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 = ipv6_add_addr(idev, &addr, tmp_plen,
1128                             ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1129                             addr_flags);
1130         if (IS_ERR(ift)) {
1131                 in6_ifa_put(ifp);
1132                 in6_dev_put(idev);
1133                 pr_info("%s: retry temporary address regeneration\n", __func__);
1134                 tmpaddr = &addr;
1135                 write_lock(&idev->lock);
1136                 goto retry;
1137         }
1138
1139         spin_lock_bh(&ift->lock);
1140         ift->ifpub = ifp;
1141         ift->valid_lft = tmp_valid_lft;
1142         ift->prefered_lft = tmp_prefered_lft;
1143         ift->cstamp = now;
1144         ift->tstamp = tmp_tstamp;
1145         spin_unlock_bh(&ift->lock);
1146
1147         addrconf_dad_start(ift);
1148         in6_ifa_put(ift);
1149         in6_dev_put(idev);
1150 out:
1151         return ret;
1152 }
1153 #endif
1154
1155 /*
1156  *      Choose an appropriate source address (RFC3484)
1157  */
1158 enum {
1159         IPV6_SADDR_RULE_INIT = 0,
1160         IPV6_SADDR_RULE_LOCAL,
1161         IPV6_SADDR_RULE_SCOPE,
1162         IPV6_SADDR_RULE_PREFERRED,
1163 #ifdef CONFIG_IPV6_MIP6
1164         IPV6_SADDR_RULE_HOA,
1165 #endif
1166         IPV6_SADDR_RULE_OIF,
1167         IPV6_SADDR_RULE_LABEL,
1168 #ifdef CONFIG_IPV6_PRIVACY
1169         IPV6_SADDR_RULE_PRIVACY,
1170 #endif
1171         IPV6_SADDR_RULE_ORCHID,
1172         IPV6_SADDR_RULE_PREFIX,
1173         IPV6_SADDR_RULE_MAX
1174 };
1175
1176 struct ipv6_saddr_score {
1177         int                     rule;
1178         int                     addr_type;
1179         struct inet6_ifaddr     *ifa;
1180         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1181         int                     scopedist;
1182         int                     matchlen;
1183 };
1184
1185 struct ipv6_saddr_dst {
1186         const struct in6_addr *addr;
1187         int ifindex;
1188         int scope;
1189         int label;
1190         unsigned int prefs;
1191 };
1192
1193 static inline int ipv6_saddr_preferred(int type)
1194 {
1195         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1196                 return 1;
1197         return 0;
1198 }
1199
1200 static int ipv6_get_saddr_eval(struct net *net,
1201                                struct ipv6_saddr_score *score,
1202                                struct ipv6_saddr_dst *dst,
1203                                int i)
1204 {
1205         int ret;
1206
1207         if (i <= score->rule) {
1208                 switch (i) {
1209                 case IPV6_SADDR_RULE_SCOPE:
1210                         ret = score->scopedist;
1211                         break;
1212                 case IPV6_SADDR_RULE_PREFIX:
1213                         ret = score->matchlen;
1214                         break;
1215                 default:
1216                         ret = !!test_bit(i, score->scorebits);
1217                 }
1218                 goto out;
1219         }
1220
1221         switch (i) {
1222         case IPV6_SADDR_RULE_INIT:
1223                 /* Rule 0: remember if hiscore is not ready yet */
1224                 ret = !!score->ifa;
1225                 break;
1226         case IPV6_SADDR_RULE_LOCAL:
1227                 /* Rule 1: Prefer same address */
1228                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1229                 break;
1230         case IPV6_SADDR_RULE_SCOPE:
1231                 /* Rule 2: Prefer appropriate scope
1232                  *
1233                  *      ret
1234                  *       ^
1235                  *    -1 |  d 15
1236                  *    ---+--+-+---> scope
1237                  *       |
1238                  *       |             d is scope of the destination.
1239                  *  B-d  |  \
1240                  *       |   \      <- smaller scope is better if
1241                  *  B-15 |    \        if scope is enough for destinaion.
1242                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1243                  * d-C-1 | /
1244                  *       |/         <- greater is better
1245                  *   -C  /             if scope is not enough for destination.
1246                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1247                  *
1248                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1249                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1250                  * Assume B = 0 and we get C > 29.
1251                  */
1252                 ret = __ipv6_addr_src_scope(score->addr_type);
1253                 if (ret >= dst->scope)
1254                         ret = -ret;
1255                 else
1256                         ret -= 128;     /* 30 is enough */
1257                 score->scopedist = ret;
1258                 break;
1259         case IPV6_SADDR_RULE_PREFERRED:
1260                 /* Rule 3: Avoid deprecated and optimistic addresses */
1261                 ret = ipv6_saddr_preferred(score->addr_type) ||
1262                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1263                 break;
1264 #ifdef CONFIG_IPV6_MIP6
1265         case IPV6_SADDR_RULE_HOA:
1266             {
1267                 /* Rule 4: Prefer home address */
1268                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1269                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1270                 break;
1271             }
1272 #endif
1273         case IPV6_SADDR_RULE_OIF:
1274                 /* Rule 5: Prefer outgoing interface */
1275                 ret = (!dst->ifindex ||
1276                        dst->ifindex == score->ifa->idev->dev->ifindex);
1277                 break;
1278         case IPV6_SADDR_RULE_LABEL:
1279                 /* Rule 6: Prefer matching label */
1280                 ret = ipv6_addr_label(net,
1281                                       &score->ifa->addr, score->addr_type,
1282                                       score->ifa->idev->dev->ifindex) == dst->label;
1283                 break;
1284 #ifdef CONFIG_IPV6_PRIVACY
1285         case IPV6_SADDR_RULE_PRIVACY:
1286             {
1287                 /* Rule 7: Prefer public address
1288                  * Note: prefer temporary address if use_tempaddr >= 2
1289                  */
1290                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1291                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1292                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1293                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1294                 break;
1295             }
1296 #endif
1297         case IPV6_SADDR_RULE_ORCHID:
1298                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1299                  *          non-ORCHID vs non-ORCHID
1300                  */
1301                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1302                         ipv6_addr_orchid(dst->addr));
1303                 break;
1304         case IPV6_SADDR_RULE_PREFIX:
1305                 /* Rule 8: Use longest matching prefix */
1306                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1307                 if (ret > score->ifa->prefix_len)
1308                         ret = score->ifa->prefix_len;
1309                 score->matchlen = ret;
1310                 break;
1311         default:
1312                 ret = 0;
1313         }
1314
1315         if (ret)
1316                 __set_bit(i, score->scorebits);
1317         score->rule = i;
1318 out:
1319         return ret;
1320 }
1321
1322 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1323                        const struct in6_addr *daddr, unsigned int prefs,
1324                        struct in6_addr *saddr)
1325 {
1326         struct ipv6_saddr_score scores[2],
1327                                 *score = &scores[0], *hiscore = &scores[1];
1328         struct ipv6_saddr_dst dst;
1329         struct net_device *dev;
1330         int dst_type;
1331
1332         dst_type = __ipv6_addr_type(daddr);
1333         dst.addr = daddr;
1334         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1335         dst.scope = __ipv6_addr_src_scope(dst_type);
1336         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1337         dst.prefs = prefs;
1338
1339         hiscore->rule = -1;
1340         hiscore->ifa = NULL;
1341
1342         rcu_read_lock();
1343
1344         for_each_netdev_rcu(net, dev) {
1345                 struct inet6_dev *idev;
1346
1347                 /* Candidate Source Address (section 4)
1348                  *  - multicast and link-local destination address,
1349                  *    the set of candidate source address MUST only
1350                  *    include addresses assigned to interfaces
1351                  *    belonging to the same link as the outgoing
1352                  *    interface.
1353                  * (- For site-local destination addresses, the
1354                  *    set of candidate source addresses MUST only
1355                  *    include addresses assigned to interfaces
1356                  *    belonging to the same site as the outgoing
1357                  *    interface.)
1358                  */
1359                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1360                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1361                     dst.ifindex && dev->ifindex != dst.ifindex)
1362                         continue;
1363
1364                 idev = __in6_dev_get(dev);
1365                 if (!idev)
1366                         continue;
1367
1368                 read_lock_bh(&idev->lock);
1369                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1370                         int i;
1371
1372                         /*
1373                          * - Tentative Address (RFC2462 section 5.4)
1374                          *  - A tentative address is not considered
1375                          *    "assigned to an interface" in the traditional
1376                          *    sense, unless it is also flagged as optimistic.
1377                          * - Candidate Source Address (section 4)
1378                          *  - In any case, anycast addresses, multicast
1379                          *    addresses, and the unspecified address MUST
1380                          *    NOT be included in a candidate set.
1381                          */
1382                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1383                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1384                                 continue;
1385
1386                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1387
1388                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1389                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1390                                 LIMIT_NETDEBUG(KERN_DEBUG
1391                                                "ADDRCONF: unspecified / multicast address "
1392                                                "assigned as unicast address on %s",
1393                                                dev->name);
1394                                 continue;
1395                         }
1396
1397                         score->rule = -1;
1398                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1399
1400                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1401                                 int minihiscore, miniscore;
1402
1403                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1404                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1405
1406                                 if (minihiscore > miniscore) {
1407                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1408                                             score->scopedist > 0) {
1409                                                 /*
1410                                                  * special case:
1411                                                  * each remaining entry
1412                                                  * has too small (not enough)
1413                                                  * scope, because ifa entries
1414                                                  * are sorted by their scope
1415                                                  * values.
1416                                                  */
1417                                                 goto try_nextdev;
1418                                         }
1419                                         break;
1420                                 } else if (minihiscore < miniscore) {
1421                                         if (hiscore->ifa)
1422                                                 in6_ifa_put(hiscore->ifa);
1423
1424                                         in6_ifa_hold(score->ifa);
1425
1426                                         swap(hiscore, score);
1427
1428                                         /* restore our iterator */
1429                                         score->ifa = hiscore->ifa;
1430
1431                                         break;
1432                                 }
1433                         }
1434                 }
1435 try_nextdev:
1436                 read_unlock_bh(&idev->lock);
1437         }
1438         rcu_read_unlock();
1439
1440         if (!hiscore->ifa)
1441                 return -EADDRNOTAVAIL;
1442
1443         *saddr = hiscore->ifa->addr;
1444         in6_ifa_put(hiscore->ifa);
1445         return 0;
1446 }
1447 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1448
1449 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1450                       unsigned char banned_flags)
1451 {
1452         struct inet6_ifaddr *ifp;
1453         int err = -EADDRNOTAVAIL;
1454
1455         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1456                 if (ifp->scope == IFA_LINK &&
1457                     !(ifp->flags & banned_flags)) {
1458                         *addr = ifp->addr;
1459                         err = 0;
1460                         break;
1461                 }
1462         }
1463         return err;
1464 }
1465
1466 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1467                     unsigned char banned_flags)
1468 {
1469         struct inet6_dev *idev;
1470         int err = -EADDRNOTAVAIL;
1471
1472         rcu_read_lock();
1473         idev = __in6_dev_get(dev);
1474         if (idev) {
1475                 read_lock_bh(&idev->lock);
1476                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1477                 read_unlock_bh(&idev->lock);
1478         }
1479         rcu_read_unlock();
1480         return err;
1481 }
1482
1483 static int ipv6_count_addresses(struct inet6_dev *idev)
1484 {
1485         int cnt = 0;
1486         struct inet6_ifaddr *ifp;
1487
1488         read_lock_bh(&idev->lock);
1489         list_for_each_entry(ifp, &idev->addr_list, if_list)
1490                 cnt++;
1491         read_unlock_bh(&idev->lock);
1492         return cnt;
1493 }
1494
1495 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1496                   const struct net_device *dev, int strict)
1497 {
1498         struct inet6_ifaddr *ifp;
1499         unsigned int hash = inet6_addr_hash(addr);
1500
1501         rcu_read_lock_bh();
1502         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1503                 if (!net_eq(dev_net(ifp->idev->dev), net))
1504                         continue;
1505                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1506                     !(ifp->flags&IFA_F_TENTATIVE) &&
1507                     (dev == NULL || ifp->idev->dev == dev ||
1508                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1509                         rcu_read_unlock_bh();
1510                         return 1;
1511                 }
1512         }
1513
1514         rcu_read_unlock_bh();
1515         return 0;
1516 }
1517 EXPORT_SYMBOL(ipv6_chk_addr);
1518
1519 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1520                                struct net_device *dev)
1521 {
1522         unsigned int hash = inet6_addr_hash(addr);
1523         struct inet6_ifaddr *ifp;
1524
1525         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1526                 if (!net_eq(dev_net(ifp->idev->dev), net))
1527                         continue;
1528                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1529                         if (dev == NULL || ifp->idev->dev == dev)
1530                                 return true;
1531                 }
1532         }
1533         return false;
1534 }
1535
1536 /* Compares an address/prefix_len with addresses on device @dev.
1537  * If one is found it returns true.
