net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86 #include <linux/random.h>
  87 #include <linux/uaccess.h>
  88 #include <asm/unaligned.h>
  89
  90 #include <linux/proc_fs.h>
  91 #include <linux/seq_file.h>
  92 #include <linux/export.h>
  93
  94 /* Set to 3 to get tracing... */
  95 #define ACONF_DEBUG 2
  96
  97 #if ACONF_DEBUG >= 3
  98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
  99 #else
 100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 101 #endif
 102
 103 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 104
 105 static inline u32 cstamp_delta(unsigned long cstamp)
 106 {
 107         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 108 }
 109
 110 #ifdef CONFIG_SYSCTL
 111 static void addrconf_sysctl_register(struct inet6_dev *idev);
 112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 113 #else
 114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 115 {
 116 }
 117
 118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 119 {
 120 }
 121 #endif
 122
 123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 125 static void ipv6_regen_rndid(unsigned long data);
 126
 127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 128 static int ipv6_count_addresses(struct inet6_dev *idev);
 129
 130 /*
 131  *      Configured unicast address hash table
 132  */
 133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 134 static DEFINE_SPINLOCK(addrconf_hash_lock);
 135
 136 static void addrconf_verify(unsigned long);
 137
 138 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 139 static DEFINE_SPINLOCK(addrconf_verify_lock);
 140
 141 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 142 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 143
 144 static void addrconf_type_change(struct net_device *dev,
 145                                  unsigned long event);
 146 static int addrconf_ifdown(struct net_device *dev, int how);
 147
 148 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 149                                                   int plen,
 150                                                   const struct net_device *dev,
 151                                                   u32 flags, u32 noflags);
 152
 153 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_timer(unsigned long data);
 155 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 156 static void addrconf_dad_run(struct inet6_dev *idev);
 157 static void addrconf_rs_timer(unsigned long data);
 158 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 160
 161 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 162                                 struct prefix_info *pinfo);
 163 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 164                                struct net_device *dev);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .mldv1_unsolicited_report_interval = 10 * HZ,
 175         .mldv2_unsolicited_report_interval = HZ,
 176         .dad_transmits          = 1,
 177         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 178         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 179         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199         .suppress_frag_ndisc    = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .force_mld_version      = 0,
 210         .mldv1_unsolicited_report_interval = 10 * HZ,
 211         .mldv2_unsolicited_report_interval = HZ,
 212         .dad_transmits          = 1,
 213         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 214         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 215         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 216         .use_tempaddr           = 0,
 217         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 218         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 219         .regen_max_retry        = REGEN_MAX_RETRY,
 220         .max_desync_factor      = MAX_DESYNC_FACTOR,
 221         .max_addresses          = IPV6_MAX_ADDRESSES,
 222         .accept_ra_defrtr       = 1,
 223         .accept_ra_pinfo        = 1,
 224 #ifdef CONFIG_IPV6_ROUTER_PREF
 225         .accept_ra_rtr_pref     = 1,
 226         .rtr_probe_interval     = 60 * HZ,
 227 #ifdef CONFIG_IPV6_ROUTE_INFO
 228         .accept_ra_rt_info_max_plen = 0,
 229 #endif
 230 #endif
 231         .proxy_ndp              = 0,
 232         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 233         .disable_ipv6           = 0,
 234         .accept_dad             = 1,
 235         .suppress_frag_ndisc    = 1,
 236 };
 237
 238 /* Check if a valid qdisc is available */
 239 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 240 {
 241         return !qdisc_tx_is_noop(dev);
 242 }
 243
 244 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 245 {
 246         if (del_timer(&idev->rs_timer))
 247                 __in6_dev_put(idev);
 248 }
 249
 250 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 251 {
 252         if (del_timer(&ifp->dad_timer))
 253                 __in6_ifa_put(ifp);
 254 }
 255
 256 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 257                                   unsigned long when)
 258 {
 259         if (!timer_pending(&idev->rs_timer))
 260                 in6_dev_hold(idev);
 261         mod_timer(&idev->rs_timer, jiffies + when);
 262 }
 263
 264 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 265                                    unsigned long when)
 266 {
 267         if (!timer_pending(&ifp->dad_timer))
 268                 in6_ifa_hold(ifp);
 269         mod_timer(&ifp->dad_timer, jiffies + when);
 270 }
 271
 272 static int snmp6_alloc_dev(struct inet6_dev *idev)
 273 {
 274         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 275                           sizeof(struct ipstats_mib),
 276                           __alignof__(struct ipstats_mib)) < 0)
 277                 goto err_ip;
 278         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 279                                         GFP_KERNEL);
 280         if (!idev->stats.icmpv6dev)
 281                 goto err_icmp;
 282         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 283                                            GFP_KERNEL);
 284         if (!idev->stats.icmpv6msgdev)
 285                 goto err_icmpmsg;
 286
 287         return 0;
 288
 289 err_icmpmsg:
 290         kfree(idev->stats.icmpv6dev);
 291 err_icmp:
 292         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 293 err_ip:
 294         return -ENOMEM;
 295 }
 296
 297 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 298 {
 299         struct inet6_dev *ndev;
 300
 301         ASSERT_RTNL();
 302
 303         if (dev->mtu < IPV6_MIN_MTU)
 304                 return NULL;
 305
 306         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 307
 308         if (ndev == NULL)
 309                 return NULL;
 310
 311         rwlock_init(&ndev->lock);
 312         ndev->dev = dev;
 313         INIT_LIST_HEAD(&ndev->addr_list);
 314         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 315                     (unsigned long)ndev);
 316         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 317         ndev->cnf.mtu6 = dev->mtu;
 318         ndev->cnf.sysctl = NULL;
 319         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 320         if (ndev->nd_parms == NULL) {
 321                 kfree(ndev);
 322                 return NULL;
 323         }
 324         if (ndev->cnf.forwarding)
 325                 dev_disable_lro(dev);
 326         /* We refer to the device */
 327         dev_hold(dev);
 328
 329         if (snmp6_alloc_dev(ndev) < 0) {
 330                 ADBG(KERN_WARNING
 331                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 332                         __func__, dev->name);
 333                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 334                 dev_put(dev);
 335                 kfree(ndev);
 336                 return NULL;
 337         }
 338
 339         if (snmp6_register_dev(ndev) < 0) {
 340                 ADBG(KERN_WARNING
 341                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 342                         __func__, dev->name);
 343                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 344                 ndev->dead = 1;
 345                 in6_dev_finish_destroy(ndev);
 346                 return NULL;
 347         }
 348
 349         /* One reference from device.  We must do this before
 350          * we invoke __ipv6_regen_rndid().
 351          */
 352         in6_dev_hold(ndev);
 353
 354         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 355                 ndev->cnf.accept_dad = -1;
 356
 357 #if IS_ENABLED(CONFIG_IPV6_SIT)
 358         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 359                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 360                 ndev->cnf.rtr_solicits = 0;
 361         }
 362 #endif
 363
 364         INIT_LIST_HEAD(&ndev->tempaddr_list);
 365         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 366         if ((dev->flags&IFF_LOOPBACK) ||
 367             dev->type == ARPHRD_TUNNEL ||
 368             dev->type == ARPHRD_TUNNEL6 ||
 369             dev->type == ARPHRD_SIT ||
 370             dev->type == ARPHRD_NONE) {
 371                 ndev->cnf.use_tempaddr = -1;
 372         } else {
 373                 in6_dev_hold(ndev);
 374                 ipv6_regen_rndid((unsigned long) ndev);
 375         }
 376
 377         ndev->token = in6addr_any;
 378
 379         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 380                 ndev->if_flags |= IF_READY;
 381
 382         ipv6_mc_init_dev(ndev);
 383         ndev->tstamp = jiffies;
 384         addrconf_sysctl_register(ndev);
 385         /* protected by rtnl_lock */
 386         rcu_assign_pointer(dev->ip6_ptr, ndev);
 387
 388         /* Join interface-local all-node multicast group */
 389         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 390
 391         /* Join all-node multicast group */
 392         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 393
 394         /* Join all-router multicast group if forwarding is set */
 395         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 396                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 397
 398         return ndev;
 399 }
 400
 401 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 402 {
 403         struct inet6_dev *idev;
 404
 405         ASSERT_RTNL();
 406
 407         idev = __in6_dev_get(dev);
 408         if (!idev) {
 409                 idev = ipv6_add_dev(dev);
 410                 if (!idev)
 411                         return NULL;
 412         }
 413
 414         if (dev->flags&IFF_UP)
 415                 ipv6_mc_up(idev);
 416         return idev;
 417 }
 418
 419 static int inet6_netconf_msgsize_devconf(int type)
 420 {
 421         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 422                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 423
 424         /* type -1 is used for ALL */
 425         if (type == -1 || type == NETCONFA_FORWARDING)
 426                 size += nla_total_size(4);
 427 #ifdef CONFIG_IPV6_MROUTE
 428         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 429                 size += nla_total_size(4);
 430 #endif
 431
 432         return size;
 433 }
 434
 435 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 436                                       struct ipv6_devconf *devconf, u32 portid,
 437                                       u32 seq, int event, unsigned int flags,
 438                                       int type)
 439 {
 440         struct nlmsghdr  *nlh;
 441         struct netconfmsg *ncm;
 442
 443         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 444                         flags);
 445         if (nlh == NULL)
 446                 return -EMSGSIZE;
 447
 448         ncm = nlmsg_data(nlh);
 449         ncm->ncm_family = AF_INET6;
 450
 451         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 452                 goto nla_put_failure;
 453
 454         /* type -1 is used for ALL */
 455         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 456             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 457                 goto nla_put_failure;
 458 #ifdef CONFIG_IPV6_MROUTE
 459         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 460             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 461                         devconf->mc_forwarding) < 0)
 462                 goto nla_put_failure;
 463 #endif
 464         return nlmsg_end(skb, nlh);
 465
 466 nla_put_failure:
 467         nlmsg_cancel(skb, nlh);
 468         return -EMSGSIZE;
 469 }
 470
 471 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 472                                   struct ipv6_devconf *devconf)
 473 {
 474         struct sk_buff *skb;
 475         int err = -ENOBUFS;
 476
 477         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 478         if (skb == NULL)
 479                 goto errout;
 480
 481         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 482                                          RTM_NEWNETCONF, 0, type);
 483         if (err < 0) {
 484                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 485                 WARN_ON(err == -EMSGSIZE);
 486                 kfree_skb(skb);
 487                 goto errout;
 488         }
 489         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 490         return;
 491 errout:
 492         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 493 }
 494
 495 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 496         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 497         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 498 };
 499
 500 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 501                                      struct nlmsghdr *nlh)
 502 {
 503         struct net *net = sock_net(in_skb->sk);
 504         struct nlattr *tb[NETCONFA_MAX+1];
 505         struct netconfmsg *ncm;
 506         struct sk_buff *skb;
 507         struct ipv6_devconf *devconf;
 508         struct inet6_dev *in6_dev;
 509         struct net_device *dev;
 510         int ifindex;
 511         int err;
 512
 513         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 514                           devconf_ipv6_policy);
 515         if (err < 0)
 516                 goto errout;
 517
 518         err = EINVAL;
 519         if (!tb[NETCONFA_IFINDEX])
 520                 goto errout;
 521
 522         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 523         switch (ifindex) {
 524         case NETCONFA_IFINDEX_ALL:
 525                 devconf = net->ipv6.devconf_all;
 526                 break;
 527         case NETCONFA_IFINDEX_DEFAULT:
 528                 devconf = net->ipv6.devconf_dflt;
 529                 break;
 530         default:
 531                 dev = __dev_get_by_index(net, ifindex);
 532                 if (dev == NULL)
 533                         goto errout;
 534                 in6_dev = __in6_dev_get(dev);
 535                 if (in6_dev == NULL)
 536                         goto errout;
 537                 devconf = &in6_dev->cnf;
 538                 break;
 539         }
 540
 541         err = -ENOBUFS;
 542         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 543         if (skb == NULL)
 544                 goto errout;
 545
 546         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 547                                          NETLINK_CB(in_skb).portid,
 548                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 549                                          -1);
 550         if (err < 0) {
 551                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 552                 WARN_ON(err == -EMSGSIZE);
 553                 kfree_skb(skb);
 554                 goto errout;
 555         }
 556         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 557 errout:
 558         return err;
 559 }
 560
 561 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 562                                       struct netlink_callback *cb)
 563 {
 564         struct net *net = sock_net(skb->sk);
 565         int h, s_h;
 566         int idx, s_idx;
 567         struct net_device *dev;
 568         struct inet6_dev *idev;
 569         struct hlist_head *head;
 570
 571         s_h = cb->args[0];
 572         s_idx = idx = cb->args[1];
 573
 574         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 575                 idx = 0;
 576                 head = &net->dev_index_head[h];
 577                 rcu_read_lock();
 578                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 579                           net->dev_base_seq;
 580                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 581                         if (idx < s_idx)
 582                                 goto cont;
 583                         idev = __in6_dev_get(dev);
 584                         if (!idev)
 585                                 goto cont;
 586
 587                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 588                                                        &idev->cnf,
 589                                                        NETLINK_CB(cb->skb).portid,
 590                                                        cb->nlh->nlmsg_seq,
 591                                                        RTM_NEWNETCONF,
 592                                                        NLM_F_MULTI,
 593                                                        -1) <= 0) {
 594                                 rcu_read_unlock();
 595                                 goto done;
 596                         }
 597                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 598 cont:
 599                         idx++;
 600                 }
 601                 rcu_read_unlock();
 602         }
 603         if (h == NETDEV_HASHENTRIES) {
 604                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 605                                                net->ipv6.devconf_all,
 606                                                NETLINK_CB(cb->skb).portid,
 607                                                cb->nlh->nlmsg_seq,
 608                                                RTM_NEWNETCONF, NLM_F_MULTI,
 609                                                -1) <= 0)
 610                         goto done;
 611                 else
 612                         h++;
 613         }
 614         if (h == NETDEV_HASHENTRIES + 1) {
 615                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 616                                                net->ipv6.devconf_dflt,
 617                                                NETLINK_CB(cb->skb).portid,
 618                                                cb->nlh->nlmsg_seq,
 619                                                RTM_NEWNETCONF, NLM_F_MULTI,
 620                                                -1) <= 0)
 621                         goto done;
 622                 else
 623                         h++;
 624         }
 625 done:
 626         cb->args[0] = h;
 627         cb->args[1] = idx;
 628
 629         return skb->len;
 630 }
 631
 632 #ifdef CONFIG_SYSCTL
 633 static void dev_forward_change(struct inet6_dev *idev)
 634 {
 635         struct net_device *dev;
 636         struct inet6_ifaddr *ifa;
 637
 638         if (!idev)
 639                 return;
 640         dev = idev->dev;
 641         if (idev->cnf.forwarding)
 642                 dev_disable_lro(dev);
 643         if (dev->flags & IFF_MULTICAST) {
 644                 if (idev->cnf.forwarding) {
 645                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 646                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 647                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 648                 } else {
 649                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 650                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 651                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 652                 }
 653         }
 654
 655         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 656                 if (ifa->flags&IFA_F_TENTATIVE)
 657                         continue;
 658                 if (idev->cnf.forwarding)
 659                         addrconf_join_anycast(ifa);
 660                 else
 661                         addrconf_leave_anycast(ifa);
 662         }
 663         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 664                                      dev->ifindex, &idev->cnf);
 665 }
 666
 667
 668 static void addrconf_forward_change(struct net *net, __s32 newf)
 669 {
 670         struct net_device *dev;
 671         struct inet6_dev *idev;
 672
 673         for_each_netdev(net, dev) {
 674                 idev = __in6_dev_get(dev);
 675                 if (idev) {
 676                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 677                         idev->cnf.forwarding = newf;
 678                         if (changed)
 679                                 dev_forward_change(idev);
 680                 }
 681         }
 682 }
 683
 684 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 685 {
 686         struct net *net;
 687         int old;
 688
 689         if (!rtnl_trylock())
 690                 return restart_syscall();
 691
 692         net = (struct net *)table->extra2;
 693         old = *p;
 694         *p = newf;
 695
 696         if (p == &net->ipv6.devconf_dflt->forwarding) {
 697                 if ((!newf) ^ (!old))
 698                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 699                                                      NETCONFA_IFINDEX_DEFAULT,
 700                                                      net->ipv6.devconf_dflt);
 701                 rtnl_unlock();
 702                 return 0;
 703         }
 704
 705         if (p == &net->ipv6.devconf_all->forwarding) {
 706                 net->ipv6.devconf_dflt->forwarding = newf;
 707                 addrconf_forward_change(net, newf);
 708                 if ((!newf) ^ (!old))
 709                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 710                                                      NETCONFA_IFINDEX_ALL,
 711                                                      net->ipv6.devconf_all);
 712         } else if ((!newf) ^ (!old))
 713                 dev_forward_change((struct inet6_dev *)table->extra1);
 714         rtnl_unlock();
 715
 716         if (newf)
 717                 rt6_purge_dflt_routers(net);
 718         return 1;
 719 }
 720 #endif
 721
 722 /* Nobody refers to this ifaddr, destroy it */
 723 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 724 {
 725         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 726
 727 #ifdef NET_REFCNT_DEBUG
 728         pr_debug("%s\n", __func__);
 729 #endif
 730
 731         in6_dev_put(ifp->idev);
 732
 733         if (del_timer(&ifp->dad_timer))
 734                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 735
 736         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 737                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 738                 return;
 739         }
 740         ip6_rt_put(ifp->rt);
 741
 742         kfree_rcu(ifp, rcu);
 743 }
 744
 745 static void
 746 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 747 {
 748         struct list_head *p;
 749         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 750
 751         /*
 752          * Each device address list is sorted in order of scope -
 753          * global before linklocal.
