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