1538  */
1539 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1540         const unsigned int prefix_len, struct net_device *dev)
1541 {
1542         struct inet6_dev *idev;
1543         struct inet6_ifaddr *ifa;
1544         bool ret = false;
1545
1546         rcu_read_lock();
1547         idev = __in6_dev_get(dev);
1548         if (idev) {
1549                 read_lock_bh(&idev->lock);
1550                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1551                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1552                         if (ret)
1553                                 break;
1554                 }
1555                 read_unlock_bh(&idev->lock);
1556         }
1557         rcu_read_unlock();
1558
1559         return ret;
1560 }
1561 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1562
1563 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1564 {
1565         struct inet6_dev *idev;
1566         struct inet6_ifaddr *ifa;
1567         int     onlink;
1568
1569         onlink = 0;
1570         rcu_read_lock();
1571         idev = __in6_dev_get(dev);
1572         if (idev) {
1573                 read_lock_bh(&idev->lock);
1574                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1575                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1576                                                    ifa->prefix_len);
1577                         if (onlink)
1578                                 break;
1579                 }
1580                 read_unlock_bh(&idev->lock);
1581         }
1582         rcu_read_unlock();
1583         return onlink;
1584 }
1585 EXPORT_SYMBOL(ipv6_chk_prefix);
1586
1587 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1588                                      struct net_device *dev, int strict)
1589 {
1590         struct inet6_ifaddr *ifp, *result = NULL;
1591         unsigned int hash = inet6_addr_hash(addr);
1592
1593         rcu_read_lock_bh();
1594         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1595                 if (!net_eq(dev_net(ifp->idev->dev), net))
1596                         continue;
1597                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1598                         if (dev == NULL || ifp->idev->dev == dev ||
1599                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1600                                 result = ifp;
1601                                 in6_ifa_hold(ifp);
1602                                 break;
1603                         }
1604                 }
1605         }
1606         rcu_read_unlock_bh();
1607
1608         return result;
1609 }
1610
1611 /* Gets referenced address, destroys ifaddr */
1612
1613 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1614 {
1615         if (ifp->flags&IFA_F_PERMANENT) {
1616                 spin_lock_bh(&ifp->lock);
1617                 addrconf_del_timer(ifp);
1618                 ifp->flags |= IFA_F_TENTATIVE;
1619                 if (dad_failed)
1620                         ifp->flags |= IFA_F_DADFAILED;
1621                 spin_unlock_bh(&ifp->lock);
1622                 if (dad_failed)
1623                         ipv6_ifa_notify(0, ifp);
1624                 in6_ifa_put(ifp);
1625 #ifdef CONFIG_IPV6_PRIVACY
1626         } else if (ifp->flags&IFA_F_TEMPORARY) {
1627                 struct inet6_ifaddr *ifpub;
1628                 spin_lock_bh(&ifp->lock);
1629                 ifpub = ifp->ifpub;
1630                 if (ifpub) {
1631                         in6_ifa_hold(ifpub);
1632                         spin_unlock_bh(&ifp->lock);
1633                         ipv6_create_tempaddr(ifpub, ifp);
1634                         in6_ifa_put(ifpub);
1635                 } else {
1636                         spin_unlock_bh(&ifp->lock);
1637                 }
1638                 ipv6_del_addr(ifp);
1639 #endif
1640         } else
1641                 ipv6_del_addr(ifp);
1642 }
1643
1644 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1645 {
1646         int err = -ENOENT;
1647
1648         spin_lock(&ifp->state_lock);
1649         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1650                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1651                 err = 0;
1652         }
1653         spin_unlock(&ifp->state_lock);
1654
1655         return err;
1656 }
1657
1658 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1659 {
1660         struct inet6_dev *idev = ifp->idev;
1661
1662         if (addrconf_dad_end(ifp)) {
1663                 in6_ifa_put(ifp);
1664                 return;
1665         }
1666
1667         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1668                              ifp->idev->dev->name, &ifp->addr);
1669
1670         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1671                 struct in6_addr addr;
1672
1673                 addr.s6_addr32[0] = htonl(0xfe800000);
1674                 addr.s6_addr32[1] = 0;
1675
1676                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1677                     ipv6_addr_equal(&ifp->addr, &addr)) {
1678                         /* DAD failed for link-local based on MAC address */
1679                         idev->cnf.disable_ipv6 = 1;
1680
1681                         pr_info("%s: IPv6 being disabled!\n",
1682                                 ifp->idev->dev->name);
1683                 }
1684         }
1685
1686         addrconf_dad_stop(ifp, 1);
1687 }
1688
1689 /* Join to solicited addr multicast group. */
1690
1691 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1692 {
1693         struct in6_addr maddr;
1694
1695         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1696                 return;
1697
1698         addrconf_addr_solict_mult(addr, &maddr);
1699         ipv6_dev_mc_inc(dev, &maddr);
1700 }
1701
1702 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1703 {
1704         struct in6_addr maddr;
1705
1706         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1707                 return;
1708
1709         addrconf_addr_solict_mult(addr, &maddr);
1710         __ipv6_dev_mc_dec(idev, &maddr);
1711 }
1712
1713 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1714 {
1715         struct in6_addr addr;
1716         if (ifp->prefix_len == 127) /* RFC 6164 */
1717                 return;
1718         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1719         if (ipv6_addr_any(&addr))
1720                 return;
1721         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1722 }
1723
1724 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1725 {
1726         struct in6_addr addr;
1727         if (ifp->prefix_len == 127) /* RFC 6164 */
1728                 return;
1729         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1730         if (ipv6_addr_any(&addr))
1731                 return;
1732         __ipv6_dev_ac_dec(ifp->idev, &addr);
1733 }
1734
1735 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1736 {
1737         if (dev->addr_len != ETH_ALEN)
1738                 return -1;
1739         memcpy(eui, dev->dev_addr, 3);
1740         memcpy(eui + 5, dev->dev_addr + 3, 3);
1741
1742         /*
1743          * The zSeries OSA network cards can be shared among various
1744          * OS instances, but the OSA cards have only one MAC address.
1745          * This leads to duplicate address conflicts in conjunction
1746          * with IPv6 if more than one instance uses the same card.
1747          *
1748          * The driver for these cards can deliver a unique 16-bit
1749          * identifier for each instance sharing the same card.  It is
1750          * placed instead of 0xFFFE in the interface identifier.  The
1751          * "u" bit of the interface identifier is not inverted in this
1752          * case.  Hence the resulting interface identifier has local
1753          * scope according to RFC2373.
1754          */
1755         if (dev->dev_id) {
1756                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1757                 eui[4] = dev->dev_id & 0xFF;
1758         } else {
1759                 eui[3] = 0xFF;
1760                 eui[4] = 0xFE;
1761                 eui[0] ^= 2;
1762         }
1763         return 0;
1764 }
1765
1766 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1767 {
1768         if (dev->addr_len != IEEE802154_ADDR_LEN)
1769                 return -1;
1770         memcpy(eui, dev->dev_addr, 8);
1771         eui[0] ^= 2;
1772         return 0;
1773 }
1774
1775 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1776 {
1777         union fwnet_hwaddr *ha;
1778
1779         if (dev->addr_len != FWNET_ALEN)
1780                 return -1;
1781
1782         ha = (union fwnet_hwaddr *)dev->dev_addr;
1783
1784         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1785         eui[0] ^= 2;
1786         return 0;
1787 }
1788
1789 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1790 {
1791         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1792         if (dev->addr_len != ARCNET_ALEN)
1793                 return -1;
1794         memset(eui, 0, 7);
1795         eui[7] = *(u8 *)dev->dev_addr;
1796         return 0;
1797 }
1798
1799 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1800 {
1801         if (dev->addr_len != INFINIBAND_ALEN)
1802                 return -1;
1803         memcpy(eui, dev->dev_addr + 12, 8);
1804         eui[0] |= 2;
1805         return 0;
1806 }
1807
1808 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1809 {
1810         if (addr == 0)
1811                 return -1;
1812         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1813                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1814                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1815                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1816                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1817                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1818         eui[1] = 0;
1819         eui[2] = 0x5E;
1820         eui[3] = 0xFE;
1821         memcpy(eui + 4, &addr, 4);
1822         return 0;
1823 }
1824
1825 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1826 {
1827         if (dev->priv_flags & IFF_ISATAP)
1828                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1829         return -1;
1830 }
1831
1832 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1833 {
1834         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1835 }
1836
1837 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1838 {
1839         switch (dev->type) {
1840         case ARPHRD_ETHER:
1841         case ARPHRD_FDDI:
1842                 return addrconf_ifid_eui48(eui, dev);
1843         case ARPHRD_ARCNET:
1844                 return addrconf_ifid_arcnet(eui, dev);
1845         case ARPHRD_INFINIBAND:
1846                 return addrconf_ifid_infiniband(eui, dev);
1847         case ARPHRD_SIT:
1848                 return addrconf_ifid_sit(eui, dev);
1849         case ARPHRD_IPGRE:
1850                 return addrconf_ifid_gre(eui, dev);
1851         case ARPHRD_IEEE802154:
1852                 return addrconf_ifid_eui64(eui, dev);
1853         case ARPHRD_IEEE1394:
1854                 return addrconf_ifid_ieee1394(eui, dev);
1855         }
1856         return -1;
1857 }
1858
1859 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1860 {
1861         int err = -1;
1862         struct inet6_ifaddr *ifp;
1863
1864         read_lock_bh(&idev->lock);
1865         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1866                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1867                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1868                         err = 0;
1869                         break;
1870                 }
1871         }
1872         read_unlock_bh(&idev->lock);
1873         return err;
1874 }
1875
1876 #ifdef CONFIG_IPV6_PRIVACY
1877 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1878 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1879 {
1880 regen:
1881         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1882         idev->rndid[0] &= ~0x02;
1883
1884         /*
1885          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1886          * check if generated address is not inappropriate
1887          *
1888          *  - Reserved subnet anycast (RFC 2526)
1889          *      11111101 11....11 1xxxxxxx
1890          *  - ISATAP (RFC4214) 6.1
1891          *      00-00-5E-FE-xx-xx-xx-xx
1892          *  - value 0
1893          *  - XXX: already assigned to an address on the device
1894          */
1895         if (idev->rndid[0] == 0xfd &&
1896             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1897             (idev->rndid[7]&0x80))
1898                 goto regen;
1899         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1900                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1901                         goto regen;
1902                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1903                         goto regen;
1904         }
1905 }
1906
1907 static void ipv6_regen_rndid(unsigned long data)
1908 {
1909         struct inet6_dev *idev = (struct inet6_dev *) data;
1910         unsigned long expires;
1911
1912         rcu_read_lock_bh();
1913         write_lock_bh(&idev->lock);
1914
1915         if (idev->dead)
1916                 goto out;
1917
1918         __ipv6_regen_rndid(idev);
1919
1920         expires = jiffies +
1921                 idev->cnf.temp_prefered_lft * HZ -
1922                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1923                 idev->cnf.max_desync_factor * HZ;
1924         if (time_before(expires, jiffies)) {
1925                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1926                         __func__, idev->dev->name);
1927                 goto out;
1928         }
1929
1930         if (!mod_timer(&idev->regen_timer, expires))
1931                 in6_dev_hold(idev);
1932
1933 out:
1934         write_unlock_bh(&idev->lock);
1935         rcu_read_unlock_bh();
1936         in6_dev_put(idev);
1937 }
1938
1939 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1940 {
1941         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1942                 __ipv6_regen_rndid(idev);
1943 }
1944 #endif
1945
1946 /*
1947  *      Add prefix route.
1948  */
1949
1950 static void
1951 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1952                       unsigned long expires, u32 flags)
1953 {
1954         struct fib6_config cfg = {
1955                 .fc_table = RT6_TABLE_PREFIX,
1956                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1957                 .fc_ifindex = dev->ifindex,
1958                 .fc_expires = expires,
1959                 .fc_dst_len = plen,
1960                 .fc_flags = RTF_UP | flags,
1961                 .fc_nlinfo.nl_net = dev_net(dev),
1962                 .fc_protocol = RTPROT_KERNEL,
1963         };
1964
1965         cfg.fc_dst = *pfx;
1966
1967         /* Prevent useless cloning on PtP SIT.
1968            This thing is done here expecting that the whole
1969            class of non-broadcast devices need not cloning.
1970          */
1971 #if IS_ENABLED(CONFIG_IPV6_SIT)
1972         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1973                 cfg.fc_flags |= RTF_NONEXTHOP;
1974 #endif
1975
1976         ip6_route_add(&cfg);
1977 }
1978
1979
1980 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1981                                                   int plen,
1982                                                   const struct net_device *dev,
1983                                                   u32 flags, u32 noflags)
1984 {
1985         struct fib6_node *fn;
1986         struct rt6_info *rt = NULL;
1987         struct fib6_table *table;
1988
1989         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1990         if (table == NULL)
1991                 return NULL;
1992
1993         read_lock_bh(&table->tb6_lock);
1994         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1995         if (!fn)
1996                 goto out;
1997         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1998                 if (rt->dst.dev->ifindex != dev->ifindex)
1999                         continue;
2000                 if ((rt->rt6i_flags & flags) != flags)
2001                         continue;
2002                 if ((rt->rt6i_flags & noflags) != 0)
2003                         continue;
2004                 dst_hold(&rt->dst);
2005                 break;
2006         }
2007 out:
2008         read_unlock_bh(&table->tb6_lock);
2009         return rt;
2010 }
2011
2012
2013 /* Create "default" multicast route to the interface */
2014
2015 static void addrconf_add_mroute(struct net_device *dev)
2016 {
2017         struct fib6_config cfg = {
2018                 .fc_table = RT6_TABLE_LOCAL,
2019                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2020                 .fc_ifindex = dev->ifindex,
2021                 .fc_dst_len = 8,
2022                 .fc_flags = RTF_UP,
2023                 .fc_nlinfo.nl_net = dev_net(dev),
2024         };
2025
2026         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2027
2028         ip6_route_add(&cfg);
2029 }
2030
2031 #if IS_ENABLED(CONFIG_IPV6_SIT)
2032 static void sit_route_add(struct net_device *dev)
2033 {
2034         struct fib6_config cfg = {
2035                 .fc_table = RT6_TABLE_MAIN,
2036                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2037                 .fc_ifindex = dev->ifindex,
2038                 .fc_dst_len = 96,
2039                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2040                 .fc_nlinfo.nl_net = dev_net(dev),
2041         };
2042
2043         /* prefix length - 96 bits "::d.d.d.d" */
2044         ip6_route_add(&cfg);
2045 }
2046 #endif
2047
2048 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2049 {
2050         struct inet6_dev *idev;
2051
2052         ASSERT_RTNL();
2053
2054         idev = ipv6_find_idev(dev);
2055         if (!idev)
2056                 return ERR_PTR(-ENOBUFS);
2057
2058         if (idev->cnf.disable_ipv6)
2059                 return ERR_PTR(-EACCES);
2060
2061         /* Add default multicast route */
2062         if (!(dev->flags & IFF_LOOPBACK))
2063                 addrconf_add_mroute(dev);
2064
2065         return idev;
2066 }
2067
2068 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2069 {
2070         struct prefix_info *pinfo;
2071         __u32 valid_lft;
2072         __u32 prefered_lft;
2073         int addr_type;
2074         struct inet6_dev *in6_dev;
2075         struct net *net = dev_net(dev);
2076
2077         pinfo = (struct prefix_info *) opt;
2078
2079         if (len < sizeof(struct prefix_info)) {
2080                 ADBG(("addrconf: prefix option too short\n"));
2081                 return;
2082         }
2083
2084         /*
2085          *      Validation checks ([ADDRCONF], page 19)
2086          */
2087
2088         addr_type = ipv6_addr_type(&pinfo->prefix);
2089
2090         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2091                 return;
2092
2093         valid_lft = ntohl(pinfo->valid);
2094         prefered_lft = ntohl(pinfo->prefered);
2095
2096         if (prefered_lft > valid_lft) {
2097                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2098                 return;
2099         }
2100
2101         in6_dev = in6_dev_get(dev);
2102
2103         if (in6_dev == NULL) {
2104                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2105                                     dev->name);
2106                 return;
2107         }
2108
2109         /*
2110          *      Two things going on here:
2111          *      1) Add routes for on-link prefixes
2112          *      2) Configure prefixes with the auto flag set
2113          */
2114
2115         if (pinfo->onlink) {
2116                 struct rt6_info *rt;
2117                 unsigned long rt_expires;
2118
2119                 /* Avoid arithmetic overflow. Really, we could
2120                  * save rt_expires in seconds, likely valid_lft,
2121                  * but it would require division in fib gc, that it
2122                  * not good.