 754          */
 755         list_for_each(p, &idev->addr_list) {
 756                 struct inet6_ifaddr *ifa
 757                         = list_entry(p, struct inet6_ifaddr, if_list);
 758                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 759                         break;
 760         }
 761
 762         list_add_tail(&ifp->if_list, p);
 763 }
 764
 765 static u32 inet6_addr_hash(const struct in6_addr *addr)
 766 {
 767         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 768 }
 769
 770 /* On success it returns ifp with increased reference count */
 771
 772 static struct inet6_ifaddr *
 773 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 774               const struct in6_addr *peer_addr, int pfxlen,
 775               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 776 {
 777         struct inet6_ifaddr *ifa = NULL;
 778         struct rt6_info *rt;
 779         unsigned int hash;
 780         int err = 0;
 781         int addr_type = ipv6_addr_type(addr);
 782
 783         if (addr_type == IPV6_ADDR_ANY ||
 784             addr_type & IPV6_ADDR_MULTICAST ||
 785             (!(idev->dev->flags & IFF_LOOPBACK) &&
 786              addr_type & IPV6_ADDR_LOOPBACK))
 787                 return ERR_PTR(-EADDRNOTAVAIL);
 788
 789         rcu_read_lock_bh();
 790         if (idev->dead) {
 791                 err = -ENODEV;                  /*XXX*/
 792                 goto out2;
 793         }
 794
 795         if (idev->cnf.disable_ipv6) {
 796                 err = -EACCES;
 797                 goto out2;
 798         }
 799
 800         spin_lock(&addrconf_hash_lock);
 801
 802         /* Ignore adding duplicate addresses on an interface */
 803         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 804                 ADBG("ipv6_add_addr: already assigned\n");
 805                 err = -EEXIST;
 806                 goto out;
 807         }
 808
 809         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 810
 811         if (ifa == NULL) {
 812                 ADBG("ipv6_add_addr: malloc failed\n");
 813                 err = -ENOBUFS;
 814                 goto out;
 815         }
 816
 817         rt = addrconf_dst_alloc(idev, addr, false);
 818         if (IS_ERR(rt)) {
 819                 err = PTR_ERR(rt);
 820                 goto out;
 821         }
 822
 823         ifa->addr = *addr;
 824         if (peer_addr)
 825                 ifa->peer_addr = *peer_addr;
 826
 827         spin_lock_init(&ifa->lock);
 828         spin_lock_init(&ifa->state_lock);
 829         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 830                     (unsigned long)ifa);
 831         INIT_HLIST_NODE(&ifa->addr_lst);
 832         ifa->scope = scope;
 833         ifa->prefix_len = pfxlen;
 834         ifa->flags = flags | IFA_F_TENTATIVE;
 835         ifa->valid_lft = valid_lft;
 836         ifa->prefered_lft = prefered_lft;
 837         ifa->cstamp = ifa->tstamp = jiffies;
 838         ifa->tokenized = false;
 839
 840         ifa->rt = rt;
 841
 842         ifa->idev = idev;
 843         in6_dev_hold(idev);
 844         /* For caller */
 845         in6_ifa_hold(ifa);
 846
 847         /* Add to big hash table */
 848         hash = inet6_addr_hash(addr);
 849
 850         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 851         spin_unlock(&addrconf_hash_lock);
 852
 853         write_lock(&idev->lock);
 854         /* Add to inet6_dev unicast addr list. */
 855         ipv6_link_dev_addr(idev, ifa);
 856
 857         if (ifa->flags&IFA_F_TEMPORARY) {
 858                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 859                 in6_ifa_hold(ifa);
 860         }
 861
 862         in6_ifa_hold(ifa);
 863         write_unlock(&idev->lock);
 864 out2:
 865         rcu_read_unlock_bh();
 866
 867         if (likely(err == 0))
 868                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 869         else {
 870                 kfree(ifa);
 871                 ifa = ERR_PTR(err);
 872         }
 873
 874         return ifa;
 875 out:
 876         spin_unlock(&addrconf_hash_lock);
 877         goto out2;
 878 }
 879
 880 /* This function wants to get referenced ifp and releases it before return */
 881
 882 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 883 {
 884         struct inet6_ifaddr *ifa, *ifn;
 885         struct inet6_dev *idev = ifp->idev;
 886         int state;
 887         int deleted = 0, onlink = 0;
 888         unsigned long expires = jiffies;
 889
 890         spin_lock_bh(&ifp->state_lock);
 891         state = ifp->state;
 892         ifp->state = INET6_IFADDR_STATE_DEAD;
 893         spin_unlock_bh(&ifp->state_lock);
 894
 895         if (state == INET6_IFADDR_STATE_DEAD)
 896                 goto out;
 897
 898         spin_lock_bh(&addrconf_hash_lock);
 899         hlist_del_init_rcu(&ifp->addr_lst);
 900         spin_unlock_bh(&addrconf_hash_lock);
 901
 902         write_lock_bh(&idev->lock);
 903
 904         if (ifp->flags&IFA_F_TEMPORARY) {
 905                 list_del(&ifp->tmp_list);
 906                 if (ifp->ifpub) {
 907                         in6_ifa_put(ifp->ifpub);
 908                         ifp->ifpub = NULL;
 909                 }
 910                 __in6_ifa_put(ifp);
 911         }
 912
 913         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 914                 if (ifa == ifp) {
 915                         list_del_init(&ifp->if_list);
 916                         __in6_ifa_put(ifp);
 917
 918                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 919                                 break;
 920                         deleted = 1;
 921                         continue;
 922                 } else if (ifp->flags & IFA_F_PERMANENT) {
 923                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 924                                               ifp->prefix_len)) {
 925                                 if (ifa->flags & IFA_F_PERMANENT) {
 926                                         onlink = 1;
 927                                         if (deleted)
 928                                                 break;
 929                                 } else {
 930                                         unsigned long lifetime;
 931
 932                                         if (!onlink)
 933                                                 onlink = -1;
 934
 935                                         spin_lock(&ifa->lock);
 936
 937                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 938                                         /*
 939                                          * Note: Because this address is
 940                                          * not permanent, lifetime <
 941                                          * LONG_MAX / HZ here.
 942                                          */
 943                                         if (time_before(expires,
 944                                                         ifa->tstamp + lifetime * HZ))
 945                                                 expires = ifa->tstamp + lifetime * HZ;
 946                                         spin_unlock(&ifa->lock);
 947                                 }
 948                         }
 949                 }
 950         }
 951         write_unlock_bh(&idev->lock);
 952
 953         addrconf_del_dad_timer(ifp);
 954
 955         ipv6_ifa_notify(RTM_DELADDR, ifp);
 956
 957         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
 958
 959         /*
 960          * Purge or update corresponding prefix
 961          *
 962          * 1) we don't purge prefix here if address was not permanent.
 963          *    prefix is managed by its own lifetime.
 964          * 2) if there're no addresses, delete prefix.
 965          * 3) if there're still other permanent address(es),
 966          *    corresponding prefix is still permanent.
 967          * 4) otherwise, update prefix lifetime to the
 968          *    longest valid lifetime among the corresponding
 969          *    addresses on the device.
 970          *    Note: subsequent RA will update lifetime.
 971          *
 972          * --yoshfuji
 973          */
 974         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 975                 struct in6_addr prefix;
 976                 struct rt6_info *rt;
 977
 978                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 979
 980                 rt = addrconf_get_prefix_route(&prefix,
 981                                                ifp->prefix_len,
 982                                                ifp->idev->dev,
 983                                                0, RTF_GATEWAY | RTF_DEFAULT);
 984
 985                 if (rt) {
 986                         if (onlink == 0) {
 987                                 ip6_del_rt(rt);
 988                                 rt = NULL;
 989                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 990                                 rt6_set_expires(rt, expires);
 991                         }
 992                 }
 993                 ip6_rt_put(rt);
 994         }
 995
 996         /* clean up prefsrc entries */
 997         rt6_remove_prefsrc(ifp);
 998 out:
 999         in6_ifa_put(ifp);
1000 }
1001
1002 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1003 {
1004         struct inet6_dev *idev = ifp->idev;
1005         struct in6_addr addr, *tmpaddr;
1006         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1007         unsigned long regen_advance;
1008         int tmp_plen;
1009         int ret = 0;
1010         u32 addr_flags;
1011         unsigned long now = jiffies;
1012
1013         write_lock(&idev->lock);
1014         if (ift) {
1015                 spin_lock_bh(&ift->lock);
1016                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1017                 spin_unlock_bh(&ift->lock);
1018                 tmpaddr = &addr;
1019         } else {
1020                 tmpaddr = NULL;
1021         }
1022 retry:
1023         in6_dev_hold(idev);
1024         if (idev->cnf.use_tempaddr <= 0) {
1025                 write_unlock(&idev->lock);
1026                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1027                 in6_dev_put(idev);
1028                 ret = -1;
1029                 goto out;
1030         }
1031         spin_lock_bh(&ifp->lock);
1032         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1033                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1034                 spin_unlock_bh(&ifp->lock);
1035                 write_unlock(&idev->lock);
1036                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1037                         __func__);
1038                 in6_dev_put(idev);
1039                 ret = -1;
1040                 goto out;
1041         }
1042         in6_ifa_hold(ifp);
1043         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1044         __ipv6_try_regen_rndid(idev, tmpaddr);
1045         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1046         age = (now - ifp->tstamp) / HZ;
1047         tmp_valid_lft = min_t(__u32,
1048                               ifp->valid_lft,
1049                               idev->cnf.temp_valid_lft + age);
1050         tmp_prefered_lft = min_t(__u32,
1051                                  ifp->prefered_lft,
1052                                  idev->cnf.temp_prefered_lft + age -
1053                                  idev->cnf.max_desync_factor);
1054         tmp_plen = ifp->prefix_len;
1055         tmp_tstamp = ifp->tstamp;
1056         spin_unlock_bh(&ifp->lock);
1057
1058         regen_advance = idev->cnf.regen_max_retry *
1059                         idev->cnf.dad_transmits *
1060                         idev->nd_parms->retrans_time / HZ;
1061         write_unlock(&idev->lock);
1062
1063         /* A temporary address is created only if this calculated Preferred
1064          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1065          * an implementation must not create a temporary address with a zero
1066          * Preferred Lifetime.
1067          */
1068         if (tmp_prefered_lft <= regen_advance) {
1069                 in6_ifa_put(ifp);
1070                 in6_dev_put(idev);
1071                 ret = -1;
1072                 goto out;
1073         }
1074
1075         addr_flags = IFA_F_TEMPORARY;
1076         /* set in addrconf_prefix_rcv() */
1077         if (ifp->flags & IFA_F_OPTIMISTIC)
1078                 addr_flags |= IFA_F_OPTIMISTIC;
1079
1080         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1081                             ipv6_addr_scope(&addr), addr_flags,
1082                             tmp_valid_lft, tmp_prefered_lft);
1083         if (IS_ERR(ift)) {
1084                 in6_ifa_put(ifp);
1085                 in6_dev_put(idev);
1086                 pr_info("%s: retry temporary address regeneration\n", __func__);
1087                 tmpaddr = &addr;
1088                 write_lock(&idev->lock);
1089                 goto retry;
1090         }
1091
1092         spin_lock_bh(&ift->lock);
1093         ift->ifpub = ifp;
1094         ift->cstamp = now;
1095         ift->tstamp = tmp_tstamp;
1096         spin_unlock_bh(&ift->lock);
1097
1098         addrconf_dad_start(ift);
1099         in6_ifa_put(ift);
1100         in6_dev_put(idev);
1101 out:
1102         return ret;
1103 }
1104
1105 /*
1106  *      Choose an appropriate source address (RFC3484)
1107  */
1108 enum {
1109         IPV6_SADDR_RULE_INIT = 0,
1110         IPV6_SADDR_RULE_LOCAL,
1111         IPV6_SADDR_RULE_SCOPE,
1112         IPV6_SADDR_RULE_PREFERRED,
1113 #ifdef CONFIG_IPV6_MIP6
1114         IPV6_SADDR_RULE_HOA,
1115 #endif
1116         IPV6_SADDR_RULE_OIF,
1117         IPV6_SADDR_RULE_LABEL,
1118         IPV6_SADDR_RULE_PRIVACY,
1119         IPV6_SADDR_RULE_ORCHID,
1120         IPV6_SADDR_RULE_PREFIX,
1121         IPV6_SADDR_RULE_MAX
1122 };
1123
1124 struct ipv6_saddr_score {
1125         int                     rule;
1126         int                     addr_type;
1127         struct inet6_ifaddr     *ifa;
1128         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1129         int                     scopedist;
1130         int                     matchlen;
1131 };
1132
1133 struct ipv6_saddr_dst {
1134         const struct in6_addr *addr;
1135         int ifindex;
1136         int scope;
1137         int label;
1138         unsigned int prefs;
1139 };
1140
1141 static inline int ipv6_saddr_preferred(int type)
1142 {
1143         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1144                 return 1;
1145         return 0;
1146 }
1147
1148 static int ipv6_get_saddr_eval(struct net *net,
1149                                struct ipv6_saddr_score *score,
1150                                struct ipv6_saddr_dst *dst,
1151                                int i)
1152 {
1153         int ret;
1154
1155         if (i <= score->rule) {
1156                 switch (i) {
1157                 case IPV6_SADDR_RULE_SCOPE:
1158                         ret = score->scopedist;
1159                         break;
1160                 case IPV6_SADDR_RULE_PREFIX:
1161                         ret = score->matchlen;
1162                         break;
1163                 default:
1164                         ret = !!test_bit(i, score->scorebits);
1165                 }
1166                 goto out;
1167         }
1168
1169         switch (i) {
1170         case IPV6_SADDR_RULE_INIT:
1171                 /* Rule 0: remember if hiscore is not ready yet */
1172                 ret = !!score->ifa;
1173                 break;
1174         case IPV6_SADDR_RULE_LOCAL:
1175                 /* Rule 1: Prefer same address */
1176                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1177                 break;
1178         case IPV6_SADDR_RULE_SCOPE:
1179                 /* Rule 2: Prefer appropriate scope
1180                  *
1181                  *      ret
1182                  *       ^
1183                  *    -1 |  d 15
1184                  *    ---+--+-+---> scope
1185                  *       |
1186                  *       |             d is scope of the destination.
1187                  *  B-d  |  \
1188                  *       |   \      <- smaller scope is better if
1189                  *  B-15 |    \        if scope is enough for destinaion.
1190                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1191                  * d-C-1 | /
1192                  *       |/         <- greater is better
1193                  *   -C  /             if scope is not enough for destination.
1194                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1195                  *
1196                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1197                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1198                  * Assume B = 0 and we get C > 29.
1199                  */
1200                 ret = __ipv6_addr_src_scope(score->addr_type);
1201                 if (ret >= dst->scope)
1202                         ret = -ret;
1203                 else
1204                         ret -= 128;     /* 30 is enough */
1205                 score->scopedist = ret;
1206                 break;
1207         case IPV6_SADDR_RULE_PREFERRED:
1208                 /* Rule 3: Avoid deprecated and optimistic addresses */
1209                 ret = ipv6_saddr_preferred(score->addr_type) ||
1210                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1211                 break;
1212 #ifdef CONFIG_IPV6_MIP6
1213         case IPV6_SADDR_RULE_HOA:
1214             {
1215                 /* Rule 4: Prefer home address */
1216                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1217                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1218                 break;
1219             }
1220 #endif
1221         case IPV6_SADDR_RULE_OIF:
1222                 /* Rule 5: Prefer outgoing interface */
1223                 ret = (!dst->ifindex ||
1224                        dst->ifindex == score->ifa->idev->dev->ifindex);
1225                 break;
1226         case IPV6_SADDR_RULE_LABEL:
1227                 /* Rule 6: Prefer matching label */
1228                 ret = ipv6_addr_label(net,
1229                                       &score->ifa->addr, score->addr_type,
1230                                       score->ifa->idev->dev->ifindex) == dst->label;
1231                 break;
1232         case IPV6_SADDR_RULE_PRIVACY:
1233             {
1234                 /* Rule 7: Prefer public address
1235                  * Note: prefer temporary address if use_tempaddr >= 2
1236                  */
1237                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1238                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1239                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1240                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1241                 break;
1242             }
1243         case IPV6_SADDR_RULE_ORCHID:
1244                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1245                  *          non-ORCHID vs non-ORCHID
1246                  */
1247                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1248                         ipv6_addr_orchid(dst->addr));
1249                 break;
1250         case IPV6_SADDR_RULE_PREFIX:
1251                 /* Rule 8: Use longest matching prefix */
1252                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1253                 if (ret > score->ifa->prefix_len)
1254                         ret = score->ifa->prefix_len;
1255                 score->matchlen = ret;
1256                 break;
1257         default:
1258                 ret = 0;
1259         }
1260
1261         if (ret)
1262                 __set_bit(i, score->scorebits);
1263         score->rule = i;
1264 out:
1265         return ret;
1266 }
1267
1268 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1269                        const struct in6_addr *daddr, unsigned int prefs,
1270                        struct in6_addr *saddr)
1271 {
1272         struct ipv6_saddr_score scores[2],
1273                                 *score = &scores[0], *hiscore = &scores[1];
1274         struct ipv6_saddr_dst dst;
1275         struct net_device *dev;
1276         int dst_type;
1277
1278         dst_type = __ipv6_addr_type(daddr);
1279         dst.addr = daddr;
1280         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1281         dst.scope = __ipv6_addr_src_scope(dst_type);
1282         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1283         dst.prefs = prefs;
1284
1285         hiscore->rule = -1;
1286         hiscore->ifa = NULL;
1287
1288         rcu_read_lock();
1289
1290         for_each_netdev_rcu(net, dev) {
1291                 struct inet6_dev *idev;
1292
1293                 /* Candidate Source Address (section 4)
1294                  *  - multicast and link-local destination address,
1295                  *    the set of candidate source address MUST only
1296                  *    include addresses assigned to interfaces
1297                  *    belonging to the same link as the outgoing
1298                  *    interface.
1299                  * (- For site-local destination addresses, the
1300                  *    set of candidate source addresses MUST only
1301                  *    include addresses assigned to interfaces
1302                  *    belonging to the same site as the outgoing
1303                  *    interface.)
1304                  */
1305                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1306                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1307                     dst.ifindex && dev->ifindex != dst.ifindex)
1308                         continue;
1309
1310                 idev = __in6_dev_get(dev);
1311                 if (!idev)
1312                         continue;
1313
1314                 read_lock_bh(&idev->lock);
1315                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1316                         int i;
1317
1318                         /*
1319                          * - Tentative Address (RFC2462 section 5.4)
1320                          *  - A tentative address is not considered
1321                          *    "assigned to an interface" in the traditional
1322                          *    sense, unless it is also flagged as optimistic.
1323                          * - Candidate Source Address (section 4)
1324                          *  - In any case, anycast addresses, multicast
1325                          *    addresses, and the unspecified address MUST
1326                          *    NOT be included in a candidate set.
1327                          */
1328                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1329                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1330                                 continue;
1331
1332                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1333
1334                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1335                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1336                                 LIMIT_NETDEBUG(KERN_DEBUG
1337                                                "ADDRCONF: unspecified / multicast address "
1338                                                "assigned as unicast address on %s",
1339                                                dev->name);
1340                                 continue;
1341                         }
1342
1343                         score->rule = -1;
1344                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1345
1346                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1347                                 int minihiscore, miniscore;
1348
1349                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1350                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1351
1352                                 if (minihiscore > miniscore) {
1353                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1354                                             score->scopedist > 0) {
1355                                                 /*
1356                                                  * special case:
1357                                                  * each remaining entry
1358                                                  * has too small (not enough)
1359                                                  * scope, because ifa entries
1360                                                  * are sorted by their scope
1361                                                  * values.
