int hash;
 823        int err = 0;
 824        int addr_type = ipv6_addr_type(addr);
 825
 826        if (addr_type == IPV6_ADDR_ANY ||
 827            addr_type & IPV6_ADDR_MULTICAST ||
 828            (!(idev->dev->flags & IFF_LOOPBACK) &&
 829             addr_type & IPV6_ADDR_LOOPBACK))
 830                return ERR_PTR(-EADDRNOTAVAIL);
 831
 832        rcu_read_lock_bh();
 833        if (idev->dead) {
 834                err = -ENODEV;                  /*XXX*/
 835                goto out2;
 836        }
 837
 838        if (idev->cnf.disable_ipv6) {
 839                err = -EACCES;
 840                goto out2;
 841        }
 842
 843        spin_lock(&addrconf_hash_lock);
 844
 845        /* Ignore adding duplicate addresses on an interface */
 846        if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 847                ADBG(("ipv6_add_addr: already assigned\n"));
 848                err = -EEXIST;
 849                goto out;
 850        }
 851
 852        ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 853
 854        if (ifa == NULL) {
 855                ADBG(("ipv6_add_addr: malloc failed\n"));
 856                err = -ENOBUFS;
 857                goto out;
 858        }
 859
 860        rt = addrconf_dst_alloc(idev, addr, false);
 861        if (IS_ERR(rt)) {
 862                err = PTR_ERR(rt);
 863                goto out;
 864        }
 865
 866        ifa->addr = *addr;
 867        if (peer_addr)
 868                ifa->peer_addr = *peer_addr;
 869
 870        spin_lock_init(&ifa->lock);
 871        spin_lock_init(&ifa->state_lock);
 872        setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 873                    (unsigned long)ifa);
 874        INIT_HLIST_NODE(&ifa->addr_lst);
 875        ifa->scope = scope;
 876        ifa->prefix_len = pfxlen;
 877        ifa->flags = flags | IFA_F_TENTATIVE;
 878        ifa->valid_lft = valid_lft;
 879        ifa->prefered_lft = prefered_lft;
 880        ifa->cstamp = ifa->tstamp = jiffies;
 881        ifa->tokenized = false;
 882
 883        ifa->rt = rt;
 884
 885        ifa->idev = idev;
 886        in6_dev_hold(idev);
 887        /* For caller */
 888        in6_ifa_hold(ifa);
 889
 890        /* Add to big hash table */
 891        hash = inet6_addr_hash(addr);
 892
 893        hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 894        spin_unlock(&addrconf_hash_lock);
 895
 896        write_lock(&idev->lock);
 897        /* Add to inet6_dev unicast addr list. */
 898        ipv6_link_dev_addr(idev, ifa);
 899
 900#ifdef CONFIG_IPV6_PRIVACY
 901        if (ifa->flags&IFA_F_TEMPORARY) {
 902                list_add(&ifa->tmp_list, &idev->tempaddr_list);
 903                in6_ifa_hold(ifa);
 904        }
 905#endif
 906
 907        in6_ifa_hold(ifa);
 908        write_unlock(&idev->lock);
 909out2:
 910        rcu_read_unlock_bh();
 911
 912        if (likely(err == 0))
 913                inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 914        else {
 915                kfree(ifa);
 916                ifa = ERR_PTR(err);
 917        }
 918
 919        return ifa;
 920out:
 921        spin_unlock(&addrconf_hash_lock);
 922        goto out2;
 923}
 924
 925/* This function wants to get referenced ifp and releases it before return */
 926
 927static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 928{
 929        struct inet6_ifaddr *ifa, *ifn;
 930        struct inet6_dev *idev = ifp->idev;
 931        int state;
 932        int deleted = 0, onlink = 0;
 933        unsigned long expires = jiffies;
 934
 935        spin_lock_bh(&ifp->state_lock);
 936        state = ifp->state;
 937        ifp->state = INET6_IFADDR_STATE_DEAD;
 938        spin_unlock_bh(&ifp->state_lock);
 939
 940        if (state == INET6_IFADDR_STATE_DEAD)
 941                goto out;
 942
 943        spin_lock_bh(&addrconf_hash_lock);
 944        hlist_del_init_rcu(&ifp->addr_lst);
 945        spin_unlock_bh(&addrconf_hash_lock);
 946
 947        write_lock_bh(&idev->lock);
 948#ifdef CONFIG_IPV6_PRIVACY
 949        if (ifp->flags&IFA_F_TEMPORARY) {
 950                list_del(&ifp->tmp_list);
 951                if (ifp->ifpub) {
 952                        in6_ifa_put(ifp->ifpub);
 953                        ifp->ifpub = NULL;
 954                }
 955                __in6_ifa_put(ifp);
 956        }
 957#endif
 958
 959        list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 960                if (ifa == ifp) {
 961                        list_del_init(&ifp->if_list);
 962                        __in6_ifa_put(ifp);
 963
 964                        if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 965                                break;
 966                        deleted = 1;
 967                        continue;
 968                } else if (ifp->flags & IFA_F_PERMANENT) {
 969                        if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 970                                              ifp->prefix_len)) {
 971                                if (ifa->flags & IFA_F_PERMANENT) {
 972                                        onlink = 1;
 973                                        if (deleted)
 974                                                break;
 975                                } else {
 976                                        unsigned long lifetime;
 977
 978                                        if (!onlink)
 979                                                onlink = -1;
 980
 981                                        spin_lock(&ifa->lock);
 982
 983                                        lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 984                                        /*
 985                                         * Note: Because this address is
 986                                         * not permanent, lifetime <
 987                                         * LONG_MAX / HZ here.
 988                                         */
 989                                        if (time_before(expires,
 990                                                        ifa->tstamp + lifetime * HZ))
 991                                                expires = ifa->tstamp + lifetime * HZ;
 992                                        spin_unlock(&ifa->lock);
 993                                }
 994                        }
 995                }
 996        }
 997        write_unlock_bh(&idev->lock);
 998
 999        addrconf_del_dad_timer(ifp);
1000
1001        ipv6_ifa_notify(RTM_DELADDR, ifp);
1002
1003        inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1004
1005        /*
1006         * Purge or update corresponding prefix
1007         *
1008         * 1) we don't purge prefix here if address was not permanent.
1009         *    prefix is managed by its own lifetime.
1010         * 2) if there're no addresses, delete prefix.
1011         * 3) if there're still other permanent address(es),
1012         *    corresponding prefix is still permanent.
1013         * 4) otherwise, update prefix lifetime to the
1014         *    longest valid lifetime among the corresponding
1015         *    addresses on the device.
1016         *    Note: subsequent RA will update lifetime.
1017         *
1018         * --yoshfuji
1019         */
1020        if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1021                struct in6_addr prefix;
1022                struct rt6_info *rt;
1023
1024                ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1025
1026                rt = addrconf_get_prefix_route(&prefix,
1027                                               ifp->prefix_len,
1028                                               ifp->idev->dev,
1029                                               0, RTF_GATEWAY | RTF_DEFAULT);
1030
1031                if (rt) {
1032                        if (onlink == 0) {
1033                                ip6_del_rt(rt);
1034                                rt = NULL;
1035                        } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1036                                rt6_set_expires(rt, expires);
1037                        }
1038                }
1039                ip6_rt_put(rt);
1040        }
1041
1042        /* clean up prefsrc entries */
1043        rt6_remove_prefsrc(ifp);
1044out:
1045        in6_ifa_put(ifp);
1046}
1047
1048#ifdef CONFIG_IPV6_PRIVACY
1049static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1050{
1051        struct inet6_dev *idev = ifp->idev;
1052        struct in6_addr addr, *tmpaddr;
1053        unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1054        unsigned long regen_advance;
1055        int tmp_plen;
1056        int ret = 0;
1057        int max_addresses;
1058        u32 addr_flags;
1059        unsigned long now = jiffies;
1060
1061        write_lock(&idev->lock);
1062        if (ift) {
1063                spin_lock_bh(&ift->lock);
1064                memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1065                spin_unlock_bh(&ift->lock);
1066                tmpaddr = &addr;
1067        } else {
1068                tmpaddr = NULL;
1069        }
1070retry:
1071        in6_dev_hold(idev);
1072        if (idev->cnf.use_tempaddr <= 0) {
1073                write_unlock(&idev->lock);
1074                pr_info("%s: use_tempaddr is disabled\n", __func__);
1075                in6_dev_put(idev);
1076                ret = -1;
1077                goto out;
1078        }
1079        spin_lock_bh(&ifp->lock);
1080        if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1081                idev->cnf.use_tempaddr = -1;    /*XXX*/
1082                spin_unlock_bh(&ifp->lock);
1083                write_unlock(&idev->lock);
1084                pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1085                        __func__);
1086                in6_dev_put(idev);
1087                ret = -1;
1088                goto out;
1089        }
1090        in6_ifa_hold(ifp);
1091        memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1092        __ipv6_try_regen_rndid(idev, tmpaddr);
1093        memcpy(&addr.s6_addr[8], idev->rndid, 8);
1094        age = (now - ifp->tstamp) / HZ;
1095        tmp_valid_lft = min_t(__u32,
1096                              ifp->valid_lft,
1097                              idev->cnf.temp_valid_lft + age);
1098        tmp_prefered_lft = min_t(__u32,
1099                                 ifp->prefered_lft,
1100                                 idev->cnf.temp_prefered_lft + age -
1101                                 idev->cnf.max_desync_factor);
1102        tmp_plen = ifp->prefix_len;
1103        max_addresses = idev->cnf.max_addresses;
1104        tmp_tstamp = ifp->tstamp;
1105        spin_unlock_bh(&ifp->lock);
1106
1107        regen_advance = idev->cnf.regen_max_retry *
1108                        idev->cnf.dad_transmits *
1109                        idev->nd_parms->retrans_time / HZ;
1110        write_unlock(&idev->lock);
1111
1112        /* A temporary address is created only if this calculated Preferred
1113         * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1114         * an implementation must not create a temporary address with a zero
1115         * Preferred Lifetime.
1116         */
1117        if (tmp_prefered_lft <= regen_advance) {
1118                in6_ifa_put(ifp);
1119                in6_dev_put(idev);
1120                ret = -1;
1121                goto out;
1122        }
1123
1124        addr_flags = IFA_F_TEMPORARY;
1125        /* set in addrconf_prefix_rcv() */
1126        if (ifp->flags & IFA_F_OPTIMISTIC)
1127                addr_flags |= IFA_F_OPTIMISTIC;
1128
1129        ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1130                            ipv6_addr_scope(&addr), addr_flags,
1131                            tmp_valid_lft, tmp_prefered_lft);
1132        if (IS_ERR(ift)) {
1133                in6_ifa_put(ifp);
1134                in6_dev_put(idev);
1135                pr_info("%s: retry temporary address regeneration\n", __func__);
1136                tmpaddr = &addr;
1137                write_lock(&idev->lock);
1138                goto retry;
1139        }
1140
1141        spin_lock_bh(&ift->lock);
1142        ift->ifpub = ifp;
1143        ift->cstamp = now;
1144        ift->tstamp = tmp_tstamp;
1145        spin_unlock_bh(&ift->lock);
1146
1147        addrconf_dad_start(ift);
1148        in6_ifa_put(ift);
1149        in6_dev_put(idev);
1150out:
1151        return ret;
1152}
1153#endif
1154
1155/*
1156 *      Choose an appropriate source address (RFC3484)
1157 */
1158enum {
1159        IPV6_SADDR_RULE_INIT = 0,
1160        IPV6_SADDR_RULE_LOCAL,
1161        IPV6_SADDR_RULE_SCOPE,
1162        IPV6_SADDR_RULE_PREFERRED,
1163#ifdef CONFIG_IPV6_MIP6
1164        IPV6_SADDR_RULE_HOA,
1165#endif
1166        IPV6_SADDR_RULE_OIF,
1167        IPV6_SADDR_RULE_LABEL,
1168#ifdef CONFIG_IPV6_PRIVACY
1169        IPV6_SADDR_RULE_PRIVACY,
1170#endif
1171        IPV6_SADDR_RULE_ORCHID,
1172        IPV6_SADDR_RULE_PREFIX,
1173        IPV6_SADDR_RULE_MAX
1174};
1175
1176struct ipv6_saddr_score {
1177        int                     rule;
1178        int                     addr_type;
1179        struct inet6_ifaddr     *ifa;
1180        DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1181        int                     scopedist;
1182        int                     matchlen;
1183};
1184
1185struct ipv6_saddr_dst {
1186        const struct in6_addr *addr;
1187        int ifindex;
1188        int scope;
1189        int label;
1190        unsigned int prefs;
1191};
1192
1193static inline int ipv6_saddr_preferred(int type)
1194{
1195        if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1196                return 1;
1197        return 0;
1198}
1199
1200static int ipv6_get_saddr_eval(struct net *net,
1201                               struct ipv6_saddr_score *score,
1202                               struct ipv6_saddr_dst *dst,
1203                               int i)
1204{
1205        int ret;
1206
1207        if (i <= score->rule) {
1208                switch (i) {
1209                case IPV6_SADDR_RULE_SCOPE:
1210                        ret = score->scopedist;
1211                        break;
1212                case IPV6_SADDR_RULE_PREFIX:
1213                        ret = score->matchlen;
1214                        break;
1215                default:
1216                        ret = !!test_bit(i, score->scorebits);
1217                }
1218                goto out;
1219        }
1220
1221        switch (i) {
1222        case IPV6_SADDR_RULE_INIT:
1223                /* Rule 0: remember if hiscore is not ready yet */
1224                ret = !!score->ifa;
1225                break;
1226        case IPV6_SADDR_RULE_LOCAL:
1227                /* Rule 1: Prefer same address */
1228                ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1229                break;
1230        case IPV6_SADDR_RULE_SCOPE:
1231                /* Rule 2: Prefer appropriate scope
1232                 *
1233                 *      ret
1234                 *       ^
1235                 *    -1 |  d 15
1236                 *    ---+--+-+---> scope
1237                 *       |
1238                 *       |             d is scope of the destination.
1239                 *  B-d  |  \
1240                 *       |   \      <- smaller scope is better if
1241                 *  B-15 |    \        if scope is enough for destinaion.
1242                 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1243                 * d-C-1 | /
1244                 *       |/         <- greater is better
1245                 *   -C  /             if scope is not enough for destination.
1246                 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1247                 *
1248                 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1249                 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1250                 * Assume B = 0 and we get C > 29.
1251                 */
1252                ret = __ipv6_addr_src_scope(score->addr_type);
1253                if (ret >= dst->scope)
1254                        ret = -ret;
1255                else
1256                        ret -= 128;     /* 30 is enough */
1257                score->scopedist = ret;
1258                break;
1259        case IPV6_SADDR_RULE_PREFERRED:
1260                /* Rule 3: Avoid deprecated and optimistic addresses */
1261                ret = ipv6_saddr_preferred(score->addr_type) ||
1262                      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1263                break;
1264#ifdef CONFIG_IPV6_MIP6
1265        case IPV6_SADDR_RULE_HOA:
1266            {
1267                /* Rule 4: Prefer home address */
1268                int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1269                ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1270                break;
1271            }
1272#endif
1273        case IPV6_SADDR_RULE_OIF:
1274                /* Rule 5: Prefer outgoing interface */
1275                ret = (!dst->ifindex ||
1276                       dst->ifindex == score->ifa->idev->dev->ifindex);
1277                break;
1278        case IPV6_SADDR_RULE_LABEL:
1279                /* Rule 6: Prefer matching label */
1280                ret = ipv6_addr_label(net,
1281                                      &score->ifa->addr, score->addr_type,
1282                                      score->ifa->idev->dev->ifindex) == dst->label;
1283                break;
1284#ifdef CONFIG_IPV6_PRIVACY
1285        case IPV6_SADDR_RULE_PRIVACY:
1286            {
1287                /* Rule 7: Prefer public address
1288                 * Note: prefer temporary address if use_tempaddr >= 2
1289                 */
1290                int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1291                                !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1292                                score->ifa->idev->cnf.use_tempaddr >= 2;
1293                ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1294                break;
1295            }
1296#endif
1297        case IPV6_SADDR_RULE_ORCHID:
1298                /* Rule 8-: Prefer ORCHID vs ORCHID or
1299                 *          non-ORCHID vs non-ORCHID
1300                 */
1301                ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1302                        ipv6_addr_orchid(dst->addr));
1303                break;
1304        case IPV6_SADDR_RULE_PREFIX:
1305                /* Rule 8: Use longest matching prefix */
1306                ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1307                if (ret > score->ifa->prefix_len)
1308                        ret = score->ifa->prefix_len;
1309                score->matchlen = ret;
1310                break;
1311        default:
1312                ret = 0;
1313        }
1314
1315        if (ret)
1316                __set_bit(i, score->scorebits);
1317        score->rule = i;
1318out:
1319        return ret;
1320}
1321
1322int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1323                       const struct in6_addr *daddr, unsigned int prefs,
1324                       struct in6_addr *saddr)
1325{
1326        struct ipv6_saddr_score scores[2],
1327                                *score = &scores[0], *hiscore = &scores[1];
1328        struct ipv6_saddr_dst dst;
1329        struct net_device *dev;
1330        int dst_type;
1331
1332        dst_type = __ipv6_addr_type(daddr);
1333        dst.addr = daddr;
1334        dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1335        dst.scope = __ipv6_addr_src_scope(dst_type);
1336        dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1337        dst.prefs = prefs;
1338
1339        hiscore->rule = -1;
1340        hiscore->ifa = NULL;
1341
1342        rcu_read_lock();
1343
1344        for_each_netdev_rcu(net, dev) {
1345                struct inet6_dev *idev;
1346
1347                /* Candidate Source Address (section 4)
1348                 *  - multicast and link-local destination address,
1349                 *    the set of candidate source address MUST only
1350                 *    include addresses assigned to interfaces
1351                 *    belonging to the same link as the outgoing
1352                 *    interface.