2123                  */
2124                 if (HZ > USER_HZ)
2125                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2126                 else
2127                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2128
2129                 if (addrconf_finite_timeout(rt_expires))
2130                         rt_expires *= HZ;
2131
2132                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2133                                                pinfo->prefix_len,
2134                                                dev,
2135                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2136                                                RTF_GATEWAY | RTF_DEFAULT);
2137
2138                 if (rt) {
2139                         /* Autoconf prefix route */
2140                         if (valid_lft == 0) {
2141                                 ip6_del_rt(rt);
2142                                 rt = NULL;
2143                         } else if (addrconf_finite_timeout(rt_expires)) {
2144                                 /* not infinity */
2145                                 rt6_set_expires(rt, jiffies + rt_expires);
2146                         } else {
2147                                 rt6_clean_expires(rt);
2148                         }
2149                 } else if (valid_lft) {
2150                         clock_t expires = 0;
2151                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2152                         if (addrconf_finite_timeout(rt_expires)) {
2153                                 /* not infinity */
2154                                 flags |= RTF_EXPIRES;
2155                                 expires = jiffies_to_clock_t(rt_expires);
2156                         }
2157                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2158                                               dev, expires, flags);
2159                 }
2160                 ip6_rt_put(rt);
2161         }
2162
2163         /* Try to figure out our local address for this prefix */
2164
2165         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2166                 struct inet6_ifaddr *ifp;
2167                 struct in6_addr addr;
2168                 int create = 0, update_lft = 0;
2169                 bool tokenized = false;
2170
2171                 if (pinfo->prefix_len == 64) {
2172                         memcpy(&addr, &pinfo->prefix, 8);
2173
2174                         if (!ipv6_addr_any(&in6_dev->token)) {
2175                                 read_lock_bh(&in6_dev->lock);
2176                                 memcpy(addr.s6_addr + 8,
2177                                        in6_dev->token.s6_addr + 8, 8);
2178                                 read_unlock_bh(&in6_dev->lock);
2179                                 tokenized = true;
2180                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2181                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2182                                 in6_dev_put(in6_dev);
2183                                 return;
2184                         }
2185                         goto ok;
2186                 }
2187                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2188                                     pinfo->prefix_len);
2189                 in6_dev_put(in6_dev);
2190                 return;
2191
2192 ok:
2193
2194                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2195
2196                 if (ifp == NULL && valid_lft) {
2197                         int max_addresses = in6_dev->cnf.max_addresses;
2198                         u32 addr_flags = 0;
2199
2200 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2201                         if (in6_dev->cnf.optimistic_dad &&
2202                             !net->ipv6.devconf_all->forwarding && sllao)
2203                                 addr_flags = IFA_F_OPTIMISTIC;
2204 #endif
2205
2206                         /* Do not allow to create too much of autoconfigured
2207                          * addresses; this would be too easy way to crash kernel.
2208                          */
2209                         if (!max_addresses ||
2210                             ipv6_count_addresses(in6_dev) < max_addresses)
2211                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2212                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2213                                                     addr_flags);
2214
2215                         if (IS_ERR_OR_NULL(ifp)) {
2216                                 in6_dev_put(in6_dev);
2217                                 return;
2218                         }
2219
2220                         update_lft = create = 1;
2221                         ifp->cstamp = jiffies;
2222                         ifp->tokenized = tokenized;
2223                         addrconf_dad_start(ifp);
2224                 }
2225
2226                 if (ifp) {
2227                         int flags;
2228                         unsigned long now;
2229 #ifdef CONFIG_IPV6_PRIVACY
2230                         struct inet6_ifaddr *ift;
2231 #endif
2232                         u32 stored_lft;
2233
2234                         /* update lifetime (RFC2462 5.5.3 e) */
2235                         spin_lock(&ifp->lock);
2236                         now = jiffies;
2237                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2238                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2239                         else
2240                                 stored_lft = 0;
2241                         if (!update_lft && stored_lft) {
2242                                 if (valid_lft > MIN_VALID_LIFETIME ||
2243                                     valid_lft > stored_lft)
2244                                         update_lft = 1;
2245                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2246                                         /* valid_lft <= stored_lft is always true */
2247                                         /*
2248                                          * RFC 4862 Section 5.5.3e:
2249                                          * "Note that the preferred lifetime of
2250                                          *  the corresponding address is always
2251                                          *  reset to the Preferred Lifetime in
2252                                          *  the received Prefix Information
2253                                          *  option, regardless of whether the
2254                                          *  valid lifetime is also reset or
2255                                          *  ignored."
2256                                          *
2257                                          *  So if the preferred lifetime in
2258                                          *  this advertisement is different
2259                                          *  than what we have stored, but the
2260                                          *  valid lifetime is invalid, just
2261                                          *  reset prefered_lft.
2262                                          *
2263                                          *  We must set the valid lifetime
2264                                          *  to the stored lifetime since we'll
2265                                          *  be updating the timestamp below,
2266                                          *  else we'll set it back to the
2267                                          *  minimum.
2268                                          */
2269                                         if (prefered_lft != ifp->prefered_lft) {
2270                                                 valid_lft = stored_lft;
2271                                                 update_lft = 1;
2272                                         }
2273                                 } else {
2274                                         valid_lft = MIN_VALID_LIFETIME;
2275                                         if (valid_lft < prefered_lft)
2276                                                 prefered_lft = valid_lft;
2277                                         update_lft = 1;
2278                                 }
2279                         }
2280
2281                         if (update_lft) {
2282                                 ifp->valid_lft = valid_lft;
2283                                 ifp->prefered_lft = prefered_lft;
2284                                 ifp->tstamp = now;
2285                                 flags = ifp->flags;
2286                                 ifp->flags &= ~IFA_F_DEPRECATED;
2287                                 spin_unlock(&ifp->lock);
2288
2289                                 if (!(flags&IFA_F_TENTATIVE))
2290                                         ipv6_ifa_notify(0, ifp);
2291                         } else
2292                                 spin_unlock(&ifp->lock);
2293
2294 #ifdef CONFIG_IPV6_PRIVACY
2295                         read_lock_bh(&in6_dev->lock);
2296                         /* update all temporary addresses in the list */
2297                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2298                                             tmp_list) {
2299                                 int age, max_valid, max_prefered;
2300
2301                                 if (ifp != ift->ifpub)
2302                                         continue;
2303
2304                                 /*
2305                                  * RFC 4941 section 3.3:
2306                                  * If a received option will extend the lifetime
2307                                  * of a public address, the lifetimes of
2308                                  * temporary addresses should be extended,
2309                                  * subject to the overall constraint that no
2310                                  * temporary addresses should ever remain
2311                                  * "valid" or "preferred" for a time longer than
2312                                  * (TEMP_VALID_LIFETIME) or
2313                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2314                                  * respectively.
2315                                  */
2316                                 age = (now - ift->cstamp) / HZ;
2317                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2318                                 if (max_valid < 0)
2319                                         max_valid = 0;
2320
2321                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2322                                                in6_dev->cnf.max_desync_factor -
2323                                                age;
2324                                 if (max_prefered < 0)
2325                                         max_prefered = 0;
2326
2327                                 if (valid_lft > max_valid)
2328                                         valid_lft = max_valid;
2329
2330                                 if (prefered_lft > max_prefered)
2331                                         prefered_lft = max_prefered;
2332
2333                                 spin_lock(&ift->lock);
2334                                 flags = ift->flags;
2335                                 ift->valid_lft = valid_lft;
2336                                 ift->prefered_lft = prefered_lft;
2337                                 ift->tstamp = now;
2338                                 if (prefered_lft > 0)
2339                                         ift->flags &= ~IFA_F_DEPRECATED;
2340
2341                                 spin_unlock(&ift->lock);
2342                                 if (!(flags&IFA_F_TENTATIVE))
2343                                         ipv6_ifa_notify(0, ift);
2344                         }
2345
2346                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2347                                 /*
2348                                  * When a new public address is created as
2349                                  * described in [ADDRCONF], also create a new
2350                                  * temporary address. Also create a temporary
2351                                  * address if it's enabled but no temporary
2352                                  * address currently exists.
2353                                  */
2354                                 read_unlock_bh(&in6_dev->lock);
2355                                 ipv6_create_tempaddr(ifp, NULL);
2356                         } else {
2357                                 read_unlock_bh(&in6_dev->lock);
2358                         }
2359 #endif
2360                         in6_ifa_put(ifp);
2361                         addrconf_verify(0);
2362                 }
2363         }
2364         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2365         in6_dev_put(in6_dev);
2366 }
2367
2368 /*
2369  *      Set destination address.
2370  *      Special case for SIT interfaces where we create a new "virtual"
2371  *      device.
2372  */
2373 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2374 {
2375         struct in6_ifreq ireq;
2376         struct net_device *dev;
2377         int err = -EINVAL;
2378
2379         rtnl_lock();
2380
2381         err = -EFAULT;
2382         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2383                 goto err_exit;
2384
2385         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2386
2387         err = -ENODEV;
2388         if (dev == NULL)
2389                 goto err_exit;
2390
2391 #if IS_ENABLED(CONFIG_IPV6_SIT)
2392         if (dev->type == ARPHRD_SIT) {
2393                 const struct net_device_ops *ops = dev->netdev_ops;
2394                 struct ifreq ifr;
2395                 struct ip_tunnel_parm p;
2396
2397                 err = -EADDRNOTAVAIL;
2398                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2399                         goto err_exit;
2400
2401                 memset(&p, 0, sizeof(p));
2402                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2403                 p.iph.saddr = 0;
2404                 p.iph.version = 4;
2405                 p.iph.ihl = 5;
2406                 p.iph.protocol = IPPROTO_IPV6;
2407                 p.iph.ttl = 64;
2408                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2409
2410                 if (ops->ndo_do_ioctl) {
2411                         mm_segment_t oldfs = get_fs();
2412
2413                         set_fs(KERNEL_DS);
2414                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2415                         set_fs(oldfs);
2416                 } else
2417                         err = -EOPNOTSUPP;
2418
2419                 if (err == 0) {
2420                         err = -ENOBUFS;
2421                         dev = __dev_get_by_name(net, p.name);
2422                         if (!dev)
2423                                 goto err_exit;
2424                         err = dev_open(dev);
2425                 }
2426         }
2427 #endif
2428
2429 err_exit:
2430         rtnl_unlock();
2431         return err;
2432 }
2433
2434 /*
2435  *      Manual configuration of address on an interface
2436  */
2437 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2438                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2439                           __u32 valid_lft)
2440 {
2441         struct inet6_ifaddr *ifp;
2442         struct inet6_dev *idev;
2443         struct net_device *dev;
2444         int scope;
2445         u32 flags;
2446         clock_t expires;
2447         unsigned long timeout;
2448
2449         ASSERT_RTNL();
2450
2451         if (plen > 128)
2452                 return -EINVAL;
2453
2454         /* check the lifetime */
2455         if (!valid_lft || prefered_lft > valid_lft)
2456                 return -EINVAL;
2457
2458         dev = __dev_get_by_index(net, ifindex);
2459         if (!dev)
2460                 return -ENODEV;
2461
2462         idev = addrconf_add_dev(dev);
2463         if (IS_ERR(idev))
2464                 return PTR_ERR(idev);
2465
2466         scope = ipv6_addr_scope(pfx);
2467
2468         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2469         if (addrconf_finite_timeout(timeout)) {
2470                 expires = jiffies_to_clock_t(timeout * HZ);
2471                 valid_lft = timeout;
2472                 flags = RTF_EXPIRES;
2473         } else {
2474                 expires = 0;
2475                 flags = 0;
2476                 ifa_flags |= IFA_F_PERMANENT;
2477         }
2478
2479         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2480         if (addrconf_finite_timeout(timeout)) {
2481                 if (timeout == 0)
2482                         ifa_flags |= IFA_F_DEPRECATED;
2483                 prefered_lft = timeout;
2484         }
2485
2486         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2487
2488         if (!IS_ERR(ifp)) {
2489                 spin_lock_bh(&ifp->lock);
2490                 ifp->valid_lft = valid_lft;
2491                 ifp->prefered_lft = prefered_lft;
2492                 ifp->tstamp = jiffies;
2493                 spin_unlock_bh(&ifp->lock);
2494
2495                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2496                                       expires, flags);
2497                 /*
2498                  * Note that section 3.1 of RFC 4429 indicates
2499                  * that the Optimistic flag should not be set for
2500                  * manually configured addresses
2501                  */
2502                 addrconf_dad_start(ifp);
2503                 in6_ifa_put(ifp);
2504                 addrconf_verify(0);
2505                 return 0;
2506         }
2507
2508         return PTR_ERR(ifp);
2509 }
2510
2511 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2512                           unsigned int plen)
2513 {
2514         struct inet6_ifaddr *ifp;
2515         struct inet6_dev *idev;
2516         struct net_device *dev;
2517
2518         if (plen > 128)
2519                 return -EINVAL;
2520
2521         dev = __dev_get_by_index(net, ifindex);
2522         if (!dev)
2523                 return -ENODEV;
2524
2525         if ((idev = __in6_dev_get(dev)) == NULL)
2526                 return -ENXIO;
2527
2528         read_lock_bh(&idev->lock);
2529         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2530                 if (ifp->prefix_len == plen &&
2531                     ipv6_addr_equal(pfx, &ifp->addr)) {
2532                         in6_ifa_hold(ifp);
2533                         read_unlock_bh(&idev->lock);
2534
2535                         ipv6_del_addr(ifp);
2536
2537                         /* If the last address is deleted administratively,
2538                            disable IPv6 on this interface.