1362                                                  */
1363                                                 goto try_nextdev;
1364                                         }
1365                                         break;
1366                                 } else if (minihiscore < miniscore) {
1367                                         if (hiscore->ifa)
1368                                                 in6_ifa_put(hiscore->ifa);
1369
1370                                         in6_ifa_hold(score->ifa);
1371
1372                                         swap(hiscore, score);
1373
1374                                         /* restore our iterator */
1375                                         score->ifa = hiscore->ifa;
1376
1377                                         break;
1378                                 }
1379                         }
1380                 }
1381 try_nextdev:
1382                 read_unlock_bh(&idev->lock);
1383         }
1384         rcu_read_unlock();
1385
1386         if (!hiscore->ifa)
1387                 return -EADDRNOTAVAIL;
1388
1389         *saddr = hiscore->ifa->addr;
1390         in6_ifa_put(hiscore->ifa);
1391         return 0;
1392 }
1393 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1394
1395 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1396                       unsigned char banned_flags)
1397 {
1398         struct inet6_ifaddr *ifp;
1399         int err = -EADDRNOTAVAIL;
1400
1401         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1402                 if (ifp->scope == IFA_LINK &&
1403                     !(ifp->flags & banned_flags)) {
1404                         *addr = ifp->addr;
1405                         err = 0;
1406                         break;
1407                 }
1408         }
1409         return err;
1410 }
1411
1412 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1413                     unsigned char banned_flags)
1414 {
1415         struct inet6_dev *idev;
1416         int err = -EADDRNOTAVAIL;
1417
1418         rcu_read_lock();
1419         idev = __in6_dev_get(dev);
1420         if (idev) {
1421                 read_lock_bh(&idev->lock);
1422                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1423                 read_unlock_bh(&idev->lock);
1424         }
1425         rcu_read_unlock();
1426         return err;
1427 }
1428
1429 static int ipv6_count_addresses(struct inet6_dev *idev)
1430 {
1431         int cnt = 0;
1432         struct inet6_ifaddr *ifp;
1433
1434         read_lock_bh(&idev->lock);
1435         list_for_each_entry(ifp, &idev->addr_list, if_list)
1436                 cnt++;
1437         read_unlock_bh(&idev->lock);
1438         return cnt;
1439 }
1440
1441 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1442                   const struct net_device *dev, int strict)
1443 {
1444         struct inet6_ifaddr *ifp;
1445         unsigned int hash = inet6_addr_hash(addr);
1446
1447         rcu_read_lock_bh();
1448         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1449                 if (!net_eq(dev_net(ifp->idev->dev), net))
1450                         continue;
1451                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1452                     !(ifp->flags&IFA_F_TENTATIVE) &&
1453                     (dev == NULL || ifp->idev->dev == dev ||
1454                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1455                         rcu_read_unlock_bh();
1456                         return 1;
1457                 }
1458         }
1459
1460         rcu_read_unlock_bh();
1461         return 0;
1462 }
1463 EXPORT_SYMBOL(ipv6_chk_addr);
1464
1465 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1466                                struct net_device *dev)
1467 {
1468         unsigned int hash = inet6_addr_hash(addr);
1469         struct inet6_ifaddr *ifp;
1470
1471         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1472                 if (!net_eq(dev_net(ifp->idev->dev), net))
1473                         continue;
1474                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1475                         if (dev == NULL || ifp->idev->dev == dev)
1476                                 return true;
1477                 }
1478         }
1479         return false;
1480 }
1481
1482 /* Compares an address/prefix_len with addresses on device @dev.
1483  * If one is found it returns true.
1484  */
1485 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1486         const unsigned int prefix_len, struct net_device *dev)
1487 {
1488         struct inet6_dev *idev;
1489         struct inet6_ifaddr *ifa;
1490         bool ret = false;
1491
1492         rcu_read_lock();
1493         idev = __in6_dev_get(dev);
1494         if (idev) {
1495                 read_lock_bh(&idev->lock);
1496                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1497                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1498                         if (ret)
1499                                 break;
1500                 }
1501                 read_unlock_bh(&idev->lock);
1502         }
1503         rcu_read_unlock();
1504
1505         return ret;
1506 }
1507 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1508
1509 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1510 {
1511         struct inet6_dev *idev;
1512         struct inet6_ifaddr *ifa;
1513         int     onlink;
1514
1515         onlink = 0;
1516         rcu_read_lock();
1517         idev = __in6_dev_get(dev);
1518         if (idev) {
1519                 read_lock_bh(&idev->lock);
1520                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1521                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1522                                                    ifa->prefix_len);
1523                         if (onlink)
1524                                 break;
1525                 }
1526                 read_unlock_bh(&idev->lock);
1527         }
1528         rcu_read_unlock();
1529         return onlink;
1530 }
1531 EXPORT_SYMBOL(ipv6_chk_prefix);
1532
1533 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1534                                      struct net_device *dev, int strict)
1535 {
1536         struct inet6_ifaddr *ifp, *result = NULL;
1537         unsigned int hash = inet6_addr_hash(addr);
1538
1539         rcu_read_lock_bh();
1540         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1541                 if (!net_eq(dev_net(ifp->idev->dev), net))
1542                         continue;
1543                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1544                         if (dev == NULL || ifp->idev->dev == dev ||
1545                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1546                                 result = ifp;
1547                                 in6_ifa_hold(ifp);
1548                                 break;
1549                         }
1550                 }
1551         }
1552         rcu_read_unlock_bh();
1553
1554         return result;
1555 }
1556
1557 /* Gets referenced address, destroys ifaddr */
1558
1559 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1560 {
1561         if (ifp->flags&IFA_F_PERMANENT) {
1562                 spin_lock_bh(&ifp->lock);
1563                 addrconf_del_dad_timer(ifp);
1564                 ifp->flags |= IFA_F_TENTATIVE;
1565                 if (dad_failed)
1566                         ifp->flags |= IFA_F_DADFAILED;
1567                 spin_unlock_bh(&ifp->lock);
1568                 if (dad_failed)
1569                         ipv6_ifa_notify(0, ifp);
1570                 in6_ifa_put(ifp);
1571         } else if (ifp->flags&IFA_F_TEMPORARY) {
1572                 struct inet6_ifaddr *ifpub;
1573                 spin_lock_bh(&ifp->lock);
1574                 ifpub = ifp->ifpub;
1575                 if (ifpub) {
1576                         in6_ifa_hold(ifpub);
1577                         spin_unlock_bh(&ifp->lock);
1578                         ipv6_create_tempaddr(ifpub, ifp);
1579                         in6_ifa_put(ifpub);
1580                 } else {
1581                         spin_unlock_bh(&ifp->lock);
1582                 }
1583                 ipv6_del_addr(ifp);
1584         } else
1585                 ipv6_del_addr(ifp);
1586 }
1587
1588 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1589 {
1590         int err = -ENOENT;
1591
1592         spin_lock(&ifp->state_lock);
1593         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1594                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1595                 err = 0;
1596         }
1597         spin_unlock(&ifp->state_lock);
1598
1599         return err;
1600 }
1601
1602 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1603 {
1604         struct inet6_dev *idev = ifp->idev;
1605
1606         if (addrconf_dad_end(ifp)) {
1607                 in6_ifa_put(ifp);
1608                 return;
1609         }
1610
1611         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1612                              ifp->idev->dev->name, &ifp->addr);
1613
1614         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1615                 struct in6_addr addr;
1616
1617                 addr.s6_addr32[0] = htonl(0xfe800000);
1618                 addr.s6_addr32[1] = 0;
1619
1620                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1621                     ipv6_addr_equal(&ifp->addr, &addr)) {
1622                         /* DAD failed for link-local based on MAC address */
1623                         idev->cnf.disable_ipv6 = 1;
1624
1625                         pr_info("%s: IPv6 being disabled!\n",
1626                                 ifp->idev->dev->name);
1627                 }
1628         }
1629
1630         addrconf_dad_stop(ifp, 1);
1631 }
1632
1633 /* Join to solicited addr multicast group. */
1634
1635 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1636 {
1637         struct in6_addr maddr;
1638
1639         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1640                 return;
1641
1642         addrconf_addr_solict_mult(addr, &maddr);
1643         ipv6_dev_mc_inc(dev, &maddr);
1644 }
1645
1646 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1647 {
1648         struct in6_addr maddr;
1649
1650         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1651                 return;
1652
1653         addrconf_addr_solict_mult(addr, &maddr);
1654         __ipv6_dev_mc_dec(idev, &maddr);
1655 }
1656
1657 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1658 {
1659         struct in6_addr addr;
1660         if (ifp->prefix_len == 127) /* RFC 6164 */
1661                 return;
1662         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1663         if (ipv6_addr_any(&addr))
1664                 return;
1665         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1666 }
1667
1668 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1669 {
1670         struct in6_addr addr;
1671         if (ifp->prefix_len == 127) /* RFC 6164 */
1672                 return;
1673         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1674         if (ipv6_addr_any(&addr))
1675                 return;
1676         __ipv6_dev_ac_dec(ifp->idev, &addr);
1677 }
1678
1679 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1680 {
1681         if (dev->addr_len != ETH_ALEN)
1682                 return -1;
1683         memcpy(eui, dev->dev_addr, 3);
1684         memcpy(eui + 5, dev->dev_addr + 3, 3);
1685
1686         /*
1687          * The zSeries OSA network cards can be shared among various
1688          * OS instances, but the OSA cards have only one MAC address.
1689          * This leads to duplicate address conflicts in conjunction
1690          * with IPv6 if more than one instance uses the same card.
1691          *
1692          * The driver for these cards can deliver a unique 16-bit
1693          * identifier for each instance sharing the same card.  It is
1694          * placed instead of 0xFFFE in the interface identifier.  The
1695          * "u" bit of the interface identifier is not inverted in this
1696          * case.  Hence the resulting interface identifier has local
1697          * scope according to RFC2373.
1698          */
1699         if (dev->dev_id) {
1700                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1701                 eui[4] = dev->dev_id & 0xFF;
1702         } else {
1703                 eui[3] = 0xFF;
1704                 eui[4] = 0xFE;
1705                 eui[0] ^= 2;
1706         }
1707         return 0;
1708 }
1709
1710 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1711 {
1712         if (dev->addr_len != IEEE802154_ADDR_LEN)
1713                 return -1;
1714         memcpy(eui, dev->dev_addr, 8);
1715         eui[0] ^= 2;
1716         return 0;
1717 }
1718
1719 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1720 {
1721         union fwnet_hwaddr *ha;
1722
1723         if (dev->addr_len != FWNET_ALEN)
1724                 return -1;
1725
1726         ha = (union fwnet_hwaddr *)dev->dev_addr;
1727
1728         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1729         eui[0] ^= 2;
1730         return 0;
1731 }
1732
1733 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1734 {
1735         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1736         if (dev->addr_len != ARCNET_ALEN)
1737                 return -1;
1738         memset(eui, 0, 7);
1739         eui[7] = *(u8 *)dev->dev_addr;
1740         return 0;
1741 }
1742
1743 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1744 {
1745         if (dev->addr_len != INFINIBAND_ALEN)
1746                 return -1;
1747         memcpy(eui, dev->dev_addr + 12, 8);
1748         eui[0] |= 2;
1749         return 0;
1750 }
1751
1752 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1753 {
1754         if (addr == 0)
1755                 return -1;
1756         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1757                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1758                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1759                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1760                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1761                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1762         eui[1] = 0;
1763         eui[2] = 0x5E;
1764         eui[3] = 0xFE;
1765         memcpy(eui + 4, &addr, 4);
1766         return 0;
1767 }
1768
1769 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1770 {
1771         if (dev->priv_flags & IFF_ISATAP)
1772                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1773         return -1;
1774 }
1775
1776 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1777 {
1778         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1779 }
1780
1781 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1782 {
1783         memcpy(eui, dev->perm_addr, 3);
1784         memcpy(eui + 5, dev->perm_addr + 3, 3);
1785         eui[3] = 0xFF;
1786         eui[4] = 0xFE;
1787         eui[0] ^= 2;
1788         return 0;
1789 }
1790
1791 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1792 {
1793         switch (dev->type) {
1794         case ARPHRD_ETHER:
1795         case ARPHRD_FDDI:
1796                 return addrconf_ifid_eui48(eui, dev);
1797         case ARPHRD_ARCNET:
1798                 return addrconf_ifid_arcnet(eui, dev);
1799         case ARPHRD_INFINIBAND:
1800                 return addrconf_ifid_infiniband(eui, dev);
1801         case ARPHRD_SIT:
1802                 return addrconf_ifid_sit(eui, dev);
1803         case ARPHRD_IPGRE:
1804                 return addrconf_ifid_gre(eui, dev);
1805         case ARPHRD_IEEE802154:
1806                 return addrconf_ifid_eui64(eui, dev);
1807         case ARPHRD_IEEE1394:
1808                 return addrconf_ifid_ieee1394(eui, dev);
1809         case ARPHRD_TUNNEL6:
1810                 return addrconf_ifid_ip6tnl(eui, dev);
1811         }
1812         return -1;
1813 }
1814
1815 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1816 {
1817         int err = -1;
1818         struct inet6_ifaddr *ifp;
1819
1820         read_lock_bh(&idev->lock);
1821         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1822                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1823                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1824                         err = 0;
1825                         break;
1826                 }
1827         }
1828         read_unlock_bh(&idev->lock);
1829         return err;
1830 }
1831
1832 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1833 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1834 {
1835 regen:
1836         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1837         idev->rndid[0] &= ~0x02;
1838
1839         /*
1840          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1841          * check if generated address is not inappropriate
1842          *
1843          *  - Reserved subnet anycast (RFC 2526)
1844          *      11111101 11....11 1xxxxxxx
1845          *  - ISATAP (RFC4214) 6.1
1846          *      00-00-5E-FE-xx-xx-xx-xx
1847          *  - value 0
1848          *  - XXX: already assigned to an address on the device
1849          */
1850         if (idev->rndid[0] == 0xfd &&
1851             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1852             (idev->rndid[7]&0x80))
1853                 goto regen;
1854         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1855                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1856                         goto regen;
1857                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1858                         goto regen;
1859         }
1860 }
1861
1862 static void ipv6_regen_rndid(unsigned long data)
1863 {
1864         struct inet6_dev *idev = (struct inet6_dev *) data;
1865         unsigned long expires;
1866
1867         rcu_read_lock_bh();
1868         write_lock_bh(&idev->lock);
1869
1870         if (idev->dead)
1871                 goto out;
1872
1873         __ipv6_regen_rndid(idev);
1874
1875         expires = jiffies +
1876                 idev->cnf.temp_prefered_lft * HZ -
1877                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1878                 idev->cnf.max_desync_factor * HZ;
1879         if (time_before(expires, jiffies)) {
1880                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1881                         __func__, idev->dev->name);
1882                 goto out;
1883         }
1884
1885         if (!mod_timer(&idev->regen_timer, expires))
1886                 in6_dev_hold(idev);
1887
1888 out:
1889         write_unlock_bh(&idev->lock);
1890         rcu_read_unlock_bh();
1891         in6_dev_put(idev);
1892 }
1893
1894 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1895 {
1896         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1897                 __ipv6_regen_rndid(idev);
1898 }
1899
1900 /*
1901  *      Add prefix route.
1902  */
1903
1904 static void
1905 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1906                       unsigned long expires, u32 flags)
1907 {
1908         struct fib6_config cfg = {
1909                 .fc_table = RT6_TABLE_PREFIX,
1910                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1911                 .fc_ifindex = dev->ifindex,
1912                 .fc_expires = expires,
1913                 .fc_dst_len = plen,
1914                 .fc_flags = RTF_UP | flags,
1915                 .fc_nlinfo.nl_net = dev_net(dev),
1916                 .fc_protocol = RTPROT_KERNEL,
1917         };
1918
1919         cfg.fc_dst = *pfx;
1920
1921         /* Prevent useless cloning on PtP SIT.
1922            This thing is done here expecting that the whole
1923            class of non-broadcast devices need not cloning.
1924          */
1925 #if IS_ENABLED(CONFIG_IPV6_SIT)
1926         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1927                 cfg.fc_flags |= RTF_NONEXTHOP;
1928 #endif
1929
1930         ip6_route_add(&cfg);
1931 }
1932
1933
1934 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1935                                                   int plen,
1936                                                   const struct net_device *dev,
1937                                                   u32 flags, u32 noflags)
1938 {
1939         struct fib6_node *fn;
1940         struct rt6_info *rt = NULL;
1941         struct fib6_table *table;
1942
1943         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1944         if (table == NULL)
1945                 return NULL;
1946
1947         read_lock_bh(&table->tb6_lock);
1948         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1949         if (!fn)
1950                 goto out;
1951         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1952                 if (rt->dst.dev->ifindex != dev->ifindex)
1953                         continue;
1954                 if ((rt->rt6i_flags & flags) != flags)
1955                         continue;
1956                 if ((rt->rt6i_flags & noflags) != 0)
1957                         continue;
1958                 dst_hold(&rt->dst);
1959                 break;
1960         }
1961 out:
1962         read_unlock_bh(&table->tb6_lock);
1963         return rt;
1964 }
1965
1966
1967 /* Create "default" multicast route to the interface */
1968
1969 static void addrconf_add_mroute(struct net_device *dev)
1970 {
1971         struct fib6_config cfg = {
1972                 .fc_table = RT6_TABLE_LOCAL,
1973                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1974                 .fc_ifindex = dev->ifindex,
1975                 .fc_dst_len = 8,
1976                 .fc_flags = RTF_UP,
1977                 .fc_nlinfo.nl_net = dev_net(dev),
1978         };
1979
1980         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1981
1982         ip6_route_add(&cfg);
1983 }
1984
1985 #if IS_ENABLED(CONFIG_IPV6_SIT)
1986 static void sit_route_add(struct net_device *dev)
1987 {
1988         struct fib6_config cfg = {
1989                 .fc_table = RT6_TABLE_MAIN,
1990                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1991                 .fc_ifindex = dev->ifindex,
1992                 .fc_dst_len = 96,
1993                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1994                 .fc_nlinfo.nl_net = dev_net(dev),
1995         };
1996
1997         /* prefix length - 96 bits "::d.d.d.d" */
1998         ip6_route_add(&cfg);
1999 }
2000 #endif
2001
2002 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2003 {
2004         struct inet6_dev *idev;
2005
2006         ASSERT_RTNL();
2007
2008         idev = ipv6_find_idev(dev);
2009         if (!idev)
2010                 return ERR_PTR(-ENOBUFS);
2011
2012         if (idev->cnf.disable_ipv6)
2013                 return ERR_PTR(-EACCES);
2014
2015         /* Add default multicast route */
2016         if (!(dev->flags & IFF_LOOPBACK))
2017                 addrconf_add_mroute(dev);
2018
2019         return idev;
2020 }
2021
2022 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2023 {
2024         struct prefix_info *pinfo;
2025         __u32 valid_lft;
2026         __u32 prefered_lft;
2027         int addr_type;
2028         struct inet6_dev *in6_dev;
2029         struct net *net = dev_net(dev);
2030
2031         pinfo = (struct prefix_info *) opt;
2032
2033         if (len < sizeof(struct prefix_info)) {
2034                 ADBG("addrconf: prefix option too short\n");
2035                 return;
2036         }
2037
2038         /*
2039          *      Validation checks ([ADDRCONF], page 19)
2040          */
2041
2042         addr_type = ipv6_addr_type(&pinfo->prefix);
2043
2044         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2045                 return;
2046
2047         valid_lft = ntohl(pinfo->valid);
2048         prefered_lft = ntohl(pinfo->prefered);
2049
2050         if (prefered_lft > valid_lft) {
2051                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2052                 return;
2053         }
2054
2055         in6_dev = in6_dev_get(dev);
2056
2057         if (in6_dev == NULL) {
2058                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2059                                     dev->name);
2060                 return;
2061         }
2062
2063         /*
2064          *      Two things going on here:
2065          *      1) Add routes for on-link prefixes
2066          *      2) Configure prefixes with the auto flag set
2067          */
2068
2069         if (pinfo->onlink) {
2070                 struct rt6_info *rt;
2071                 unsigned long rt_expires;
2072
2073                 /* Avoid arithmetic overflow. Really, we could
2074                  * save rt_expires in seconds, likely valid_lft,
2075                  * but it would require division in fib gc, that it
2076                  * not good.