1353                 * (- For site-local destination addresses, the
1354                 *    set of candidate source addresses MUST only
1355                 *    include addresses assigned to interfaces
1356                 *    belonging to the same site as the outgoing
1357                 *    interface.)
1358                 */
1359                if (((dst_type & IPV6_ADDR_MULTICAST) ||
1360                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1361                    dst.ifindex && dev->ifindex != dst.ifindex)
1362                        continue;
1363
1364                idev = __in6_dev_get(dev);
1365                if (!idev)
1366                        continue;
1367
1368                read_lock_bh(&idev->lock);
1369                list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1370                        int i;
1371
1372                        /*
1373                         * - Tentative Address (RFC2462 section 5.4)
1374                         *  - A tentative address is not considered
1375                         *    "assigned to an interface" in the traditional
1376                         *    sense, unless it is also flagged as optimistic.
1377                         * - Candidate Source Address (section 4)
1378                         *  - In any case, anycast addresses, multicast
1379                         *    addresses, and the unspecified address MUST
1380                         *    NOT be included in a candidate set.
1381                         */
1382                        if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1383                            (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1384                                continue;
1385
1386                        score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1387
1388                        if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1389                                     score->addr_type & IPV6_ADDR_MULTICAST)) {
1390                                LIMIT_NETDEBUG(KERN_DEBUG
1391                                               "ADDRCONF: unspecified / multicast address "
1392                                               "assigned as unicast address on %s",
1393                                               dev->name);
1394                                continue;
1395                        }
1396
1397                        score->rule = -1;
1398                        bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1399
1400                        for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1401                                int minihiscore, miniscore;
1402
1403                                minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1404                                miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1405
1406                                if (minihiscore > miniscore) {
1407                                        if (i == IPV6_SADDR_RULE_SCOPE &&
1408                                            score->scopedist > 0) {
1409                                                /*
1410                                                 * special case:
1411                                                 * each remaining entry
1412                                                 * has too small (not enough)
1413                                                 * scope, because ifa entries
1414                                                 * are sorted by their scope
1415                                                 * values.
1416                                                 */
1417                                                goto try_nextdev;
1418                                        }
1419                                        break;
1420                                } else if (minihiscore < miniscore) {
1421                                        if (hiscore->ifa)
1422                                                in6_ifa_put(hiscore->ifa);
1423
1424                                        in6_ifa_hold(score->ifa);
1425
1426                                        swap(hiscore, score);
1427
1428                                        /* restore our iterator */
1429                                        score->ifa = hiscore->ifa;
1430
1431                                        break;
1432                                }
1433                        }
1434                }
1435try_nextdev:
1436                read_unlock_bh(&idev->lock);
1437        }
1438        rcu_read_unlock();
1439
1440        if (!hiscore->ifa)
1441                return -EADDRNOTAVAIL;
1442
1443        *saddr = hiscore->ifa->addr;
1444        in6_ifa_put(hiscore->ifa);
1445        return 0;
1446}
1447EXPORT_SYMBOL(ipv6_dev_get_saddr);
1448
1449int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1450                      unsigned char banned_flags)
1451{
1452        struct inet6_ifaddr *ifp;
1453        int err = -EADDRNOTAVAIL;
1454
1455        list_for_each_entry(ifp, &idev->addr_list, if_list) {
1456                if (ifp->scope == IFA_LINK &&
1457                    !(ifp->flags & banned_flags)) {
1458                        *addr = ifp->addr;
1459                        err = 0;
1460                        break;
1461                }
1462        }
1463        return err;
1464}
1465
1466int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1467                    unsigned char banned_flags)
1468{
1469        struct inet6_dev *idev;
1470        int err = -EADDRNOTAVAIL;
1471
1472        rcu_read_lock();
1473        idev = __in6_dev_get(dev);
1474        if (idev) {
1475                read_lock_bh(&idev->lock);
1476                err = __ipv6_get_lladdr(idev, addr, banned_flags);
1477                read_unlock_bh(&idev->lock);
1478        }
1479        rcu_read_unlock();
1480        return err;
1481}
1482
1483static int ipv6_count_addresses(struct inet6_dev *idev)
1484{
1485        int cnt = 0;
1486        struct inet6_ifaddr *ifp;
1487
1488        read_lock_bh(&idev->lock);
1489        list_for_each_entry(ifp, &idev->addr_list, if_list)
1490                cnt++;
1491        read_unlock_bh(&idev->lock);
1492        return cnt;
1493}
1494
1495int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1496                  const struct net_device *dev, int strict)
1497{
1498        struct inet6_ifaddr *ifp;
1499        unsigned int hash = inet6_addr_hash(addr);
1500
1501        rcu_read_lock_bh();
1502        hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1503                if (!net_eq(dev_net(ifp->idev->dev), net))
1504                        continue;
1505                if (ipv6_addr_equal(&ifp->addr, addr) &&
1506                    !(ifp->flags&IFA_F_TENTATIVE) &&
1507                    (dev == NULL || ifp->idev->dev == dev ||
1508                     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1509                        rcu_read_unlock_bh();
1510                        return 1;
1511                }
1512        }
1513
1514        rcu_read_unlock_bh();
1515        return 0;
1516}
1517EXPORT_SYMBOL(ipv6_chk_addr);
1518
1519static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1520                               struct net_device *dev)
1521{
1522        unsigned int hash = inet6_addr_hash(addr);
1523        struct inet6_ifaddr *ifp;
1524
1525        hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1526                if (!net_eq(dev_net(ifp->idev->dev), net))
1527                        continue;
1528                if (ipv6_addr_equal(&ifp->addr, addr)) {
1529                        if (dev == NULL || ifp->idev->dev == dev)
1530                                return true;
1531                }
1532        }
1533        return false;
1534}
1535
1536int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1537{
1538        struct inet6_dev *idev;
1539        struct inet6_ifaddr *ifa;
1540        int     onlink;
1541
1542        onlink = 0;
1543        rcu_read_lock();
1544        idev = __in6_dev_get(dev);
1545        if (idev) {
1546                read_lock_bh(&idev->lock);
1547                list_for_each_entry(ifa, &idev->addr_list, if_list) {
1548                        onlink = ipv6_prefix_equal(addr, &ifa->addr,
1549                                                   ifa->prefix_len);
1550                        if (onlink)
1551                                break;
1552                }
1553                read_unlock_bh(&idev->lock);
1554        }
1555        rcu_read_unlock();
1556        return onlink;
1557}
1558EXPORT_SYMBOL(ipv6_chk_prefix);
1559
1560struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1561                                     struct net_device *dev, int strict)
1562{
1563        struct inet6_ifaddr *ifp, *result = NULL;
1564        unsigned int hash = inet6_addr_hash(addr);
1565
1566        rcu_read_lock_bh();
1567        hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1568                if (!net_eq(dev_net(ifp->idev->dev), net))
1569                        continue;
1570                if (ipv6_addr_equal(&ifp->addr, addr)) {
1571                        if (dev == NULL || ifp->idev->dev == dev ||
1572                            !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1573                                result = ifp;
1574                                in6_ifa_hold(ifp);
1575                                break;
1576                        }
1577                }
1578        }
1579        rcu_read_unlock_bh();
1580
1581        return result;
1582}
1583
1584/* Gets referenced address, destroys ifaddr */
1585
1586static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1587{
1588        if (ifp->flags&IFA_F_PERMANENT) {
1589                spin_lock_bh(&ifp->lock);
1590                addrconf_del_dad_timer(ifp);
1591                ifp->flags |= IFA_F_TENTATIVE;
1592                if (dad_failed)
1593                        ifp->flags |= IFA_F_DADFAILED;
1594                spin_unlock_bh(&ifp->lock);
1595                if (dad_failed)
1596                        ipv6_ifa_notify(0, ifp);
1597                in6_ifa_put(ifp);
1598#ifdef CONFIG_IPV6_PRIVACY
1599        } else if (ifp->flags&IFA_F_TEMPORARY) {
1600                struct inet6_ifaddr *ifpub;
1601                spin_lock_bh(&ifp->lock);
1602                ifpub = ifp->ifpub;
1603                if (ifpub) {
1604                        in6_ifa_hold(ifpub);
1605                        spin_unlock_bh(&ifp->lock);
1606                        ipv6_create_tempaddr(ifpub, ifp);
1607                        in6_ifa_put(ifpub);
1608                } else {
1609                        spin_unlock_bh(&ifp->lock);
1610                }
1611                ipv6_del_addr(ifp);
1612#endif
1613        } else
1614                ipv6_del_addr(ifp);
1615}
1616
1617static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1618{
1619        int err = -ENOENT;
1620
1621        spin_lock(&ifp->state_lock);
1622        if (ifp->state == INET6_IFADDR_STATE_DAD) {
1623                ifp->state = INET6_IFADDR_STATE_POSTDAD;
1624                err = 0;
1625        }
1626        spin_unlock(&ifp->state_lock);
1627
1628        return err;
1629}
1630
1631void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1632{
1633        struct inet6_dev *idev = ifp->idev;
1634
1635        if (addrconf_dad_end(ifp)) {
1636                in6_ifa_put(ifp);
1637                return;
1638        }
1639
1640        net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1641                             ifp->idev->dev->name, &ifp->addr);
1642
1643        if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1644                struct in6_addr addr;
1645
1646                addr.s6_addr32[0] = htonl(0xfe800000);
1647                addr.s6_addr32[1] = 0;
1648
1649                if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1650                    ipv6_addr_equal(&ifp->addr, &addr)) {
1651                        /* DAD failed for link-local based on MAC address */
1652                        idev->cnf.disable_ipv6 = 1;
1653
1654                        pr_info("%s: IPv6 being disabled!\n",
1655                                ifp->idev->dev->name);
1656                }
1657        }
1658
1659        addrconf_dad_stop(ifp, 1);
1660}
1661
1662/* Join to solicited addr multicast group. */
1663
1664void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1665{
1666        struct in6_addr maddr;
1667
1668        if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1669                return;
1670
1671        addrconf_addr_solict_mult(addr, &maddr);
1672        ipv6_dev_mc_inc(dev, &maddr);
1673}
1674
1675void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1676{
1677        struct in6_addr maddr;
1678
1679        if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1680                return;
1681
1682        addrconf_addr_solict_mult(addr, &maddr);
1683        __ipv6_dev_mc_dec(idev, &maddr);
1684}
1685
1686static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1687{
1688        struct in6_addr addr;
1689        if (ifp->prefix_len == 127) /* RFC 6164 */
1690                return;
1691        ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1692        if (ipv6_addr_any(&addr))
1693                return;
1694        ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1695}
1696
1697static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1698{
1699        struct in6_addr addr;
1700        if (ifp->prefix_len == 127) /* RFC 6164 */
1701                return;
1702        ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1703        if (ipv6_addr_any(&addr))
1704                return;
1705        __ipv6_dev_ac_dec(ifp->idev, &addr);
1706}
1707
1708static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1709{
1710        if (dev->addr_len != ETH_ALEN)
1711                return -1;
1712        memcpy(eui, dev->dev_addr, 3);
1713        memcpy(eui + 5, dev->dev_addr + 3, 3);
1714
1715        /*
1716         * The zSeries OSA network cards can be shared among various
1717         * OS instances, but the OSA cards have only one MAC address.
1718         * This leads to duplicate address conflicts in conjunction
1719         * with IPv6 if more than one instance uses the same card.
1720         *
1721         * The driver for these cards can deliver a unique 16-bit
1722         * identifier for each instance sharing the same card.  It is
1723         * placed instead of 0xFFFE in the interface identifier.  The
1724         * "u" bit of the interface identifier is not inverted in this
1725         * case.  Hence the resulting interface identifier has local
1726         * scope according to RFC2373.
1727         */
1728        if (dev->dev_id) {
1729                eui[3] = (dev->dev_id >> 8) & 0xFF;
1730                eui[4] = dev->dev_id & 0xFF;
1731        } else {
1732                eui[3] = 0xFF;
1733                eui[4] = 0xFE;
1734                eui[0] ^= 2;
1735        }
1736        return 0;
1737}
1738
1739static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1740{
1741        if (dev->addr_len != IEEE802154_ADDR_LEN)
1742                return -1;
1743        memcpy(eui, dev->dev_addr, 8);
1744        eui[0] ^= 2;
1745        return 0;
1746}
1747
1748static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1749{
1750        union fwnet_hwaddr *ha;
1751
1752        if (dev->addr_len != FWNET_ALEN)
1753                return -1;
1754
1755        ha = (union fwnet_hwaddr *)dev->dev_addr;
1756
1757        memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1758        eui[0] ^= 2;
1759        return 0;
1760}
1761
1762static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1763{
1764        /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1765        if (dev->addr_len != ARCNET_ALEN)
1766                return -1;
1767        memset(eui, 0, 7);
1768        eui[7] = *(u8 *)dev->dev_addr;
1769        return 0;
1770}
1771
1772static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1773{
1774        if (dev->addr_len != INFINIBAND_ALEN)
1775                return -1;
1776        memcpy(eui, dev->dev_addr + 12, 8);
1777        eui[0] |= 2;
1778        return 0;
1779}
1780
1781static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1782{
1783        if (addr == 0)
1784                return -1;
1785        eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1786                  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1787                  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1788                  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1789                  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1790                  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1791        eui[1] = 0;
1792        eui[2] = 0x5E;
1793        eui[3] = 0xFE;
1794        memcpy(eui + 4, &addr, 4);
1795        return 0;
1796}
1797
1798static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1799{
1800        if (dev->priv_flags & IFF_ISATAP)
1801                return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1802        return -1;
1803}
1804
1805static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1806{
1807        return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1808}
1809
1810static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1811{
1812        switch (dev->type) {
1813        case ARPHRD_ETHER:
1814        case ARPHRD_FDDI:
1815                return addrconf_ifid_eui48(eui, dev);
1816        case ARPHRD_ARCNET:
1817                return addrconf_ifid_arcnet(eui, dev);
1818        case ARPHRD_INFINIBAND:
1819                return addrconf_ifid_infiniband(eui, dev);
1820        case ARPHRD_SIT:
1821                return addrconf_ifid_sit(eui, dev);
1822        case ARPHRD_IPGRE:
1823                return addrconf_ifid_gre(eui, dev);
1824        case ARPHRD_IEEE802154:
1825                return addrconf_ifid_eui64(eui, dev);
1826        case ARPHRD_IEEE1394:
1827                return addrconf_ifid_ieee1394(eui, dev);
1828        }
1829        return -1;
1830}
1831
1832static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1833{
1834        int err = -1;
1835        struct inet6_ifaddr *ifp;
1836
1837        read_lock_bh(&idev->lock);
1838        list_for_each_entry(ifp, &idev->addr_list, if_list) {
1839                if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1840                        memcpy(eui, ifp->addr.s6_addr+8, 8);
1841                        err = 0;
1842                        break;
1843                }
1844        }
1845        read_unlock_bh(&idev->lock);
1846        return err;
1847}
1848
1849#ifdef CONFIG_IPV6_PRIVACY
1850/* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1851static void __ipv6_regen_rndid(struct inet6_dev *idev)
1852{
1853regen:
1854        get_random_bytes(idev->rndid, sizeof(idev->rndid));
1855        idev->rndid[0] &= ~0x02;
1856
1857        /*
1858         * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1859         * check if generated address is not inappropriate
1860         *
1861         *  - Reserved subnet anycast (RFC 2526)
1862         *      11111101 11....11 1xxxxxxx
1863         *  - ISATAP (RFC4214) 6.1
1864         *      00-00-5E-FE-xx-xx-xx-xx
1865         *  - value 0
1866         *  - XXX: already assigned to an address on the device
1867         */
1868        if (idev->rndid[0] == 0xfd &&
1869            (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1870            (idev->rndid[7]&0x80))
1871                goto regen;
1872        if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1873                if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1874                        goto regen;
1875                if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1876                        goto regen;
1877        }
1878}
1879
1880static void ipv6_regen_rndid(unsigned long data)
1881{
1882        struct inet6_dev *idev = (struct inet6_dev *) data;
1883        unsigned long expires;
1884
1885        rcu_read_lock_bh();
1886        write_lock_bh(&idev->lock);
1887
1888        if (idev->dead)
1889                goto out;
1890
1891        __ipv6_regen_rndid(idev);
1892
1893        expires = jiffies +
1894                idev->cnf.temp_prefered_lft * HZ -
1895                idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1896                idev->cnf.max_desync_factor * HZ;
1897        if (time_before(expires, jiffies)) {
1898                pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1899                        __func__, idev->dev->name);
1900                goto out;
1901        }
1902
1903        if (!mod_timer(&idev->regen_timer, expires))
1904                in6_dev_hold(idev);
1905
1906out:
1907        write_unlock_bh(&idev->lock);
1908        rcu_read_unlock_bh();
1909        in6_dev_put(idev);
1910}
1911
1912static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1913{
1914        if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1915                __ipv6_regen_rndid(idev);
1916}
1917#endif
1918
1919/*
1920 *      Add prefix route.