2539                          */
2540                         if (list_empty(&idev->addr_list))
2541                                 addrconf_ifdown(idev->dev, 1);
2542                         return 0;
2543                 }
2544         }
2545         read_unlock_bh(&idev->lock);
2546         return -EADDRNOTAVAIL;
2547 }
2548
2549
2550 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2551 {
2552         struct in6_ifreq ireq;
2553         int err;
2554
2555         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2556                 return -EPERM;
2557
2558         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2559                 return -EFAULT;
2560
2561         rtnl_lock();
2562         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2563                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2564                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2565         rtnl_unlock();
2566         return err;
2567 }
2568
2569 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2570 {
2571         struct in6_ifreq ireq;
2572         int err;
2573
2574         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2575                 return -EPERM;
2576
2577         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2578                 return -EFAULT;
2579
2580         rtnl_lock();
2581         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2582                              ireq.ifr6_prefixlen);
2583         rtnl_unlock();
2584         return err;
2585 }
2586
2587 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2588                      int plen, int scope)
2589 {
2590         struct inet6_ifaddr *ifp;
2591
2592         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2593         if (!IS_ERR(ifp)) {
2594                 spin_lock_bh(&ifp->lock);
2595                 ifp->flags &= ~IFA_F_TENTATIVE;
2596                 spin_unlock_bh(&ifp->lock);
2597                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2598                 in6_ifa_put(ifp);
2599         }
2600 }
2601
2602 #if IS_ENABLED(CONFIG_IPV6_SIT)
2603 static void sit_add_v4_addrs(struct inet6_dev *idev)
2604 {
2605         struct in6_addr addr;
2606         struct net_device *dev;
2607         struct net *net = dev_net(idev->dev);
2608         int scope;
2609
2610         ASSERT_RTNL();
2611
2612         memset(&addr, 0, sizeof(struct in6_addr));
2613         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2614
2615         if (idev->dev->flags&IFF_POINTOPOINT) {
2616                 addr.s6_addr32[0] = htonl(0xfe800000);
2617                 scope = IFA_LINK;
2618         } else {
2619                 scope = IPV6_ADDR_COMPATv4;
2620         }
2621
2622         if (addr.s6_addr32[3]) {
2623                 add_addr(idev, &addr, 128, scope);
2624                 return;
2625         }
2626
2627         for_each_netdev(net, dev) {
2628                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2629                 if (in_dev && (dev->flags & IFF_UP)) {
2630                         struct in_ifaddr *ifa;
2631
2632                         int flag = scope;
2633
2634                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2635                                 int plen;
2636
2637                                 addr.s6_addr32[3] = ifa->ifa_local;
2638
2639                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2640                                         continue;
2641                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2642                                         if (idev->dev->flags&IFF_POINTOPOINT)
2643                                                 continue;
2644                                         flag |= IFA_HOST;
2645                                 }
2646                                 if (idev->dev->flags&IFF_POINTOPOINT)
2647                                         plen = 64;
2648                                 else
2649                                         plen = 96;
2650
2651                                 add_addr(idev, &addr, plen, flag);
2652                         }
2653                 }
2654         }
2655 }
2656 #endif
2657
2658 static void init_loopback(struct net_device *dev)
2659 {
2660         struct inet6_dev  *idev;
2661         struct net_device *sp_dev;
2662         struct inet6_ifaddr *sp_ifa;
2663         struct rt6_info *sp_rt;
2664
2665         /* ::1 */
2666
2667         ASSERT_RTNL();
2668
2669         if ((idev = ipv6_find_idev(dev)) == NULL) {
2670                 pr_debug("%s: add_dev failed\n", __func__);
2671                 return;
2672         }
2673
2674         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2675
2676         /* Add routes to other interface's IPv6 addresses */
2677         for_each_netdev(dev_net(dev), sp_dev) {
2678                 if (!strcmp(sp_dev->name, dev->name))
2679                         continue;
2680
2681                 idev = __in6_dev_get(sp_dev);
2682                 if (!idev)
2683                         continue;
2684
2685                 read_lock_bh(&idev->lock);
2686                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2687
2688                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2689                                 continue;
2690
2691                         if (sp_ifa->rt)
2692                                 continue;
2693
2694                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2695
2696                         /* Failure cases are ignored */
2697                         if (!IS_ERR(sp_rt)) {
2698                                 sp_ifa->rt = sp_rt;
2699                                 ip6_ins_rt(sp_rt);
2700                         }
2701                 }
2702                 read_unlock_bh(&idev->lock);
2703         }
2704 }
2705
2706 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2707 {
2708         struct inet6_ifaddr *ifp;
2709         u32 addr_flags = IFA_F_PERMANENT;
2710
2711 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2712         if (idev->cnf.optimistic_dad &&
2713             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2714                 addr_flags |= IFA_F_OPTIMISTIC;
2715 #endif
2716
2717
2718         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2719         if (!IS_ERR(ifp)) {
2720                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2721                 addrconf_dad_start(ifp);
2722                 in6_ifa_put(ifp);
2723         }
2724 }
2725
2726 static void addrconf_dev_config(struct net_device *dev)
2727 {
2728         struct in6_addr addr;
2729         struct inet6_dev *idev;
2730
2731         ASSERT_RTNL();
2732
2733         if ((dev->type != ARPHRD_ETHER) &&
2734             (dev->type != ARPHRD_FDDI) &&
2735             (dev->type != ARPHRD_ARCNET) &&
2736             (dev->type != ARPHRD_INFINIBAND) &&
2737             (dev->type != ARPHRD_IEEE802154) &&
2738             (dev->type != ARPHRD_IEEE1394)) {
2739                 /* Alas, we support only Ethernet autoconfiguration. */
2740                 return;
2741         }
2742
2743         idev = addrconf_add_dev(dev);
2744         if (IS_ERR(idev))
2745                 return;
2746
2747         memset(&addr, 0, sizeof(struct in6_addr));
2748         addr.s6_addr32[0] = htonl(0xFE800000);
2749
2750         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2751                 addrconf_add_linklocal(idev, &addr);
2752 }
2753
2754 #if IS_ENABLED(CONFIG_IPV6_SIT)
2755 static void addrconf_sit_config(struct net_device *dev)
2756 {
2757         struct inet6_dev *idev;
2758
2759         ASSERT_RTNL();
2760
2761         /*
2762          * Configure the tunnel with one of our IPv4
2763          * addresses... we should configure all of
2764          * our v4 addrs in the tunnel
2765          */
2766
2767         if ((idev = ipv6_find_idev(dev)) == NULL) {
2768                 pr_debug("%s: add_dev failed\n", __func__);
2769                 return;
2770         }
2771
2772         if (dev->priv_flags & IFF_ISATAP) {
2773                 struct in6_addr addr;
2774
2775                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2776                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2777                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2778                         addrconf_add_linklocal(idev, &addr);
2779                 return;
2780         }
2781
2782         sit_add_v4_addrs(idev);
2783
2784         if (dev->flags&IFF_POINTOPOINT)
2785                 addrconf_add_mroute(dev);
2786         else
2787                 sit_route_add(dev);
2788 }
2789 #endif
2790
2791 #if IS_ENABLED(CONFIG_NET_IPGRE)
2792 static void addrconf_gre_config(struct net_device *dev)
2793 {
2794         struct inet6_dev *idev;
2795         struct in6_addr addr;
2796
2797         pr_info("%s(%s)\n", __func__, dev->name);
2798
2799         ASSERT_RTNL();
2800
2801         if ((idev = ipv6_find_idev(dev)) == NULL) {
2802                 pr_debug("%s: add_dev failed\n", __func__);
2803                 return;
2804         }
2805
2806         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2807         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2808
2809         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2810                 addrconf_add_linklocal(idev, &addr);
2811 }
2812 #endif
2813
2814 static inline int
2815 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2816 {
2817         struct in6_addr lladdr;
2818
2819         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2820                 addrconf_add_linklocal(idev, &lladdr);
2821                 return 0;
2822         }
2823         return -1;
2824 }
2825
2826 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2827 {
2828         struct net_device *link_dev;
2829         struct net *net = dev_net(idev->dev);
2830
2831         /* first try to inherit the link-local address from the link device */
2832         if (idev->dev->iflink &&
2833             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2834                 if (!ipv6_inherit_linklocal(idev, link_dev))
2835                         return;
2836         }
2837         /* then try to inherit it from any device */
2838         for_each_netdev(net, link_dev) {
2839                 if (!ipv6_inherit_linklocal(idev, link_dev))
2840                         return;
2841         }
2842         pr_debug("init ip6-ip6: add_linklocal failed\n");
2843 }
2844
2845 /*
2846  * Autoconfigure tunnel with a link-local address so routing protocols,
2847  * DHCPv6, MLD etc. can be run over the virtual link
2848  */
2849
2850 static void addrconf_ip6_tnl_config(struct net_device *dev)
2851 {
2852         struct inet6_dev *idev;
2853
2854         ASSERT_RTNL();
2855
2856         idev = addrconf_add_dev(dev);
2857         if (IS_ERR(idev)) {
2858                 pr_debug("init ip6-ip6: add_dev failed\n");
2859                 return;
2860         }
2861         ip6_tnl_add_linklocal(idev);
2862 }
2863
2864 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2865                            void *data)
2866 {
2867         struct net_device *dev = (struct net_device *) data;
2868         struct inet6_dev *idev = __in6_dev_get(dev);
2869         int run_pending = 0;
2870         int err;
2871
2872         switch (event) {
2873         case NETDEV_REGISTER:
2874                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2875                         idev = ipv6_add_dev(dev);
2876                         if (!idev)
2877                                 return notifier_from_errno(-ENOMEM);
2878                 }
2879                 break;
2880
2881         case NETDEV_UP:
2882         case NETDEV_CHANGE:
2883                 if (dev->flags & IFF_SLAVE)
2884                         break;
2885
2886                 if (event == NETDEV_UP) {
2887                         if (!addrconf_qdisc_ok(dev)) {
2888                                 /* device is not ready yet. */
2889                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2890                                         dev->name);
2891                                 break;
2892                         }
2893
2894                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2895                                 idev = ipv6_add_dev(dev);
2896
2897                         if (idev) {
2898                                 idev->if_flags |= IF_READY;
2899                                 run_pending = 1;
2900                         }
2901                 } else {
2902                         if (!addrconf_qdisc_ok(dev)) {
2903                                 /* device is still not ready. */
2904                                 break;
2905                         }
2906
2907                         if (idev) {
2908                                 if (idev->if_flags & IF_READY)
2909                                         /* device is already configured. */
2910                                         break;
2911                                 idev->if_flags |= IF_READY;
2912                         }
2913
2914                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2915                                 dev->name);
2916
2917                         run_pending = 1;
2918                 }
2919
2920                 switch (dev->type) {
2921 #if IS_ENABLED(CONFIG_IPV6_SIT)
2922                 case ARPHRD_SIT:
2923                         addrconf_sit_config(dev);
2924                         break;
2925 #endif
2926 #if IS_ENABLED(CONFIG_NET_IPGRE)
2927                 case ARPHRD_IPGRE:
2928                         addrconf_gre_config(dev);
2929                         break;
2930 #endif
2931                 case ARPHRD_TUNNEL6:
2932                         addrconf_ip6_tnl_config(dev);
2933                         break;
2934                 case ARPHRD_LOOPBACK:
2935                         init_loopback(dev);
2936                         break;
2937
2938                 default:
2939                         addrconf_dev_config(dev);
2940                         break;
2941                 }
2942
2943                 if (idev) {
2944                         if (run_pending)
2945                                 addrconf_dad_run(idev);
2946
2947                         /*
2948                          * If the MTU changed during the interface down,
2949                          * when the interface up, the changed MTU must be
2950                          * reflected in the idev as well as routers.
2951                          */
2952                         if (idev->cnf.mtu6 != dev->mtu &&
2953                             dev->mtu >= IPV6_MIN_MTU) {
2954                                 rt6_mtu_change(dev, dev->mtu);
2955                                 idev->cnf.mtu6 = dev->mtu;
2956                         }
2957                         idev->tstamp = jiffies;
2958                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2959
2960                         /*
2961                          * If the changed mtu during down is lower than
2962                          * IPV6_MIN_MTU stop IPv6 on this interface.
2963                          */
2964                         if (dev->mtu < IPV6_MIN_MTU)
2965                                 addrconf_ifdown(dev, 1);
2966                 }
2967                 break;
2968
2969         case NETDEV_CHANGEMTU:
2970                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2971                         rt6_mtu_change(dev, dev->mtu);
2972                         idev->cnf.mtu6 = dev->mtu;
2973                         break;
2974                 }
2975
2976                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2977                         idev = ipv6_add_dev(dev);
2978                         if (idev)
2979                                 break;
2980                 }
2981
2982                 /*
2983                  * MTU falled under IPV6_MIN_MTU.
2984                  * Stop IPv6 on this interface.
2985                  */
2986
2987         case NETDEV_DOWN:
2988         case NETDEV_UNREGISTER:
2989                 /*
2990                  *      Remove all addresses from this interface.
2991                  */
2992                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2993                 break;
2994
2995         case NETDEV_CHANGENAME:
2996                 if (idev) {
2997                         snmp6_unregister_dev(idev);
2998                         addrconf_sysctl_unregister(idev);
2999                         addrconf_sysctl_register(idev);
3000                         err = snmp6_register_dev(idev);
3001                         if (err)
3002                                 return notifier_from_errno(err);
3003                 }
3004                 break;
3005
3006         case NETDEV_PRE_TYPE_CHANGE:
3007         case NETDEV_POST_TYPE_CHANGE:
3008                 addrconf_type_change(dev, event);
3009                 break;
3010         }
3011
3012         return NOTIFY_OK;
3013 }
3014
3015 /*
3016  *      addrconf module should be notified of a device going up
3017  */
3018 static struct notifier_block ipv6_dev_notf = {
3019         .notifier_call = addrconf_notify,
3020 };
3021
3022 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3023 {
3024         struct inet6_dev *idev;
3025         ASSERT_RTNL();
3026
3027         idev = __in6_dev_get(dev);
3028
3029         if (event == NETDEV_POST_TYPE_CHANGE)
3030                 ipv6_mc_remap(idev);
3031         else if (event == NETDEV_PRE_TYPE_CHANGE)
3032                 ipv6_mc_unmap(idev);
3033 }
3034
3035 static int addrconf_ifdown(struct net_device *dev, int how)
3036 {
3037         struct net *net = dev_net(dev);
3038         struct inet6_dev *idev;
3039         struct inet6_ifaddr *ifa;
3040         int state, i;
3041
3042         ASSERT_RTNL();
3043
3044         rt6_ifdown(net, dev);
3045         neigh_ifdown(&nd_tbl, dev);
3046
3047         idev = __in6_dev_get(dev);
3048         if (idev == NULL)
3049                 return -ENODEV;
3050
3051         /*
3052          * Step 1: remove reference to ipv6 device from parent device.
3053          *         Do not dev_put!
3054          */
3055         if (how) {
3056                 idev->dead = 1;
3057
3058                 /* protected by rtnl_lock */
3059                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3060
3061                 /* Step 1.5: remove snmp6 entry */
3062                 snmp6_unregister_dev(idev);
3063
3064         }
3065
3066         /* Step 2: clear hash table */
3067         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3068                 struct hlist_head *h = &inet6_addr_lst[i];
3069
3070                 spin_lock_bh(&addrconf_hash_lock);
3071         restart:
3072                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3073                         if (ifa->idev == idev) {
3074                                 hlist_del_init_rcu(&ifa->addr_lst);
3075                                 addrconf_del_timer(ifa);
3076                                 goto restart;
3077                         }
3078                 }
3079                 spin_unlock_bh(&addrconf_hash_lock);
3080         }
3081
3082         write_lock_bh(&idev->lock);
3083
3084         /* Step 2: clear flags for stateless addrconf */
3085         if (!how)
3086                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3087
3088 #ifdef CONFIG_IPV6_PRIVACY
3089         if (how && del_timer(&idev->regen_timer))
3090                 in6_dev_put(idev);
3091
3092         /* Step 3: clear tempaddr list */
3093         while (!list_empty(&idev->tempaddr_list)) {
3094                 ifa = list_first_entry(&idev->tempaddr_list,
3095                                        struct inet6_ifaddr, tmp_list);
3096                 list_del(&ifa->tmp_list);
3097                 write_unlock_bh(&idev->lock);
3098                 spin_lock_bh(&ifa->lock);
3099
3100                 if (ifa->ifpub) {
3101                         in6_ifa_put(ifa->ifpub);
3102                         ifa->ifpub = NULL;
3103                 }
3104                 spin_unlock_bh(&ifa->lock);
3105                 in6_ifa_put(ifa);
3106                 write_lock_bh(&idev->lock);
3107         }
3108 #endif
3109
3110         while (!list_empty(&idev->addr_list)) {
3111                 ifa = list_first_entry(&idev->addr_list,
3112                                        struct inet6_ifaddr, if_list);
3113                 addrconf_del_timer(ifa);
3114
3115                 list_del(&ifa->if_list);
3116
3117                 write_unlock_bh(&idev->lock);
3118
3119                 spin_lock_bh(&ifa->state_lock);
3120                 state = ifa->state;
3121                 ifa->state = INET6_IFADDR_STATE_DEAD;
3122                 spin_unlock_bh(&ifa->state_lock);
3123
3124                 if (state != INET6_IFADDR_STATE_DEAD) {
3125                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3126                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3127                 }
3128                 in6_ifa_put(ifa);
3129
3130                 write_lock_bh(&idev->lock);
3131         }
3132
3133         write_unlock_bh(&idev->lock);
3134
3135         /* Step 5: Discard multicast list */
3136         if (how)
3137                 ipv6_mc_destroy_dev(idev);
3138         else
3139                 ipv6_mc_down(idev);
3140
3141         idev->tstamp = jiffies;
3142
3143         /* Last: Shot the device (if unregistered) */
3144         if (how) {
3145                 addrconf_sysctl_unregister(idev);
3146                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3147                 neigh_ifdown(&nd_tbl, dev);
3148                 in6_dev_put(idev);
3149         }
3150         return 0;
3151 }
3152
3153 static void addrconf_rs_timer(unsigned long data)
3154 {
3155         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3156         struct inet6_dev *idev = ifp->idev;
3157
3158         read_lock(&idev->lock);
3159         if (idev->dead || !(idev->if_flags & IF_READY))
3160                 goto out;
3161
3162         if (!ipv6_accept_ra(idev))
3163                 goto out;
3164
3165         /* Announcement received after solicitation was sent */
3166         if (idev->if_flags & IF_RA_RCVD)
3167                 goto out;
3168
3169         spin_lock(&ifp->lock);
3170         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3171                 /* The wait after the last probe can be shorter */
3172                 addrconf_mod_timer(ifp, AC_RS,
3173                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3174                                    idev->cnf.rtr_solicit_delay :
3175                                    idev->cnf.rtr_solicit_interval);
3176                 spin_unlock(&ifp->lock);
3177
3178                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3179         } else {
3180                 spin_unlock(&ifp->lock);
3181                 /*
3182                  * Note: we do not support deprecated "all on-link"
3183                  * assumption any longer.