2077                  */
2078                 if (HZ > USER_HZ)
2079                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2080                 else
2081                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2082
2083                 if (addrconf_finite_timeout(rt_expires))
2084                         rt_expires *= HZ;
2085
2086                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2087                                                pinfo->prefix_len,
2088                                                dev,
2089                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2090                                                RTF_GATEWAY | RTF_DEFAULT);
2091
2092                 if (rt) {
2093                         /* Autoconf prefix route */
2094                         if (valid_lft == 0) {
2095                                 ip6_del_rt(rt);
2096                                 rt = NULL;
2097                         } else if (addrconf_finite_timeout(rt_expires)) {
2098                                 /* not infinity */
2099                                 rt6_set_expires(rt, jiffies + rt_expires);
2100                         } else {
2101                                 rt6_clean_expires(rt);
2102                         }
2103                 } else if (valid_lft) {
2104                         clock_t expires = 0;
2105                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2106                         if (addrconf_finite_timeout(rt_expires)) {
2107                                 /* not infinity */
2108                                 flags |= RTF_EXPIRES;
2109                                 expires = jiffies_to_clock_t(rt_expires);
2110                         }
2111                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2112                                               dev, expires, flags);
2113                 }
2114                 ip6_rt_put(rt);
2115         }
2116
2117         /* Try to figure out our local address for this prefix */
2118
2119         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2120                 struct inet6_ifaddr *ifp;
2121                 struct in6_addr addr;
2122                 int create = 0, update_lft = 0;
2123                 bool tokenized = false;
2124
2125                 if (pinfo->prefix_len == 64) {
2126                         memcpy(&addr, &pinfo->prefix, 8);
2127
2128                         if (!ipv6_addr_any(&in6_dev->token)) {
2129                                 read_lock_bh(&in6_dev->lock);
2130                                 memcpy(addr.s6_addr + 8,
2131                                        in6_dev->token.s6_addr + 8, 8);
2132                                 read_unlock_bh(&in6_dev->lock);
2133                                 tokenized = true;
2134                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2135                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2136                                 in6_dev_put(in6_dev);
2137                                 return;
2138                         }
2139                         goto ok;
2140                 }
2141                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2142                                     pinfo->prefix_len);
2143                 in6_dev_put(in6_dev);
2144                 return;
2145
2146 ok:
2147
2148                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2149
2150                 if (ifp == NULL && valid_lft) {
2151                         int max_addresses = in6_dev->cnf.max_addresses;
2152                         u32 addr_flags = 0;
2153
2154 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2155                         if (in6_dev->cnf.optimistic_dad &&
2156                             !net->ipv6.devconf_all->forwarding && sllao)
2157                                 addr_flags = IFA_F_OPTIMISTIC;
2158 #endif
2159
2160                         /* Do not allow to create too much of autoconfigured
2161                          * addresses; this would be too easy way to crash kernel.
2162                          */
2163                         if (!max_addresses ||
2164                             ipv6_count_addresses(in6_dev) < max_addresses)
2165                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2166                                                     pinfo->prefix_len,
2167                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2168                                                     addr_flags, valid_lft,
2169                                                     prefered_lft);
2170
2171                         if (IS_ERR_OR_NULL(ifp)) {
2172                                 in6_dev_put(in6_dev);
2173                                 return;
2174                         }
2175
2176                         update_lft = 0;
2177                         create = 1;
2178                         ifp->cstamp = jiffies;
2179                         ifp->tokenized = tokenized;
2180                         addrconf_dad_start(ifp);
2181                 }
2182
2183                 if (ifp) {
2184                         int flags;
2185                         unsigned long now;
2186                         struct inet6_ifaddr *ift;
2187                         u32 stored_lft;
2188
2189                         /* update lifetime (RFC2462 5.5.3 e) */
2190                         spin_lock(&ifp->lock);
2191                         now = jiffies;
2192                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2193                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2194                         else
2195                                 stored_lft = 0;
2196                         if (!update_lft && !create && stored_lft) {
2197                                 const u32 minimum_lft = min(
2198                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2199                                 valid_lft = max(valid_lft, minimum_lft);
2200
2201                                 /* RFC4862 Section 5.5.3e:
2202                                  * "Note that the preferred lifetime of the
2203                                  *  corresponding address is always reset to
2204                                  *  the Preferred Lifetime in the received
2205                                  *  Prefix Information option, regardless of
2206                                  *  whether the valid lifetime is also reset or
2207                                  *  ignored."
2208                                  *
2209                                  * So we should always update prefered_lft here.
2210                                  */
2211                                 update_lft = 1;
2212                         }
2213
2214                         if (update_lft) {
2215                                 ifp->valid_lft = valid_lft;
2216                                 ifp->prefered_lft = prefered_lft;
2217                                 ifp->tstamp = now;
2218                                 flags = ifp->flags;
2219                                 ifp->flags &= ~IFA_F_DEPRECATED;
2220                                 spin_unlock(&ifp->lock);
2221
2222                                 if (!(flags&IFA_F_TENTATIVE))
2223                                         ipv6_ifa_notify(0, ifp);
2224                         } else
2225                                 spin_unlock(&ifp->lock);
2226
2227                         read_lock_bh(&in6_dev->lock);
2228                         /* update all temporary addresses in the list */
2229                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2230                                             tmp_list) {
2231                                 int age, max_valid, max_prefered;
2232
2233                                 if (ifp != ift->ifpub)
2234                                         continue;
2235
2236                                 /*
2237                                  * RFC 4941 section 3.3:
2238                                  * If a received option will extend the lifetime
2239                                  * of a public address, the lifetimes of
2240                                  * temporary addresses should be extended,
2241                                  * subject to the overall constraint that no
2242                                  * temporary addresses should ever remain
2243                                  * "valid" or "preferred" for a time longer than
2244                                  * (TEMP_VALID_LIFETIME) or
2245                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2246                                  * respectively.
2247                                  */
2248                                 age = (now - ift->cstamp) / HZ;
2249                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2250                                 if (max_valid < 0)
2251                                         max_valid = 0;
2252
2253                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2254                                                in6_dev->cnf.max_desync_factor -
2255                                                age;
2256                                 if (max_prefered < 0)
2257                                         max_prefered = 0;
2258
2259                                 if (valid_lft > max_valid)
2260                                         valid_lft = max_valid;
2261
2262                                 if (prefered_lft > max_prefered)
2263                                         prefered_lft = max_prefered;
2264
2265                                 spin_lock(&ift->lock);
2266                                 flags = ift->flags;
2267                                 ift->valid_lft = valid_lft;
2268                                 ift->prefered_lft = prefered_lft;
2269                                 ift->tstamp = now;
2270                                 if (prefered_lft > 0)
2271                                         ift->flags &= ~IFA_F_DEPRECATED;
2272
2273                                 spin_unlock(&ift->lock);
2274                                 if (!(flags&IFA_F_TENTATIVE))
2275                                         ipv6_ifa_notify(0, ift);
2276                         }
2277
2278                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2279                                 /*
2280                                  * When a new public address is created as
2281                                  * described in [ADDRCONF], also create a new
2282                                  * temporary address. Also create a temporary
2283                                  * address if it's enabled but no temporary
2284                                  * address currently exists.
2285                                  */
2286                                 read_unlock_bh(&in6_dev->lock);
2287                                 ipv6_create_tempaddr(ifp, NULL);
2288                         } else {
2289                                 read_unlock_bh(&in6_dev->lock);
2290                         }
2291
2292                         in6_ifa_put(ifp);
2293                         addrconf_verify(0);
2294                 }
2295         }
2296         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2297         in6_dev_put(in6_dev);
2298 }
2299
2300 /*
2301  *      Set destination address.
2302  *      Special case for SIT interfaces where we create a new "virtual"
2303  *      device.
2304  */
2305 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2306 {
2307         struct in6_ifreq ireq;
2308         struct net_device *dev;
2309         int err = -EINVAL;
2310
2311         rtnl_lock();
2312
2313         err = -EFAULT;
2314         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2315                 goto err_exit;
2316
2317         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2318
2319         err = -ENODEV;
2320         if (dev == NULL)
2321                 goto err_exit;
2322
2323 #if IS_ENABLED(CONFIG_IPV6_SIT)
2324         if (dev->type == ARPHRD_SIT) {
2325                 const struct net_device_ops *ops = dev->netdev_ops;
2326                 struct ifreq ifr;
2327                 struct ip_tunnel_parm p;
2328
2329                 err = -EADDRNOTAVAIL;
2330                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2331                         goto err_exit;
2332
2333                 memset(&p, 0, sizeof(p));
2334                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2335                 p.iph.saddr = 0;
2336                 p.iph.version = 4;
2337                 p.iph.ihl = 5;
2338                 p.iph.protocol = IPPROTO_IPV6;
2339                 p.iph.ttl = 64;
2340                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2341
2342                 if (ops->ndo_do_ioctl) {
2343                         mm_segment_t oldfs = get_fs();
2344
2345                         set_fs(KERNEL_DS);
2346                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2347                         set_fs(oldfs);
2348                 } else
2349                         err = -EOPNOTSUPP;
2350
2351                 if (err == 0) {
2352                         err = -ENOBUFS;
2353                         dev = __dev_get_by_name(net, p.name);
2354                         if (!dev)
2355                                 goto err_exit;
2356                         err = dev_open(dev);
2357                 }
2358         }
2359 #endif
2360
2361 err_exit:
2362         rtnl_unlock();
2363         return err;
2364 }
2365
2366 /*
2367  *      Manual configuration of address on an interface
2368  */
2369 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2370                           const struct in6_addr *peer_pfx,
2371                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2372                           __u32 valid_lft)
2373 {
2374         struct inet6_ifaddr *ifp;
2375         struct inet6_dev *idev;
2376         struct net_device *dev;
2377         int scope;
2378         u32 flags;
2379         clock_t expires;
2380         unsigned long timeout;
2381
2382         ASSERT_RTNL();
2383
2384         if (plen > 128)
2385                 return -EINVAL;
2386
2387         /* check the lifetime */
2388         if (!valid_lft || prefered_lft > valid_lft)
2389                 return -EINVAL;
2390
2391         dev = __dev_get_by_index(net, ifindex);
2392         if (!dev)
2393                 return -ENODEV;
2394
2395         idev = addrconf_add_dev(dev);
2396         if (IS_ERR(idev))
2397                 return PTR_ERR(idev);
2398
2399         scope = ipv6_addr_scope(pfx);
2400
2401         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2402         if (addrconf_finite_timeout(timeout)) {
2403                 expires = jiffies_to_clock_t(timeout * HZ);
2404                 valid_lft = timeout;
2405                 flags = RTF_EXPIRES;
2406         } else {
2407                 expires = 0;
2408                 flags = 0;
2409                 ifa_flags |= IFA_F_PERMANENT;
2410         }
2411
2412         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2413         if (addrconf_finite_timeout(timeout)) {
2414                 if (timeout == 0)
2415                         ifa_flags |= IFA_F_DEPRECATED;
2416                 prefered_lft = timeout;
2417         }
2418
2419         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2420                             valid_lft, prefered_lft);
2421
2422         if (!IS_ERR(ifp)) {
2423                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2424                                       expires, flags);
2425                 /*
2426                  * Note that section 3.1 of RFC 4429 indicates
2427                  * that the Optimistic flag should not be set for
2428                  * manually configured addresses
2429                  */
2430                 addrconf_dad_start(ifp);
2431                 in6_ifa_put(ifp);
2432                 addrconf_verify(0);
2433                 return 0;
2434         }
2435
2436         return PTR_ERR(ifp);
2437 }
2438
2439 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2440                           unsigned int plen)
2441 {
2442         struct inet6_ifaddr *ifp;
2443         struct inet6_dev *idev;
2444         struct net_device *dev;
2445
2446         if (plen > 128)
2447                 return -EINVAL;
2448
2449         dev = __dev_get_by_index(net, ifindex);
2450         if (!dev)
2451                 return -ENODEV;
2452
2453         if ((idev = __in6_dev_get(dev)) == NULL)
2454                 return -ENXIO;
2455
2456         read_lock_bh(&idev->lock);
2457         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2458                 if (ifp->prefix_len == plen &&
2459                     ipv6_addr_equal(pfx, &ifp->addr)) {
2460                         in6_ifa_hold(ifp);
2461                         read_unlock_bh(&idev->lock);
2462
2463                         ipv6_del_addr(ifp);
2464                         return 0;
2465                 }
2466         }
2467         read_unlock_bh(&idev->lock);
2468         return -EADDRNOTAVAIL;
2469 }
2470
2471
2472 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2473 {
2474         struct in6_ifreq ireq;
2475         int err;
2476
2477         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2478                 return -EPERM;
2479
2480         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2481                 return -EFAULT;
2482
2483         rtnl_lock();
2484         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2485                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2486                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2487         rtnl_unlock();
2488         return err;
2489 }
2490
2491 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2492 {
2493         struct in6_ifreq ireq;
2494         int err;
2495
2496         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2497                 return -EPERM;
2498
2499         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2500                 return -EFAULT;
2501
2502         rtnl_lock();
2503         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2504                              ireq.ifr6_prefixlen);
2505         rtnl_unlock();
2506         return err;
2507 }
2508
2509 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2510                      int plen, int scope)
2511 {
2512         struct inet6_ifaddr *ifp;
2513
2514         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2515                             scope, IFA_F_PERMANENT, 0, 0);
2516         if (!IS_ERR(ifp)) {
2517                 spin_lock_bh(&ifp->lock);
2518                 ifp->flags &= ~IFA_F_TENTATIVE;
2519                 spin_unlock_bh(&ifp->lock);
2520                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2521                 in6_ifa_put(ifp);
2522         }
2523 }
2524
2525 #if IS_ENABLED(CONFIG_IPV6_SIT)
2526 static void sit_add_v4_addrs(struct inet6_dev *idev)
2527 {
2528         struct in6_addr addr;
2529         struct net_device *dev;
2530         struct net *net = dev_net(idev->dev);
2531         int scope;
2532
2533         ASSERT_RTNL();
2534
2535         memset(&addr, 0, sizeof(struct in6_addr));
2536         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2537
2538         if (idev->dev->flags&IFF_POINTOPOINT) {
2539                 addr.s6_addr32[0] = htonl(0xfe800000);
2540                 scope = IFA_LINK;
2541         } else {
2542                 scope = IPV6_ADDR_COMPATv4;
2543         }
2544
2545         if (addr.s6_addr32[3]) {
2546                 add_addr(idev, &addr, 128, scope);
2547                 return;
2548         }
2549
2550         for_each_netdev(net, dev) {
2551                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2552                 if (in_dev && (dev->flags & IFF_UP)) {
2553                         struct in_ifaddr *ifa;
2554
2555                         int flag = scope;
2556
2557                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2558                                 int plen;
2559
2560                                 addr.s6_addr32[3] = ifa->ifa_local;
2561
2562                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2563                                         continue;
2564                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2565                                         if (idev->dev->flags&IFF_POINTOPOINT)
2566                                                 continue;
2567                                         flag |= IFA_HOST;
2568                                 }
2569                                 if (idev->dev->flags&IFF_POINTOPOINT)
2570                                         plen = 64;
2571                                 else
2572                                         plen = 96;
2573
2574                                 add_addr(idev, &addr, plen, flag);
2575                         }
2576                 }
2577         }
2578 }
2579 #endif
2580
2581 static void init_loopback(struct net_device *dev)
2582 {
2583         struct inet6_dev  *idev;
2584         struct net_device *sp_dev;
2585         struct inet6_ifaddr *sp_ifa;
2586         struct rt6_info *sp_rt;
2587
2588         /* ::1 */
2589
2590         ASSERT_RTNL();
2591
2592         if ((idev = ipv6_find_idev(dev)) == NULL) {
2593                 pr_debug("%s: add_dev failed\n", __func__);
2594                 return;
2595         }
2596
2597         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2598
2599         /* Add routes to other interface's IPv6 addresses */
2600         for_each_netdev(dev_net(dev), sp_dev) {
2601                 if (!strcmp(sp_dev->name, dev->name))
2602                         continue;
2603
2604                 idev = __in6_dev_get(sp_dev);
2605                 if (!idev)
2606                         continue;
2607
2608                 read_lock_bh(&idev->lock);
2609                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2610
2611                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2612                                 continue;
2613
2614                         if (sp_ifa->rt)
2615                                 continue;
2616
2617                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2618
2619                         /* Failure cases are ignored */
2620                         if (!IS_ERR(sp_rt)) {
2621                                 sp_ifa->rt = sp_rt;
2622                                 ip6_ins_rt(sp_rt);
2623                         }
2624                 }
2625                 read_unlock_bh(&idev->lock);
2626         }
2627 }
2628
2629 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2630 {
2631         struct inet6_ifaddr *ifp;
2632         u32 addr_flags = IFA_F_PERMANENT;
2633
2634 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2635         if (idev->cnf.optimistic_dad &&
2636             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2637                 addr_flags |= IFA_F_OPTIMISTIC;
2638 #endif
2639
2640
2641         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2642         if (!IS_ERR(ifp)) {
2643                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2644                 addrconf_dad_start(ifp);
2645                 in6_ifa_put(ifp);
2646         }
2647 }
2648
2649 static void addrconf_dev_config(struct net_device *dev)
2650 {
2651         struct in6_addr addr;
2652         struct inet6_dev *idev;
2653
2654         ASSERT_RTNL();
2655
2656         if ((dev->type != ARPHRD_ETHER) &&
2657             (dev->type != ARPHRD_FDDI) &&
2658             (dev->type != ARPHRD_ARCNET) &&
2659             (dev->type != ARPHRD_INFINIBAND) &&
2660             (dev->type != ARPHRD_IEEE802154) &&
2661             (dev->type != ARPHRD_IEEE1394) &&
2662             (dev->type != ARPHRD_TUNNEL6)) {
2663                 /* Alas, we support only Ethernet autoconfiguration. */
2664                 return;
2665         }
2666
2667         idev = addrconf_add_dev(dev);
2668         if (IS_ERR(idev))
2669                 return;
2670
2671         memset(&addr, 0, sizeof(struct in6_addr));
2672         addr.s6_addr32[0] = htonl(0xFE800000);
2673
2674         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2675                 addrconf_add_linklocal(idev, &addr);
2676 }
2677
2678 #if IS_ENABLED(CONFIG_IPV6_SIT)
2679 static void addrconf_sit_config(struct net_device *dev)
2680 {
2681         struct inet6_dev *idev;
2682
2683         ASSERT_RTNL();
2684
2685         /*
2686          * Configure the tunnel with one of our IPv4
2687          * addresses... we should configure all of
2688          * our v4 addrs in the tunnel
2689          */
2690
2691         if ((idev = ipv6_find_idev(dev)) == NULL) {
2692                 pr_debug("%s: add_dev failed\n", __func__);
2693                 return;
2694         }
2695
2696         if (dev->priv_flags & IFF_ISATAP) {
2697                 struct in6_addr addr;
2698
2699                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2700                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2701                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2702                         addrconf_add_linklocal(idev, &addr);
2703                 return;
2704         }
2705
2706         sit_add_v4_addrs(idev);
2707
2708         if (dev->flags&IFF_POINTOPOINT)
2709                 addrconf_add_mroute(dev);
2710         else
2711                 sit_route_add(dev);
2712 }
2713 #endif
2714
2715 #if IS_ENABLED(CONFIG_NET_IPGRE)
2716 static void addrconf_gre_config(struct net_device *dev)
2717 {
2718         struct inet6_dev *idev;
2719         struct in6_addr addr;
2720
2721         ASSERT_RTNL();
2722
2723         if ((idev = ipv6_find_idev(dev)) == NULL) {
2724                 pr_debug("%s: add_dev failed\n", __func__);
2725                 return;
2726         }
2727
2728         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2729         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2730
2731         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2732                 addrconf_add_linklocal(idev, &addr);
2733 }
2734 #endif
2735
2736 static inline int
2737 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2738 {
2739         struct in6_addr lladdr;
2740
2741         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2742                 addrconf_add_linklocal(idev, &lladdr);
2743                 return 0;
2744         }
2745         return -1;
2746 }
2747
2748 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2749                            void *ptr)
2750 {
2751         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2752         struct inet6_dev *idev = __in6_dev_get(dev);
2753         int run_pending = 0;
2754         int err;
2755
2756         switch (event) {
2757         case NETDEV_REGISTER:
2758                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2759                         idev = ipv6_add_dev(dev);
2760                         if (!idev)
2761                                 return notifier_from_errno(-ENOMEM);
2762                 }
2763                 break;
2764
2765         case NETDEV_UP:
2766         case NETDEV_CHANGE:
2767                 if (dev->flags & IFF_SLAVE)
2768                         break;
2769
2770                 if (event == NETDEV_UP) {
2771                         if (!addrconf_qdisc_ok(dev)) {
2772                                 /* device is not ready yet. */
2773                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2774                                         dev->name);
2775                                 break;
2776                         }
2777
2778                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2779                                 idev = ipv6_add_dev(dev);
2780
2781                         if (idev) {
2782                                 idev->if_flags |= IF_READY;
2783                                 run_pending = 1;
2784                         }
2785                 } else {
2786                         if (!addrconf_qdisc_ok(dev)) {
2787                                 /* device is still not ready. */
2788                                 break;
2789                         }
2790
2791                         if (idev) {
2792                                 if (idev->if_flags & IF_READY)
2793                                         /* device is already configured. */
2794                                         break;
2795                                 idev->if_flags |= IF_READY;
2796                         }
2797
2798                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2799                                 dev->name);
2800
2801                         run_pending = 1;
2802                 }
2803
2804                 switch (dev->type) {
2805 #if IS_ENABLED(CONFIG_IPV6_SIT)
2806                 case ARPHRD_SIT:
2807                         addrconf_sit_config(dev);
2808                         break;
2809 #endif
2810 #if IS_ENABLED(CONFIG_NET_IPGRE)
2811                 case ARPHRD_IPGRE:
2812                         addrconf_gre_config(dev);
2813                         break;
2814 #endif
2815                 case ARPHRD_LOOPBACK:
2816                         init_loopback(dev);
2817                         break;
2818
2819                 default:
2820                         addrconf_dev_config(dev);
2821                         break;
2822                 }
2823
2824                 if (idev) {
2825                         if (run_pending)
2826                                 addrconf_dad_run(idev);
2827
2828                         /*
2829                          * If the MTU changed during the interface down,
2830                          * when the interface up, the changed MTU must be
2831                          * reflected in the idev as well as routers.