1921 */
1922
1923static void
1924addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1925                      unsigned long expires, u32 flags)
1926{
1927        struct fib6_config cfg = {
1928                .fc_table = RT6_TABLE_PREFIX,
1929                .fc_metric = IP6_RT_PRIO_ADDRCONF,
1930                .fc_ifindex = dev->ifindex,
1931                .fc_expires = expires,
1932                .fc_dst_len = plen,
1933                .fc_flags = RTF_UP | flags,
1934                .fc_nlinfo.nl_net = dev_net(dev),
1935                .fc_protocol = RTPROT_KERNEL,
1936        };
1937
1938        cfg.fc_dst = *pfx;
1939
1940        /* Prevent useless cloning on PtP SIT.
1941           This thing is done here expecting that the whole
1942           class of non-broadcast devices need not cloning.
1943         */
1944#if IS_ENABLED(CONFIG_IPV6_SIT)
1945        if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1946                cfg.fc_flags |= RTF_NONEXTHOP;
1947#endif
1948
1949        ip6_route_add(&cfg);
1950}
1951
1952
1953static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1954                                                  int plen,
1955                                                  const struct net_device *dev,
1956                                                  u32 flags, u32 noflags)
1957{
1958        struct fib6_node *fn;
1959        struct rt6_info *rt = NULL;
1960        struct fib6_table *table;
1961
1962        table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1963        if (table == NULL)
1964                return NULL;
1965
1966        read_lock_bh(&table->tb6_lock);
1967        fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1968        if (!fn)
1969                goto out;
1970        for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1971                if (rt->dst.dev->ifindex != dev->ifindex)
1972                        continue;
1973                if ((rt->rt6i_flags & flags) != flags)
1974                        continue;
1975                if ((rt->rt6i_flags & noflags) != 0)
1976                        continue;
1977                dst_hold(&rt->dst);
1978                break;
1979        }
1980out:
1981        read_unlock_bh(&table->tb6_lock);
1982        return rt;
1983}
1984
1985
1986/* Create "default" multicast route to the interface */
1987
1988static void addrconf_add_mroute(struct net_device *dev)
1989{
1990        struct fib6_config cfg = {
1991                .fc_table = RT6_TABLE_LOCAL,
1992                .fc_metric = IP6_RT_PRIO_ADDRCONF,
1993                .fc_ifindex = dev->ifindex,
1994                .fc_dst_len = 8,
1995                .fc_flags = RTF_UP,
1996                .fc_nlinfo.nl_net = dev_net(dev),
1997        };
1998
1999        ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2000
2001        ip6_route_add(&cfg);
2002}
2003
2004#if IS_ENABLED(CONFIG_IPV6_SIT)
2005static void sit_route_add(struct net_device *dev)
2006{
2007        struct fib6_config cfg = {
2008                .fc_table = RT6_TABLE_MAIN,
2009                .fc_metric = IP6_RT_PRIO_ADDRCONF,
2010                .fc_ifindex = dev->ifindex,
2011                .fc_dst_len = 96,
2012                .fc_flags = RTF_UP | RTF_NONEXTHOP,
2013                .fc_nlinfo.nl_net = dev_net(dev),
2014        };
2015
2016        /* prefix length - 96 bits "::d.d.d.d" */
2017        ip6_route_add(&cfg);
2018}
2019#endif
2020
2021static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2022{
2023        struct inet6_dev *idev;
2024
2025        ASSERT_RTNL();
2026
2027        idev = ipv6_find_idev(dev);
2028        if (!idev)
2029                return ERR_PTR(-ENOBUFS);
2030
2031        if (idev->cnf.disable_ipv6)
2032                return ERR_PTR(-EACCES);
2033
2034        /* Add default multicast route */
2035        if (!(dev->flags & IFF_LOOPBACK))
2036                addrconf_add_mroute(dev);
2037
2038        return idev;
2039}
2040
2041void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2042{
2043        struct prefix_info *pinfo;
2044        __u32 valid_lft;
2045        __u32 prefered_lft;
2046        int addr_type;
2047        struct inet6_dev *in6_dev;
2048        struct net *net = dev_net(dev);
2049
2050        pinfo = (struct prefix_info *) opt;
2051
2052        if (len < sizeof(struct prefix_info)) {
2053                ADBG(("addrconf: prefix option too short\n"));
2054                return;
2055        }
2056
2057        /*
2058         *      Validation checks ([ADDRCONF], page 19)
2059         */
2060
2061        addr_type = ipv6_addr_type(&pinfo->prefix);
2062
2063        if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2064                return;
2065
2066        valid_lft = ntohl(pinfo->valid);
2067        prefered_lft = ntohl(pinfo->prefered);
2068
2069        if (prefered_lft > valid_lft) {
2070                net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2071                return;
2072        }
2073
2074        in6_dev = in6_dev_get(dev);
2075
2076        if (in6_dev == NULL) {
2077                net_dbg_ratelimited("addrconf: device %s not configured\n",
2078                                    dev->name);
2079                return;
2080        }
2081
2082        /*
2083         *      Two things going on here:
2084         *      1) Add routes for on-link prefixes
2085         *      2) Configure prefixes with the auto flag set
2086         */
2087
2088        if (pinfo->onlink) {
2089                struct rt6_info *rt;
2090                unsigned long rt_expires;
2091
2092                /* Avoid arithmetic overflow. Really, we could
2093                 * save rt_expires in seconds, likely valid_lft,
2094                 * but it would require division in fib gc, that it
2095                 * not good.
2096                 */
2097                if (HZ > USER_HZ)
2098                        rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2099                else
2100                        rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2101
2102                if (addrconf_finite_timeout(rt_expires))
2103                        rt_expires *= HZ;
2104
2105                rt = addrconf_get_prefix_route(&pinfo->prefix,
2106                                               pinfo->prefix_len,
2107                                               dev,
2108                                               RTF_ADDRCONF | RTF_PREFIX_RT,
2109                                               RTF_GATEWAY | RTF_DEFAULT);
2110
2111                if (rt) {
2112                        /* Autoconf prefix route */
2113                        if (valid_lft == 0) {
2114                                ip6_del_rt(rt);
2115                                rt = NULL;
2116                        } else if (addrconf_finite_timeout(rt_expires)) {
2117                                /* not infinity */
2118                                rt6_set_expires(rt, jiffies + rt_expires);
2119                        } else {
2120                                rt6_clean_expires(rt);
2121                        }
2122                } else if (valid_lft) {
2123                        clock_t expires = 0;
2124                        int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2125                        if (addrconf_finite_timeout(rt_expires)) {
2126                                /* not infinity */
2127                                flags |= RTF_EXPIRES;
2128                                expires = jiffies_to_clock_t(rt_expires);
2129                        }
2130                        addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2131                                              dev, expires, flags);
2132                }
2133                ip6_rt_put(rt);
2134        }
2135
2136        /* Try to figure out our local address for this prefix */
2137
2138        if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2139                struct inet6_ifaddr *ifp;
2140                struct in6_addr addr;
2141                int create = 0, update_lft = 0;
2142                bool tokenized = false;
2143
2144                if (pinfo->prefix_len == 64) {
2145                        memcpy(&addr, &pinfo->prefix, 8);
2146
2147                        if (!ipv6_addr_any(&in6_dev->token)) {
2148                                read_lock_bh(&in6_dev->lock);
2149                                memcpy(addr.s6_addr + 8,
2150                                       in6_dev->token.s6_addr + 8, 8);
2151                                read_unlock_bh(&in6_dev->lock);
2152                                tokenized = true;
2153                        } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2154                                   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2155                                in6_dev_put(in6_dev);
2156                                return;
2157                        }
2158                        goto ok;
2159                }
2160                net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2161                                    pinfo->prefix_len);
2162                in6_dev_put(in6_dev);
2163                return;
2164
2165ok:
2166
2167                ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2168
2169                if (ifp == NULL && valid_lft) {
2170                        int max_addresses = in6_dev->cnf.max_addresses;
2171                        u32 addr_flags = 0;
2172
2173#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2174                        if (in6_dev->cnf.optimistic_dad &&
2175                            !net->ipv6.devconf_all->forwarding && sllao)
2176                                addr_flags = IFA_F_OPTIMISTIC;
2177#endif
2178
2179                        /* Do not allow to create too much of autoconfigured
2180                         * addresses; this would be too easy way to crash kernel.
2181                         */
2182                        if (!max_addresses ||
2183                            ipv6_count_addresses(in6_dev) < max_addresses)
2184                                ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2185                                                    pinfo->prefix_len,
2186                                                    addr_type&IPV6_ADDR_SCOPE_MASK,
2187                                                    addr_flags, valid_lft,
2188                                                    prefered_lft);
2189
2190                        if (IS_ERR_OR_NULL(ifp)) {
2191                                in6_dev_put(in6_dev);
2192                                return;
2193                        }
2194
2195                        update_lft = 0;
2196                        create = 1;
2197                        ifp->cstamp = jiffies;
2198                        ifp->tokenized = tokenized;
2199                        addrconf_dad_start(ifp);
2200                }
2201
2202                if (ifp) {
2203                        int flags;
2204                        unsigned long now;
2205#ifdef CONFIG_IPV6_PRIVACY
2206                        struct inet6_ifaddr *ift;
2207#endif
2208                        u32 stored_lft;
2209
2210                        /* update lifetime (RFC2462 5.5.3 e) */
2211                        spin_lock(&ifp->lock);
2212                        now = jiffies;
2213                        if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2214                                stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2215                        else
2216                                stored_lft = 0;
2217                        if (!update_lft && !create && stored_lft) {
2218                                if (valid_lft > MIN_VALID_LIFETIME ||
2219                                    valid_lft > stored_lft)
2220                                        update_lft = 1;
2221                                else if (stored_lft <= MIN_VALID_LIFETIME) {
2222                                        /* valid_lft <= stored_lft is always true */
2223                                        /*
2224                                         * RFC 4862 Section 5.5.3e:
2225                                         * "Note that the preferred lifetime of
2226                                         *  the corresponding address is always
2227                                         *  reset to the Preferred Lifetime in
2228                                         *  the received Prefix Information
2229                                         *  option, regardless of whether the
2230                                         *  valid lifetime is also reset or
2231                                         *  ignored."
2232                                         *
2233                                         *  So if the preferred lifetime in
2234                                         *  this advertisement is different
2235                                         *  than what we have stored, but the
2236                                         *  valid lifetime is invalid, just
2237                                         *  reset prefered_lft.
2238                                         *
2239                                         *  We must set the valid lifetime
2240                                         *  to the stored lifetime since we'll
2241                                         *  be updating the timestamp below,
2242                                         *  else we'll set it back to the
2243                                         *  minimum.
2244                                         */
2245                                        if (prefered_lft != ifp->prefered_lft) {
2246                                                valid_lft = stored_lft;
2247                                                update_lft = 1;
2248                                        }
2249                                } else {
2250                                        valid_lft = MIN_VALID_LIFETIME;
2251                                        if (valid_lft < prefered_lft)
2252                                                prefered_lft = valid_lft;
2253                                        update_lft = 1;
2254                                }
2255                        }
2256
2257                        if (update_lft) {
2258                                ifp->valid_lft = valid_lft;
2259                                ifp->prefered_lft = prefered_lft;
2260                                ifp->tstamp = now;
2261                                flags = ifp->flags;
2262                                ifp->flags &= ~IFA_F_DEPRECATED;
2263                                spin_unlock(&ifp->lock);
2264
2265                                if (!(flags&IFA_F_TENTATIVE))
2266                                        ipv6_ifa_notify(0, ifp);
2267                        } else
2268                                spin_unlock(&ifp->lock);
2269
2270#ifdef CONFIG_IPV6_PRIVACY
2271                        read_lock_bh(&in6_dev->lock);
2272                        /* update all temporary addresses in the list */
2273                        list_for_each_entry(ift, &in6_dev->tempaddr_list,
2274                                            tmp_list) {
2275                                int age, max_valid, max_prefered;
2276
2277                                if (ifp != ift->ifpub)
2278                                        continue;
2279
2280                                /*
2281                                 * RFC 4941 section 3.3:
2282                                 * If a received option will extend the lifetime
2283                                 * of a public address, the lifetimes of
2284                                 * temporary addresses should be extended,
2285                                 * subject to the overall constraint that no
2286                                 * temporary addresses should ever remain
2287                                 * "valid" or "preferred" for a time longer than
2288                                 * (TEMP_VALID_LIFETIME) or
2289                                 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2290                                 * respectively.
2291                                 */
2292                                age = (now - ift->cstamp) / HZ;
2293                                max_valid = in6_dev->cnf.temp_valid_lft - age;
2294                                if (max_valid < 0)
2295                                        max_valid = 0;
2296
2297                                max_prefered = in6_dev->cnf.temp_prefered_lft -
2298                                               in6_dev->cnf.max_desync_factor -
2299                                               age;
2300                                if (max_prefered < 0)
2301                                        max_prefered = 0;
2302
2303                                if (valid_lft > max_valid)
2304                                        valid_lft = max_valid;
2305
2306                                if (prefered_lft > max_prefered)
2307                                        prefered_lft = max_prefered;
2308
2309                                spin_lock(&ift->lock);
2310                                flags = ift->flags;
2311                                ift->valid_lft = valid_lft;
2312                                ift->prefered_lft = prefered_lft;
2313                                ift->tstamp = now;
2314                                if (prefered_lft > 0)
2315                                        ift->flags &= ~IFA_F_DEPRECATED;
2316
2317                                spin_unlock(&ift->lock);
2318                                if (!(flags&IFA_F_TENTATIVE))
2319                                        ipv6_ifa_notify(0, ift);
2320                        }
2321
2322                        if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2323                                /*
2324                                 * When a new public address is created as
2325                                 * described in [ADDRCONF], also create a new
2326                                 * temporary address. Also create a temporary
2327                                 * address if it's enabled but no temporary
2328                                 * address currently exists.
2329                                 */
2330                                read_unlock_bh(&in6_dev->lock);
2331                                ipv6_create_tempaddr(ifp, NULL);
2332                        } else {
2333                                read_unlock_bh(&in6_dev->lock);
2334                        }
2335#endif
2336                        in6_ifa_put(ifp);
2337                        addrconf_verify(0);
2338                }
2339        }
2340        inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2341        in6_dev_put(in6_dev);
2342}
2343
2344/*
2345 *      Set destination address.
2346 *      Special case for SIT interfaces where we create a new "virtual"
2347 *      device.