3184                  */
3185                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3186         }
3187
3188 out:
3189         read_unlock(&idev->lock);
3190         in6_ifa_put(ifp);
3191 }
3192
3193 /*
3194  *      Duplicate Address Detection
3195  */
3196 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3197 {
3198         unsigned long rand_num;
3199         struct inet6_dev *idev = ifp->idev;
3200
3201         if (ifp->flags & IFA_F_OPTIMISTIC)
3202                 rand_num = 0;
3203         else
3204                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3205
3206         ifp->probes = idev->cnf.dad_transmits;
3207         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3208 }
3209
3210 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3211 {
3212         struct inet6_dev *idev = ifp->idev;
3213         struct net_device *dev = idev->dev;
3214
3215         addrconf_join_solict(dev, &ifp->addr);
3216
3217         net_srandom(ifp->addr.s6_addr32[3]);
3218
3219         read_lock_bh(&idev->lock);
3220         spin_lock(&ifp->lock);
3221         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3222                 goto out;
3223
3224         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3225             idev->cnf.accept_dad < 1 ||
3226             !(ifp->flags&IFA_F_TENTATIVE) ||
3227             ifp->flags & IFA_F_NODAD) {
3228                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3229                 spin_unlock(&ifp->lock);
3230                 read_unlock_bh(&idev->lock);
3231
3232                 addrconf_dad_completed(ifp);
3233                 return;
3234         }
3235
3236         if (!(idev->if_flags & IF_READY)) {
3237                 spin_unlock(&ifp->lock);
3238                 read_unlock_bh(&idev->lock);
3239                 /*
3240                  * If the device is not ready:
3241                  * - keep it tentative if it is a permanent address.
3242                  * - otherwise, kill it.
3243                  */
3244                 in6_ifa_hold(ifp);
3245                 addrconf_dad_stop(ifp, 0);
3246                 return;
3247         }
3248
3249         /*
3250          * Optimistic nodes can start receiving
3251          * Frames right away
3252          */
3253         if (ifp->flags & IFA_F_OPTIMISTIC)
3254                 ip6_ins_rt(ifp->rt);
3255
3256         addrconf_dad_kick(ifp);
3257 out:
3258         spin_unlock(&ifp->lock);
3259         read_unlock_bh(&idev->lock);
3260 }
3261
3262 static void addrconf_dad_timer(unsigned long data)
3263 {
3264         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3265         struct inet6_dev *idev = ifp->idev;
3266         struct in6_addr mcaddr;
3267
3268         if (!ifp->probes && addrconf_dad_end(ifp))
3269                 goto out;
3270
3271         read_lock(&idev->lock);
3272         if (idev->dead || !(idev->if_flags & IF_READY)) {
3273                 read_unlock(&idev->lock);
3274                 goto out;
3275         }
3276
3277         spin_lock(&ifp->lock);
3278         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3279                 spin_unlock(&ifp->lock);
3280                 read_unlock(&idev->lock);
3281                 goto out;
3282         }
3283
3284         if (ifp->probes == 0) {
3285                 /*
3286                  * DAD was successful
3287                  */
3288
3289                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3290                 spin_unlock(&ifp->lock);
3291                 read_unlock(&idev->lock);
3292
3293                 addrconf_dad_completed(ifp);
3294
3295                 goto out;
3296         }
3297
3298         ifp->probes--;
3299         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3300         spin_unlock(&ifp->lock);
3301         read_unlock(&idev->lock);
3302
3303         /* send a neighbour solicitation for our addr */
3304         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3305         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3306 out:
3307         in6_ifa_put(ifp);
3308 }
3309
3310 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3311 {
3312         struct net_device *dev = ifp->idev->dev;
3313
3314         /*
3315          *      Configure the address for reception. Now it is valid.
3316          */
3317
3318         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3319
3320         /* If added prefix is link local and we are prepared to process
3321            router advertisements, start sending router solicitations.
3322          */
3323
3324         if (ipv6_accept_ra(ifp->idev) &&
3325             ifp->idev->cnf.rtr_solicits > 0 &&
3326             (dev->flags&IFF_LOOPBACK) == 0 &&
3327             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3328                 /*
3329                  *      If a host as already performed a random delay
3330                  *      [...] as part of DAD [...] there is no need
3331                  *      to delay again before sending the first RS
3332                  */
3333                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3334
3335                 spin_lock_bh(&ifp->lock);
3336                 ifp->probes = 1;
3337                 ifp->idev->if_flags |= IF_RS_SENT;
3338                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3339                 spin_unlock_bh(&ifp->lock);
3340         }
3341 }
3342
3343 static void addrconf_dad_run(struct inet6_dev *idev)
3344 {
3345         struct inet6_ifaddr *ifp;
3346
3347         read_lock_bh(&idev->lock);
3348         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3349                 spin_lock(&ifp->lock);
3350                 if (ifp->flags & IFA_F_TENTATIVE &&
3351                     ifp->state == INET6_IFADDR_STATE_DAD)
3352                         addrconf_dad_kick(ifp);
3353                 spin_unlock(&ifp->lock);
3354         }
3355         read_unlock_bh(&idev->lock);
3356 }
3357
3358 #ifdef CONFIG_PROC_FS
3359 struct if6_iter_state {
3360         struct seq_net_private p;
3361         int bucket;
3362         int offset;
3363 };
3364
3365 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3366 {
3367         struct inet6_ifaddr *ifa = NULL;
3368         struct if6_iter_state *state = seq->private;
3369         struct net *net = seq_file_net(seq);
3370         int p = 0;
3371
3372         /* initial bucket if pos is 0 */
3373         if (pos == 0) {
3374                 state->bucket = 0;
3375                 state->offset = 0;
3376         }
3377
3378         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3379                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3380                                          addr_lst) {
3381                         if (!net_eq(dev_net(ifa->idev->dev), net))
3382                                 continue;
3383                         /* sync with offset */
3384                         if (p < state->offset) {
3385                                 p++;
3386                                 continue;
3387                         }
3388                         state->offset++;
3389                         return ifa;
3390                 }
3391
3392                 /* prepare for next bucket */
3393                 state->offset = 0;
3394                 p = 0;
3395         }
3396         return NULL;
3397 }
3398
3399 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3400                                          struct inet6_ifaddr *ifa)
3401 {
3402         struct if6_iter_state *state = seq->private;
3403         struct net *net = seq_file_net(seq);
3404
3405         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3406                 if (!net_eq(dev_net(ifa->idev->dev), net))
3407                         continue;
3408                 state->offset++;
3409                 return ifa;
3410         }
3411
3412         while (++state->bucket < IN6_ADDR_HSIZE) {
3413                 state->offset = 0;
3414                 hlist_for_each_entry_rcu_bh(ifa,
3415                                      &inet6_addr_lst[state->bucket], addr_lst) {
3416                         if (!net_eq(dev_net(ifa->idev->dev), net))
3417                                 continue;
3418                         state->offset++;
3419                         return ifa;
3420                 }
3421         }
3422
3423         return NULL;
3424 }
3425
3426 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3427         __acquires(rcu_bh)
3428 {
3429         rcu_read_lock_bh();
3430         return if6_get_first(seq, *pos);
3431 }
3432
3433 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3434 {
3435         struct inet6_ifaddr *ifa;
3436
3437         ifa = if6_get_next(seq, v);
3438         ++*pos;
3439         return ifa;
3440 }
3441
3442 static void if6_seq_stop(struct seq_file *seq, void *v)
3443         __releases(rcu_bh)
3444 {
3445         rcu_read_unlock_bh();
3446 }
3447
3448 static int if6_seq_show(struct seq_file *seq, void *v)
3449 {
3450         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3451         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3452                    &ifp->addr,
3453                    ifp->idev->dev->ifindex,
3454                    ifp->prefix_len,
3455                    ifp->scope,
3456                    ifp->flags,
3457                    ifp->idev->dev->name);
3458         return 0;
3459 }
3460
3461 static const struct seq_operations if6_seq_ops = {
3462         .start  = if6_seq_start,
3463         .next   = if6_seq_next,
3464         .show   = if6_seq_show,
3465         .stop   = if6_seq_stop,
3466 };
3467
3468 static int if6_seq_open(struct inode *inode, struct file *file)
3469 {
3470         return seq_open_net(inode, file, &if6_seq_ops,
3471                             sizeof(struct if6_iter_state));
3472 }
3473
3474 static const struct file_operations if6_fops = {
3475         .owner          = THIS_MODULE,
3476         .open           = if6_seq_open,
3477         .read           = seq_read,
3478         .llseek         = seq_lseek,
3479         .release        = seq_release_net,
3480 };
3481
3482 static int __net_init if6_proc_net_init(struct net *net)
3483 {
3484         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3485                 return -ENOMEM;
3486         return 0;
3487 }
3488
3489 static void __net_exit if6_proc_net_exit(struct net *net)
3490 {
3491         remove_proc_entry("if_inet6", net->proc_net);
3492 }
3493
3494 static struct pernet_operations if6_proc_net_ops = {
3495        .init = if6_proc_net_init,
3496        .exit = if6_proc_net_exit,
3497 };
3498
3499 int __init if6_proc_init(void)
3500 {
3501         return register_pernet_subsys(&if6_proc_net_ops);
3502 }
3503
3504 void if6_proc_exit(void)
3505 {
3506         unregister_pernet_subsys(&if6_proc_net_ops);
3507 }
3508 #endif  /* CONFIG_PROC_FS */
3509
3510 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3511 /* Check if address is a home address configured on any interface. */
3512 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3513 {
3514         int ret = 0;
3515         struct inet6_ifaddr *ifp = NULL;
3516         unsigned int hash = inet6_addr_hash(addr);
3517
3518         rcu_read_lock_bh();
3519         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3520                 if (!net_eq(dev_net(ifp->idev->dev), net))
3521                         continue;
3522                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3523                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3524                         ret = 1;
3525                         break;
3526                 }
3527         }
3528         rcu_read_unlock_bh();
3529         return ret;
3530 }
3531 #endif
3532
3533 /*
3534  *      Periodic address status verification
3535  */
3536
3537 static void addrconf_verify(unsigned long foo)
3538 {
3539         unsigned long now, next, next_sec, next_sched;
3540         struct inet6_ifaddr *ifp;
3541         int i;
3542
3543         rcu_read_lock_bh();
3544         spin_lock(&addrconf_verify_lock);
3545         now = jiffies;
3546         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3547
3548         del_timer(&addr_chk_timer);
3549
3550         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3551 restart:
3552                 hlist_for_each_entry_rcu_bh(ifp,
3553                                          &inet6_addr_lst[i], addr_lst) {
3554                         unsigned long age;
3555
3556                         if (ifp->flags & IFA_F_PERMANENT)
3557                                 continue;
3558
3559                         spin_lock(&ifp->lock);
3560                         /* We try to batch several events at once. */
3561                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3562
3563                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3564                             age >= ifp->valid_lft) {
3565                                 spin_unlock(&ifp->lock);
3566                                 in6_ifa_hold(ifp);
3567                                 ipv6_del_addr(ifp);
3568                                 goto restart;
3569                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3570                                 spin_unlock(&ifp->lock);
3571                                 continue;
3572                         } else if (age >= ifp->prefered_lft) {
3573                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3574                                 int deprecate = 0;
3575
3576                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3577                                         deprecate = 1;
3578                                         ifp->flags |= IFA_F_DEPRECATED;
3579                                 }
3580
3581                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3582                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3583
3584                                 spin_unlock(&ifp->lock);
3585
3586                                 if (deprecate) {
3587                                         in6_ifa_hold(ifp);
3588
3589                                         ipv6_ifa_notify(0, ifp);
3590                                         in6_ifa_put(ifp);
3591                                         goto restart;
3592                                 }
3593 #ifdef CONFIG_IPV6_PRIVACY
3594                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3595                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3596                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3597                                         ifp->idev->cnf.dad_transmits *
3598                                         ifp->idev->nd_parms->retrans_time / HZ;
3599
3600                                 if (age >= ifp->prefered_lft - regen_advance) {
3601                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3602                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3603                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3604                                         if (!ifp->regen_count && ifpub) {
3605                                                 ifp->regen_count++;
3606                                                 in6_ifa_hold(ifp);
3607                                                 in6_ifa_hold(ifpub);
3608                                                 spin_unlock(&ifp->lock);
3609
3610                                                 spin_lock(&ifpub->lock);
3611                                                 ifpub->regen_count = 0;
3612                                                 spin_unlock(&ifpub->lock);
3613                                                 ipv6_create_tempaddr(ifpub, ifp);
3614                                                 in6_ifa_put(ifpub);
3615                                                 in6_ifa_put(ifp);
3616                                                 goto restart;
3617                                         }
3618                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3619                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3620                                 spin_unlock(&ifp->lock);
3621 #endif
3622                         } else {
3623                                 /* ifp->prefered_lft <= ifp->valid_lft */
3624                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3625                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3626                                 spin_unlock(&ifp->lock);
3627                         }
3628                 }
3629         }
3630
3631         next_sec = round_jiffies_up(next);
3632         next_sched = next;
3633
3634         /* If rounded timeout is accurate enough, accept it. */
3635         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3636                 next_sched = next_sec;
3637
3638         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3639         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3640                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3641
3642         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3643               now, next, next_sec, next_sched));
3644
3645         addr_chk_timer.expires = next_sched;
3646         add_timer(&addr_chk_timer);
3647         spin_unlock(&addrconf_verify_lock);
3648         rcu_read_unlock_bh();
3649 }
3650
3651 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3652 {
3653         struct in6_addr *pfx = NULL;
3654
3655         if (addr)
3656                 pfx = nla_data(addr);
3657
3658         if (local) {
3659                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3660                         pfx = NULL;
3661                 else
3662                         pfx = nla_data(local);
3663         }
3664
3665         return pfx;
3666 }
3667
3668 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3669         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3670         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3671         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3672 };
3673
3674 static int
3675 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3676 {
3677         struct net *net = sock_net(skb->sk);
3678         struct ifaddrmsg *ifm;
3679         struct nlattr *tb[IFA_MAX+1];
3680         struct in6_addr *pfx;
3681         int err;
3682
3683         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3684         if (err < 0)
3685                 return err;
3686
3687         ifm = nlmsg_data(nlh);
3688         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3689         if (pfx == NULL)
3690                 return -EINVAL;
3691
3692         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3693 }
3694
3695 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3696                              u32 prefered_lft, u32 valid_lft)
3697 {
3698         u32 flags;
3699         clock_t expires;
3700         unsigned long timeout;
3701
3702         if (!valid_lft || (prefered_lft > valid_lft))
3703                 return -EINVAL;
3704
3705         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3706         if (addrconf_finite_timeout(timeout)) {
3707                 expires = jiffies_to_clock_t(timeout * HZ);
3708                 valid_lft = timeout;
3709                 flags = RTF_EXPIRES;
3710         } else {
3711                 expires = 0;
3712                 flags = 0;
3713                 ifa_flags |= IFA_F_PERMANENT;
3714         }
3715
3716         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3717         if (addrconf_finite_timeout(timeout)) {
3718                 if (timeout == 0)
3719                         ifa_flags |= IFA_F_DEPRECATED;
3720                 prefered_lft = timeout;
3721         }
3722
3723         spin_lock_bh(&ifp->lock);
3724         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3725         ifp->tstamp = jiffies;
3726         ifp->valid_lft = valid_lft;
3727         ifp->prefered_lft = prefered_lft;
3728
3729         spin_unlock_bh(&ifp->lock);
3730         if (!(ifp->flags&IFA_F_TENTATIVE))
3731                 ipv6_ifa_notify(0, ifp);
3732
3733         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3734                               expires, flags);
3735         addrconf_verify(0);
3736
3737         return 0;
3738 }
3739
3740 static int
3741 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3742 {
3743         struct net *net = sock_net(skb->sk);
3744         struct ifaddrmsg *ifm;
3745         struct nlattr *tb[IFA_MAX+1];
3746         struct in6_addr *pfx;
3747         struct inet6_ifaddr *ifa;
3748         struct net_device *dev;
3749         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3750         u8 ifa_flags;
3751         int err;
3752
3753         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3754         if (err < 0)
3755                 return err;
3756
3757         ifm = nlmsg_data(nlh);
3758         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3759         if (pfx == NULL)
3760                 return -EINVAL;
3761
3762         if (tb[IFA_CACHEINFO]) {
3763                 struct ifa_cacheinfo *ci;
3764
3765                 ci = nla_data(tb[IFA_CACHEINFO]);
3766                 valid_lft = ci->ifa_valid;
3767                 preferred_lft = ci->ifa_prefered;
3768         } else {
3769                 preferred_lft = INFINITY_LIFE_TIME;
3770                 valid_lft = INFINITY_LIFE_TIME;
3771         }
3772
3773         dev =  __dev_get_by_index(net, ifm->ifa_index);
3774         if (dev == NULL)
3775                 return -ENODEV;
3776
3777         /* We ignore other flags so far. */
3778         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3779
3780         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3781         if (ifa == NULL) {
3782                 /*
3783                  * It would be best to check for !NLM_F_CREATE here but
3784                  * userspace alreay relies on not having to provide this.