2832                          */
2833                         if (idev->cnf.mtu6 != dev->mtu &&
2834                             dev->mtu >= IPV6_MIN_MTU) {
2835                                 rt6_mtu_change(dev, dev->mtu);
2836                                 idev->cnf.mtu6 = dev->mtu;
2837                         }
2838                         idev->tstamp = jiffies;
2839                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2840
2841                         /*
2842                          * If the changed mtu during down is lower than
2843                          * IPV6_MIN_MTU stop IPv6 on this interface.
2844                          */
2845                         if (dev->mtu < IPV6_MIN_MTU)
2846                                 addrconf_ifdown(dev, 1);
2847                 }
2848                 break;
2849
2850         case NETDEV_CHANGEMTU:
2851                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2852                         rt6_mtu_change(dev, dev->mtu);
2853                         idev->cnf.mtu6 = dev->mtu;
2854                         break;
2855                 }
2856
2857                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2858                         idev = ipv6_add_dev(dev);
2859                         if (idev)
2860                                 break;
2861                 }
2862
2863                 /*
2864                  * MTU falled under IPV6_MIN_MTU.
2865                  * Stop IPv6 on this interface.
2866                  */
2867
2868         case NETDEV_DOWN:
2869         case NETDEV_UNREGISTER:
2870                 /*
2871                  *      Remove all addresses from this interface.
2872                  */
2873                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2874                 break;
2875
2876         case NETDEV_CHANGENAME:
2877                 if (idev) {
2878                         snmp6_unregister_dev(idev);
2879                         addrconf_sysctl_unregister(idev);
2880                         addrconf_sysctl_register(idev);
2881                         err = snmp6_register_dev(idev);
2882                         if (err)
2883                                 return notifier_from_errno(err);
2884                 }
2885                 break;
2886
2887         case NETDEV_PRE_TYPE_CHANGE:
2888         case NETDEV_POST_TYPE_CHANGE:
2889                 addrconf_type_change(dev, event);
2890                 break;
2891         }
2892
2893         return NOTIFY_OK;
2894 }
2895
2896 /*
2897  *      addrconf module should be notified of a device going up
2898  */
2899 static struct notifier_block ipv6_dev_notf = {
2900         .notifier_call = addrconf_notify,
2901 };
2902
2903 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2904 {
2905         struct inet6_dev *idev;
2906         ASSERT_RTNL();
2907
2908         idev = __in6_dev_get(dev);
2909
2910         if (event == NETDEV_POST_TYPE_CHANGE)
2911                 ipv6_mc_remap(idev);
2912         else if (event == NETDEV_PRE_TYPE_CHANGE)
2913                 ipv6_mc_unmap(idev);
2914 }
2915
2916 static int addrconf_ifdown(struct net_device *dev, int how)
2917 {
2918         struct net *net = dev_net(dev);
2919         struct inet6_dev *idev;
2920         struct inet6_ifaddr *ifa;
2921         int state, i;
2922
2923         ASSERT_RTNL();
2924
2925         rt6_ifdown(net, dev);
2926         neigh_ifdown(&nd_tbl, dev);
2927
2928         idev = __in6_dev_get(dev);
2929         if (idev == NULL)
2930                 return -ENODEV;
2931
2932         /*
2933          * Step 1: remove reference to ipv6 device from parent device.
2934          *         Do not dev_put!
2935          */
2936         if (how) {
2937                 idev->dead = 1;
2938
2939                 /* protected by rtnl_lock */
2940                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2941
2942                 /* Step 1.5: remove snmp6 entry */
2943                 snmp6_unregister_dev(idev);
2944
2945         }
2946
2947         /* Step 2: clear hash table */
2948         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2949                 struct hlist_head *h = &inet6_addr_lst[i];
2950
2951                 spin_lock_bh(&addrconf_hash_lock);
2952         restart:
2953                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2954                         if (ifa->idev == idev) {
2955                                 hlist_del_init_rcu(&ifa->addr_lst);
2956                                 addrconf_del_dad_timer(ifa);
2957                                 goto restart;
2958                         }
2959                 }
2960                 spin_unlock_bh(&addrconf_hash_lock);
2961         }
2962
2963         write_lock_bh(&idev->lock);
2964
2965         addrconf_del_rs_timer(idev);
2966
2967         /* Step 2: clear flags for stateless addrconf */
2968         if (!how)
2969                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2970
2971         if (how && del_timer(&idev->regen_timer))
2972                 in6_dev_put(idev);
2973
2974         /* Step 3: clear tempaddr list */
2975         while (!list_empty(&idev->tempaddr_list)) {
2976                 ifa = list_first_entry(&idev->tempaddr_list,
2977                                        struct inet6_ifaddr, tmp_list);
2978                 list_del(&ifa->tmp_list);
2979                 write_unlock_bh(&idev->lock);
2980                 spin_lock_bh(&ifa->lock);
2981
2982                 if (ifa->ifpub) {
2983                         in6_ifa_put(ifa->ifpub);
2984                         ifa->ifpub = NULL;
2985                 }
2986                 spin_unlock_bh(&ifa->lock);
2987                 in6_ifa_put(ifa);
2988                 write_lock_bh(&idev->lock);
2989         }
2990
2991         while (!list_empty(&idev->addr_list)) {
2992                 ifa = list_first_entry(&idev->addr_list,
2993                                        struct inet6_ifaddr, if_list);
2994                 addrconf_del_dad_timer(ifa);
2995
2996                 list_del(&ifa->if_list);
2997
2998                 write_unlock_bh(&idev->lock);
2999
3000                 spin_lock_bh(&ifa->state_lock);
3001                 state = ifa->state;
3002                 ifa->state = INET6_IFADDR_STATE_DEAD;
3003                 spin_unlock_bh(&ifa->state_lock);
3004
3005                 if (state != INET6_IFADDR_STATE_DEAD) {
3006                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3007                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3008                 }
3009                 in6_ifa_put(ifa);
3010
3011                 write_lock_bh(&idev->lock);
3012         }
3013
3014         write_unlock_bh(&idev->lock);
3015
3016         /* Step 5: Discard multicast list */
3017         if (how)
3018                 ipv6_mc_destroy_dev(idev);
3019         else
3020                 ipv6_mc_down(idev);
3021
3022         idev->tstamp = jiffies;
3023
3024         /* Last: Shot the device (if unregistered) */
3025         if (how) {
3026                 addrconf_sysctl_unregister(idev);
3027                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3028                 neigh_ifdown(&nd_tbl, dev);
3029                 in6_dev_put(idev);
3030         }
3031         return 0;
3032 }
3033
3034 static void addrconf_rs_timer(unsigned long data)
3035 {
3036         struct inet6_dev *idev = (struct inet6_dev *)data;
3037         struct net_device *dev = idev->dev;
3038         struct in6_addr lladdr;
3039
3040         write_lock(&idev->lock);
3041         if (idev->dead || !(idev->if_flags & IF_READY))
3042                 goto out;
3043
3044         if (!ipv6_accept_ra(idev))
3045                 goto out;
3046
3047         /* Announcement received after solicitation was sent */
3048         if (idev->if_flags & IF_RA_RCVD)
3049                 goto out;
3050
3051         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3052                 write_unlock(&idev->lock);
3053                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3054                         ndisc_send_rs(dev, &lladdr,
3055                                       &in6addr_linklocal_allrouters);
3056                 else
3057                         goto put;
3058
3059                 write_lock(&idev->lock);
3060                 /* The wait after the last probe can be shorter */
3061                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3062                                              idev->cnf.rtr_solicits) ?
3063                                       idev->cnf.rtr_solicit_delay :
3064                                       idev->cnf.rtr_solicit_interval);
3065         } else {
3066                 /*
3067                  * Note: we do not support deprecated "all on-link"
3068                  * assumption any longer.
3069                  */
3070                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3071         }
3072
3073 out:
3074         write_unlock(&idev->lock);
3075 put:
3076         in6_dev_put(idev);
3077 }
3078
3079 /*
3080  *      Duplicate Address Detection
3081  */
3082 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3083 {
3084         unsigned long rand_num;
3085         struct inet6_dev *idev = ifp->idev;
3086
3087         if (ifp->flags & IFA_F_OPTIMISTIC)
3088                 rand_num = 0;
3089         else
3090                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3091
3092         ifp->dad_probes = idev->cnf.dad_transmits;
3093         addrconf_mod_dad_timer(ifp, rand_num);
3094 }
3095
3096 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3097 {
3098         struct inet6_dev *idev = ifp->idev;
3099         struct net_device *dev = idev->dev;
3100
3101         addrconf_join_solict(dev, &ifp->addr);
3102
3103         net_srandom(ifp->addr.s6_addr32[3]);
3104
3105         read_lock_bh(&idev->lock);
3106         spin_lock(&ifp->lock);
3107         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3108                 goto out;
3109
3110         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3111             idev->cnf.accept_dad < 1 ||
3112             !(ifp->flags&IFA_F_TENTATIVE) ||
3113             ifp->flags & IFA_F_NODAD) {
3114                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3115                 spin_unlock(&ifp->lock);
3116                 read_unlock_bh(&idev->lock);
3117
3118                 addrconf_dad_completed(ifp);
3119                 return;
3120         }
3121
3122         if (!(idev->if_flags & IF_READY)) {
3123                 spin_unlock(&ifp->lock);
3124                 read_unlock_bh(&idev->lock);
3125                 /*
3126                  * If the device is not ready:
3127                  * - keep it tentative if it is a permanent address.
3128                  * - otherwise, kill it.
3129                  */
3130                 in6_ifa_hold(ifp);
3131                 addrconf_dad_stop(ifp, 0);
3132                 return;
3133         }
3134
3135         /*
3136          * Optimistic nodes can start receiving
3137          * Frames right away
3138          */
3139         if (ifp->flags & IFA_F_OPTIMISTIC)
3140                 ip6_ins_rt(ifp->rt);
3141
3142         addrconf_dad_kick(ifp);
3143 out:
3144         spin_unlock(&ifp->lock);
3145         read_unlock_bh(&idev->lock);
3146 }
3147
3148 static void addrconf_dad_timer(unsigned long data)
3149 {
3150         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3151         struct inet6_dev *idev = ifp->idev;
3152         struct in6_addr mcaddr;
3153
3154         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3155                 goto out;
3156
3157         write_lock(&idev->lock);
3158         if (idev->dead || !(idev->if_flags & IF_READY)) {
3159                 write_unlock(&idev->lock);
3160                 goto out;
3161         }
3162
3163         spin_lock(&ifp->lock);
3164         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3165                 spin_unlock(&ifp->lock);
3166                 write_unlock(&idev->lock);
3167                 goto out;
3168         }
3169
3170         if (ifp->dad_probes == 0) {
3171                 /*
3172                  * DAD was successful
3173                  */
3174
3175                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3176                 spin_unlock(&ifp->lock);
3177                 write_unlock(&idev->lock);
3178
3179                 addrconf_dad_completed(ifp);
3180
3181                 goto out;
3182         }
3183
3184         ifp->dad_probes--;
3185         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3186         spin_unlock(&ifp->lock);
3187         write_unlock(&idev->lock);
3188
3189         /* send a neighbour solicitation for our addr */
3190         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3191         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3192 out:
3193         in6_ifa_put(ifp);
3194 }
3195
3196 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3197 {
3198         struct net_device *dev = ifp->idev->dev;
3199         struct in6_addr lladdr;
3200         bool send_rs, send_mld;
3201
3202         addrconf_del_dad_timer(ifp);
3203
3204         /*
3205          *      Configure the address for reception. Now it is valid.
3206          */
3207
3208         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3209
3210         /* If added prefix is link local and we are prepared to process
3211            router advertisements, start sending router solicitations.
3212          */
3213
3214         read_lock_bh(&ifp->idev->lock);
3215         spin_lock(&ifp->lock);
3216         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3217                    ifp->idev->valid_ll_addr_cnt == 1;
3218         send_rs = send_mld &&
3219                   ipv6_accept_ra(ifp->idev) &&
3220                   ifp->idev->cnf.rtr_solicits > 0 &&
3221                   (dev->flags&IFF_LOOPBACK) == 0;
3222         spin_unlock(&ifp->lock);
3223         read_unlock_bh(&ifp->idev->lock);
3224
3225         /* While dad is in progress mld report's source address is in6_addrany.
3226          * Resend with proper ll now.