2348 */
2349int addrconf_set_dstaddr(struct net *net, void __user *arg)
2350{
2351        struct in6_ifreq ireq;
2352        struct net_device *dev;
2353        int err = -EINVAL;
2354
2355        rtnl_lock();
2356
2357        err = -EFAULT;
2358        if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2359                goto err_exit;
2360
2361        dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2362
2363        err = -ENODEV;
2364        if (dev == NULL)
2365                goto err_exit;
2366
2367#if IS_ENABLED(CONFIG_IPV6_SIT)
2368        if (dev->type == ARPHRD_SIT) {
2369                const struct net_device_ops *ops = dev->netdev_ops;
2370                struct ifreq ifr;
2371                struct ip_tunnel_parm p;
2372
2373                err = -EADDRNOTAVAIL;
2374                if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2375                        goto err_exit;
2376
2377                memset(&p, 0, sizeof(p));
2378                p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2379                p.iph.saddr = 0;
2380                p.iph.version = 4;
2381                p.iph.ihl = 5;
2382                p.iph.protocol = IPPROTO_IPV6;
2383                p.iph.ttl = 64;
2384                ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2385
2386                if (ops->ndo_do_ioctl) {
2387                        mm_segment_t oldfs = get_fs();
2388
2389                        set_fs(KERNEL_DS);
2390                        err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2391                        set_fs(oldfs);
2392                } else
2393                        err = -EOPNOTSUPP;
2394
2395                if (err == 0) {
2396                        err = -ENOBUFS;
2397                        dev = __dev_get_by_name(net, p.name);
2398                        if (!dev)
2399                                goto err_exit;
2400                        err = dev_open(dev);
2401                }
2402        }
2403#endif
2404
2405err_exit:
2406        rtnl_unlock();
2407        return err;
2408}
2409
2410/*
2411 *      Manual configuration of address on an interface
2412 */
2413static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2414                          const struct in6_addr *peer_pfx,
2415                          unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2416                          __u32 valid_lft)
2417{
2418        struct inet6_ifaddr *ifp;
2419        struct inet6_dev *idev;
2420        struct net_device *dev;
2421        int scope;
2422        u32 flags;
2423        clock_t expires;
2424        unsigned long timeout;
2425
2426        ASSERT_RTNL();
2427
2428        if (plen > 128)
2429                return -EINVAL;
2430
2431        /* check the lifetime */
2432        if (!valid_lft || prefered_lft > valid_lft)
2433                return -EINVAL;
2434
2435        dev = __dev_get_by_index(net, ifindex);
2436        if (!dev)
2437                return -ENODEV;
2438
2439        idev = addrconf_add_dev(dev);
2440        if (IS_ERR(idev))
2441                return PTR_ERR(idev);
2442
2443        scope = ipv6_addr_scope(pfx);
2444
2445        timeout = addrconf_timeout_fixup(valid_lft, HZ);
2446        if (addrconf_finite_timeout(timeout)) {
2447                expires = jiffies_to_clock_t(timeout * HZ);
2448                valid_lft = timeout;
2449                flags = RTF_EXPIRES;
2450        } else {
2451                expires = 0;
2452                flags = 0;
2453                ifa_flags |= IFA_F_PERMANENT;
2454        }
2455
2456        timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2457        if (addrconf_finite_timeout(timeout)) {
2458                if (timeout == 0)
2459                        ifa_flags |= IFA_F_DEPRECATED;
2460                prefered_lft = timeout;
2461        }
2462
2463        ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2464                            valid_lft, prefered_lft);
2465
2466        if (!IS_ERR(ifp)) {
2467                addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2468                                      expires, flags);
2469                /*
2470                 * Note that section 3.1 of RFC 4429 indicates
2471                 * that the Optimistic flag should not be set for
2472                 * manually configured addresses
2473                 */
2474                addrconf_dad_start(ifp);
2475                in6_ifa_put(ifp);
2476                addrconf_verify(0);
2477                return 0;
2478        }
2479
2480        return PTR_ERR(ifp);
2481}
2482
2483static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2484                          unsigned int plen)
2485{
2486        struct inet6_ifaddr *ifp;
2487        struct inet6_dev *idev;
2488        struct net_device *dev;
2489
2490        if (plen > 128)
2491                return -EINVAL;
2492
2493        dev = __dev_get_by_index(net, ifindex);
2494        if (!dev)
2495                return -ENODEV;
2496
2497        if ((idev = __in6_dev_get(dev)) == NULL)
2498                return -ENXIO;
2499
2500        read_lock_bh(&idev->lock);
2501        list_for_each_entry(ifp, &idev->addr_list, if_list) {
2502                if (ifp->prefix_len == plen &&
2503                    ipv6_addr_equal(pfx, &ifp->addr)) {
2504                        in6_ifa_hold(ifp);
2505                        read_unlock_bh(&idev->lock);
2506
2507                        ipv6_del_addr(ifp);
2508                        return 0;
2509                }
2510        }
2511        read_unlock_bh(&idev->lock);
2512        return -EADDRNOTAVAIL;
2513}
2514
2515
2516int addrconf_add_ifaddr(struct net *net, void __user *arg)
2517{
2518        struct in6_ifreq ireq;
2519        int err;
2520
2521        if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2522                return -EPERM;
2523
2524        if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2525                return -EFAULT;
2526
2527        rtnl_lock();
2528        err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2529                             ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2530                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2531        rtnl_unlock();
2532        return err;
2533}
2534
2535int addrconf_del_ifaddr(struct net *net, void __user *arg)
2536{
2537        struct in6_ifreq ireq;
2538        int err;
2539
2540        if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2541                return -EPERM;
2542
2543        if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2544                return -EFAULT;
2545
2546        rtnl_lock();
2547        err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2548                             ireq.ifr6_prefixlen);
2549        rtnl_unlock();
2550        return err;
2551}
2552
2553static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2554                     int plen, int scope)
2555{
2556        struct inet6_ifaddr *ifp;
2557
2558        ifp = ipv6_add_addr(idev, addr, NULL, plen,
2559                            scope, IFA_F_PERMANENT, 0, 0);
2560        if (!IS_ERR(ifp)) {
2561                spin_lock_bh(&ifp->lock);
2562                ifp->flags &= ~IFA_F_TENTATIVE;
2563                spin_unlock_bh(&ifp->lock);
2564                ipv6_ifa_notify(RTM_NEWADDR, ifp);
2565                in6_ifa_put(ifp);
2566        }
2567}
2568
2569#if IS_ENABLED(CONFIG_IPV6_SIT)
2570static void sit_add_v4_addrs(struct inet6_dev *idev)
2571{
2572        struct in6_addr addr;
2573        struct net_device *dev;
2574        struct net *net = dev_net(idev->dev);
2575        int scope;
2576
2577        ASSERT_RTNL();
2578
2579        memset(&addr, 0, sizeof(struct in6_addr));
2580        memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2581
2582        if (idev->dev->flags&IFF_POINTOPOINT) {
2583                addr.s6_addr32[0] = htonl(0xfe800000);
2584                scope = IFA_LINK;
2585        } else {
2586                scope = IPV6_ADDR_COMPATv4;
2587        }
2588
2589        if (addr.s6_addr32[3]) {
2590                add_addr(idev, &addr, 128, scope);
2591                return;
2592        }
2593
2594        for_each_netdev(net, dev) {
2595                struct in_device *in_dev = __in_dev_get_rtnl(dev);
2596                if (in_dev && (dev->flags & IFF_UP)) {
2597                        struct in_ifaddr *ifa;
2598
2599                        int flag = scope;
2600
2601                        for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2602                                int plen;
2603
2604                                addr.s6_addr32[3] = ifa->ifa_local;
2605
2606                                if (ifa->ifa_scope == RT_SCOPE_LINK)
2607                                        continue;
2608                                if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2609                                        if (idev->dev->flags&IFF_POINTOPOINT)
2610                                                continue;
2611                                        flag |= IFA_HOST;
2612                                }
2613                                if (idev->dev->flags&IFF_POINTOPOINT)
2614                                        plen = 64;
2615                                else
2616                                        plen = 96;
2617
2618                                add_addr(idev, &addr, plen, flag);
2619                        }
2620                }
2621        }
2622}
2623#endif
2624
2625static void init_loopback(struct net_device *dev)
2626{
2627        struct inet6_dev  *idev;
2628        struct net_device *sp_dev;
2629        struct inet6_ifaddr *sp_ifa;
2630        struct rt6_info *sp_rt;
2631
2632        /* ::1 */
2633
2634        ASSERT_RTNL();
2635
2636        if ((idev = ipv6_find_idev(dev)) == NULL) {
2637                pr_debug("%s: add_dev failed\n", __func__);
2638                return;
2639        }
2640
2641        add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2642
2643        /* Add routes to other interface's IPv6 addresses */
2644        for_each_netdev(dev_net(dev), sp_dev) {
2645                if (!strcmp(sp_dev->name, dev->name))
2646                        continue;
2647
2648                idev = __in6_dev_get(sp_dev);
2649                if (!idev)
2650                        continue;
2651
2652                read_lock_bh(&idev->lock);
2653                list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2654
2655                        if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2656                                continue;
2657
2658                        if (sp_ifa->rt)
2659                                continue;
2660
2661                        sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2662
2663                        /* Failure cases are ignored */
2664                        if (!IS_ERR(sp_rt)) {
2665                                sp_ifa->rt = sp_rt;
2666                                ip6_ins_rt(sp_rt);
2667                        }
2668                }
2669                read_unlock_bh(&idev->lock);
2670        }
2671}
2672
2673static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2674{
2675        struct inet6_ifaddr *ifp;
2676        u32 addr_flags = IFA_F_PERMANENT;
2677
2678#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2679        if (idev->cnf.optimistic_dad &&
2680            !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2681                addr_flags |= IFA_F_OPTIMISTIC;
2682#endif
2683
2684
2685        ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2686        if (!IS_ERR(ifp)) {
2687                addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2688                addrconf_dad_start(ifp);
2689                in6_ifa_put(ifp);
2690        }
2691}
2692
2693static void addrconf_dev_config(struct net_device *dev)
2694{
2695        struct in6_addr addr;
2696        struct inet6_dev *idev;
2697
2698        ASSERT_RTNL();
2699
2700        if ((dev->type != ARPHRD_ETHER) &&
2701            (dev->type != ARPHRD_FDDI) &&
2702            (dev->type != ARPHRD_ARCNET) &&
2703            (dev->type != ARPHRD_INFINIBAND) &&
2704            (dev->type != ARPHRD_IEEE802154) &&
2705            (dev->type != ARPHRD_IEEE1394)) {
2706                /* Alas, we support only Ethernet autoconfiguration. */
2707                return;
2708        }
2709
2710        idev = addrconf_add_dev(dev);
2711        if (IS_ERR(idev))
2712                return;
2713
2714        memset(&addr, 0, sizeof(struct in6_addr));
2715        addr.s6_addr32[0] = htonl(0xFE800000);
2716
2717        if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2718                addrconf_add_linklocal(idev, &addr);
2719}
2720
2721#if IS_ENABLED(CONFIG_IPV6_SIT)
2722static void addrconf_sit_config(struct net_device *dev)
2723{
2724        struct inet6_dev *idev;
2725
2726        ASSERT_RTNL();
2727
2728        /*
2729         * Configure the tunnel with one of our IPv4
2730         * addresses... we should configure all of
2731         * our v4 addrs in the tunnel
2732         */
2733
2734        if ((idev = ipv6_find_idev(dev)) == NULL) {
2735                pr_debug("%s: add_dev failed\n", __func__);
2736                return;
2737        }
2738
2739        if (dev->priv_flags & IFF_ISATAP) {
2740                struct in6_addr addr;
2741
2742                ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2743                addrconf_prefix_route(&addr, 64, dev, 0, 0);
2744                if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2745                        addrconf_add_linklocal(idev, &addr);
2746                return;
2747        }
2748
2749        sit_add_v4_addrs(idev);
2750
2751        if (dev->flags&IFF_POINTOPOINT)
2752                addrconf_add_mroute(dev);
2753        else
2754                sit_route_add(dev);
2755}
2756#endif
2757
2758#if IS_ENABLED(CONFIG_NET_IPGRE)
2759static void addrconf_gre_config(struct net_device *dev)
2760{
2761        struct inet6_dev *idev;
2762        struct in6_addr addr;
2763
2764        ASSERT_RTNL();
2765
2766        if ((idev = ipv6_find_idev(dev)) == NULL) {
2767                pr_debug("%s: add_dev failed\n", __func__);
2768                return;
2769        }
2770
2771        ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2772        addrconf_prefix_route(&addr, 64, dev, 0, 0);
2773
2774        if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2775                addrconf_add_linklocal(idev, &addr);
2776}
2777#endif
2778
2779static inline int
2780ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2781{
2782        struct in6_addr lladdr;
2783
2784        if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2785                addrconf_add_linklocal(idev, &lladdr);
2786                return 0;
2787        }
2788        return -1;
2789}
2790
2791static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2792{
2793        struct net_device *link_dev;
2794        struct net *net = dev_net(idev->dev);
2795
2796        /* first try to inherit the link-local address from the link device */
2797        if (idev->dev->iflink &&
2798            (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2799                if (!ipv6_inherit_linklocal(idev, link_dev))
2800                        return;
2801        }
2802        /* then try to inherit it from any device */
2803        for_each_netdev(net, link_dev) {
2804                if (!ipv6_inherit_linklocal(idev, link_dev))
2805                        return;
2806        }
2807        pr_debug("init ip6-ip6: add_linklocal failed\n");
2808}
2809
2810/*
2811 * Autoconfigure tunnel with a link-local address so routing protocols,
2812 * DHCPv6, MLD etc. can be run over the virtual link
2813 */
2814
2815static void addrconf_ip6_tnl_config(struct net_device *dev)
2816{
2817        struct inet6_dev *idev;
2818
2819        ASSERT_RTNL();
2820
2821        idev = addrconf_add_dev(dev);
2822        if (IS_ERR(idev)) {
2823                pr_debug("init ip6-ip6: add_dev failed\n");
2824                return;
2825        }
2826        ip6_tnl_add_linklocal(idev);
2827}
2828
2829static int addrconf_notify(struct notifier_block *this, unsigned long event,
2830                           void *ptr)
2831{
2832        struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2833        struct inet6_dev *idev = __in6_dev_get(dev);
2834        int run_pending = 0;
2835        int err;
2836
2837        switch (event) {
2838        case NETDEV_REGISTER:
2839                if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2840                        idev = ipv6_add_dev(dev);
2841                        if (!idev)
2842                                return notifier_from_errno(-ENOMEM);
2843                }
2844                break;
2845
2846        case NETDEV_UP:
2847        case NETDEV_CHANGE:
2848                if (dev->flags & IFF_SLAVE)
2849                        break;
2850
2851                if (event == NETDEV_UP) {
2852                        if (!addrconf_qdisc_ok(dev)) {
2853                                /* device is not ready yet. */
2854                                pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2855                                        dev->name);
2856                                break;
2857                        }
2858
2859                        if (!idev && dev->mtu >= IPV6_MIN_MTU)
2860                                idev = ipv6_add_dev(dev);
2861
2862                        if (idev) {
2863                                idev->if_flags |= IF_READY;
2864                                run_pending = 1;
2865                        }
2866                } else {
2867                        if (!addrconf_qdisc_ok(dev)) {
2868                                /* device is still not ready. */
2869                                break;
2870                        }
2871
2872                        if (idev) {
2873                                if (idev->if_flags & IF_READY)
2874                                        /* device is already configured. */
2875                                        break;
2876                                idev->if_flags |= IF_READY;
2877                        }
2878
2879                        pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2880                                dev->name);
2881
2882                        run_pending = 1;
2883                }
2884
2885                switch (dev->type) {
2886#if IS_ENABLED(CONFIG_IPV6_SIT)
2887                case ARPHRD_SIT:
2888                        addrconf_sit_config(dev);
2889                        break;
2890#endif
2891#if IS_ENABLED(CONFIG_NET_IPGRE)
2892                case ARPHRD_IPGRE:
2893                        addrconf_gre_config(dev);
2894                        break;
2895#endif
2896                case ARPHRD_TUNNEL6:
2897                        addrconf_ip6_tnl_config(dev);
2898                        break;
2899                case ARPHRD_LOOPBACK:
2900                        init_loopback(dev);
2901                        break;
2902
2903                default:
2904                        addrconf_dev_config(dev);
2905                        break;
2906                }
2907
2908                if (idev) {
2909                        if (run_pending)
2910                                addrconf_dad_run(idev);
2911
2912                        /*
2913                         * If the MTU changed during the interface down,
2914                         * when the interface up, the changed MTU must be
2915                         * reflected in the idev as well as routers.
2916                         */
2917                        if (idev->cnf.mtu6 != dev->mtu &&
2918                            dev->mtu >= IPV6_MIN_MTU) {
2919                                rt6_mtu_change(dev, dev->mtu);
2920                                idev->cnf.mtu6 = dev->mtu;
2921                        }
2922                        idev->tstamp = jiffies;
2923                        inet6_ifinfo_notify(RTM_NEWLINK, idev);
2924
2925                        /*
2926                         * If the changed mtu during down is lower than
2927                         * IPV6_MIN_MTU stop IPv6 on this interface.
2928                         */
2929                        if (dev->mtu < IPV6_MIN_MTU)
2930                                addrconf_ifdown(dev, 1);
2931                }
2932                break;
2933
2934        case NETDEV_CHANGEMTU:
2935                if (idev && dev->mtu >= IPV6_MIN_MTU) {
2936                        rt6_mtu_change(dev, dev->mtu);
2937                        idev->cnf.mtu6 = dev->mtu;
2938                        break;
2939                }
2940
2941                if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2942                        idev = ipv6_add_dev(dev);
2943                        if (idev)
2944                                break;
2945                }
2946
2947                /*
2948                 * MTU falled under IPV6_MIN_MTU.
2949                 * Stop IPv6 on this interface.
2950                 */
2951
2952        case NETDEV_DOWN:
2953        case NETDEV_UNREGISTER:
2954                /*
2955                 *      Remove all addresses from this interface.
2956                 */
2957                addrconf_ifdown(dev, event != NETDEV_DOWN);
2958                break;
2959
2960        case NETDEV_CHANGENAME:
2961                if (idev) {
2962                        snmp6_unregister_dev(idev);
2963                        addrconf_sysctl_unregister(idev);
2964                        addrconf_sysctl_register(idev);
2965                        err = snmp6_register_dev(idev);
2966                        if (err)
2967                                return notifier_from_errno(err);
2968                }
2969                break;
2970
2971        case NETDEV_PRE_TYPE_CHANGE:
2972        case NETDEV_POST_TYPE_CHANGE:
2973                addrconf_type_change(dev, event);
2974                break;
2975        }
2976
2977        return NOTIFY_OK;
2978}
2979
2980/*
2981 *      addrconf module should be notified of a device going up
2982 */
2983static struct notifier_block ipv6_dev_notf = {
2984        .notifier_call = addrconf_notify,
2985};
2986
2987static void addrconf_type_change(struct net_device *dev, unsigned long event)
2988{
2989        struct inet6_dev *idev;
2990        ASSERT_RTNL();
2991
2992        idev = __in6_dev_get(dev);
2993
2994        if (event == NETDEV_POST_TYPE_CHANGE)
2995                ipv6_mc_remap(idev);
2996        else if (event == NETDEV_PRE_TYPE_CHANGE)
2997                ipv6_mc_unmap(idev);
2998}
2999
3000static int addrconf_ifdown(struct net_device *dev, int how)
3001{
3002        struct net *net = dev_net(dev);
3003        struct inet6_dev *idev;
3004        struct inet6_ifaddr *ifa;
3005        int state, i;
3006
3007        ASSERT_RTNL();
3008
3009        rt6_ifdown(net, dev);
3010        neigh_ifdown(&nd_tbl, dev);
3011
3012        idev = __in6_dev_get(dev);
3013        if (idev == NULL)
3014                return -ENODEV;
3015
3016        /*
3017         * Step 1: remove reference to ipv6 device from parent device.