3785                  */
3786                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3787                                       ifm->ifa_prefixlen, ifa_flags,
3788                                       preferred_lft, valid_lft);
3789         }
3790
3791         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3792             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3793                 err = -EEXIST;
3794         else
3795                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3796
3797         in6_ifa_put(ifa);
3798
3799         return err;
3800 }
3801
3802 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3803                           u8 scope, int ifindex)
3804 {
3805         struct ifaddrmsg *ifm;
3806
3807         ifm = nlmsg_data(nlh);
3808         ifm->ifa_family = AF_INET6;
3809         ifm->ifa_prefixlen = prefixlen;
3810         ifm->ifa_flags = flags;
3811         ifm->ifa_scope = scope;
3812         ifm->ifa_index = ifindex;
3813 }
3814
3815 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3816                          unsigned long tstamp, u32 preferred, u32 valid)
3817 {
3818         struct ifa_cacheinfo ci;
3819
3820         ci.cstamp = cstamp_delta(cstamp);
3821         ci.tstamp = cstamp_delta(tstamp);
3822         ci.ifa_prefered = preferred;
3823         ci.ifa_valid = valid;
3824
3825         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3826 }
3827
3828 static inline int rt_scope(int ifa_scope)
3829 {
3830         if (ifa_scope & IFA_HOST)
3831                 return RT_SCOPE_HOST;
3832         else if (ifa_scope & IFA_LINK)
3833                 return RT_SCOPE_LINK;
3834         else if (ifa_scope & IFA_SITE)
3835                 return RT_SCOPE_SITE;
3836         else
3837                 return RT_SCOPE_UNIVERSE;
3838 }
3839
3840 static inline int inet6_ifaddr_msgsize(void)
3841 {
3842         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3843                + nla_total_size(16) /* IFA_ADDRESS */
3844                + nla_total_size(sizeof(struct ifa_cacheinfo));
3845 }
3846
3847 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3848                              u32 portid, u32 seq, int event, unsigned int flags)
3849 {
3850         struct nlmsghdr  *nlh;
3851         u32 preferred, valid;
3852
3853         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3854         if (nlh == NULL)
3855                 return -EMSGSIZE;
3856
3857         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3858                       ifa->idev->dev->ifindex);
3859
3860         if (!(ifa->flags&IFA_F_PERMANENT)) {
3861                 preferred = ifa->prefered_lft;
3862                 valid = ifa->valid_lft;
3863                 if (preferred != INFINITY_LIFE_TIME) {
3864                         long tval = (jiffies - ifa->tstamp)/HZ;
3865                         if (preferred > tval)
3866                                 preferred -= tval;
3867                         else
3868                                 preferred = 0;
3869                         if (valid != INFINITY_LIFE_TIME) {
3870                                 if (valid > tval)
3871                                         valid -= tval;
3872                                 else
3873                                         valid = 0;
3874                         }
3875                 }
3876         } else {
3877                 preferred = INFINITY_LIFE_TIME;
3878                 valid = INFINITY_LIFE_TIME;
3879         }
3880
3881         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3882             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3883                 nlmsg_cancel(skb, nlh);
3884                 return -EMSGSIZE;
3885         }
3886
3887         return nlmsg_end(skb, nlh);
3888 }
3889
3890 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3891                                 u32 portid, u32 seq, int event, u16 flags)
3892 {
3893         struct nlmsghdr  *nlh;
3894         u8 scope = RT_SCOPE_UNIVERSE;
3895         int ifindex = ifmca->idev->dev->ifindex;
3896
3897         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3898                 scope = RT_SCOPE_SITE;
3899
3900         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3901         if (nlh == NULL)
3902                 return -EMSGSIZE;
3903
3904         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3905         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3906             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3907                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3908                 nlmsg_cancel(skb, nlh);
3909                 return -EMSGSIZE;
3910         }
3911
3912         return nlmsg_end(skb, nlh);
3913 }
3914
3915 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3916                                 u32 portid, u32 seq, int event, unsigned int flags)
3917 {
3918         struct nlmsghdr  *nlh;
3919         u8 scope = RT_SCOPE_UNIVERSE;
3920         int ifindex = ifaca->aca_idev->dev->ifindex;
3921
3922         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3923                 scope = RT_SCOPE_SITE;
3924
3925         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3926         if (nlh == NULL)
3927                 return -EMSGSIZE;
3928
3929         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3930         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3931             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3932                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3933                 nlmsg_cancel(skb, nlh);
3934                 return -EMSGSIZE;
3935         }
3936
3937         return nlmsg_end(skb, nlh);
3938 }
3939
3940 enum addr_type_t {
3941         UNICAST_ADDR,
3942         MULTICAST_ADDR,
3943         ANYCAST_ADDR,
3944 };
3945
3946 /* called with rcu_read_lock() */
3947 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3948                           struct netlink_callback *cb, enum addr_type_t type,
3949                           int s_ip_idx, int *p_ip_idx)
3950 {
3951         struct ifmcaddr6 *ifmca;
3952         struct ifacaddr6 *ifaca;
3953         int err = 1;
3954         int ip_idx = *p_ip_idx;
3955
3956         read_lock_bh(&idev->lock);
3957         switch (type) {
3958         case UNICAST_ADDR: {
3959                 struct inet6_ifaddr *ifa;
3960
3961                 /* unicast address incl. temp addr */
3962                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3963                         if (++ip_idx < s_ip_idx)
3964                                 continue;
3965                         err = inet6_fill_ifaddr(skb, ifa,
3966                                                 NETLINK_CB(cb->skb).portid,
3967                                                 cb->nlh->nlmsg_seq,
3968                                                 RTM_NEWADDR,
3969                                                 NLM_F_MULTI);
3970                         if (err <= 0)
3971                                 break;
3972                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3973                 }
3974                 break;
3975         }
3976         case MULTICAST_ADDR:
3977                 /* multicast address */
3978                 for (ifmca = idev->mc_list; ifmca;
3979                      ifmca = ifmca->next, ip_idx++) {
3980                         if (ip_idx < s_ip_idx)
3981                                 continue;
3982                         err = inet6_fill_ifmcaddr(skb, ifmca,
3983                                                   NETLINK_CB(cb->skb).portid,
3984                                                   cb->nlh->nlmsg_seq,
3985                                                   RTM_GETMULTICAST,
3986                                                   NLM_F_MULTI);
3987                         if (err <= 0)
3988                                 break;
3989                 }
3990                 break;
3991         case ANYCAST_ADDR:
3992                 /* anycast address */
3993                 for (ifaca = idev->ac_list; ifaca;
3994                      ifaca = ifaca->aca_next, ip_idx++) {
3995                         if (ip_idx < s_ip_idx)
3996                                 continue;
3997                         err = inet6_fill_ifacaddr(skb, ifaca,
3998                                                   NETLINK_CB(cb->skb).portid,
3999                                                   cb->nlh->nlmsg_seq,
4000                                                   RTM_GETANYCAST,
4001                                                   NLM_F_MULTI);
4002                         if (err <= 0)
4003                                 break;
4004                 }
4005                 break;
4006         default:
4007                 break;
4008         }
4009         read_unlock_bh(&idev->lock);
4010         *p_ip_idx = ip_idx;
4011         return err;
4012 }
4013
4014 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4015                            enum addr_type_t type)
4016 {
4017         struct net *net = sock_net(skb->sk);
4018         int h, s_h;
4019         int idx, ip_idx;
4020         int s_idx, s_ip_idx;
4021         struct net_device *dev;
4022         struct inet6_dev *idev;
4023         struct hlist_head *head;
4024
4025         s_h = cb->args[0];
4026         s_idx = idx = cb->args[1];
4027         s_ip_idx = ip_idx = cb->args[2];
4028
4029         rcu_read_lock();
4030         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4031         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4032                 idx = 0;
4033                 head = &net->dev_index_head[h];
4034                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4035                         if (idx < s_idx)
4036                                 goto cont;
4037                         if (h > s_h || idx > s_idx)
4038                                 s_ip_idx = 0;
4039                         ip_idx = 0;
4040                         idev = __in6_dev_get(dev);
4041                         if (!idev)
4042                                 goto cont;
4043
4044                         if (in6_dump_addrs(idev, skb, cb, type,
4045                                            s_ip_idx, &ip_idx) <= 0)
4046                                 goto done;
4047 cont:
4048                         idx++;
4049                 }
4050         }
4051 done:
4052         rcu_read_unlock();
4053         cb->args[0] = h;
4054         cb->args[1] = idx;
4055         cb->args[2] = ip_idx;
4056
4057         return skb->len;
4058 }
4059
4060 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4061 {
4062         enum addr_type_t type = UNICAST_ADDR;
4063
4064         return inet6_dump_addr(skb, cb, type);
4065 }
4066
4067 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4068 {
4069         enum addr_type_t type = MULTICAST_ADDR;
4070
4071         return inet6_dump_addr(skb, cb, type);
4072 }
4073
4074
4075 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4076 {
4077         enum addr_type_t type = ANYCAST_ADDR;
4078
4079         return inet6_dump_addr(skb, cb, type);
4080 }
4081
4082 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4083 {
4084         struct net *net = sock_net(in_skb->sk);
4085         struct ifaddrmsg *ifm;
4086         struct nlattr *tb[IFA_MAX+1];
4087         struct in6_addr *addr = NULL;
4088         struct net_device *dev = NULL;
4089         struct inet6_ifaddr *ifa;
4090         struct sk_buff *skb;
4091         int err;
4092
4093         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4094         if (err < 0)
4095                 goto errout;
4096
4097         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4098         if (addr == NULL) {
4099                 err = -EINVAL;
4100                 goto errout;
4101         }
4102
4103         ifm = nlmsg_data(nlh);
4104         if (ifm->ifa_index)
4105                 dev = __dev_get_by_index(net, ifm->ifa_index);
4106
4107         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4108         if (!ifa) {
4109                 err = -EADDRNOTAVAIL;
4110                 goto errout;
4111         }
4112
4113         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4114         if (!skb) {
4115                 err = -ENOBUFS;
4116                 goto errout_ifa;
4117         }
4118
4119         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4120                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4121         if (err < 0) {
4122                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4123                 WARN_ON(err == -EMSGSIZE);
4124                 kfree_skb(skb);
4125                 goto errout_ifa;
4126         }
4127         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4128 errout_ifa:
4129         in6_ifa_put(ifa);
4130 errout:
4131         return err;
4132 }
4133
4134 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4135 {
4136         struct sk_buff *skb;
4137         struct net *net = dev_net(ifa->idev->dev);
4138         int err = -ENOBUFS;
4139
4140         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4141         if (skb == NULL)
4142                 goto errout;
4143
4144         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4145         if (err < 0) {
4146                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4147                 WARN_ON(err == -EMSGSIZE);
4148                 kfree_skb(skb);
4149                 goto errout;
4150         }
4151         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4152         return;
4153 errout:
4154         if (err < 0)
4155                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4156 }
4157
4158 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4159                                 __s32 *array, int bytes)
4160 {
4161         BUG_ON(bytes < (DEVCONF_MAX * 4));
4162
4163         memset(array, 0, bytes);
4164         array[DEVCONF_FORWARDING] = cnf->forwarding;
4165         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4166         array[DEVCONF_MTU6] = cnf->mtu6;
4167         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4168         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4169         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4170         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4171         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4172         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4173                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4174         array[DEVCONF_RTR_SOLICIT_DELAY] =
4175                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4176         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4177 #ifdef CONFIG_IPV6_PRIVACY
4178         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4179         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4180         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4181         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4182         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4183 #endif
4184         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4185         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4186         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4187 #ifdef CONFIG_IPV6_ROUTER_PREF
4188         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4189         array[DEVCONF_RTR_PROBE_INTERVAL] =
4190                 jiffies_to_msecs(cnf->rtr_probe_interval);
4191 #ifdef CONFIG_IPV6_ROUTE_INFO
4192         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4193 #endif
4194 #endif
4195         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4196         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4197 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4198         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4199 #endif
4200 #ifdef CONFIG_IPV6_MROUTE
4201         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4202 #endif
4203         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4204         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4205         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4206         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4207 }
4208
4209 static inline size_t inet6_ifla6_size(void)
4210 {
4211         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4212              + nla_total_size(sizeof(struct ifla_cacheinfo))
4213              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4214              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4215              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4216              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4217 }
4218
4219 static inline size_t inet6_if_nlmsg_size(void)
4220 {
4221         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4222                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4223                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4224                + nla_total_size(4) /* IFLA_MTU */
4225                + nla_total_size(4) /* IFLA_LINK */
4226                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4227 }
4228