3227          */
3228         if (send_mld)
3229                 ipv6_mc_dad_complete(ifp->idev);
3230
3231         if (send_rs) {
3232                 /*
3233                  *      If a host as already performed a random delay
3234                  *      [...] as part of DAD [...] there is no need
3235                  *      to delay again before sending the first RS
3236                  */
3237                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3238                         return;
3239                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3240
3241                 write_lock_bh(&ifp->idev->lock);
3242                 spin_lock(&ifp->lock);
3243                 ifp->idev->rs_probes = 1;
3244                 ifp->idev->if_flags |= IF_RS_SENT;
3245                 addrconf_mod_rs_timer(ifp->idev,
3246                                       ifp->idev->cnf.rtr_solicit_interval);
3247                 spin_unlock(&ifp->lock);
3248                 write_unlock_bh(&ifp->idev->lock);
3249         }
3250 }
3251
3252 static void addrconf_dad_run(struct inet6_dev *idev)
3253 {
3254         struct inet6_ifaddr *ifp;
3255
3256         read_lock_bh(&idev->lock);
3257         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3258                 spin_lock(&ifp->lock);
3259                 if (ifp->flags & IFA_F_TENTATIVE &&
3260                     ifp->state == INET6_IFADDR_STATE_DAD)
3261                         addrconf_dad_kick(ifp);
3262                 spin_unlock(&ifp->lock);
3263         }
3264         read_unlock_bh(&idev->lock);
3265 }
3266
3267 #ifdef CONFIG_PROC_FS
3268 struct if6_iter_state {
3269         struct seq_net_private p;
3270         int bucket;
3271         int offset;
3272 };
3273
3274 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3275 {
3276         struct inet6_ifaddr *ifa = NULL;
3277         struct if6_iter_state *state = seq->private;
3278         struct net *net = seq_file_net(seq);
3279         int p = 0;
3280
3281         /* initial bucket if pos is 0 */
3282         if (pos == 0) {
3283                 state->bucket = 0;
3284                 state->offset = 0;
3285         }
3286
3287         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3288                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3289                                          addr_lst) {
3290                         if (!net_eq(dev_net(ifa->idev->dev), net))
3291                                 continue;
3292                         /* sync with offset */
3293                         if (p < state->offset) {
3294                                 p++;
3295                                 continue;
3296                         }
3297                         state->offset++;
3298                         return ifa;
3299                 }
3300
3301                 /* prepare for next bucket */
3302                 state->offset = 0;
3303                 p = 0;
3304         }
3305         return NULL;
3306 }
3307
3308 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3309                                          struct inet6_ifaddr *ifa)
3310 {
3311         struct if6_iter_state *state = seq->private;
3312         struct net *net = seq_file_net(seq);
3313
3314         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3315                 if (!net_eq(dev_net(ifa->idev->dev), net))
3316                         continue;
3317                 state->offset++;
3318                 return ifa;
3319         }
3320
3321         while (++state->bucket < IN6_ADDR_HSIZE) {
3322                 state->offset = 0;
3323                 hlist_for_each_entry_rcu_bh(ifa,
3324                                      &inet6_addr_lst[state->bucket], addr_lst) {
3325                         if (!net_eq(dev_net(ifa->idev->dev), net))
3326                                 continue;
3327                         state->offset++;
3328                         return ifa;
3329                 }
3330         }
3331
3332         return NULL;
3333 }
3334
3335 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3336         __acquires(rcu_bh)
3337 {
3338         rcu_read_lock_bh();
3339         return if6_get_first(seq, *pos);
3340 }
3341
3342 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3343 {
3344         struct inet6_ifaddr *ifa;
3345
3346         ifa = if6_get_next(seq, v);
3347         ++*pos;
3348         return ifa;
3349 }
3350
3351 static void if6_seq_stop(struct seq_file *seq, void *v)
3352         __releases(rcu_bh)
3353 {
3354         rcu_read_unlock_bh();
3355 }
3356
3357 static int if6_seq_show(struct seq_file *seq, void *v)
3358 {
3359         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3360         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3361                    &ifp->addr,
3362                    ifp->idev->dev->ifindex,
3363                    ifp->prefix_len,
3364                    ifp->scope,
3365                    ifp->flags,
3366                    ifp->idev->dev->name);
3367         return 0;
3368 }
3369
3370 static const struct seq_operations if6_seq_ops = {
3371         .start  = if6_seq_start,
3372         .next   = if6_seq_next,
3373         .show   = if6_seq_show,
3374         .stop   = if6_seq_stop,
3375 };
3376
3377 static int if6_seq_open(struct inode *inode, struct file *file)
3378 {
3379         return seq_open_net(inode, file, &if6_seq_ops,
3380                             sizeof(struct if6_iter_state));
3381 }
3382
3383 static const struct file_operations if6_fops = {
3384         .owner          = THIS_MODULE,
3385         .open           = if6_seq_open,
3386         .read           = seq_read,
3387         .llseek         = seq_lseek,
3388         .release        = seq_release_net,
3389 };
3390
3391 static int __net_init if6_proc_net_init(struct net *net)
3392 {
3393         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3394                 return -ENOMEM;
3395         return 0;
3396 }
3397
3398 static void __net_exit if6_proc_net_exit(struct net *net)
3399 {
3400         remove_proc_entry("if_inet6", net->proc_net);
3401 }
3402
3403 static struct pernet_operations if6_proc_net_ops = {
3404        .init = if6_proc_net_init,
3405        .exit = if6_proc_net_exit,
3406 };
3407
3408 int __init if6_proc_init(void)
3409 {
3410         return register_pernet_subsys(&if6_proc_net_ops);
3411 }
3412
3413 void if6_proc_exit(void)
3414 {
3415         unregister_pernet_subsys(&if6_proc_net_ops);
3416 }
3417 #endif  /* CONFIG_PROC_FS */
3418
3419 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3420 /* Check if address is a home address configured on any interface. */
3421 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3422 {
3423         int ret = 0;
3424         struct inet6_ifaddr *ifp = NULL;
3425         unsigned int hash = inet6_addr_hash(addr);
3426
3427         rcu_read_lock_bh();
3428         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3429                 if (!net_eq(dev_net(ifp->idev->dev), net))
3430                         continue;
3431                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3432                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3433                         ret = 1;
3434                         break;
3435                 }
3436         }
3437         rcu_read_unlock_bh();
3438         return ret;
3439 }
3440 #endif
3441
3442 /*
3443  *      Periodic address status verification
3444  */
3445
3446 static void addrconf_verify(unsigned long foo)
3447 {
3448         unsigned long now, next, next_sec, next_sched;
3449         struct inet6_ifaddr *ifp;
3450         int i;
3451
3452         rcu_read_lock_bh();
3453         spin_lock(&addrconf_verify_lock);
3454         now = jiffies;
3455         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3456
3457         del_timer(&addr_chk_timer);
3458
3459         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3460 restart:
3461                 hlist_for_each_entry_rcu_bh(ifp,
3462                                          &inet6_addr_lst[i], addr_lst) {
3463                         unsigned long age;
3464
3465                         if (ifp->flags & IFA_F_PERMANENT)
3466                                 continue;
3467
3468                         spin_lock(&ifp->lock);
3469                         /* We try to batch several events at once. */
3470                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3471
3472                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3473                             age >= ifp->valid_lft) {
3474                                 spin_unlock(&ifp->lock);
3475                                 in6_ifa_hold(ifp);
3476                                 ipv6_del_addr(ifp);
3477                                 goto restart;
3478                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3479                                 spin_unlock(&ifp->lock);
3480                                 continue;
3481                         } else if (age >= ifp->prefered_lft) {
3482                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3483                                 int deprecate = 0;
3484
3485                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3486                                         deprecate = 1;
3487                                         ifp->flags |= IFA_F_DEPRECATED;
3488                                 }
3489
3490                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3491                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3492
3493                                 spin_unlock(&ifp->lock);
3494
3495                                 if (deprecate) {
3496                                         in6_ifa_hold(ifp);
3497
3498                                         ipv6_ifa_notify(0, ifp);
3499                                         in6_ifa_put(ifp);
3500                                         goto restart;
3501                                 }
3502                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3503                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3504                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3505                                         ifp->idev->cnf.dad_transmits *
3506                                         ifp->idev->nd_parms->retrans_time / HZ;
3507
3508                                 if (age >= ifp->prefered_lft - regen_advance) {
3509                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3510                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3511                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3512                                         if (!ifp->regen_count && ifpub) {
3513                                                 ifp->regen_count++;
3514                                                 in6_ifa_hold(ifp);
3515                                                 in6_ifa_hold(ifpub);
3516                                                 spin_unlock(&ifp->lock);
3517
3518                                                 spin_lock(&ifpub->lock);
3519                                                 ifpub->regen_count = 0;
3520                                                 spin_unlock(&ifpub->lock);
3521                                                 ipv6_create_tempaddr(ifpub, ifp);
3522                                                 in6_ifa_put(ifpub);
3523                                                 in6_ifa_put(ifp);
3524                                                 goto restart;
3525                                         }
3526                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3527                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3528                                 spin_unlock(&ifp->lock);
3529                         } else {
3530                                 /* ifp->prefered_lft <= ifp->valid_lft */
3531                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3532                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3533                                 spin_unlock(&ifp->lock);
3534                         }
3535                 }
3536         }
3537
3538         next_sec = round_jiffies_up(next);
3539         next_sched = next;
3540
3541         /* If rounded timeout is accurate enough, accept it. */
3542         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3543                 next_sched = next_sec;
3544
3545         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3546         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3547                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3548
3549         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3550               now, next, next_sec, next_sched);
3551
3552         addr_chk_timer.expires = next_sched;
3553         add_timer(&addr_chk_timer);
3554         spin_unlock(&addrconf_verify_lock);
3555         rcu_read_unlock_bh();
3556 }
3557
3558 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3559                                      struct in6_addr **peer_pfx)
3560 {
3561         struct in6_addr *pfx = NULL;
3562
3563         *peer_pfx = NULL;
3564
3565         if (addr)
3566                 pfx = nla_data(addr);
3567
3568         if (local) {
3569                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3570                         *peer_pfx = pfx;
3571                 pfx = nla_data(local);
3572         }
3573
3574         return pfx;
3575 }
3576
3577 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3578         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3579         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3580         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3581 };
3582
3583 static int
3584 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3585 {
3586         struct net *net = sock_net(skb->sk);
3587         struct ifaddrmsg *ifm;
3588         struct nlattr *tb[IFA_MAX+1];
3589         struct in6_addr *pfx, *peer_pfx;
3590         int err;
3591
3592         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3593         if (err < 0)
3594                 return err;
3595
3596         ifm = nlmsg_data(nlh);
3597         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3598         if (pfx == NULL)
3599                 return -EINVAL;
3600
3601         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3602 }
3603
3604 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3605                              u32 prefered_lft, u32 valid_lft)
3606 {
3607         u32 flags;
3608         clock_t expires;
3609         unsigned long timeout;
3610
3611         if (!valid_lft || (prefered_lft > valid_lft))
3612                 return -EINVAL;
3613
3614         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3615         if (addrconf_finite_timeout(timeout)) {
3616                 expires = jiffies_to_clock_t(timeout * HZ);
3617                 valid_lft = timeout;
3618                 flags = RTF_EXPIRES;
3619         } else {
3620                 expires = 0;
3621                 flags = 0;
3622                 ifa_flags |= IFA_F_PERMANENT;
3623         }
3624
3625         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3626         if (addrconf_finite_timeout(timeout)) {
3627                 if (timeout == 0)
3628                         ifa_flags |= IFA_F_DEPRECATED;
3629                 prefered_lft = timeout;
3630         }
3631
3632         spin_lock_bh(&ifp->lock);
3633         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3634         ifp->tstamp = jiffies;
3635         ifp->valid_lft = valid_lft;
3636         ifp->prefered_lft = prefered_lft;
3637
3638         spin_unlock_bh(&ifp->lock);
3639         if (!(ifp->flags&IFA_F_TENTATIVE))
3640                 ipv6_ifa_notify(0, ifp);
3641
3642         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3643                               expires, flags);
3644         addrconf_verify(0);
3645
3646         return 0;
3647 }
3648
3649 static int
3650 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3651 {
3652         struct net *net = sock_net(skb->sk);
3653         struct ifaddrmsg *ifm;
3654         struct nlattr *tb[IFA_MAX+1];
3655         struct in6_addr *pfx, *peer_pfx;
3656         struct inet6_ifaddr *ifa;
3657         struct net_device *dev;
3658         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3659         u8 ifa_flags;
3660         int err;
3661
3662         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3663         if (err < 0)
3664                 return err;
3665
3666         ifm = nlmsg_data(nlh);
3667         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3668         if (pfx == NULL)
3669                 return -EINVAL;
3670
3671         if (tb[IFA_CACHEINFO]) {
3672                 struct ifa_cacheinfo *ci;
3673
3674                 ci = nla_data(tb[IFA_CACHEINFO]);
3675                 valid_lft = ci->ifa_valid;
3676                 preferred_lft = ci->ifa_prefered;
3677         } else {
3678                 preferred_lft = INFINITY_LIFE_TIME;
3679                 valid_lft = INFINITY_LIFE_TIME;
3680         }
3681
3682         dev =  __dev_get_by_index(net, ifm->ifa_index);
3683         if (dev == NULL)
3684                 return -ENODEV;
3685
3686         /* We ignore other flags so far. */
3687         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3688
3689         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3690         if (ifa == NULL) {
3691                 /*
3692                  * It would be best to check for !NLM_F_CREATE here but
3693                  * userspace alreay relies on not having to provide this.
3694                  */
3695                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3696                                       ifm->ifa_prefixlen, ifa_flags,
3697                                       preferred_lft, valid_lft);
3698         }
3699
3700         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3701             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3702                 err = -EEXIST;
3703         else
3704                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3705
3706         in6_ifa_put(ifa);
3707
3708         return err;
3709 }
3710
3711 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3712                           u8 scope, int ifindex)
3713 {
3714         struct ifaddrmsg *ifm;
3715
3716         ifm = nlmsg_data(nlh);
3717         ifm->ifa_family = AF_INET6;
3718         ifm->ifa_prefixlen = prefixlen;
3719         ifm->ifa_flags = flags;
3720         ifm->ifa_scope = scope;
3721         ifm->ifa_index = ifindex;
3722 }
3723
3724 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3725                          unsigned long tstamp, u32 preferred, u32 valid)
3726 {
3727         struct ifa_cacheinfo ci;
3728
3729         ci.cstamp = cstamp_delta(cstamp);
3730         ci.tstamp = cstamp_delta(tstamp);
3731         ci.ifa_prefered = preferred;
3732         ci.ifa_valid = valid;
3733
3734         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3735 }
3736
3737 static inline int rt_scope(int ifa_scope)
3738 {
3739         if (ifa_scope & IFA_HOST)
3740                 return RT_SCOPE_HOST;
3741         else if (ifa_scope & IFA_LINK)
3742                 return RT_SCOPE_LINK;
3743         else if (ifa_scope & IFA_SITE)
3744                 return RT_SCOPE_SITE;
3745         else
3746                 return RT_SCOPE_UNIVERSE;
3747 }
3748
3749 static inline int inet6_ifaddr_msgsize(void)
3750 {
3751         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3752                + nla_total_size(16) /* IFA_LOCAL */
3753                + nla_total_size(16) /* IFA_ADDRESS */
3754                + nla_total_size(sizeof(struct ifa_cacheinfo));
3755 }
3756
3757 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3758                              u32 portid, u32 seq, int event, unsigned int flags)
3759 {
3760         struct nlmsghdr  *nlh;
3761         u32 preferred, valid;
3762
3763         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3764         if (nlh == NULL)
3765                 return -EMSGSIZE;
3766
3767         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3768                       ifa->idev->dev->ifindex);
3769
3770         if (!(ifa->flags&IFA_F_PERMANENT)) {
3771                 preferred = ifa->prefered_lft;
3772                 valid = ifa->valid_lft;
3773                 if (preferred != INFINITY_LIFE_TIME) {
3774                         long tval = (jiffies - ifa->tstamp)/HZ;
3775                         if (preferred > tval)
3776                                 preferred -= tval;
3777                         else
3778                                 preferred = 0;
3779                         if (valid != INFINITY_LIFE_TIME) {
3780                                 if (valid > tval)
3781                                         valid -= tval;
3782                                 else
3783                                         valid = 0;
3784                         }
3785                 }
3786         } else {
3787                 preferred = INFINITY_LIFE_TIME;
3788                 valid = INFINITY_LIFE_TIME;
3789         }
3790
3791         if (!ipv6_addr_any(&ifa->peer_addr)) {
3792                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3793                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3794                         goto error;
3795         } else
3796                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3797                         goto error;
3798
3799         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3800                 goto error;
3801
3802         return nlmsg_end(skb, nlh);
3803
3804 error:
3805         nlmsg_cancel(skb, nlh);
3806         return -EMSGSIZE;
3807 }
3808
3809 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3810                                 u32 portid, u32 seq, int event, u16 flags)
3811 {
3812         struct nlmsghdr  *nlh;
3813         u8 scope = RT_SCOPE_UNIVERSE;
3814         int ifindex = ifmca->idev->dev->ifindex;
3815
3816         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3817                 scope = RT_SCOPE_SITE;
3818
3819         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3820         if (nlh == NULL)
3821                 return -EMSGSIZE;
3822
3823         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3824         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3825             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3826                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3827                 nlmsg_cancel(skb, nlh);
3828                 return -EMSGSIZE;
3829         }
3830
3831         return nlmsg_end(skb, nlh);
3832 }
3833
3834 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3835                                 u32 portid, u32 seq, int event, unsigned int flags)
3836 {
3837         struct nlmsghdr  *nlh;
3838         u8 scope = RT_SCOPE_UNIVERSE;
3839         int ifindex = ifaca->aca_idev->dev->ifindex;
3840
3841         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3842                 scope = RT_SCOPE_SITE;
3843
3844         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3845         if (nlh == NULL)
3846                 return -EMSGSIZE;
3847
3848         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3849         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3850             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3851                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3852                 nlmsg_cancel(skb, nlh);
3853                 return -EMSGSIZE;
3854         }
3855
3856         return nlmsg_end(skb, nlh);
3857 }
3858
3859 enum addr_type_t {
3860         UNICAST_ADDR,
3861         MULTICAST_ADDR,
3862         ANYCAST_ADDR,
3863 };
3864
3865 /* called with rcu_read_lock() */
3866 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3867                           struct netlink_callback *cb, enum addr_type_t type,
3868                           int s_ip_idx, int *p_ip_idx)
3869 {
3870         struct ifmcaddr6 *ifmca;
3871         struct ifacaddr6 *ifaca;
3872         int err = 1;
3873         int ip_idx = *p_ip_idx;
3874
3875         read_lock_bh(&idev->lock);
3876         switch (type) {
3877         case UNICAST_ADDR: {
3878                 struct inet6_ifaddr *ifa;
3879
3880                 /* unicast address incl. temp addr */
3881                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3882                         if (++ip_idx < s_ip_idx)
3883                                 continue;
3884                         err = inet6_fill_ifaddr(skb, ifa,
3885                                                 NETLINK_CB(cb->skb).portid,
3886                                                 cb->nlh->nlmsg_seq,
3887                                                 RTM_NEWADDR,
3888                                                 NLM_F_MULTI);
3889                         if (err <= 0)
3890                                 break;
3891                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3892                 }
3893                 break;
3894         }
3895         case MULTICAST_ADDR:
3896                 /* multicast address */
3897                 for (ifmca = idev->mc_list; ifmca;
3898                      ifmca = ifmca->next, ip_idx++) {
3899                         if (ip_idx < s_ip_idx)
3900                                 continue;
3901                         err = inet6_fill_ifmcaddr(skb, ifmca,
3902                                                   NETLINK_CB(cb->skb).portid,
3903                                                   cb->nlh->nlmsg_seq,
3904                                                   RTM_GETMULTICAST,
3905                                                   NLM_F_MULTI);
3906                         if (err <= 0)
3907                                 break;
3908                 }