3018         *         Do not dev_put!
3019         */
3020        if (how) {
3021                idev->dead = 1;
3022
3023                /* protected by rtnl_lock */
3024                RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3025
3026                /* Step 1.5: remove snmp6 entry */
3027                snmp6_unregister_dev(idev);
3028
3029        }
3030
3031        /* Step 2: clear hash table */
3032        for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3033                struct hlist_head *h = &inet6_addr_lst[i];
3034
3035                spin_lock_bh(&addrconf_hash_lock);
3036        restart:
3037                hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3038                        if (ifa->idev == idev) {
3039                                hlist_del_init_rcu(&ifa->addr_lst);
3040                                addrconf_del_dad_timer(ifa);
3041                                goto restart;
3042                        }
3043                }
3044                spin_unlock_bh(&addrconf_hash_lock);
3045        }
3046
3047        write_lock_bh(&idev->lock);
3048
3049        addrconf_del_rs_timer(idev);
3050
3051        /* Step 2: clear flags for stateless addrconf */
3052        if (!how)
3053                idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3054
3055#ifdef CONFIG_IPV6_PRIVACY
3056        if (how && del_timer(&idev->regen_timer))
3057                in6_dev_put(idev);
3058
3059        /* Step 3: clear tempaddr list */
3060        while (!list_empty(&idev->tempaddr_list)) {
3061                ifa = list_first_entry(&idev->tempaddr_list,
3062                                       struct inet6_ifaddr, tmp_list);
3063                list_del(&ifa->tmp_list);
3064                write_unlock_bh(&idev->lock);
3065                spin_lock_bh(&ifa->lock);
3066
3067                if (ifa->ifpub) {
3068                        in6_ifa_put(ifa->ifpub);
3069                        ifa->ifpub = NULL;
3070                }
3071                spin_unlock_bh(&ifa->lock);
3072                in6_ifa_put(ifa);
3073                write_lock_bh(&idev->lock);
3074        }
3075#endif
3076
3077        while (!list_empty(&idev->addr_list)) {
3078                ifa = list_first_entry(&idev->addr_list,
3079                                       struct inet6_ifaddr, if_list);
3080                addrconf_del_dad_timer(ifa);
3081
3082                list_del(&ifa->if_list);
3083
3084                write_unlock_bh(&idev->lock);
3085
3086                spin_lock_bh(&ifa->state_lock);
3087                state = ifa->state;
3088                ifa->state = INET6_IFADDR_STATE_DEAD;
3089                spin_unlock_bh(&ifa->state_lock);
3090
3091                if (state != INET6_IFADDR_STATE_DEAD) {
3092                        __ipv6_ifa_notify(RTM_DELADDR, ifa);
3093                        inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3094                }
3095                in6_ifa_put(ifa);
3096
3097                write_lock_bh(&idev->lock);
3098        }
3099
3100        write_unlock_bh(&idev->lock);
3101
3102        /* Step 5: Discard multicast list */
3103        if (how)
3104                ipv6_mc_destroy_dev(idev);
3105        else
3106                ipv6_mc_down(idev);
3107
3108        idev->tstamp = jiffies;
3109
3110        /* Last: Shot the device (if unregistered) */
3111        if (how) {
3112                addrconf_sysctl_unregister(idev);
3113                neigh_parms_release(&nd_tbl, idev->nd_parms);
3114                neigh_ifdown(&nd_tbl, dev);
3115                in6_dev_put(idev);
3116        }
3117        return 0;
3118}
3119
3120static void addrconf_rs_timer(unsigned long data)
3121{
3122        struct inet6_dev *idev = (struct inet6_dev *)data;
3123        struct in6_addr lladdr;
3124
3125        write_lock(&idev->lock);
3126        if (idev->dead || !(idev->if_flags & IF_READY))
3127                goto out;
3128
3129        if (!ipv6_accept_ra(idev))
3130                goto out;
3131
3132        /* Announcement received after solicitation was sent */
3133        if (idev->if_flags & IF_RA_RCVD)
3134                goto out;
3135
3136        if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3137                if (!__ipv6_get_lladdr(idev, &lladdr, IFA_F_TENTATIVE))
3138                        ndisc_send_rs(idev->dev, &lladdr,
3139                                      &in6addr_linklocal_allrouters);
3140                else
3141                        goto out;
3142
3143                /* The wait after the last probe can be shorter */
3144                addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3145                                             idev->cnf.rtr_solicits) ?
3146                                      idev->cnf.rtr_solicit_delay :
3147                                      idev->cnf.rtr_solicit_interval);
3148        } else {
3149                /*
3150                 * Note: we do not support deprecated "all on-link"
3151                 * assumption any longer.
3152                 */
3153                pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3154        }
3155
3156out:
3157        write_unlock(&idev->lock);
3158        in6_dev_put(idev);
3159}
3160
3161/*
3162 *      Duplicate Address Detection
3163 */
3164static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3165{
3166        unsigned long rand_num;
3167        struct inet6_dev *idev = ifp->idev;
3168
3169        if (ifp->flags & IFA_F_OPTIMISTIC)
3170                rand_num = 0;
3171        else
3172                rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3173
3174        ifp->dad_probes = idev->cnf.dad_transmits;
3175        addrconf_mod_dad_timer(ifp, rand_num);
3176}
3177
3178static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3179{
3180        struct inet6_dev *idev = ifp->idev;
3181        struct net_device *dev = idev->dev;
3182
3183        addrconf_join_solict(dev, &ifp->addr);
3184
3185        net_srandom(ifp->addr.s6_addr32[3]);
3186
3187        read_lock_bh(&idev->lock);
3188        spin_lock(&ifp->lock);
3189        if (ifp->state == INET6_IFADDR_STATE_DEAD)
3190                goto out;
3191
3192        if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3193            idev->cnf.accept_dad < 1 ||
3194            !(ifp->flags&IFA_F_TENTATIVE) ||
3195            ifp->flags & IFA_F_NODAD) {
3196                ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3197                spin_unlock(&ifp->lock);
3198                read_unlock_bh(&idev->lock);
3199
3200                addrconf_dad_completed(ifp);
3201                return;
3202        }
3203
3204        if (!(idev->if_flags & IF_READY)) {
3205                spin_unlock(&ifp->lock);
3206                read_unlock_bh(&idev->lock);
3207                /*
3208                 * If the device is not ready:
3209                 * - keep it tentative if it is a permanent address.
3210                 * - otherwise, kill it.
3211                 */
3212                in6_ifa_hold(ifp);
3213                addrconf_dad_stop(ifp, 0);
3214                return;
3215        }
3216
3217        /*
3218         * Optimistic nodes can start receiving
3219         * Frames right away
3220         */
3221        if (ifp->flags & IFA_F_OPTIMISTIC)
3222                ip6_ins_rt(ifp->rt);
3223
3224        addrconf_dad_kick(ifp);
3225out:
3226        spin_unlock(&ifp->lock);
3227        read_unlock_bh(&idev->lock);
3228}
3229
3230static void addrconf_dad_timer(unsigned long data)
3231{
3232        struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3233        struct inet6_dev *idev = ifp->idev;
3234        struct in6_addr mcaddr;
3235
3236        if (!ifp->dad_probes && addrconf_dad_end(ifp))
3237                goto out;
3238
3239        write_lock(&idev->lock);
3240        if (idev->dead || !(idev->if_flags & IF_READY)) {
3241                write_unlock(&idev->lock);
3242                goto out;
3243        }
3244
3245        spin_lock(&ifp->lock);
3246        if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3247                spin_unlock(&ifp->lock);
3248                write_unlock(&idev->lock);
3249                goto out;
3250        }
3251
3252        if (ifp->dad_probes == 0) {
3253                /*
3254                 * DAD was successful
3255                 */
3256
3257                ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3258                spin_unlock(&ifp->lock);
3259                write_unlock(&idev->lock);
3260
3261                addrconf_dad_completed(ifp);
3262
3263                goto out;
3264        }
3265
3266        ifp->dad_probes--;
3267        addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3268        spin_unlock(&ifp->lock);
3269        write_unlock(&idev->lock);
3270
3271        /* send a neighbour solicitation for our addr */
3272        addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3273        ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3274out:
3275        in6_ifa_put(ifp);
3276}
3277
3278static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3279{
3280        struct net_device *dev = ifp->idev->dev;
3281        struct in6_addr lladdr;
3282        bool send_rs, send_mld;
3283
3284        addrconf_del_dad_timer(ifp);
3285
3286        /*
3287         *      Configure the address for reception. Now it is valid.
3288         */
3289
3290        ipv6_ifa_notify(RTM_NEWADDR, ifp);
3291
3292        /* If added prefix is link local and we are prepared to process
3293           router advertisements, start sending router solicitations.
3294         */
3295
3296        read_lock_bh(&ifp->idev->lock);
3297        spin_lock(&ifp->lock);
3298        send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3299                   ifp->idev->valid_ll_addr_cnt == 1;
3300        send_rs = send_mld &&
3301                  ipv6_accept_ra(ifp->idev) &&
3302                  ifp->idev->cnf.rtr_solicits > 0 &&
3303                  (dev->flags&IFF_LOOPBACK) == 0;
3304        spin_unlock(&ifp->lock);
3305        read_unlock_bh(&ifp->idev->lock);
3306
3307        /* While dad is in progress mld report's source address is in6_addrany.
3308         * Resend with proper ll now.
3309         */
3310        if (send_mld)
3311                ipv6_mc_dad_complete(ifp->idev);
3312
3313        if (send_rs) {
3314                /*
3315                 *      If a host as already performed a random delay
3316                 *      [...] as part of DAD [...] there is no need
3317                 *      to delay again before sending the first RS
3318                 */
3319                if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3320                        return;
3321                ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3322
3323                write_lock_bh(&ifp->idev->lock);
3324                spin_lock(&ifp->lock);
3325                ifp->idev->rs_probes = 1;
3326                ifp->idev->if_flags |= IF_RS_SENT;
3327                addrconf_mod_rs_timer(ifp->idev,
3328                                      ifp->idev->cnf.rtr_solicit_interval);
3329                spin_unlock(&ifp->lock);
3330                write_unlock_bh(&ifp->idev->lock);
3331        }
3332}
3333
3334static void addrconf_dad_run(struct inet6_dev *idev)
3335{
3336        struct inet6_ifaddr *ifp;
3337
3338        read_lock_bh(&idev->lock);
3339        list_for_each_entry(ifp, &idev->addr_list, if_list) {
3340                spin_lock(&ifp->lock);
3341                if (ifp->flags & IFA_F_TENTATIVE &&
3342                    ifp->state == INET6_IFADDR_STATE_DAD)
3343                        addrconf_dad_kick(ifp);
3344                spin_unlock(&ifp->lock);
3345        }
3346        read_unlock_bh(&idev->lock);
3347}
3348
3349#ifdef CONFIG_PROC_FS
3350struct if6_iter_state {
3351        struct seq_net_private p;
3352        int bucket;
3353        int offset;
3354};
3355
3356static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3357{
3358        struct inet6_ifaddr *ifa = NULL;
3359        struct if6_iter_state *state = seq->private;
3360        struct net *net = seq_file_net(seq);
3361        int p = 0;
3362
3363        /* initial bucket if pos is 0 */
3364        if (pos == 0) {
3365                state->bucket = 0;
3366                state->offset = 0;
3367        }
3368
3369        for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3370                hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3371                                         addr_lst) {
3372                        if (!net_eq(dev_net(ifa->idev->dev), net))
3373                                continue;
3374                        /* sync with offset */
3375                        if (p < state->offset) {
3376                                p++;
3377                                continue;
3378                        }
3379                        state->offset++;
3380                        return ifa;
3381                }
3382
3383                /* prepare for next bucket */
3384                state->offset = 0;
3385                p = 0;
3386        }
3387        return NULL;
3388}
3389
3390static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3391                                         struct inet6_ifaddr *ifa)
3392{
3393        struct if6_iter_state *state = seq->private;
3394        struct net *net = seq_file_net(seq);
3395
3396        hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3397                if (!net_eq(dev_net(ifa->idev->dev), net))
3398                        continue;
3399                state->offset++;
3400                return ifa;
3401        }
3402
3403        while (++state->bucket < IN6_ADDR_HSIZE) {
3404                state->offset = 0;
3405                hlist_for_each_entry_rcu_bh(ifa,
3406                                     &inet6_addr_lst[state->bucket], addr_lst) {
3407                        if (!net_eq(dev_net(ifa->idev->dev), net))
3408                                continue;
3409                        state->offset++;
3410                        return ifa;
3411                }
3412        }
3413
3414        return NULL;
3415}
3416
3417static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3418        __acquires(rcu_bh)
3419{
3420        rcu_read_lock_bh();
3421        return if6_get_first(seq, *pos);
3422}
3423
3424static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3425{
3426        struct inet6_ifaddr *ifa;
3427
3428        ifa = if6_get_next(seq, v);
3429        ++*pos;
3430        return ifa;
3431}
3432
3433static void if6_seq_stop(struct seq_file *seq, void *v)
3434        __releases(rcu_bh)
3435{
3436        rcu_read_unlock_bh();
3437}
3438
3439static int if6_seq_show(struct seq_file *seq, void *v)
3440{
3441        struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3442        seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3443                   &ifp->addr,
3444                   ifp->idev->dev->ifindex,
3445                   ifp->prefix_len,
3446                   ifp->scope,
3447                   ifp->flags,
3448                   ifp->idev->dev->name);
3449        return 0;
3450}
3451
3452static const struct seq_operations if6_seq_ops = {
3453        .start  = if6_seq_start,
3454        .next   = if6_seq_next,
3455        .show   = if6_seq_show,
3456        .stop   = if6_seq_stop,
3457};
3458
3459static int if6_seq_open(struct inode *inode, struct file *file)
3460{
3461        return seq_open_net(inode, file, &if6_seq_ops,
3462                            sizeof(struct if6_iter_state));
3463}
3464
3465static const struct file_operations if6_fops = {
3466        .owner          = THIS_MODULE,
3467        .open           = if6_seq_open,
3468        .read           = seq_read,
3469        .llseek         = seq_lseek,
3470        .release        = seq_release_net,
3471};
3472
3473static int __net_init if6_proc_net_init(struct net *net)
3474{
3475        if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3476                return -ENOMEM;
3477        return 0;
3478}
3479
3480static void __net_exit if6_proc_net_exit(struct net *net)
3481{
3482        remove_proc_entry("if_inet6", net->proc_net);
3483}
3484
3485static struct pernet_operations if6_proc_net_ops = {
3486       .init = if6_proc_net_init,
3487       .exit = if6_proc_net_exit,
3488};
3489
3490int __init if6_proc_init(void)
3491{
3492        return register_pernet_subsys(&if6_proc_net_ops);
3493}
3494
3495void if6_proc_exit(void)
3496{
3497        unregister_pernet_subsys(&if6_proc_net_ops);
3498}
3499#endif  /* CONFIG_PROC_FS */
3500
3501#if IS_ENABLED(CONFIG_IPV6_MIP6)
3502/* Check if address is a home address configured on any interface. */
3503int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3504{
3505        int ret = 0;
3506        struct inet6_ifaddr *ifp = NULL;
3507        unsigned int hash = inet6_addr_hash(addr);
3508
3509        rcu_read_lock_bh();
3510        hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3511                if (!net_eq(dev_net(ifp->idev->dev), net))
3512                        continue;
3513                if (ipv6_addr_equal(&ifp->addr, addr) &&
3514                    (ifp->flags & IFA_F_HOMEADDRESS)) {
3515                        ret = 1;
3516                        break;
3517                }
3518        }
3519        rcu_read_unlock_bh();
3520        return ret;
3521}
3522#endif
3523
3524/*
3525 *      Periodic address status verification
3526 */
3527
3528static void addrconf_verify(unsigned long foo)
3529{
3530        unsigned long now, next, next_sec, next_sched;
3531        struct inet6_ifaddr *ifp;
3532        int i;
3533
3534        rcu_read_lock_bh();
3535        spin_lock(&addrconf_verify_lock);
3536        now = jiffies;
3537        next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3538
3539        del_timer(&addr_chk_timer);
3540
3541        for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3542restart:
3543                hlist_for_each_entry_rcu_bh(ifp,
3544                                         &inet6_addr_lst[i], addr_lst) {
3545                        unsigned long age;
3546
3547                        if (ifp->flags & IFA_F_PERMANENT)
3548                                continue;
3549
3550                        spin_lock(&ifp->lock);
3551                        /* We try to batch several events at once. */
3552                        age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3553
3554                        if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3555                            age >= ifp->valid_lft) {
3556                                spin_unlock(&ifp->lock);
3557                                in6_ifa_hold(ifp);
3558                                ipv6_del_addr(ifp);
3559                                goto restart;
3560                        } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3561                                spin_unlock(&ifp->lock);
3562                                continue;
3563                        } else if (age >= ifp->prefered_lft) {
3564                                /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3565                                int deprecate = 0;
3566
3567                                if (!(ifp->flags&IFA_F_DEPRECATED)) {
3568                                        deprecate = 1;
3569                                        ifp->flags |= IFA_F_DEPRECATED;
3570                                }
3571
3572                                if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3573                                        next = ifp->tstamp + ifp->valid_lft * HZ;
3574
3575                                spin_unlock(&ifp->lock);
3576
3577                                if (deprecate) {
3578                                        in6_ifa_hold(ifp);
3579
3580                                        ipv6_ifa_notify(0, ifp);
3581                                        in6_ifa_put(ifp);
3582                                        goto restart;
3583                                }
3584#ifdef CONFIG_IPV6_PRIVACY
3585                        } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3586                                   !(ifp->flags&IFA_F_TENTATIVE)) {
3587                                unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3588                                        ifp->idev->cnf.dad_transmits *
3589                                        ifp->idev->nd_parms->retrans_time / HZ;
3590
3591                                if (age >= ifp->prefered_lft - regen_advance) {
3592                                        struct inet6_ifaddr *ifpub = ifp->ifpub;
3593                                        if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3594                                                next = ifp->tstamp + ifp->prefered_lft * HZ;
3595                                        if (!ifp->regen_count && ifpub) {
3596                                                ifp->regen_count++;
3597                                                in6_ifa_hold(ifp);
3598                                                in6_ifa_hold(ifpub);
3599                                                spin_unlock(&ifp->lock);
3600
3601                                                spin_lock(&ifpub->lock);
3602                                                ifpub->regen_count = 0;
3603                                                spin_unlock(&ifpub->lock);