4229 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4230                                       int items, int bytes)
4231 {
4232         int i;
4233         int pad = bytes - sizeof(u64) * items;
4234         BUG_ON(pad < 0);
4235
4236         /* Use put_unaligned() because stats may not be aligned for u64. */
4237         put_unaligned(items, &stats[0]);
4238         for (i = 1; i < items; i++)
4239                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4240
4241         memset(&stats[items], 0, pad);
4242 }
4243
4244 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4245                                       int items, int bytes, size_t syncpoff)
4246 {
4247         int i;
4248         int pad = bytes - sizeof(u64) * items;
4249         BUG_ON(pad < 0);
4250
4251         /* Use put_unaligned() because stats may not be aligned for u64. */
4252         put_unaligned(items, &stats[0]);
4253         for (i = 1; i < items; i++)
4254                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4255
4256         memset(&stats[items], 0, pad);
4257 }
4258
4259 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4260                              int bytes)
4261 {
4262         switch (attrtype) {
4263         case IFLA_INET6_STATS:
4264                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4265                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4266                 break;
4267         case IFLA_INET6_ICMP6STATS:
4268                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4269                 break;
4270         }
4271 }
4272
4273 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4274 {
4275         struct nlattr *nla;
4276         struct ifla_cacheinfo ci;
4277
4278         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4279                 goto nla_put_failure;
4280         ci.max_reasm_len = IPV6_MAXPLEN;
4281         ci.tstamp = cstamp_delta(idev->tstamp);
4282         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4283         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4284         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4285                 goto nla_put_failure;
4286         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4287         if (nla == NULL)
4288                 goto nla_put_failure;
4289         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4290
4291         /* XXX - MC not implemented */
4292
4293         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4294         if (nla == NULL)
4295                 goto nla_put_failure;
4296         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4297
4298         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4299         if (nla == NULL)
4300                 goto nla_put_failure;
4301         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4302
4303         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4304         if (nla == NULL)
4305                 goto nla_put_failure;
4306         read_lock_bh(&idev->lock);
4307         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4308         read_unlock_bh(&idev->lock);
4309
4310         return 0;
4311
4312 nla_put_failure:
4313         return -EMSGSIZE;
4314 }
4315
4316 static size_t inet6_get_link_af_size(const struct net_device *dev)
4317 {
4318         if (!__in6_dev_get(dev))
4319                 return 0;
4320
4321         return inet6_ifla6_size();
4322 }
4323
4324 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4325 {
4326         struct inet6_dev *idev = __in6_dev_get(dev);
4327
4328         if (!idev)
4329                 return -ENODATA;
4330
4331         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4332                 return -EMSGSIZE;
4333
4334         return 0;
4335 }
4336
4337 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4338 {
4339         struct inet6_ifaddr *ifp;
4340         struct net_device *dev = idev->dev;
4341         bool update_rs = false;
4342         struct in6_addr ll_addr;
4343
4344         if (token == NULL)
4345                 return -EINVAL;
4346         if (ipv6_addr_any(token))
4347                 return -EINVAL;
4348         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4349                 return -EINVAL;
4350         if (!ipv6_accept_ra(idev))
4351                 return -EINVAL;
4352         if (idev->cnf.rtr_solicits <= 0)
4353                 return -EINVAL;
4354
4355         write_lock_bh(&idev->lock);
4356
4357         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4358         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4359
4360         write_unlock_bh(&idev->lock);
4361
4362         if (!idev->dead && (idev->if_flags & IF_READY) &&
4363             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4364                              IFA_F_OPTIMISTIC)) {
4365
4366                 /* If we're not ready, then normal ifup will take care
4367                  * of this. Otherwise, we need to request our rs here.
4368                  */
4369                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4370                 update_rs = true;
4371         }
4372
4373         write_lock_bh(&idev->lock);
4374
4375         if (update_rs)
4376                 idev->if_flags |= IF_RS_SENT;
4377
4378         /* Well, that's kinda nasty ... */
4379         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4380                 spin_lock(&ifp->lock);
4381                 if (ifp->tokenized) {
4382                         ifp->valid_lft = 0;
4383                         ifp->prefered_lft = 0;
4384                 }
4385                 spin_unlock(&ifp->lock);
4386         }
4387
4388         write_unlock_bh(&idev->lock);
4389         return 0;
4390 }
4391
4392 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4393 {
4394         int err = -EINVAL;
4395         struct inet6_dev *idev = __in6_dev_get(dev);
4396         struct nlattr *tb[IFLA_INET6_MAX + 1];
4397
4398         if (!idev)
4399                 return -EAFNOSUPPORT;
4400
4401         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4402                 BUG();
4403
4404         if (tb[IFLA_INET6_TOKEN])
4405                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4406
4407         return err;
4408 }
4409
4410 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4411                              u32 portid, u32 seq, int event, unsigned int flags)
4412 {
4413         struct net_device *dev = idev->dev;
4414         struct ifinfomsg *hdr;
4415         struct nlmsghdr *nlh;
4416         void *protoinfo;
4417
4418         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4419         if (nlh == NULL)
4420                 return -EMSGSIZE;
4421
4422         hdr = nlmsg_data(nlh);
4423         hdr->ifi_family = AF_INET6;
4424         hdr->__ifi_pad = 0;
4425         hdr->ifi_type = dev->type;
4426         hdr->ifi_index = dev->ifindex;
4427         hdr->ifi_flags = dev_get_flags(dev);
4428         hdr->ifi_change = 0;
4429
4430         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4431             (dev->addr_len &&
4432              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4433             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4434             (dev->ifindex != dev->iflink &&
4435              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4436                 goto nla_put_failure;
4437         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4438         if (protoinfo == NULL)
4439                 goto nla_put_failure;
4440
4441         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4442                 goto nla_put_failure;
4443
4444         nla_nest_end(skb, protoinfo);
4445         return nlmsg_end(skb, nlh);
4446
4447 nla_put_failure:
4448         nlmsg_cancel(skb, nlh);
4449         return -EMSGSIZE;
4450 }
4451
4452 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4453 {
4454         struct net *net = sock_net(skb->sk);
4455         int h, s_h;
4456         int idx = 0, s_idx;
4457         struct net_device *dev;
4458         struct inet6_dev *idev;
4459         struct hlist_head *head;
4460
4461         s_h = cb->args[0];
4462         s_idx = cb->args[1];
4463
4464         rcu_read_lock();
4465         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4466                 idx = 0;
4467                 head = &net->dev_index_head[h];
4468                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4469                         if (idx < s_idx)
4470                                 goto cont;
4471                         idev = __in6_dev_get(dev);
4472                         if (!idev)
4473                                 goto cont;
4474                         if (inet6_fill_ifinfo(skb, idev,
4475                                               NETLINK_CB(cb->skb).portid,
4476                                               cb->nlh->nlmsg_seq,
4477                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4478                                 goto out;
4479 cont:
4480                         idx++;
4481                 }
4482         }
4483 out:
4484         rcu_read_unlock();
4485         cb->args[1] = idx;
4486         cb->args[0] = h;
4487
4488         return skb->len;
4489 }
4490
4491 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4492 {
4493         struct sk_buff *skb;
4494         struct net *net = dev_net(idev->dev);
4495         int err = -ENOBUFS;
4496
4497         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4498         if (skb == NULL)
4499                 goto errout;
4500
4501         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4502         if (err < 0) {
4503                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4504                 WARN_ON(err == -EMSGSIZE);
4505                 kfree_skb(skb);
4506                 goto errout;
4507         }
4508         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4509         return;
4510 errout:
4511         if (err < 0)
4512                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4513 }
4514
4515 static inline size_t inet6_prefix_nlmsg_size(void)
4516 {
4517         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4518                + nla_total_size(sizeof(struct in6_addr))
4519                + nla_total_size(sizeof(struct prefix_cacheinfo));
4520 }
4521
4522 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4523                              struct prefix_info *pinfo, u32 portid, u32 seq,
4524                              int event, unsigned int flags)
4525 {
4526         struct prefixmsg *pmsg;
4527         struct nlmsghdr *nlh;
4528         struct prefix_cacheinfo ci;
4529
4530         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4531         if (nlh == NULL)
4532                 return -EMSGSIZE;
4533
4534         pmsg = nlmsg_data(nlh);
4535         pmsg->prefix_family = AF_INET6;
4536         pmsg->prefix_pad1 = 0;
4537         pmsg->prefix_pad2 = 0;
4538         pmsg->prefix_ifindex = idev->dev->ifindex;
4539         pmsg->prefix_len = pinfo->prefix_len;
4540         pmsg->prefix_type = pinfo->type;
4541         pmsg->prefix_pad3 = 0;
4542         pmsg->prefix_flags = 0;
4543         if (pinfo->onlink)
4544                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4545         if (pinfo->autoconf)
4546                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4547
4548         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4549                 goto nla_put_failure;
4550         ci.preferred_time = ntohl(pinfo->prefered);
4551         ci.valid_time = ntohl(pinfo->valid);
4552         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4553                 goto nla_put_failure;
4554         return nlmsg_end(skb, nlh);
4555
4556 nla_put_failure:
4557         nlmsg_cancel(skb, nlh);
4558         return -EMSGSIZE;
4559 }
4560
4561 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4562                          struct prefix_info *pinfo)
4563 {
4564         struct sk_buff *skb;
4565         struct net *net = dev_net(idev->dev);
4566         int err = -ENOBUFS;
4567
4568         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4569         if (skb == NULL)
4570                 goto errout;
4571
4572         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4573         if (err < 0) {
4574                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4575                 WARN_ON(err == -EMSGSIZE);
4576                 kfree_skb(skb);
4577                 goto errout;
4578         }
4579         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4580         return;
4581 errout:
4582         if (err < 0)
4583                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4584 }
4585
4586 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4587 {
4588         struct net *net = dev_net(ifp->idev->dev);
4589
4590         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4591
4592         switch (event) {
4593         case RTM_NEWADDR:
4594                 /*
4595                  * If the address was optimistic
4596                  * we inserted the route at the start of
4597                  * our DAD process, so we don't need
4598                  * to do it again
4599                  */
4600                 if (!(ifp->rt->rt6i_node))
4601                         ip6_ins_rt(ifp->rt);
4602                 if (ifp->idev->cnf.forwarding)
4603                         addrconf_join_anycast(ifp);
4604                 break;
4605         case RTM_DELADDR:
4606                 if (ifp->idev->cnf.forwarding)
4607                         addrconf_leave_anycast(ifp);
4608                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4609                 dst_hold(&ifp->rt->dst);
4610
4611                 if (ip6_del_rt(ifp->rt))
4612                         dst_free(&ifp->rt->dst);
4613                 break;
4614         }
4615         atomic_inc(&net->ipv6.dev_addr_genid);
4616 }
4617
4618 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4619 {
4620         rcu_read_lock_bh();
4621         if (likely(ifp->idev->dead == 0))
4622                 __ipv6_ifa_notify(event, ifp);
4623         rcu_read_unlock_bh();
4624 }
4625
4626 #ifdef CONFIG_SYSCTL
4627
4628 static
4629 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4630                            void __user *buffer, size_t *lenp, loff_t *ppos)
4631 {
4632         int *valp = ctl->data;
4633         int val = *valp;
4634         loff_t pos = *ppos;
4635         ctl_table lctl;
4636         int ret;
4637
4638         /*
4639          * ctl->data points to idev->cnf.forwarding, we should
4640          * not modify it until we get the rtnl lock.
4641          */
4642         lctl = *ctl;
4643         lctl.data = &val;
4644
4645         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4646
4647         if (write)
4648                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4649         if (ret)
4650                 *ppos = pos;
4651         return ret;
4652 }
4653
4654 static void dev_disable_change(struct inet6_dev *idev)
4655 {
4656         if (!idev || !idev->dev)
4657                 return;
4658
4659         if (idev->cnf.disable_ipv6)
4660                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4661         else
4662                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4663 }
4664
4665 static void addrconf_disable_change(struct net *net, __s32 newf)
4666 {
4667         struct net_device *dev;
4668         struct inet6_dev *idev;
4669
4670         rcu_read_lock();
4671         for_each_netdev_rcu(net, dev) {
4672                 idev = __in6_dev_get(dev);
4673                 if (idev) {
4674                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4675                         idev->cnf.disable_ipv6 = newf;
4676                         if (changed)
4677                                 dev_disable_change(idev);
4678                 }
4679         }
4680         rcu_read_unlock();
4681 }
4682
4683 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4684 {
4685         struct net *net;
4686         int old;
4687
4688         if (!rtnl_trylock())
4689                 return restart_syscall();
4690
4691         net = (struct net *)table->extra2;
4692         old = *p;
4693         *p = newf;
4694
4695         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4696                 rtnl_unlock();
4697                 return 0;
4698         }
4699
4700         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4701                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4702                 addrconf_disable_change(net, newf);
4703         } else if ((!newf) ^ (!old))
4704                 dev_disable_change((struct inet6_dev *)table->extra1);
4705
4706         rtnl_unlock();
4707         return 0;
4708 }
4709
4710 static
4711 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4712                             void __user *buffer, size_t *lenp, loff_t *ppos)
4713 {
4714         int *valp = ctl->data;
4715         int val = *valp;
4716         loff_t pos = *ppos;
4717         ctl_table lctl;
4718         int ret;
4719
4720         /*
4721          * ctl->data points to idev->cnf.disable_ipv6, we should
4722          * not modify it until we get the rtnl lock.