3909                 break;
3910         case ANYCAST_ADDR:
3911                 /* anycast address */
3912                 for (ifaca = idev->ac_list; ifaca;
3913                      ifaca = ifaca->aca_next, ip_idx++) {
3914                         if (ip_idx < s_ip_idx)
3915                                 continue;
3916                         err = inet6_fill_ifacaddr(skb, ifaca,
3917                                                   NETLINK_CB(cb->skb).portid,
3918                                                   cb->nlh->nlmsg_seq,
3919                                                   RTM_GETANYCAST,
3920                                                   NLM_F_MULTI);
3921                         if (err <= 0)
3922                                 break;
3923                 }
3924                 break;
3925         default:
3926                 break;
3927         }
3928         read_unlock_bh(&idev->lock);
3929         *p_ip_idx = ip_idx;
3930         return err;
3931 }
3932
3933 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3934                            enum addr_type_t type)
3935 {
3936         struct net *net = sock_net(skb->sk);
3937         int h, s_h;
3938         int idx, ip_idx;
3939         int s_idx, s_ip_idx;
3940         struct net_device *dev;
3941         struct inet6_dev *idev;
3942         struct hlist_head *head;
3943
3944         s_h = cb->args[0];
3945         s_idx = idx = cb->args[1];
3946         s_ip_idx = ip_idx = cb->args[2];
3947
3948         rcu_read_lock();
3949         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3950         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3951                 idx = 0;
3952                 head = &net->dev_index_head[h];
3953                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3954                         if (idx < s_idx)
3955                                 goto cont;
3956                         if (h > s_h || idx > s_idx)
3957                                 s_ip_idx = 0;
3958                         ip_idx = 0;
3959                         idev = __in6_dev_get(dev);
3960                         if (!idev)
3961                                 goto cont;
3962
3963                         if (in6_dump_addrs(idev, skb, cb, type,
3964                                            s_ip_idx, &ip_idx) <= 0)
3965                                 goto done;
3966 cont:
3967                         idx++;
3968                 }
3969         }
3970 done:
3971         rcu_read_unlock();
3972         cb->args[0] = h;
3973         cb->args[1] = idx;
3974         cb->args[2] = ip_idx;
3975
3976         return skb->len;
3977 }
3978
3979 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3980 {
3981         enum addr_type_t type = UNICAST_ADDR;
3982
3983         return inet6_dump_addr(skb, cb, type);
3984 }
3985
3986 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3987 {
3988         enum addr_type_t type = MULTICAST_ADDR;
3989
3990         return inet6_dump_addr(skb, cb, type);
3991 }
3992
3993
3994 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3995 {
3996         enum addr_type_t type = ANYCAST_ADDR;
3997
3998         return inet6_dump_addr(skb, cb, type);
3999 }
4000
4001 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4002 {
4003         struct net *net = sock_net(in_skb->sk);
4004         struct ifaddrmsg *ifm;
4005         struct nlattr *tb[IFA_MAX+1];
4006         struct in6_addr *addr = NULL, *peer;
4007         struct net_device *dev = NULL;
4008         struct inet6_ifaddr *ifa;
4009         struct sk_buff *skb;
4010         int err;
4011
4012         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4013         if (err < 0)
4014                 goto errout;
4015
4016         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4017         if (addr == NULL) {
4018                 err = -EINVAL;
4019                 goto errout;
4020         }
4021
4022         ifm = nlmsg_data(nlh);
4023         if (ifm->ifa_index)
4024                 dev = __dev_get_by_index(net, ifm->ifa_index);
4025
4026         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4027         if (!ifa) {
4028                 err = -EADDRNOTAVAIL;
4029                 goto errout;
4030         }
4031
4032         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4033         if (!skb) {
4034                 err = -ENOBUFS;
4035                 goto errout_ifa;
4036         }
4037
4038         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4039                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4040         if (err < 0) {
4041                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4042                 WARN_ON(err == -EMSGSIZE);
4043                 kfree_skb(skb);
4044                 goto errout_ifa;
4045         }
4046         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4047 errout_ifa:
4048         in6_ifa_put(ifa);
4049 errout:
4050         return err;
4051 }
4052
4053 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4054 {
4055         struct sk_buff *skb;
4056         struct net *net = dev_net(ifa->idev->dev);
4057         int err = -ENOBUFS;
4058
4059         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4060         if (skb == NULL)
4061                 goto errout;
4062
4063         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4064         if (err < 0) {
4065                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4066                 WARN_ON(err == -EMSGSIZE);
4067                 kfree_skb(skb);
4068                 goto errout;
4069         }
4070         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4071         return;
4072 errout:
4073         if (err < 0)
4074                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4075 }
4076
4077 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4078                                 __s32 *array, int bytes)
4079 {
4080         BUG_ON(bytes < (DEVCONF_MAX * 4));
4081
4082         memset(array, 0, bytes);
4083         array[DEVCONF_FORWARDING] = cnf->forwarding;
4084         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4085         array[DEVCONF_MTU6] = cnf->mtu6;
4086         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4087         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4088         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4089         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4090         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4091         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4092                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4093         array[DEVCONF_RTR_SOLICIT_DELAY] =
4094                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4095         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4096         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4097                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4098         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4099                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4100         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4101         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4102         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4103         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4104         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4105         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4106         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4107         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4108 #ifdef CONFIG_IPV6_ROUTER_PREF
4109         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4110         array[DEVCONF_RTR_PROBE_INTERVAL] =
4111                 jiffies_to_msecs(cnf->rtr_probe_interval);
4112 #ifdef CONFIG_IPV6_ROUTE_INFO
4113         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4114 #endif
4115 #endif
4116         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4117         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4118 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4119         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4120 #endif
4121 #ifdef CONFIG_IPV6_MROUTE
4122         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4123 #endif
4124         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4125         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4126         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4127         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4128         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4129 }
4130
4131 static inline size_t inet6_ifla6_size(void)
4132 {
4133         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4134              + nla_total_size(sizeof(struct ifla_cacheinfo))
4135              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4136              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4137              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4138              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4139 }
4140
4141 static inline size_t inet6_if_nlmsg_size(void)
4142 {
4143         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4144                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4145                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4146                + nla_total_size(4) /* IFLA_MTU */
4147                + nla_total_size(4) /* IFLA_LINK */
4148                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4149 }
4150
4151 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4152                                       int items, int bytes)
4153 {
4154         int i;
4155         int pad = bytes - sizeof(u64) * items;
4156         BUG_ON(pad < 0);
4157
4158         /* Use put_unaligned() because stats may not be aligned for u64. */
4159         put_unaligned(items, &stats[0]);
4160         for (i = 1; i < items; i++)
4161                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4162
4163         memset(&stats[items], 0, pad);
4164 }
4165
4166 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4167                                       int items, int bytes, size_t syncpoff)
4168 {
4169         int i;
4170         int pad = bytes - sizeof(u64) * items;
4171         BUG_ON(pad < 0);
4172
4173         /* Use put_unaligned() because stats may not be aligned for u64. */
4174         put_unaligned(items, &stats[0]);
4175         for (i = 1; i < items; i++)
4176                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4177
4178         memset(&stats[items], 0, pad);
4179 }
4180
4181 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4182                              int bytes)
4183 {
4184         switch (attrtype) {
4185         case IFLA_INET6_STATS:
4186                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4187                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4188                 break;
4189         case IFLA_INET6_ICMP6STATS:
4190                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4191                 break;
4192         }
4193 }
4194
4195 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4196 {
4197         struct nlattr *nla;
4198         struct ifla_cacheinfo ci;
4199
4200         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4201                 goto nla_put_failure;
4202         ci.max_reasm_len = IPV6_MAXPLEN;
4203         ci.tstamp = cstamp_delta(idev->tstamp);
4204         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4205         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4206         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4207                 goto nla_put_failure;
4208         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4209         if (nla == NULL)
4210                 goto nla_put_failure;
4211         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4212
4213         /* XXX - MC not implemented */
4214
4215         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4216         if (nla == NULL)
4217                 goto nla_put_failure;
4218         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4219
4220         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4221         if (nla == NULL)
4222                 goto nla_put_failure;
4223         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4224
4225         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4226         if (nla == NULL)
4227                 goto nla_put_failure;
4228         read_lock_bh(&idev->lock);
4229         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4230         read_unlock_bh(&idev->lock);
4231
4232         return 0;
4233
4234 nla_put_failure:
4235         return -EMSGSIZE;
4236 }
4237
4238 static size_t inet6_get_link_af_size(const struct net_device *dev)
4239 {
4240         if (!__in6_dev_get(dev))
4241                 return 0;
4242
4243         return inet6_ifla6_size();
4244 }
4245
4246 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4247 {
4248         struct inet6_dev *idev = __in6_dev_get(dev);
4249
4250         if (!idev)
4251                 return -ENODATA;
4252
4253         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4254                 return -EMSGSIZE;
4255
4256         return 0;
4257 }
4258
4259 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4260 {
4261         struct inet6_ifaddr *ifp;
4262         struct net_device *dev = idev->dev;
4263         bool update_rs = false;
4264         struct in6_addr ll_addr;
4265
4266         if (token == NULL)
4267                 return -EINVAL;
4268         if (ipv6_addr_any(token))
4269                 return -EINVAL;
4270         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4271                 return -EINVAL;
4272         if (!ipv6_accept_ra(idev))
4273                 return -EINVAL;
4274         if (idev->cnf.rtr_solicits <= 0)
4275                 return -EINVAL;
4276
4277         write_lock_bh(&idev->lock);
4278
4279         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4280         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4281
4282         write_unlock_bh(&idev->lock);
4283
4284         if (!idev->dead && (idev->if_flags & IF_READY) &&
4285             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4286                              IFA_F_OPTIMISTIC)) {
4287
4288                 /* If we're not ready, then normal ifup will take care
4289                  * of this. Otherwise, we need to request our rs here.
4290                  */
4291                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4292                 update_rs = true;
4293         }
4294
4295         write_lock_bh(&idev->lock);
4296
4297         if (update_rs) {
4298                 idev->if_flags |= IF_RS_SENT;
4299                 idev->rs_probes = 1;
4300                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4301         }
4302
4303         /* Well, that's kinda nasty ... */
4304         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4305                 spin_lock(&ifp->lock);
4306                 if (ifp->tokenized) {
4307                         ifp->valid_lft = 0;
4308                         ifp->prefered_lft = 0;
4309                 }
4310                 spin_unlock(&ifp->lock);
4311         }
4312
4313         write_unlock_bh(&idev->lock);
4314         addrconf_verify(0);
4315         return 0;
4316 }
4317
4318 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4319 {
4320         int err = -EINVAL;
4321         struct inet6_dev *idev = __in6_dev_get(dev);
4322         struct nlattr *tb[IFLA_INET6_MAX + 1];
4323
4324         if (!idev)
4325                 return -EAFNOSUPPORT;
4326
4327         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4328                 BUG();
4329
4330         if (tb[IFLA_INET6_TOKEN])
4331                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4332
4333         return err;
4334 }
4335
4336 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4337                              u32 portid, u32 seq, int event, unsigned int flags)
4338 {
4339         struct net_device *dev = idev->dev;
4340         struct ifinfomsg *hdr;
4341         struct nlmsghdr *nlh;
4342         void *protoinfo;
4343
4344         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4345         if (nlh == NULL)
4346                 return -EMSGSIZE;
4347
4348         hdr = nlmsg_data(nlh);
4349         hdr->ifi_family = AF_INET6;
4350         hdr->__ifi_pad = 0;
4351         hdr->ifi_type = dev->type;
4352         hdr->ifi_index = dev->ifindex;
4353         hdr->ifi_flags = dev_get_flags(dev);
4354         hdr->ifi_change = 0;
4355
4356         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4357             (dev->addr_len &&
4358              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4359             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4360             (dev->ifindex != dev->iflink &&
4361              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4362                 goto nla_put_failure;
4363         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4364         if (protoinfo == NULL)
4365                 goto nla_put_failure;
4366
4367         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4368                 goto nla_put_failure;
4369
4370         nla_nest_end(skb, protoinfo);
4371         return nlmsg_end(skb, nlh);
4372
4373 nla_put_failure:
4374         nlmsg_cancel(skb, nlh);
4375         return -EMSGSIZE;
4376 }
4377
4378 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4379 {
4380         struct net *net = sock_net(skb->sk);
4381         int h, s_h;
4382         int idx = 0, s_idx;
4383         struct net_device *dev;
4384         struct inet6_dev *idev;
4385         struct hlist_head *head;
4386
4387         s_h = cb->args[0];
4388         s_idx = cb->args[1];
4389
4390         rcu_read_lock();
4391         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4392                 idx = 0;
4393                 head = &net->dev_index_head[h];
4394                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4395                         if (idx < s_idx)
4396                                 goto cont;
4397                         idev = __in6_dev_get(dev);
4398                         if (!idev)
4399                                 goto cont;
4400                         if (inet6_fill_ifinfo(skb, idev,
4401                                               NETLINK_CB(cb->skb).portid,
4402                                               cb->nlh->nlmsg_seq,
4403                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4404                                 goto out;
4405 cont:
4406                         idx++;
4407                 }
4408         }
4409 out:
4410         rcu_read_unlock();
4411         cb->args[1] = idx;
4412         cb->args[0] = h;
4413
4414         return skb->len;
4415 }
4416
4417 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4418 {
4419         struct sk_buff *skb;
4420         struct net *net = dev_net(idev->dev);
4421         int err = -ENOBUFS;
4422
4423         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4424         if (skb == NULL)
4425                 goto errout;
4426
4427         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4428         if (err < 0) {
4429                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4430                 WARN_ON(err == -EMSGSIZE);
4431                 kfree_skb(skb);
4432                 goto errout;
4433         }
4434         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4435         return;
4436 errout:
4437         if (err < 0)
4438                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4439 }
4440
4441 static inline size_t inet6_prefix_nlmsg_size(void)
4442 {
4443         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4444                + nla_total_size(sizeof(struct in6_addr))
4445                + nla_total_size(sizeof(struct prefix_cacheinfo));
4446 }
4447
4448 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4449                              struct prefix_info *pinfo, u32 portid, u32 seq,
4450                              int event, unsigned int flags)
4451 {
4452         struct prefixmsg *pmsg;
4453         struct nlmsghdr *nlh;
4454         struct prefix_cacheinfo ci;
4455
4456         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4457         if (nlh == NULL)
4458                 return -EMSGSIZE;
4459
4460         pmsg = nlmsg_data(nlh);
4461         pmsg->prefix_family = AF_INET6;
4462         pmsg->prefix_pad1 = 0;
4463         pmsg->prefix_pad2 = 0;
4464         pmsg->prefix_ifindex = idev->dev->ifindex;
4465         pmsg->prefix_len = pinfo->prefix_len;
4466         pmsg->prefix_type = pinfo->type;
4467         pmsg->prefix_pad3 = 0;
4468         pmsg->prefix_flags = 0;
4469         if (pinfo->onlink)
4470                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4471         if (pinfo->autoconf)
4472                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4473
4474         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4475                 goto nla_put_failure;
4476         ci.preferred_time = ntohl(pinfo->prefered);
4477         ci.valid_time = ntohl(pinfo->valid);
4478         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4479                 goto nla_put_failure;
4480         return nlmsg_end(skb, nlh);
4481
4482 nla_put_failure:
4483         nlmsg_cancel(skb, nlh);
4484         return -EMSGSIZE;
4485 }
4486
4487 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4488                          struct prefix_info *pinfo)
4489 {
4490         struct sk_buff *skb;
4491         struct net *net = dev_net(idev->dev);
4492         int err = -ENOBUFS;
4493
4494         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4495         if (skb == NULL)
4496                 goto errout;
4497
4498         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4499         if (err < 0) {
4500                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4501                 WARN_ON(err == -EMSGSIZE);
4502                 kfree_skb(skb);
4503                 goto errout;
4504         }
4505         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4506         return;
4507 errout:
4508         if (err < 0)
4509                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4510 }
4511
4512 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4513 {
4514         write_lock_bh(&ifp->idev->lock);
4515         spin_lock(&ifp->lock);
4516         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4517                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4518             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4519                 ifp->idev->valid_ll_addr_cnt += count;
4520         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4521         spin_unlock(&ifp->lock);
4522         write_unlock_bh(&ifp->idev->lock);
4523 }
4524
4525 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4526 {
4527         struct net *net = dev_net(ifp->idev->dev);
4528
4529         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4530
4531         switch (event) {
4532         case RTM_NEWADDR:
4533                 update_valid_ll_addr_cnt(ifp, 1);
4534
4535                 /*
4536                  * If the address was optimistic
4537                  * we inserted the route at the start of
4538                  * our DAD process, so we don't need
4539                  * to do it again
4540                  */
4541                 if (!(ifp->rt->rt6i_node))
4542                         ip6_ins_rt(ifp->rt);
4543                 if (ifp->idev->cnf.forwarding)
4544                         addrconf_join_anycast(ifp);
4545                 if (!ipv6_addr_any(&ifp->peer_addr))
4546                         addrconf_prefix_route(&ifp->peer_addr, 128,
4547                                               ifp->idev->dev, 0, 0);
4548                 break;
4549         case RTM_DELADDR:
4550                 update_valid_ll_addr_cnt(ifp, -1);
4551
4552                 if (ifp->idev->cnf.forwarding)
4553                         addrconf_leave_anycast(ifp);
4554                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4555                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4556                         struct rt6_info *rt;
4557                         struct net_device *dev = ifp->idev->dev;
4558
4559                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4560                                         dev->ifindex, 1);
4561                         if (rt) {
4562                                 dst_hold(&rt->dst);
4563                                 if (ip6_del_rt(rt))
4564                                         dst_free(&rt->dst);
4565                         }
4566                 }
4567                 dst_hold(&ifp->rt->dst);
4568
4569                 if (ip6_del_rt(ifp->rt))
4570                         dst_free(&ifp->rt->dst);
4571                 break;
4572         }
4573         atomic_inc(&net->ipv6.dev_addr_genid);
4574         rt_genid_bump_ipv6(net);
4575 }
4576
4577 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4578 {
4579         rcu_read_lock_bh();
4580         if (likely(ifp->idev->dead == 0))
4581                 __ipv6_ifa_notify(event, ifp);
4582         rcu_read_unlock_bh();
4583 }
4584
4585 #ifdef CONFIG_SYSCTL
4586
4587 static
4588 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4589                            void __user *buffer, size_t *lenp, loff_t *ppos)
4590 {
4591         int *valp = ctl->data;
4592         int val = *valp;
4593         loff_t pos = *ppos;
4594         struct ctl_table lctl;
4595         int ret;
4596
4597         /*
4598          * ctl->data points to idev->cnf.forwarding, we should
4599          * not modify it until we get the rtnl lock.