3604                                                ipv6_create_tempaddr(ifpub, ifp);
3605                                                in6_ifa_put(ifpub);
3606                                                in6_ifa_put(ifp);
3607                                                goto restart;
3608                                        }
3609                                } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3610                                        next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3611                                spin_unlock(&ifp->lock);
3612#endif
3613                        } else {
3614                                /* ifp->prefered_lft <= ifp->valid_lft */
3615                                if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3616                                        next = ifp->tstamp + ifp->prefered_lft * HZ;
3617                                spin_unlock(&ifp->lock);
3618                        }
3619                }
3620        }
3621
3622        next_sec = round_jiffies_up(next);
3623        next_sched = next;
3624
3625        /* If rounded timeout is accurate enough, accept it. */
3626        if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3627                next_sched = next_sec;
3628
3629        /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3630        if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3631                next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3632
3633        ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3634              now, next, next_sec, next_sched));
3635
3636        addr_chk_timer.expires = next_sched;
3637        add_timer(&addr_chk_timer);
3638        spin_unlock(&addrconf_verify_lock);
3639        rcu_read_unlock_bh();
3640}
3641
3642static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3643                                     struct in6_addr **peer_pfx)
3644{
3645        struct in6_addr *pfx = NULL;
3646
3647        *peer_pfx = NULL;
3648
3649        if (addr)
3650                pfx = nla_data(addr);
3651
3652        if (local) {
3653                if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3654                        *peer_pfx = pfx;
3655                pfx = nla_data(local);
3656        }
3657
3658        return pfx;
3659}
3660
3661static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3662        [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3663        [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3664        [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3665};
3666
3667static int
3668inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3669{
3670        struct net *net = sock_net(skb->sk);
3671        struct ifaddrmsg *ifm;
3672        struct nlattr *tb[IFA_MAX+1];
3673        struct in6_addr *pfx, *peer_pfx;
3674        int err;
3675
3676        err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3677        if (err < 0)
3678                return err;
3679
3680        ifm = nlmsg_data(nlh);
3681        pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3682        if (pfx == NULL)
3683                return -EINVAL;
3684
3685        return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3686}
3687
3688static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3689                             u32 prefered_lft, u32 valid_lft)
3690{
3691        u32 flags;
3692        clock_t expires;
3693        unsigned long timeout;
3694
3695        if (!valid_lft || (prefered_lft > valid_lft))
3696                return -EINVAL;
3697
3698        timeout = addrconf_timeout_fixup(valid_lft, HZ);
3699        if (addrconf_finite_timeout(timeout)) {
3700                expires = jiffies_to_clock_t(timeout * HZ);
3701                valid_lft = timeout;
3702                flags = RTF_EXPIRES;
3703        } else {
3704                expires = 0;
3705                flags = 0;
3706                ifa_flags |= IFA_F_PERMANENT;
3707        }
3708
3709        timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3710        if (addrconf_finite_timeout(timeout)) {
3711                if (timeout == 0)
3712                        ifa_flags |= IFA_F_DEPRECATED;
3713                prefered_lft = timeout;
3714        }
3715
3716        spin_lock_bh(&ifp->lock);
3717        ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3718        ifp->tstamp = jiffies;
3719        ifp->valid_lft = valid_lft;
3720        ifp->prefered_lft = prefered_lft;
3721
3722        spin_unlock_bh(&ifp->lock);
3723        if (!(ifp->flags&IFA_F_TENTATIVE))
3724                ipv6_ifa_notify(0, ifp);
3725
3726        addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3727                              expires, flags);
3728        addrconf_verify(0);
3729
3730        return 0;
3731}
3732
3733static int
3734inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3735{
3736        struct net *net = sock_net(skb->sk);
3737        struct ifaddrmsg *ifm;
3738        struct nlattr *tb[IFA_MAX+1];
3739        struct in6_addr *pfx, *peer_pfx;
3740        struct inet6_ifaddr *ifa;
3741        struct net_device *dev;
3742        u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3743        u8 ifa_flags;
3744        int err;
3745
3746        err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3747        if (err < 0)
3748                return err;
3749
3750        ifm = nlmsg_data(nlh);
3751        pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3752        if (pfx == NULL)
3753                return -EINVAL;
3754
3755        if (tb[IFA_CACHEINFO]) {
3756                struct ifa_cacheinfo *ci;
3757
3758                ci = nla_data(tb[IFA_CACHEINFO]);
3759                valid_lft = ci->ifa_valid;
3760                preferred_lft = ci->ifa_prefered;
3761        } else {
3762                preferred_lft = INFINITY_LIFE_TIME;
3763                valid_lft = INFINITY_LIFE_TIME;
3764        }
3765
3766        dev =  __dev_get_by_index(net, ifm->ifa_index);
3767        if (dev == NULL)
3768                return -ENODEV;
3769
3770        /* We ignore other flags so far. */
3771        ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3772
3773        ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3774        if (ifa == NULL) {
3775                /*
3776                 * It would be best to check for !NLM_F_CREATE here but
3777                 * userspace alreay relies on not having to provide this.
3778                 */
3779                return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3780                                      ifm->ifa_prefixlen, ifa_flags,
3781                                      preferred_lft, valid_lft);
3782        }
3783
3784        if (nlh->nlmsg_flags & NLM_F_EXCL ||
3785            !(nlh->nlmsg_flags & NLM_F_REPLACE))
3786                err = -EEXIST;
3787        else
3788                err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3789
3790        in6_ifa_put(ifa);
3791
3792        return err;
3793}
3794
3795static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3796                          u8 scope, int ifindex)
3797{
3798        struct ifaddrmsg *ifm;
3799
3800        ifm = nlmsg_data(nlh);
3801        ifm->ifa_family = AF_INET6;
3802        ifm->ifa_prefixlen = prefixlen;
3803        ifm->ifa_flags = flags;
3804        ifm->ifa_scope = scope;
3805        ifm->ifa_index = ifindex;
3806}
3807
3808static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3809                         unsigned long tstamp, u32 preferred, u32 valid)
3810{
3811        struct ifa_cacheinfo ci;
3812
3813        ci.cstamp = cstamp_delta(cstamp);
3814        ci.tstamp = cstamp_delta(tstamp);
3815        ci.ifa_prefered = preferred;
3816        ci.ifa_valid = valid;
3817
3818        return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3819}
3820
3821static inline int rt_scope(int ifa_scope)
3822{
3823        if (ifa_scope & IFA_HOST)
3824                return RT_SCOPE_HOST;
3825        else if (ifa_scope & IFA_LINK)
3826                return RT_SCOPE_LINK;
3827        else if (ifa_scope & IFA_SITE)
3828                return RT_SCOPE_SITE;
3829        else
3830                return RT_SCOPE_UNIVERSE;
3831}
3832
3833static inline int inet6_ifaddr_msgsize(void)
3834{
3835        return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3836               + nla_total_size(16) /* IFA_LOCAL */
3837               + nla_total_size(16) /* IFA_ADDRESS */
3838               + nla_total_size(sizeof(struct ifa_cacheinfo));
3839}
3840
3841static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3842                             u32 portid, u32 seq, int event, unsigned int flags)
3843{
3844        struct nlmsghdr  *nlh;
3845        u32 preferred, valid;
3846
3847        nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3848        if (nlh == NULL)
3849                return -EMSGSIZE;
3850
3851        put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3852                      ifa->idev->dev->ifindex);
3853
3854        if (!(ifa->flags&IFA_F_PERMANENT)) {
3855                preferred = ifa->prefered_lft;
3856                valid = ifa->valid_lft;
3857                if (preferred != INFINITY_LIFE_TIME) {
3858                        long tval = (jiffies - ifa->tstamp)/HZ;
3859                        if (preferred > tval)
3860                                preferred -= tval;
3861                        else
3862                                preferred = 0;
3863                        if (valid != INFINITY_LIFE_TIME) {
3864                                if (valid > tval)
3865                                        valid -= tval;
3866                                else
3867                                        valid = 0;
3868                        }
3869                }
3870        } else {
3871                preferred = INFINITY_LIFE_TIME;
3872                valid = INFINITY_LIFE_TIME;
3873        }
3874
3875        if (!ipv6_addr_any(&ifa->peer_addr)) {
3876                if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3877                    nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3878                        goto error;
3879        } else
3880                if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3881                        goto error;
3882
3883        if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3884                goto error;
3885
3886        return nlmsg_end(skb, nlh);
3887
3888error:
3889        nlmsg_cancel(skb, nlh);
3890        return -EMSGSIZE;
3891}
3892
3893static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3894                                u32 portid, u32 seq, int event, u16 flags)
3895{
3896        struct nlmsghdr  *nlh;
3897        u8 scope = RT_SCOPE_UNIVERSE;
3898        int ifindex = ifmca->idev->dev->ifindex;
3899
3900        if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3901                scope = RT_SCOPE_SITE;
3902
3903        nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3904        if (nlh == NULL)
3905                return -EMSGSIZE;
3906
3907        put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3908        if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3909            put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3910                          INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3911                nlmsg_cancel(skb, nlh);
3912                return -EMSGSIZE;
3913        }
3914
3915        return nlmsg_end(skb, nlh);
3916}
3917
3918static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3919                                u32 portid, u32 seq, int event, unsigned int flags)
3920{
3921        struct nlmsghdr  *nlh;
3922        u8 scope = RT_SCOPE_UNIVERSE;
3923        int ifindex = ifaca->aca_idev->dev->ifindex;
3924
3925        if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3926                scope = RT_SCOPE_SITE;
3927
3928        nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3929        if (nlh == NULL)
3930                return -EMSGSIZE;
3931
3932        put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3933        if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3934            put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3935                          INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3936                nlmsg_cancel(skb, nlh);
3937                return -EMSGSIZE;
3938        }
3939
3940        return nlmsg_end(skb, nlh);
3941}
3942
3943enum addr_type_t {
3944        UNICAST_ADDR,
3945        MULTICAST_ADDR,
3946        ANYCAST_ADDR,
3947};
3948
3949/* called with rcu_read_lock() */
3950static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3951                          struct netlink_callback *cb, enum addr_type_t type,
3952                          int s_ip_idx, int *p_ip_idx)
3953{
3954        struct ifmcaddr6 *ifmca;
3955        struct ifacaddr6 *ifaca;
3956        int err = 1;
3957        int ip_idx = *p_ip_idx;
3958
3959        read_lock_bh(&idev->lock);
3960        switch (type) {
3961        case UNICAST_ADDR: {
3962                struct inet6_ifaddr *ifa;
3963
3964                /* unicast address incl. temp addr */
3965                list_for_each_entry(ifa, &idev->addr_list, if_list) {
3966                        if (++ip_idx < s_ip_idx)
3967                                continue;
3968                        err = inet6_fill_ifaddr(skb, ifa,
3969                                                NETLINK_CB(cb->skb).portid,
3970                                                cb->nlh->nlmsg_seq,
3971                                                RTM_NEWADDR,
3972                                                NLM_F_MULTI);
3973                        if (err <= 0)
3974                                break;
3975                        nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3976                }
3977                break;
3978        }
3979        case MULTICAST_ADDR:
3980                /* multicast address */
3981                for (ifmca = idev->mc_list; ifmca;
3982                     ifmca = ifmca->next, ip_idx++) {
3983                        if (ip_idx < s_ip_idx)
3984                                continue;
3985                        err = inet6_fill_ifmcaddr(skb, ifmca,
3986                                                  NETLINK_CB(cb->skb).portid,
3987                                                  cb->nlh->nlmsg_seq,
3988                                                  RTM_GETMULTICAST,
3989                                                  NLM_F_MULTI);
3990                        if (err <= 0)
3991                                break;
3992                }
3993                break;
3994        case ANYCAST_ADDR:
3995                /* anycast address */
3996                for (ifaca = idev->ac_list; ifaca;
3997                     ifaca = ifaca->aca_next, ip_idx++) {
3998                        if (ip_idx < s_ip_idx)
3999                                continue;
4000                        err = inet6_fill_ifacaddr(skb, ifaca,
4001                                                  NETLINK_CB(cb->skb).portid,
4002                                                  cb->nlh->nlmsg_seq,
4003                                                  RTM_GETANYCAST,
4004                                                  NLM_F_MULTI);
4005                        if (err <= 0)
4006                                break;
4007                }
4008                break;
4009        default:
4010                break;
4011        }
4012        read_unlock_bh(&idev->lock);
4013        *p_ip_idx = ip_idx;
4014        return err;
4015}
4016
4017static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4018                           enum addr_type_t type)
4019{
4020        struct net *net = sock_net(skb->sk);
4021        int h, s_h;
4022        int idx, ip_idx;
4023        int s_idx, s_ip_idx;
4024        struct net_device *dev;
4025        struct inet6_dev *idev;
4026        struct hlist_head *head;
4027
4028        s_h = cb->args[0];
4029        s_idx = idx = cb->args[1];
4030        s_ip_idx = ip_idx = cb->args[2];
4031
4032        rcu_read_lock();
4033        cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4034        for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4035                idx = 0;
4036                head = &net->dev_index_head[h];
4037                hlist_for_each_entry_rcu(dev, head, index_hlist) {
4038                        if (idx < s_idx)
4039                                goto cont;
4040                        if (h > s_h || idx > s_idx)
4041                                s_ip_idx = 0;
4042                        ip_idx = 0;
4043                        idev = __in6_dev_get(dev);
4044                        if (!idev)
4045                                goto cont;
4046
4047                        if (in6_dump_addrs(idev, skb, cb, type,
4048                                           s_ip_idx, &ip_idx) <= 0)
4049                                goto done;
4050cont:
4051                        idx++;
4052                }
4053        }
4054done:
4055        rcu_read_unlock();
4056        cb->args[0] = h;
4057        cb->args[1] = idx;
4058        cb->args[2] = ip_idx;
4059
4060        return skb->len;
4061}
4062
4063static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4064{
4065        enum addr_type_t type = UNICAST_ADDR;
4066
4067        return inet6_dump_addr(skb, cb, type);
4068}
4069
4070static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4071{
4072        enum addr_type_t type = MULTICAST_ADDR;
4073
4074        return inet6_dump_addr(skb, cb, type);
4075}
4076
4077
4078static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4079{
4080        enum addr_type_t type = ANYCAST_ADDR;
4081
4082        return inet6_dump_addr(skb, cb, type);
4083}
4084
4085static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4086{
4087        struct net *net = sock_net(in_skb->sk);
4088        struct ifaddrmsg *ifm;
4089        struct nlattr *tb[IFA_MAX+1];
4090        struct in6_addr *addr = NULL, *peer;
4091        struct net_device *dev = NULL;
4092        struct inet6_ifaddr *ifa;
4093        struct sk_buff *skb;
4094        int err;
4095
4096        err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4097        if (err < 0)
4098                goto errout;
4099
4100        addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4101        if (addr == NULL) {
4102                err = -EINVAL;
4103                goto errout;
4104        }
4105
4106        ifm = nlmsg_data(nlh);
4107        if (ifm->ifa_index)
4108                dev = __dev_get_by_index(net, ifm->ifa_index);
4109
4110        ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4111        if (!ifa) {
4112                err = -EADDRNOTAVAIL;
4113                goto errout;
4114        }
4115
4116        skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4117        if (!skb) {
4118                err = -ENOBUFS;
4119                goto errout_ifa;
4120        }
4121
4122        err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4123                                nlh->nlmsg_seq, RTM_NEWADDR, 0);
4124        if (err < 0) {
4125                /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4126                WARN_ON(err == -EMSGSIZE);
4127                kfree_skb(skb);
4128                goto errout_ifa;
4129        }
4130        err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4131errout_ifa:
4132        in6_ifa_put(ifa);
4133errout:
4134        return err;
4135}
4136
4137static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4138{
4139        struct sk_buff *skb;
4140        struct net *net = dev_net(ifa->idev->dev);
4141        int err = -ENOBUFS;
4142
4143        skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4144        if (skb == NULL)
4145                goto errout;
4146
4147        err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4148        if (err < 0) {
4149                /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4150                WARN_ON(err == -EMSGSIZE);
4151                kfree_skb(skb);
4152                goto errout;
4153        }
4154        rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4155        return;
4156errout:
4157        if (err < 0)
4158                rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4159}
4160
4161static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4162                                __s32 *array, int bytes)
4163{
4164        BUG_ON(bytes < (DEVCONF_MAX * 4));
4165
4166        memset(array, 0, bytes);
4167        array[DEVCONF_FORWARDING] = cnf->forwarding;
4168        array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4169        array[DEVCONF_MTU6] = cnf->mtu6;
4170        array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4171        array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4172        array[DEVCONF_AUTOCONF] = cnf->autoconf;
4173        array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4174        array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4175        array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4176                jiffies_to_msecs(cnf->rtr_solicit_interval);
4177        array[DEVCONF_RTR_SOLICIT_DELAY] =
4178                jiffies_to_msecs(cnf->rtr_solicit_delay);
4179        array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4180#ifdef CONFIG_IPV6_PRIVACY
4181        array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4182        array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4183        array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4184        array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4185        array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4186#endif
4187        array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4188        array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4189        array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4190#ifdef CONFIG_IPV6_ROUTER_PREF
4191        array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4192        array[DEVCONF_RTR_PROBE_INTERVAL] =
4193                jiffies_to_msecs(cnf->rtr_probe_interval);
4194#ifdef CONFIG_IPV6_ROUTE_INFO
4195        array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4196#endif
4197#endif
4198        array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4199        array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4200#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4201        array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4202#endif
4203#ifdef CONFIG_IPV6_MROUTE
4204        array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4205#endif
4206        array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4207        array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4208        array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4209        array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4210}
4211
4212static inline size_t inet6_ifla6_size(void)
4213{
4214        return nla_total_size(4) /* IFLA_INET6_FLAGS */
4215             + nla_total_size(sizeof(struct ifla_cacheinfo))
4216             + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4217             + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4218             + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4219             + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4220}
4221
4222static inline size_t inet6_if_nlmsg_size(void)
4223{
4224        return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4225               + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4226               + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4227               + nla_total_size(4) /* IFLA_MTU */
4228               + nla_total_size(4) /* IFLA_LINK */
4229               + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4230}
4231
4232static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4233                                      int items, int bytes)
4234{
4235        int i;
4236        int pad = bytes - sizeof(u64) * items;
4237        BUG_ON(pad < 0);
4238
4239        /* Use put_unaligned() because stats may not be aligned for u64. */
4240        put_unaligned(items, &stats[0]);
4241        for (i = 1; i < items; i++)
4242                put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4243
4244        memset(&stats[items], 0, pad);
4245}
4246
4247static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4248                                      int items, int bytes, size_t syncpoff)
4249{
4250        int i;
4251        int pad = bytes - sizeof(u64) * items;
4252        BUG_ON(pad < 0);
4253
4254        /* Use put_unaligned() because stats may not be aligned for u64. */
4255        put_unaligned(items, &stats[0]);
4256        for (i = 1; i < items; i++)
4257                put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4258
4259        memset(&stats[items], 0, pad);
4260}
4261
4262static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4263                             int bytes)
4264{
4265        switch (attrtype) {
4266        case IFLA_INET6_STATS:
4267                __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4268                                     IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4269                break;
4270        case IFLA_INET6_ICMP6STATS:
4271                __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4272                break;
4273        }
4274}
4275
4276static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4277{
4278        struct nlattr *nla;
4279        struct ifla_cacheinfo ci;
4280
4281        if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4282                goto nla_put_failure;
4283        ci.max_reasm_len = IPV6_MAXPLEN;
4284        ci.tstamp = cstamp_delta(idev->tstamp);
4285        ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4286        ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4287        if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4288                goto nla_put_failure;
4289        nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4290        if (nla == NULL)
4291                goto nla_put_failure;
4292        ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4293
4294        /* XXX - MC not implemented */
4295
4296        nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4297        if (nla == NULL)
4298                goto nla_put_failure;
4299        snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4300
4301        nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4302        if (nla == NULL)
4303                goto nla_put_failure;
4304        snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4305
4306        nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4307        if (nla == NULL)
4308                goto nla_put_failure;
4309        read_lock_bh(&idev->lock);
4310        memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4311        read_unlock_bh(&idev->lock);
4312
4313        return 0;
4314
4315nla_put_failure:
4316        return -EMSGSIZE;
4317}
4318
4319static size_t inet6_get_link_af_size(const struct net_device *dev)
4320{
4321        if (!__in6_dev_get(dev))
4322                return 0;
4323
4324        return inet6_ifla6_size();
4325}
4326
4327static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4328{
4329        struct inet6_dev *idev = __in6_dev_get(dev);
4330
4331        if (!idev)
4332                return -ENODATA;
4333
4334        if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4335                return -EMSGSIZE;
4336
4337        return 0;
4338}
4339
4340static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4341{
4342        struct inet6_ifaddr *ifp;
4343        struct net_device *dev = idev->dev;
4344        bool update_rs = false;
4345        struct in6_addr ll_addr;
4346
4347        if (token == NULL)
4348                return -EINVAL;
4349        if (ipv6_addr_any(token))
4350                return -EINVAL;
4351        if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4352                return -EINVAL;
4353        if (!ipv6_accept_ra(idev))
4354                return -EINVAL;
4355        if (idev->cnf.rtr_solicits <= 0)
4356                return -EINVAL;
4357
4358        write_lock_bh(&idev->lock);
4359
4360        BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4361        memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4362
4363        write_unlock_bh(&idev->lock);
4364
4365        if (!idev->dead && (idev->if_flags & IF_READY) &&
4366            !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4367                             IFA_F_OPTIMISTIC)) {
4368
4369                /* If we're not ready, then normal ifup will take care
4370                 * of this. Otherwise, we need to request our rs here.
4371                 */
4372                ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4373                update_rs = true;
4374        }
4375
4376        write_lock_bh(&idev->lock);
4377
4378        if (update_rs) {
4379                idev->if_flags |= IF_RS_SENT;
4380                idev->rs_probes = 1;
4381                addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4382        }
4383
4384        /* Well, that's kinda nasty ... */
4385        list_for_each_entry(ifp, &idev->addr_list, if_list) {
4386                spin_lock(&ifp->lock);
4387                if (ifp->tokenized) {
4388                        ifp->valid_lft = 0;
4389                        ifp->prefered_lft = 0;
4390                }
4391                spin_unlock(&ifp->lock);
4392        }
4393
4394        write_unlock_bh(&idev->lock);
4395        addrconf_verify(0);
4396        return 0;
4397}
4398
4399static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4400{
4401        int err = -EINVAL;
4402        struct inet6_dev *idev = __in6_dev_get(dev);
4403        struct nlattr *tb[IFLA_INET6_MAX + 1];
4404
4405        if (!idev)
4406                return -EAFNOSUPPORT;
4407
4408        if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4409                BUG();
4410
4411        if (tb[IFLA_INET6_TOKEN])
4412                err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4413
4414        return err;
4415}
4416
4417static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4418                             u32 portid, u32 seq, int event, unsigned int flags)
4419{
4420        struct net_device *dev = idev->dev;
4421        struct ifinfomsg *hdr;
4422        struct nlmsghdr *nlh;
4423        void *protoinfo;
4424
4425        nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4426        if (nlh == NULL)
4427                return -EMSGSIZE;
4428
4429        hdr = nlmsg_data(nlh);
4430        hdr->ifi_family = AF_INET6;
4431        hdr->__ifi_pad = 0;
4432        hdr->ifi_type = dev->type;
4433        hdr->ifi_index = dev->ifindex;
4434        hdr->ifi_flags = dev_get_flags(dev);
4435        hdr->ifi_change = 0;
4436
4437        if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4438            (dev->addr_len &&
4439             nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4440            nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4441            (dev->ifindex != dev->iflink &&
4442             nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4443                goto nla_put_failure;
4444        protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4445        if (protoinfo == NULL)
4446                goto nla_put_failure;
4447
4448        if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4449                goto nla_put_failure;
4450
4451        nla_nest_end(skb, protoinfo);
4452        return nlmsg_end(skb, nlh);
4453
4454nla_put_failure:
4455        nlmsg_cancel(skb, nlh);
4456        return -EMSGSIZE;
4457}
4458
4459static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4460{
4461        struct net *net = sock_net(skb->sk);
4462        int h, s_h;
4463        int idx = 0, s_idx;
4464        struct net_device *dev;
4465        struct inet6_dev *idev;
4466        struct hlist_head *head;
4467
4468        s_h = cb->args[0];
4469        s_idx = cb->args[1];
4470
4471        rcu_read_lock();
4472        for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4473                idx = 0;
4474                head = &net->dev_index_head[h];
4475                hlist_for_each_entry_rcu(dev, head, index_hlist) {
4476                        if (idx < s_idx)
4477                                goto cont;
4478                        idev = __in6_dev_get(dev);
4479                        if (!idev)
4480                                goto cont;
4481                        if (inet6_fill_ifinfo(skb, idev,
4482                                              NETLINK_CB(cb->skb).portid,
4483                                              cb->nlh->nlmsg_seq,
4484                                              RTM_NEWLINK, NLM_F_MULTI) <= 0)
4485                                goto out;
4486cont:
4487                        idx++;
4488                }
4489        }
4490out:
4491        rcu_read_unlock();
4492        cb->args[1] = idx;
4493        cb->args[0] = h;
4494
4495        return skb->len;
4496}
4497
4498void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4499{
4500        struct sk_buff *skb;
4501        struct net *net = dev_net(idev->dev);
4502        int err = -ENOBUFS;
4503
4504        skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4505        if (skb == NULL)
4506                goto errout;
4507
4508        err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4509        if (err < 0) {
4510                /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4511                WARN_ON(err == -EMSGSIZE);
4512                kfree_skb(skb);
4513                goto errout;
4514        }
4515        rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4516        return;
4517errout:
4518        if (err < 0)
4519                rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4520}
4521
4522static inline size_t inet6_prefix_nlmsg_size(void)
4523{
4524        return NLMSG_ALIGN(sizeof(struct prefixmsg))
4525               + nla_total_size(sizeof(struct in6_addr))
4526               + nla_total_size(sizeof(struct prefix_cacheinfo));
4527}
4528
4529static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4530                             struct prefix_info *pinfo, u32 portid, u32 seq,
4531                             int event, unsigned int flags)
4532{
4533        struct prefixmsg *pmsg;
4534        struct nlmsghdr *nlh;
4535        struct prefix_cacheinfo ci;
4536
4537        nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4538        if (nlh == NULL)
4539                return -EMSGSIZE;
4540
4541        pmsg = nlmsg_data(nlh);
4542        pmsg->prefix_family = AF_INET6;
4543        pmsg->prefix_pad1 = 0;
4544        pmsg->prefix_pad2 = 0;
4545        pmsg->prefix_ifindex = idev->dev->ifindex;
4546        pmsg->prefix_len = pinfo->prefix_len;
4547        pmsg->prefix_type = pinfo->type;
4548        pmsg->prefix_pad3 = 0;
4549        pmsg->prefix_flags = 0;
4550        if (pinfo->onlink)
4551                pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4552        if (pinfo->autoconf)
4553                pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4554
4555        if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4556                goto nla_put_failure;
4557        ci.preferred_time = ntohl(pinfo->prefered);
4558        ci.valid_time = ntohl(pinfo->valid);
4559        if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4560                goto nla_put_failure;
4561        return nlmsg_end(skb, nlh);
4562
4563nla_put_failure:
4564        nlmsg_cancel(skb, nlh);
4565        return -EMSGSIZE;
4566}
4567
4568static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4569                         struct prefix_info *pinfo)
4570{
4571        struct sk_buff *skb;
4572        struct net *net = dev_net(idev->dev);
4573        int err = -ENOBUFS;
4574
4575        skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4576        if (skb == NULL)
4577                goto errout;
4578
4579        err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4580        if (err < 0) {
4581                /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4582                WARN_ON(err == -EMSGSIZE);
4583                kfree_skb(skb);
4584                goto errout;
4585        }
4586        rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4587        return;
4588errout:
4589        if (err < 0)
4590                rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4591}
4592
4593static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4594{
4595        write_lock_bh(&ifp->idev->lock);
4596        spin_lock(&ifp->lock);
4597        if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4598                            IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4599            (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4600                ifp->idev->valid_ll_addr_cnt += count;
4601        WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4602        spin_unlock(&ifp->lock);
4603        write_unlock_bh(&ifp->idev->lock);
4604}
4605
4606static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4607{
4608        struct net *net = dev_net(ifp->idev->dev);
4609
4610        inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4611
4612        switch (event) {
4613        case RTM_NEWADDR:
4614                update_valid_ll_addr_cnt(ifp, 1);
4615
4616                /*
4617                 * If the address was optimistic
4618                 * we inserted the route at the start of
4619                 * our DAD process, so we don't need
4620                 * to do it again
4621                 */
4622                if (!(ifp->rt->rt6i_node))
4623                        ip6_ins_rt(ifp->rt);
4624                if (ifp->idev->cnf.forwarding)
4625                        addrconf_join_anycast(ifp);
4626                if (!ipv6_addr_any(&ifp->peer_addr))
4627                        addrconf_prefix_route(&ifp->peer_addr, 128,
4628                                              ifp->idev->dev, 0, 0);
4629                break;
4630        case RTM_DELADDR:
4631                update_valid_ll_addr_cnt(ifp, -1);
4632
4633                if (ifp->idev->cnf.forwarding)
4634                        addrconf_leave_anycast(ifp);
4635                addrconf_leave_solict(ifp->idev, &ifp->addr);
4636                if (!ipv6_addr_any(&ifp->peer_addr)) {
4637                        struct rt6_info *rt;
4638                        struct net_device *dev = ifp->idev->dev;
4639
4640                        rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4641                                        dev->ifindex, 1);
4642                        if (rt) {
4643                                dst_hold(&rt->dst);
4644                                if (ip6_del_rt(rt))
4645                                        dst_free(&rt->dst);
4646                        }
4647                }
4648                dst_hold(&ifp->rt->dst);
4649
4650                if (ip6_del_rt(ifp->rt))
4651                        dst_free(&ifp->rt->dst);
4652                break;
4653        }
4654        atomic_inc(&net->ipv6.dev_addr_genid);
4655}
4656
4657static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4658{
4659        rcu_read_lock_bh();
4660        if (likely(ifp->idev->dead == 0))
4661                __ipv6_ifa_notify(event, ifp);
4662        rcu_read_unlock_bh();
4663}
4664
4665#ifdef CONFIG_SYSCTL
4666
4667static
4668int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4669                           void __user *buffer, size_t *lenp, loff_t *ppos)
4670{
4671        int *valp = ctl->data;
4672        int val = *valp;
4673        loff_t pos = *ppos;
4674        struct ctl_table lctl;
4675        int ret;
4676
4677        /*
4678         * ctl->data points to idev->cnf.forwarding, we should
4679         * not modify it until we get the rtnl lock.
4680         */
4681        lctl = *ctl;
4682        lctl.data = &val;
4683
4684        ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4685
4686        if (write)
4687                ret = addrconf_fixup_forwarding(ctl, valp, val);
4688        if (ret)
4689                *ppos = pos;
4690        return ret;
4691}
4692
4693static void dev_disable_change(struct inet6_dev *idev)
4694{
4695        struct netdev_notifier_info info;
4696
4697        if (!idev || !idev->dev)
4698                return;
4699
4700        netdev_notifier_info_init(&info, idev->dev);
4701        if (idev->cnf.disable_ipv6)
4702                addrconf_notify(NULL, NETDEV_DOWN, &info);
4703        else
4704                addrconf_notify(NULL, NETDEV_UP, &info);
4705}
4706
4707static void addrconf_disable_change(struct net *net, __s32 newf)
4708{
4709        struct net_device *dev;
4710        struct inet6_dev *idev;
4711
4712        rcu_read_lock();
4713        for_each_netdev_rcu(net, dev) {
4714                idev = __in6_dev_get(dev);
4715                if (idev) {
4716                        int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4717                        idev->cnf.disable_ipv6 = newf;
4718                        if (changed)
4719                                dev_disable_change(idev);
4720                }
4721        }
4722        rcu_read_unlock();
4723}
4724
4725static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4726{
4727        struct net *net;
4728        int old;
4729
4730        if (!rtnl_trylock())
4731                return restart_syscall();
4732
4733        net = (struct net *)table->extra2;
4734        old = *p;
4735        *p = newf;
4736
4737        if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4738                rtnl_unlock();
4739                return 0;
4740        }
4741
4742        if (p == &net->ipv6.devconf_all->disable_ipv6) {
4743                net->ipv6.devconf_dflt->disable_ipv6 = newf;
4744                addrconf_disable_change(net, newf);
4745        } else if ((!newf) ^ (!old))
4746                dev_disable_change((struct inet6_dev *)table->extra1);
4747
4748        rtnl_unlock();
4749        return 0;
4750}
4751
4752static
4753int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4754                            void __user *buffer, size_t *lenp, loff_t *ppos)
4755{
4756        int *valp = ctl->data;
4757        int val = *valp;
4758        loff_t pos = *ppos;
4759        struct ctl_table lctl;
4760        int ret;
4761
4762        /*
4763         * ctl->data points to idev->cnf.disable_ipv6, we should
4764         * not modify it until we get the rtnl lock.