4600          */
4601         lctl = *ctl;
4602         lctl.data = &val;
4603
4604         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4605
4606         if (write)
4607                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4608         if (ret)
4609                 *ppos = pos;
4610         return ret;
4611 }
4612
4613 static void dev_disable_change(struct inet6_dev *idev)
4614 {
4615         struct netdev_notifier_info info;
4616
4617         if (!idev || !idev->dev)
4618                 return;
4619
4620         netdev_notifier_info_init(&info, idev->dev);
4621         if (idev->cnf.disable_ipv6)
4622                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4623         else
4624                 addrconf_notify(NULL, NETDEV_UP, &info);
4625 }
4626
4627 static void addrconf_disable_change(struct net *net, __s32 newf)
4628 {
4629         struct net_device *dev;
4630         struct inet6_dev *idev;
4631
4632         rcu_read_lock();
4633         for_each_netdev_rcu(net, dev) {
4634                 idev = __in6_dev_get(dev);
4635                 if (idev) {
4636                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4637                         idev->cnf.disable_ipv6 = newf;
4638                         if (changed)
4639                                 dev_disable_change(idev);
4640                 }
4641         }
4642         rcu_read_unlock();
4643 }
4644
4645 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4646 {
4647         struct net *net;
4648         int old;
4649
4650         if (!rtnl_trylock())
4651                 return restart_syscall();
4652
4653         net = (struct net *)table->extra2;
4654         old = *p;
4655         *p = newf;
4656
4657         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4658                 rtnl_unlock();
4659                 return 0;
4660         }
4661
4662         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4663                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4664                 addrconf_disable_change(net, newf);
4665         } else if ((!newf) ^ (!old))
4666                 dev_disable_change((struct inet6_dev *)table->extra1);
4667
4668         rtnl_unlock();
4669         return 0;
4670 }
4671
4672 static
4673 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4674                             void __user *buffer, size_t *lenp, loff_t *ppos)
4675 {
4676         int *valp = ctl->data;
4677         int val = *valp;
4678         loff_t pos = *ppos;
4679         struct ctl_table lctl;
4680         int ret;
4681
4682         /*
4683          * ctl->data points to idev->cnf.disable_ipv6, we should
4684          * not modify it until we get the rtnl lock.
4685          */
4686         lctl = *ctl;
4687         lctl.data = &val;
4688
4689         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4690
4691         if (write)
4692                 ret = addrconf_disable_ipv6(ctl, valp, val);
4693         if (ret)
4694                 *ppos = pos;
4695         return ret;
4696 }
4697
4698 static struct addrconf_sysctl_table
4699 {
4700         struct ctl_table_header *sysctl_header;
4701         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4702 } addrconf_sysctl __read_mostly = {
4703         .sysctl_header = NULL,
4704         .addrconf_vars = {
4705                 {
4706                         .procname       = "forwarding",
4707                         .data           = &ipv6_devconf.forwarding,
4708                         .maxlen         = sizeof(int),
4709                         .mode           = 0644,
4710                         .proc_handler   = addrconf_sysctl_forward,
4711                 },
4712                 {
4713                         .procname       = "hop_limit",
4714                         .data           = &ipv6_devconf.hop_limit,
4715                         .maxlen         = sizeof(int),
4716                         .mode           = 0644,
4717                         .proc_handler   = proc_dointvec,
4718                 },
4719                 {
4720                         .procname       = "mtu",
4721                         .data           = &ipv6_devconf.mtu6,
4722                         .maxlen         = sizeof(int),
4723                         .mode           = 0644,
4724                         .proc_handler   = proc_dointvec,
4725                 },
4726                 {
4727                         .procname       = "accept_ra",
4728                         .data           = &ipv6_devconf.accept_ra,
4729                         .maxlen         = sizeof(int),
4730                         .mode           = 0644,
4731                         .proc_handler   = proc_dointvec,
4732                 },
4733                 {
4734                         .procname       = "accept_redirects",
4735                         .data           = &ipv6_devconf.accept_redirects,
4736                         .maxlen         = sizeof(int),
4737                         .mode           = 0644,
4738                         .proc_handler   = proc_dointvec,
4739                 },
4740                 {
4741                         .procname       = "autoconf",
4742                         .data           = &ipv6_devconf.autoconf,
4743                         .maxlen         = sizeof(int),
4744                         .mode           = 0644,
4745                         .proc_handler   = proc_dointvec,
4746                 },
4747                 {
4748                         .procname       = "dad_transmits",
4749                         .data           = &ipv6_devconf.dad_transmits,
4750                         .maxlen         = sizeof(int),
4751                         .mode           = 0644,
4752                         .proc_handler   = proc_dointvec,
4753                 },
4754                 {
4755                         .procname       = "router_solicitations",
4756                         .data           = &ipv6_devconf.rtr_solicits,
4757                         .maxlen         = sizeof(int),
4758                         .mode           = 0644,
4759                         .proc_handler   = proc_dointvec,
4760                 },
4761                 {
4762                         .procname       = "router_solicitation_interval",
4763                         .data           = &ipv6_devconf.rtr_solicit_interval,
4764                         .maxlen         = sizeof(int),
4765                         .mode           = 0644,
4766                         .proc_handler   = proc_dointvec_jiffies,
4767                 },
4768                 {
4769                         .procname       = "router_solicitation_delay",
4770                         .data           = &ipv6_devconf.rtr_solicit_delay,
4771                         .maxlen         = sizeof(int),
4772                         .mode           = 0644,
4773                         .proc_handler   = proc_dointvec_jiffies,
4774                 },
4775                 {
4776                         .procname       = "force_mld_version",
4777                         .data           = &ipv6_devconf.force_mld_version,
4778                         .maxlen         = sizeof(int),
4779                         .mode           = 0644,
4780                         .proc_handler   = proc_dointvec,
4781                 },
4782                 {
4783                         .procname       = "mldv1_unsolicited_report_interval",
4784                         .data           =
4785                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4786                         .maxlen         = sizeof(int),
4787                         .mode           = 0644,
4788                         .proc_handler   = proc_dointvec_ms_jiffies,
4789                 },
4790                 {
4791                         .procname       = "mldv2_unsolicited_report_interval",
4792                         .data           =
4793                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4794                         .maxlen         = sizeof(int),
4795                         .mode           = 0644,
4796                         .proc_handler   = proc_dointvec_ms_jiffies,
4797                 },
4798                 {
4799                         .procname       = "use_tempaddr",
4800                         .data           = &ipv6_devconf.use_tempaddr,
4801                         .maxlen         = sizeof(int),
4802                         .mode           = 0644,
4803                         .proc_handler   = proc_dointvec,
4804                 },
4805                 {
4806                         .procname       = "temp_valid_lft",
4807                         .data           = &ipv6_devconf.temp_valid_lft,
4808                         .maxlen         = sizeof(int),
4809                         .mode           = 0644,
4810                         .proc_handler   = proc_dointvec,
4811                 },
4812                 {
4813                         .procname       = "temp_prefered_lft",
4814                         .data           = &ipv6_devconf.temp_prefered_lft,
4815                         .maxlen         = sizeof(int),
4816                         .mode           = 0644,
4817                         .proc_handler   = proc_dointvec,
4818                 },
4819                 {
4820                         .procname       = "regen_max_retry",
4821                         .data           = &ipv6_devconf.regen_max_retry,
4822                         .maxlen         = sizeof(int),
4823                         .mode           = 0644,
4824                         .proc_handler   = proc_dointvec,
4825                 },
4826                 {
4827                         .procname       = "max_desync_factor",
4828                         .data           = &ipv6_devconf.max_desync_factor,
4829                         .maxlen         = sizeof(int),
4830                         .mode           = 0644,
4831                         .proc_handler   = proc_dointvec,
4832                 },
4833                 {
4834                         .procname       = "max_addresses",
4835                         .data           = &ipv6_devconf.max_addresses,
4836                         .maxlen         = sizeof(int),
4837                         .mode           = 0644,
4838                         .proc_handler   = proc_dointvec,
4839                 },
4840                 {
4841                         .procname       = "accept_ra_defrtr",
4842                         .data           = &ipv6_devconf.accept_ra_defrtr,
4843                         .maxlen         = sizeof(int),
4844                         .mode           = 0644,
4845                         .proc_handler   = proc_dointvec,
4846                 },
4847                 {
4848                         .procname       = "accept_ra_pinfo",
4849                         .data           = &ipv6_devconf.accept_ra_pinfo,
4850                         .maxlen         = sizeof(int),
4851                         .mode           = 0644,
4852                         .proc_handler   = proc_dointvec,
4853                 },
4854 #ifdef CONFIG_IPV6_ROUTER_PREF
4855                 {
4856                         .procname       = "accept_ra_rtr_pref",
4857                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4858                         .maxlen         = sizeof(int),
4859                         .mode           = 0644,
4860                         .proc_handler   = proc_dointvec,
4861                 },
4862                 {
4863                         .procname       = "router_probe_interval",
4864                         .data           = &ipv6_devconf.rtr_probe_interval,
4865                         .maxlen         = sizeof(int),
4866                         .mode           = 0644,
4867                         .proc_handler   = proc_dointvec_jiffies,
4868                 },
4869 #ifdef CONFIG_IPV6_ROUTE_INFO
4870                 {
4871                         .procname       = "accept_ra_rt_info_max_plen",
4872                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4873                         .maxlen         = sizeof(int),
4874                         .mode           = 0644,
4875                         .proc_handler   = proc_dointvec,
4876                 },
4877 #endif
4878 #endif
4879                 {
4880                         .procname       = "proxy_ndp",
4881                         .data           = &ipv6_devconf.proxy_ndp,
4882                         .maxlen         = sizeof(int),
4883                         .mode           = 0644,
4884                         .proc_handler   = proc_dointvec,
4885                 },
4886                 {
4887                         .procname       = "accept_source_route",
4888                         .data           = &ipv6_devconf.accept_source_route,
4889                         .maxlen         = sizeof(int),
4890                         .mode           = 0644,
4891                         .proc_handler   = proc_dointvec,
4892                 },
4893 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4894                 {
4895                         .procname       = "optimistic_dad",
4896                         .data           = &ipv6_devconf.optimistic_dad,
4897                         .maxlen         = sizeof(int),
4898                         .mode           = 0644,
4899                         .proc_handler   = proc_dointvec,
4900
4901                 },
4902 #endif
4903 #ifdef CONFIG_IPV6_MROUTE
4904                 {
4905                         .procname       = "mc_forwarding",
4906                         .data           = &ipv6_devconf.mc_forwarding,
4907                         .maxlen         = sizeof(int),
4908                         .mode           = 0444,
4909                         .proc_handler   = proc_dointvec,
4910                 },
4911 #endif
4912                 {
4913                         .procname       = "disable_ipv6",
4914                         .data           = &ipv6_devconf.disable_ipv6,
4915                         .maxlen         = sizeof(int),
4916                         .mode           = 0644,
4917                         .proc_handler   = addrconf_sysctl_disable,
4918                 },
4919                 {
4920                         .procname       = "accept_dad",
4921                         .data           = &ipv6_devconf.accept_dad,
4922                         .maxlen         = sizeof(int),
4923                         .mode           = 0644,
4924                         .proc_handler   = proc_dointvec,
4925                 },
4926                 {
4927                         .procname       = "force_tllao",
4928                         .data           = &ipv6_devconf.force_tllao,
4929                         .maxlen         = sizeof(int),
4930                         .mode           = 0644,
4931                         .proc_handler   = proc_dointvec
4932                 },
4933                 {
4934                         .procname       = "ndisc_notify",
4935                         .data           = &ipv6_devconf.ndisc_notify,
4936                         .maxlen         = sizeof(int),
4937                         .mode           = 0644,
4938                         .proc_handler   = proc_dointvec
4939                 },
4940                 {
4941                         .procname       = "suppress_frag_ndisc",
4942                         .data           = &ipv6_devconf.suppress_frag_ndisc,
4943                         .maxlen         = sizeof(int),
4944                         .mode           = 0644,
4945                         .proc_handler   = proc_dointvec
4946                 },
4947                 {
4948                         /* sentinel */
4949                 }
4950         },
4951 };
4952
4953 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4954                 struct inet6_dev *idev, struct ipv6_devconf *p)
4955 {
4956         int i;
4957         struct addrconf_sysctl_table *t;
4958         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4959
4960         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4961         if (t == NULL)
4962                 goto out;
4963
4964         for (i = 0; t->addrconf_vars[i].data; i++) {
4965                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4966                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4967                 t->addrconf_vars[i].extra2 = net;
4968         }
4969
4970         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4971
4972         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4973         if (t->sysctl_header == NULL)
4974                 goto free;
4975
4976         p->sysctl = t;
4977         return 0;
4978
4979 free:
4980         kfree(t);
4981 out:
4982         return -ENOBUFS;
4983 }
4984
4985 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4986 {
4987         struct addrconf_sysctl_table *t;
4988
4989         if (p->sysctl == NULL)
4990                 return;
4991
4992         t = p->sysctl;
4993         p->sysctl = NULL;
4994         unregister_net_sysctl_table(t->sysctl_header);
4995         kfree(t);
4996 }
4997
4998 static void addrconf_sysctl_register(struct inet6_dev *idev)
4999 {
5000         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5001                               &ndisc_ifinfo_sysctl_change);
5002         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5003                                         idev, &idev->cnf);
5004 }
5005
5006 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5007 {
5008         __addrconf_sysctl_unregister(&idev->cnf);
5009         neigh_sysctl_unregister(idev->nd_parms);
5010 }
5011
5012
5013 #endif
5014
5015 static int __net_init addrconf_init_net(struct net *net)
5016 {
5017         int err = -ENOMEM;
5018         struct ipv6_devconf *all, *dflt;
5019
5020         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5021         if (all == NULL)
5022                 goto err_alloc_all;
5023
5024         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5025         if (dflt == NULL)
5026                 goto err_alloc_dflt;
5027
5028         /* these will be inherited by all namespaces */
5029         dflt->autoconf = ipv6_defaults.autoconf;
5030         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5031
5032         net->ipv6.devconf_all = all;
5033         net->ipv6.devconf_dflt = dflt;
5034
5035 #ifdef CONFIG_SYSCTL
5036         err = __addrconf_sysctl_register(net, "all", NULL, all);
5037         if (err < 0)
5038                 goto err_reg_all;
5039
5040         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5041         if (err < 0)
5042                 goto err_reg_dflt;
5043 #endif
5044         return 0;
5045
5046 #ifdef CONFIG_SYSCTL
5047 err_reg_dflt:
5048         __addrconf_sysctl_unregister(all);
5049 err_reg_all:
5050         kfree(dflt);
5051 #endif
5052 err_alloc_dflt:
5053         kfree(all);
5054 err_alloc_all:
5055         return err;
5056 }
5057
5058 static void __net_exit addrconf_exit_net(struct net *net)
5059 {
5060 #ifdef CONFIG_SYSCTL
5061         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5062         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5063 #endif
5064         if (!net_eq(net, &init_net)) {
5065                 kfree(net->ipv6.devconf_dflt);
5066                 kfree(net->ipv6.devconf_all);
5067         }
5068 }
5069
5070 static struct pernet_operations addrconf_ops = {
5071         .init = addrconf_init_net,
5072         .exit = addrconf_exit_net,
5073 };
5074
5075 static struct rtnl_af_ops inet6_ops = {
5076         .family           = AF_INET6,
5077         .fill_link_af     = inet6_fill_link_af,
5078         .get_link_af_size = inet6_get_link_af_size,
5079         .set_link_af      = inet6_set_link_af,
5080 };
5081
5082 /*
5083  *      Init / cleanup code
5084  */
5085
5086 int __init addrconf_init(void)
5087 {
5088         int i, err;
5089
5090         err = ipv6_addr_label_init();
5091         if (err < 0) {
5092                 pr_crit("%s: cannot initialize default policy table: %d\n",
5093                         __func__, err);
5094                 goto out;
5095         }
5096
5097         err = register_pernet_subsys(&addrconf_ops);
5098         if (err < 0)
5099                 goto out_addrlabel;
5100
5101         /* The addrconf netdev notifier requires that loopback_dev
5102          * has it's ipv6 private information allocated and setup
5103          * before it can bring up and give link-local addresses
5104          * to other devices which are up.
5105          *
5106          * Unfortunately, loopback_dev is not necessarily the first
5107          * entry in the global dev_base list of net devices.  In fact,
5108          * it is likely to be the very last entry on that list.
5109          * So this causes the notifier registry below to try and
5110          * give link-local addresses to all devices besides loopback_dev
5111          * first, then loopback_dev, which cases all the non-loopback_dev
5112          * devices to fail to get a link-local address.
5113          *
5114          * So, as a temporary fix, allocate the ipv6 structure for
5115          * loopback_dev first by hand.
5116          * Longer term, all of the dependencies ipv6 has upon the loopback
5117          * device and it being up should be removed.
5118          */
5119         rtnl_lock();
5120         if (!ipv6_add_dev(init_net.loopback_dev))
5121                 err = -ENOMEM;
5122         rtnl_unlock();
5123         if (err)
5124                 goto errlo;
5125
5126         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5127                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5128
5129         register_netdevice_notifier(&ipv6_dev_notf);
5130
5131         addrconf_verify(0);
5132
5133         err = rtnl_af_register(&inet6_ops);
5134         if (err < 0)
5135                 goto errout_af;
5136
5137         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5138                               NULL);
5139         if (err < 0)
5140                 goto errout;
5141
5142         /* Only the first call to __rtnl_register can fail */
5143         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5144         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5145         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5146                         inet6_dump_ifaddr, NULL);
5147         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5148                         inet6_dump_ifmcaddr, NULL);
5149         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5150                         inet6_dump_ifacaddr, NULL);
5151         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5152                         inet6_netconf_dump_devconf, NULL);
5153
5154         ipv6_addr_label_rtnl_register();
5155
5156         return 0;
5157 errout:
5158         rtnl_af_unregister(&inet6_ops);
5159 errout_af:
5160         unregister_netdevice_notifier(&ipv6_dev_notf);
5161 errlo:
5162         unregister_pernet_subsys(&addrconf_ops);
5163 out_addrlabel:
5164         ipv6_addr_label_cleanup();
5165 out:
5166         return err;
5167 }
5168
5169 void addrconf_cleanup(void)
5170 {
5171         struct net_device *dev;
5172         int i;
5173
5174         unregister_netdevice_notifier(&ipv6_dev_notf);
5175         unregister_pernet_subsys(&addrconf_ops);
5176         ipv6_addr_label_cleanup();
5177
5178         rtnl_lock();
5179
5180         __rtnl_af_unregister(&inet6_ops);
5181
5182         /* clean dev list */
5183         for_each_netdev(&init_net, dev) {
5184                 if (__in6_dev_get(dev) == NULL)
5185                         continue;
5186                 addrconf_ifdown(dev, 1);
5187         }
5188         addrconf_ifdown(init_net.loopback_dev, 2);
5189
5190         /*
5191          *      Check hash table.
5192          */
5193         spin_lock_bh(&addrconf_hash_lock);
5194         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5195                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5196         spin_unlock_bh(&addrconf_hash_lock);
5197
5198         del_timer(&addr_chk_timer);
5199         rtnl_unlock();
5200 }