4765         */
4766        lctl = *ctl;
4767        lctl.data = &val;
4768
4769        ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4770
4771        if (write)
4772                ret = addrconf_disable_ipv6(ctl, valp, val);
4773        if (ret)
4774                *ppos = pos;
4775        return ret;
4776}
4777
4778static struct addrconf_sysctl_table
4779{
4780        struct ctl_table_header *sysctl_header;
4781        struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4782} addrconf_sysctl __read_mostly = {
4783        .sysctl_header = NULL,
4784        .addrconf_vars = {
4785                {
4786                        .procname       = "forwarding",
4787                        .data           = &ipv6_devconf.forwarding,
4788                        .maxlen         = sizeof(int),
4789                        .mode           = 0644,
4790                        .proc_handler   = addrconf_sysctl_forward,
4791                },
4792                {
4793                        .procname       = "hop_limit",
4794                        .data           = &ipv6_devconf.hop_limit,
4795                        .maxlen         = sizeof(int),
4796                        .mode           = 0644,
4797                        .proc_handler   = proc_dointvec,
4798                },
4799                {
4800                        .procname       = "mtu",
4801                        .data           = &ipv6_devconf.mtu6,
4802                        .maxlen         = sizeof(int),
4803                        .mode           = 0644,
4804                        .proc_handler   = proc_dointvec,
4805                },
4806                {
4807                        .procname       = "accept_ra",
4808                        .data           = &ipv6_devconf.accept_ra,
4809                        .maxlen         = sizeof(int),
4810                        .mode           = 0644,
4811                        .proc_handler   = proc_dointvec,
4812                },
4813                {
4814                        .procname       = "accept_redirects",
4815                        .data           = &ipv6_devconf.accept_redirects,
4816                        .maxlen         = sizeof(int),
4817                        .mode           = 0644,
4818                        .proc_handler   = proc_dointvec,
4819                },
4820                {
4821                        .procname       = "autoconf",
4822                        .data           = &ipv6_devconf.autoconf,
4823                        .maxlen         = sizeof(int),
4824                        .mode           = 0644,
4825                        .proc_handler   = proc_dointvec,
4826                },
4827                {
4828                        .procname       = "dad_transmits",
4829                        .data           = &ipv6_devconf.dad_transmits,
4830                        .maxlen         = sizeof(int),
4831                        .mode           = 0644,
4832                        .proc_handler   = proc_dointvec,
4833                },
4834                {
4835                        .procname       = "router_solicitations",
4836                        .data           = &ipv6_devconf.rtr_solicits,
4837                        .maxlen         = sizeof(int),
4838                        .mode           = 0644,
4839                        .proc_handler   = proc_dointvec,
4840                },
4841                {
4842                        .procname       = "router_solicitation_interval",
4843                        .data           = &ipv6_devconf.rtr_solicit_interval,
4844                        .maxlen         = sizeof(int),
4845                        .mode           = 0644,
4846                        .proc_handler   = proc_dointvec_jiffies,
4847                },
4848                {
4849                        .procname       = "router_solicitation_delay",
4850                        .data           = &ipv6_devconf.rtr_solicit_delay,
4851                        .maxlen         = sizeof(int),
4852                        .mode           = 0644,
4853                        .proc_handler   = proc_dointvec_jiffies,
4854                },
4855                {
4856                        .procname       = "force_mld_version",
4857                        .data           = &ipv6_devconf.force_mld_version,
4858                        .maxlen         = sizeof(int),
4859                        .mode           = 0644,
4860                        .proc_handler   = proc_dointvec,
4861                },
4862#ifdef CONFIG_IPV6_PRIVACY
4863                {
4864                        .procname       = "use_tempaddr",
4865                        .data           = &ipv6_devconf.use_tempaddr,
4866                        .maxlen         = sizeof(int),
4867                        .mode           = 0644,
4868                        .proc_handler   = proc_dointvec,
4869                },
4870                {
4871                        .procname       = "temp_valid_lft",
4872                        .data           = &ipv6_devconf.temp_valid_lft,
4873                        .maxlen         = sizeof(int),
4874                        .mode           = 0644,
4875                        .proc_handler   = proc_dointvec,
4876                },
4877                {
4878                        .procname       = "temp_prefered_lft",
4879                        .data           = &ipv6_devconf.temp_prefered_lft,
4880                        .maxlen         = sizeof(int),
4881                        .mode           = 0644,
4882                        .proc_handler   = proc_dointvec,
4883                },
4884                {
4885                        .procname       = "regen_max_retry",
4886                        .data           = &ipv6_devconf.regen_max_retry,
4887                        .maxlen         = sizeof(int),
4888                        .mode           = 0644,
4889                        .proc_handler   = proc_dointvec,
4890                },
4891                {
4892                        .procname       = "max_desync_factor",
4893                        .data           = &ipv6_devconf.max_desync_factor,
4894                        .maxlen         = sizeof(int),
4895                        .mode           = 0644,
4896                        .proc_handler   = proc_dointvec,
4897                },
4898#endif
4899                {
4900                        .procname       = "max_addresses",
4901                        .data           = &ipv6_devconf.max_addresses,
4902                        .maxlen         = sizeof(int),
4903                        .mode           = 0644,
4904                        .proc_handler   = proc_dointvec,
4905                },
4906                {
4907                        .procname       = "accept_ra_defrtr",
4908                        .data           = &ipv6_devconf.accept_ra_defrtr,
4909                        .maxlen         = sizeof(int),
4910                        .mode           = 0644,
4911                        .proc_handler   = proc_dointvec,
4912                },
4913                {
4914                        .procname       = "accept_ra_pinfo",
4915                        .data           = &ipv6_devconf.accept_ra_pinfo,
4916                        .maxlen         = sizeof(int),
4917                        .mode           = 0644,
4918                        .proc_handler   = proc_dointvec,
4919                },
4920#ifdef CONFIG_IPV6_ROUTER_PREF
4921                {
4922                        .procname       = "accept_ra_rtr_pref",
4923                        .data           = &ipv6_devconf.accept_ra_rtr_pref,
4924                        .maxlen         = sizeof(int),
4925                        .mode           = 0644,
4926                        .proc_handler   = proc_dointvec,
4927                },
4928                {
4929                        .procname       = "router_probe_interval",
4930                        .data           = &ipv6_devconf.rtr_probe_interval,
4931                        .maxlen         = sizeof(int),
4932                        .mode           = 0644,
4933                        .proc_handler   = proc_dointvec_jiffies,
4934                },
4935#ifdef CONFIG_IPV6_ROUTE_INFO
4936                {
4937                        .procname       = "accept_ra_rt_info_max_plen",
4938                        .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4939                        .maxlen         = sizeof(int),
4940                        .mode           = 0644,
4941                        .proc_handler   = proc_dointvec,
4942                },
4943#endif
4944#endif
4945                {
4946                        .procname       = "proxy_ndp",
4947                        .data           = &ipv6_devconf.proxy_ndp,
4948                        .maxlen         = sizeof(int),
4949                        .mode           = 0644,
4950                        .proc_handler   = proc_dointvec,
4951                },
4952                {
4953                        .procname       = "accept_source_route",
4954                        .data           = &ipv6_devconf.accept_source_route,
4955                        .maxlen         = sizeof(int),
4956                        .mode           = 0644,
4957                        .proc_handler   = proc_dointvec,
4958                },
4959#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4960                {
4961                        .procname       = "optimistic_dad",
4962                        .data           = &ipv6_devconf.optimistic_dad,
4963                        .maxlen         = sizeof(int),
4964                        .mode           = 0644,
4965                        .proc_handler   = proc_dointvec,
4966
4967                },
4968#endif
4969#ifdef CONFIG_IPV6_MROUTE
4970                {
4971                        .procname       = "mc_forwarding",
4972                        .data           = &ipv6_devconf.mc_forwarding,
4973                        .maxlen         = sizeof(int),
4974                        .mode           = 0444,
4975                        .proc_handler   = proc_dointvec,
4976                },
4977#endif
4978                {
4979                        .procname       = "disable_ipv6",
4980                        .data           = &ipv6_devconf.disable_ipv6,
4981                        .maxlen         = sizeof(int),
4982                        .mode           = 0644,
4983                        .proc_handler   = addrconf_sysctl_disable,
4984                },
4985                {
4986                        .procname       = "accept_dad",
4987                        .data           = &ipv6_devconf.accept_dad,
4988                        .maxlen         = sizeof(int),
4989                        .mode           = 0644,
4990                        .proc_handler   = proc_dointvec,
4991                },
4992                {
4993                        .procname       = "force_tllao",
4994                        .data           = &ipv6_devconf.force_tllao,
4995                        .maxlen         = sizeof(int),
4996                        .mode           = 0644,
4997                        .proc_handler   = proc_dointvec
4998                },
4999                {
5000                        .procname       = "ndisc_notify",
5001                        .data           = &ipv6_devconf.ndisc_notify,
5002                        .maxlen         = sizeof(int),
5003                        .mode           = 0644,
5004                        .proc_handler   = proc_dointvec
5005                },
5006                {
5007                        /* sentinel */
5008                }
5009        },
5010};
5011
5012static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5013                struct inet6_dev *idev, struct ipv6_devconf *p)
5014{
5015        int i;
5016        struct addrconf_sysctl_table *t;
5017        char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5018
5019        t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5020        if (t == NULL)
5021                goto out;
5022
5023        for (i = 0; t->addrconf_vars[i].data; i++) {
5024                t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5025                t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5026                t->addrconf_vars[i].extra2 = net;
5027        }
5028
5029        snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5030
5031        t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5032        if (t->sysctl_header == NULL)
5033                goto free;
5034
5035        p->sysctl = t;
5036        return 0;
5037
5038free:
5039        kfree(t);
5040out:
5041        return -ENOBUFS;
5042}
5043
5044static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5045{
5046        struct addrconf_sysctl_table *t;
5047
5048        if (p->sysctl == NULL)
5049                return;
5050
5051        t = p->sysctl;
5052        p->sysctl = NULL;
5053        unregister_net_sysctl_table(t->sysctl_header);
5054        kfree(t);
5055}
5056
5057static void addrconf_sysctl_register(struct inet6_dev *idev)
5058{
5059        neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5060                              &ndisc_ifinfo_sysctl_change);
5061        __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5062                                        idev, &idev->cnf);
5063}
5064
5065static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5066{
5067        __addrconf_sysctl_unregister(&idev->cnf);
5068        neigh_sysctl_unregister(idev->nd_parms);
5069}
5070
5071
5072#endif
5073
5074static int __net_init addrconf_init_net(struct net *net)
5075{
5076        int err = -ENOMEM;
5077        struct ipv6_devconf *all, *dflt;
5078
5079        all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5080        if (all == NULL)
5081                goto err_alloc_all;
5082
5083        dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5084        if (dflt == NULL)
5085                goto err_alloc_dflt;
5086
5087        /* these will be inherited by all namespaces */
5088        dflt->autoconf = ipv6_defaults.autoconf;
5089        dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5090
5091        net->ipv6.devconf_all = all;
5092        net->ipv6.devconf_dflt = dflt;
5093
5094#ifdef CONFIG_SYSCTL
5095        err = __addrconf_sysctl_register(net, "all", NULL, all);
5096        if (err < 0)
5097                goto err_reg_all;
5098
5099        err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5100        if (err < 0)
5101                goto err_reg_dflt;
5102#endif
5103        return 0;
5104
5105#ifdef CONFIG_SYSCTL
5106err_reg_dflt:
5107        __addrconf_sysctl_unregister(all);
5108err_reg_all:
5109        kfree(dflt);
5110#endif
5111err_alloc_dflt:
5112        kfree(all);
5113err_alloc_all:
5114        return err;
5115}
5116
5117static void __net_exit addrconf_exit_net(struct net *net)
5118{
5119#ifdef CONFIG_SYSCTL
5120        __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5121        __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5122#endif
5123        if (!net_eq(net, &init_net)) {
5124                kfree(net->ipv6.devconf_dflt);
5125                kfree(net->ipv6.devconf_all);
5126        }
5127}
5128
5129static struct pernet_operations addrconf_ops = {
5130        .init = addrconf_init_net,
5131        .exit = addrconf_exit_net,
5132};
5133
5134static struct rtnl_af_ops inet6_ops = {
5135        .family           = AF_INET6,
5136        .fill_link_af     = inet6_fill_link_af,
5137        .get_link_af_size = inet6_get_link_af_size,
5138        .set_link_af      = inet6_set_link_af,
5139};
5140
5141/*
5142 *      Init / cleanup code
5143 */
5144
5145int __init addrconf_init(void)
5146{
5147        int i, err;
5148
5149        err = ipv6_addr_label_init();
5150        if (err < 0) {
5151                pr_crit("%s: cannot initialize default policy table: %d\n",
5152                        __func__, err);
5153                goto out;
5154        }
5155
5156        err = register_pernet_subsys(&addrconf_ops);
5157        if (err < 0)
5158                goto out_addrlabel;
5159
5160        /* The addrconf netdev notifier requires that loopback_dev
5161         * has it's ipv6 private information allocated and setup
5162         * before it can bring up and give link-local addresses
5163         * to other devices which are up.
5164         *
5165         * Unfortunately, loopback_dev is not necessarily the first
5166         * entry in the global dev_base list of net devices.  In fact,
5167         * it is likely to be the very last entry on that list.
5168         * So this causes the notifier registry below to try and
5169         * give link-local addresses to all devices besides loopback_dev
5170         * first, then loopback_dev, which cases all the non-loopback_dev
5171         * devices to fail to get a link-local address.
5172         *
5173         * So, as a temporary fix, allocate the ipv6 structure for
5174         * loopback_dev first by hand.
5175         * Longer term, all of the dependencies ipv6 has upon the loopback
5176         * device and it being up should be removed.
5177         */
5178        rtnl_lock();
5179        if (!ipv6_add_dev(init_net.loopback_dev))
5180                err = -ENOMEM;
5181        rtnl_unlock();
5182        if (err)
5183                goto errlo;
5184
5185        for (i = 0; i < IN6_ADDR_HSIZE; i++)
5186                INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5187
5188        register_netdevice_notifier(&ipv6_dev_notf);
5189
5190        addrconf_verify(0);
5191
5192        err = rtnl_af_register(&inet6_ops);
5193        if (err < 0)
5194                goto errout_af;
5195
5196        err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5197                              NULL);
5198        if (err < 0)
5199                goto errout;
5200
5201        /* Only the first call to __rtnl_register can fail */
5202        __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5203        __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5204        __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5205                        inet6_dump_ifaddr, NULL);
5206        __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5207                        inet6_dump_ifmcaddr, NULL);
5208        __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5209                        inet6_dump_ifacaddr, NULL);
5210        __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5211                        inet6_netconf_dump_devconf, NULL);
5212
5213        ipv6_addr_label_rtnl_register();
5214
5215        return 0;
5216errout:
5217        rtnl_af_unregister(&inet6_ops);
5218errout_af:
5219        unregister_netdevice_notifier(&ipv6_dev_notf);
5220errlo:
5221        unregister_pernet_subsys(&addrconf_ops);
5222out_addrlabel:
5223        ipv6_addr_label_cleanup();
5224out:
5225        return err;
5226}
5227
5228void addrconf_cleanup(void)
5229{
5230        struct net_device *dev;
5231        int i;
5232
5233        unregister_netdevice_notifier(&ipv6_dev_notf);
5234        unregister_pernet_subsys(&addrconf_ops);
5235        ipv6_addr_label_cleanup();
5236
5237        rtnl_lock();
5238
5239        __rtnl_af_unregister(&inet6_ops);
5240
5241        /* clean dev list */
5242        for_each_netdev(&init_net, dev) {
5243                if (__in6_dev_get(dev) == NULL)
5244                        continue;
5245                addrconf_ifdown(dev, 1);
5246        }
5247        addrconf_ifdown(init_net.loopback_dev, 2);
5248
5249        /*
5250         *      Check hash table.
5251         */
5252        spin_lock_bh(&addrconf_hash_lock);
5253        for (i = 0; i < IN6_ADDR_HSIZE; i++)
5254                WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5255        spin_unlock_bh(&addrconf_hash_lock);
5256
5257        del_timer(&addr_chk_timer);
5258        rtnl_unlock();
5259}
5260