4723          */
4724         lctl = *ctl;
4725         lctl.data = &val;
4726
4727         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4728
4729         if (write)
4730                 ret = addrconf_disable_ipv6(ctl, valp, val);
4731         if (ret)
4732                 *ppos = pos;
4733         return ret;
4734 }
4735
4736 static struct addrconf_sysctl_table
4737 {
4738         struct ctl_table_header *sysctl_header;
4739         ctl_table addrconf_vars[DEVCONF_MAX+1];
4740 } addrconf_sysctl __read_mostly = {
4741         .sysctl_header = NULL,
4742         .addrconf_vars = {
4743                 {
4744                         .procname       = "forwarding",
4745                         .data           = &ipv6_devconf.forwarding,
4746                         .maxlen         = sizeof(int),
4747                         .mode           = 0644,
4748                         .proc_handler   = addrconf_sysctl_forward,
4749                 },
4750                 {
4751                         .procname       = "hop_limit",
4752                         .data           = &ipv6_devconf.hop_limit,
4753                         .maxlen         = sizeof(int),
4754                         .mode           = 0644,
4755                         .proc_handler   = proc_dointvec,
4756                 },
4757                 {
4758                         .procname       = "mtu",
4759                         .data           = &ipv6_devconf.mtu6,
4760                         .maxlen         = sizeof(int),
4761                         .mode           = 0644,
4762                         .proc_handler   = proc_dointvec,
4763                 },
4764                 {
4765                         .procname       = "accept_ra",
4766                         .data           = &ipv6_devconf.accept_ra,
4767                         .maxlen         = sizeof(int),
4768                         .mode           = 0644,
4769                         .proc_handler   = proc_dointvec,
4770                 },
4771                 {
4772                         .procname       = "accept_redirects",
4773                         .data           = &ipv6_devconf.accept_redirects,
4774                         .maxlen         = sizeof(int),
4775                         .mode           = 0644,
4776                         .proc_handler   = proc_dointvec,
4777                 },
4778                 {
4779                         .procname       = "autoconf",
4780                         .data           = &ipv6_devconf.autoconf,
4781                         .maxlen         = sizeof(int),
4782                         .mode           = 0644,
4783                         .proc_handler   = proc_dointvec,
4784                 },
4785                 {
4786                         .procname       = "dad_transmits",
4787                         .data           = &ipv6_devconf.dad_transmits,
4788                         .maxlen         = sizeof(int),
4789                         .mode           = 0644,
4790                         .proc_handler   = proc_dointvec,
4791                 },
4792                 {
4793                         .procname       = "router_solicitations",
4794                         .data           = &ipv6_devconf.rtr_solicits,
4795                         .maxlen         = sizeof(int),
4796                         .mode           = 0644,
4797                         .proc_handler   = proc_dointvec,
4798                 },
4799                 {
4800                         .procname       = "router_solicitation_interval",
4801                         .data           = &ipv6_devconf.rtr_solicit_interval,
4802                         .maxlen         = sizeof(int),
4803                         .mode           = 0644,
4804                         .proc_handler   = proc_dointvec_jiffies,
4805                 },
4806                 {
4807                         .procname       = "router_solicitation_delay",
4808                         .data           = &ipv6_devconf.rtr_solicit_delay,
4809                         .maxlen         = sizeof(int),
4810                         .mode           = 0644,
4811                         .proc_handler   = proc_dointvec_jiffies,
4812                 },
4813                 {
4814                         .procname       = "force_mld_version",
4815                         .data           = &ipv6_devconf.force_mld_version,
4816                         .maxlen         = sizeof(int),
4817                         .mode           = 0644,
4818                         .proc_handler   = proc_dointvec,
4819                 },
4820 #ifdef CONFIG_IPV6_PRIVACY
4821                 {
4822                         .procname       = "use_tempaddr",
4823                         .data           = &ipv6_devconf.use_tempaddr,
4824                         .maxlen         = sizeof(int),
4825                         .mode           = 0644,
4826                         .proc_handler   = proc_dointvec,
4827                 },
4828                 {
4829                         .procname       = "temp_valid_lft",
4830                         .data           = &ipv6_devconf.temp_valid_lft,
4831                         .maxlen         = sizeof(int),
4832                         .mode           = 0644,
4833                         .proc_handler   = proc_dointvec,
4834                 },
4835                 {
4836                         .procname       = "temp_prefered_lft",
4837                         .data           = &ipv6_devconf.temp_prefered_lft,
4838                         .maxlen         = sizeof(int),
4839                         .mode           = 0644,
4840                         .proc_handler   = proc_dointvec,
4841                 },
4842                 {
4843                         .procname       = "regen_max_retry",
4844                         .data           = &ipv6_devconf.regen_max_retry,
4845                         .maxlen         = sizeof(int),
4846                         .mode           = 0644,
4847                         .proc_handler   = proc_dointvec,
4848                 },
4849                 {
4850                         .procname       = "max_desync_factor",
4851                         .data           = &ipv6_devconf.max_desync_factor,
4852                         .maxlen         = sizeof(int),
4853                         .mode           = 0644,
4854                         .proc_handler   = proc_dointvec,
4855                 },
4856 #endif
4857                 {
4858                         .procname       = "max_addresses",
4859                         .data           = &ipv6_devconf.max_addresses,
4860                         .maxlen         = sizeof(int),
4861                         .mode           = 0644,
4862                         .proc_handler   = proc_dointvec,
4863                 },
4864                 {
4865                         .procname       = "accept_ra_defrtr",
4866                         .data           = &ipv6_devconf.accept_ra_defrtr,
4867                         .maxlen         = sizeof(int),
4868                         .mode           = 0644,
4869                         .proc_handler   = proc_dointvec,
4870                 },
4871                 {
4872                         .procname       = "accept_ra_pinfo",
4873                         .data           = &ipv6_devconf.accept_ra_pinfo,
4874                         .maxlen         = sizeof(int),
4875                         .mode           = 0644,
4876                         .proc_handler   = proc_dointvec,
4877                 },
4878 #ifdef CONFIG_IPV6_ROUTER_PREF
4879                 {
4880                         .procname       = "accept_ra_rtr_pref",
4881                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4882                         .maxlen         = sizeof(int),
4883                         .mode           = 0644,
4884                         .proc_handler   = proc_dointvec,
4885                 },
4886                 {
4887                         .procname       = "router_probe_interval",
4888                         .data           = &ipv6_devconf.rtr_probe_interval,
4889                         .maxlen         = sizeof(int),
4890                         .mode           = 0644,
4891                         .proc_handler   = proc_dointvec_jiffies,
4892                 },
4893 #ifdef CONFIG_IPV6_ROUTE_INFO
4894                 {
4895                         .procname       = "accept_ra_rt_info_max_plen",
4896                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4897                         .maxlen         = sizeof(int),
4898                         .mode           = 0644,
4899                         .proc_handler   = proc_dointvec,
4900                 },
4901 #endif
4902 #endif
4903                 {
4904                         .procname       = "proxy_ndp",
4905                         .data           = &ipv6_devconf.proxy_ndp,
4906                         .maxlen         = sizeof(int),
4907                         .mode           = 0644,
4908                         .proc_handler   = proc_dointvec,
4909                 },
4910                 {
4911                         .procname       = "accept_source_route",
4912                         .data           = &ipv6_devconf.accept_source_route,
4913                         .maxlen         = sizeof(int),
4914                         .mode           = 0644,
4915                         .proc_handler   = proc_dointvec,
4916                 },
4917 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4918                 {
4919                         .procname       = "optimistic_dad",
4920                         .data           = &ipv6_devconf.optimistic_dad,
4921                         .maxlen         = sizeof(int),
4922                         .mode           = 0644,
4923                         .proc_handler   = proc_dointvec,
4924
4925                 },
4926 #endif
4927 #ifdef CONFIG_IPV6_MROUTE
4928                 {
4929                         .procname       = "mc_forwarding",
4930                         .data           = &ipv6_devconf.mc_forwarding,
4931                         .maxlen         = sizeof(int),
4932                         .mode           = 0444,
4933                         .proc_handler   = proc_dointvec,
4934                 },
4935 #endif
4936                 {
4937                         .procname       = "disable_ipv6",
4938                         .data           = &ipv6_devconf.disable_ipv6,
4939                         .maxlen         = sizeof(int),
4940                         .mode           = 0644,
4941                         .proc_handler   = addrconf_sysctl_disable,
4942                 },
4943                 {
4944                         .procname       = "accept_dad",
4945                         .data           = &ipv6_devconf.accept_dad,
4946                         .maxlen         = sizeof(int),
4947                         .mode           = 0644,
4948                         .proc_handler   = proc_dointvec,
4949                 },
4950                 {
4951                         .procname       = "force_tllao",
4952                         .data           = &ipv6_devconf.force_tllao,
4953                         .maxlen         = sizeof(int),
4954                         .mode           = 0644,
4955                         .proc_handler   = proc_dointvec
4956                 },
4957                 {
4958                         .procname       = "ndisc_notify",
4959                         .data           = &ipv6_devconf.ndisc_notify,
4960                         .maxlen         = sizeof(int),
4961                         .mode           = 0644,
4962                         .proc_handler   = proc_dointvec
4963                 },
4964                 {
4965                         /* sentinel */
4966                 }
4967         },
4968 };
4969
4970 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4971                 struct inet6_dev *idev, struct ipv6_devconf *p)
4972 {
4973         int i;
4974         struct addrconf_sysctl_table *t;
4975         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4976
4977         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4978         if (t == NULL)
4979                 goto out;
4980
4981         for (i = 0; t->addrconf_vars[i].data; i++) {
4982                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4983                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4984                 t->addrconf_vars[i].extra2 = net;
4985         }
4986
4987         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4988
4989         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4990         if (t->sysctl_header == NULL)
4991                 goto free;
4992
4993         p->sysctl = t;
4994         return 0;
4995
4996 free:
4997         kfree(t);
4998 out:
4999         return -ENOBUFS;
5000 }
5001
5002 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5003 {
5004         struct addrconf_sysctl_table *t;
5005
5006         if (p->sysctl == NULL)
5007                 return;
5008
5009         t = p->sysctl;
5010         p->sysctl = NULL;
5011         unregister_net_sysctl_table(t->sysctl_header);
5012         kfree(t);
5013 }
5014
5015 static void addrconf_sysctl_register(struct inet6_dev *idev)
5016 {
5017         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5018                               &ndisc_ifinfo_sysctl_change);
5019         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5020                                         idev, &idev->cnf);
5021 }
5022
5023 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5024 {
5025         __addrconf_sysctl_unregister(&idev->cnf);
5026         neigh_sysctl_unregister(idev->nd_parms);
5027 }
5028
5029
5030 #endif
5031
5032 static int __net_init addrconf_init_net(struct net *net)
5033 {
5034         int err = -ENOMEM;
5035         struct ipv6_devconf *all, *dflt;
5036
5037         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5038         if (all == NULL)
5039                 goto err_alloc_all;
5040
5041         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5042         if (dflt == NULL)
5043                 goto err_alloc_dflt;
5044
5045         /* these will be inherited by all namespaces */
5046         dflt->autoconf = ipv6_defaults.autoconf;
5047         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5048
5049         net->ipv6.devconf_all = all;
5050         net->ipv6.devconf_dflt = dflt;
5051
5052 #ifdef CONFIG_SYSCTL
5053         err = __addrconf_sysctl_register(net, "all", NULL, all);
5054         if (err < 0)
5055                 goto err_reg_all;
5056
5057         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5058         if (err < 0)
5059                 goto err_reg_dflt;
5060 #endif
5061         return 0;
5062
5063 #ifdef CONFIG_SYSCTL
5064 err_reg_dflt:
5065         __addrconf_sysctl_unregister(all);
5066 err_reg_all:
5067         kfree(dflt);
5068 #endif
5069 err_alloc_dflt:
5070         kfree(all);
5071 err_alloc_all:
5072         return err;
5073 }
5074
5075 static void __net_exit addrconf_exit_net(struct net *net)
5076 {
5077 #ifdef CONFIG_SYSCTL
5078         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5079         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5080 #endif
5081         if (!net_eq(net, &init_net)) {
5082                 kfree(net->ipv6.devconf_dflt);
5083                 kfree(net->ipv6.devconf_all);
5084         }
5085 }
5086
5087 static struct pernet_operations addrconf_ops = {
5088         .init = addrconf_init_net,
5089         .exit = addrconf_exit_net,
5090 };
5091
5092 static struct rtnl_af_ops inet6_ops = {
5093         .family           = AF_INET6,
5094         .fill_link_af     = inet6_fill_link_af,
5095         .get_link_af_size = inet6_get_link_af_size,
5096         .set_link_af      = inet6_set_link_af,
5097 };
5098
5099 /*
5100  *      Init / cleanup code
5101  */
5102
5103 int __init addrconf_init(void)
5104 {
5105         int i, err;
5106
5107         err = ipv6_addr_label_init();
5108         if (err < 0) {
5109                 pr_crit("%s: cannot initialize default policy table: %d\n",
5110                         __func__, err);
5111                 goto out;
5112         }
5113
5114         err = register_pernet_subsys(&addrconf_ops);
5115         if (err < 0)
5116                 goto out_addrlabel;
5117
5118         /* The addrconf netdev notifier requires that loopback_dev
5119          * has it's ipv6 private information allocated and setup
5120          * before it can bring up and give link-local addresses
5121          * to other devices which are up.
5122          *
5123          * Unfortunately, loopback_dev is not necessarily the first
5124          * entry in the global dev_base list of net devices.  In fact,
5125          * it is likely to be the very last entry on that list.
5126          * So this causes the notifier registry below to try and
5127          * give link-local addresses to all devices besides loopback_dev
5128          * first, then loopback_dev, which cases all the non-loopback_dev
5129          * devices to fail to get a link-local address.
5130          *
5131          * So, as a temporary fix, allocate the ipv6 structure for
5132          * loopback_dev first by hand.
5133          * Longer term, all of the dependencies ipv6 has upon the loopback
5134          * device and it being up should be removed.
5135          */
5136         rtnl_lock();
5137         if (!ipv6_add_dev(init_net.loopback_dev))
5138                 err = -ENOMEM;
5139         rtnl_unlock();
5140         if (err)
5141                 goto errlo;
5142
5143         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5144                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5145
5146         register_netdevice_notifier(&ipv6_dev_notf);
5147
5148         addrconf_verify(0);
5149
5150         err = rtnl_af_register(&inet6_ops);
5151         if (err < 0)
5152                 goto errout_af;
5153
5154         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5155                               NULL);
5156         if (err < 0)
5157                 goto errout;
5158
5159         /* Only the first call to __rtnl_register can fail */
5160         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5161         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5162         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5163                         inet6_dump_ifaddr, NULL);
5164         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5165                         inet6_dump_ifmcaddr, NULL);
5166         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5167                         inet6_dump_ifacaddr, NULL);
5168         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5169                         inet6_netconf_dump_devconf, NULL);
5170
5171         ipv6_addr_label_rtnl_register();
5172
5173         return 0;
5174 errout:
5175         rtnl_af_unregister(&inet6_ops);
5176 errout_af:
5177         unregister_netdevice_notifier(&ipv6_dev_notf);
5178 errlo:
5179         unregister_pernet_subsys(&addrconf_ops);
5180 out_addrlabel:
5181         ipv6_addr_label_cleanup();
5182 out:
5183         return err;
5184 }
5185
5186 void addrconf_cleanup(void)
5187 {
5188         struct net_device *dev;
5189         int i;
5190
5191         unregister_netdevice_notifier(&ipv6_dev_notf);
5192         unregister_pernet_subsys(&addrconf_ops);
5193         ipv6_addr_label_cleanup();
5194
5195         rtnl_lock();
5196
5197         __rtnl_af_unregister(&inet6_ops);
5198
5199         /* clean dev list */
5200         for_each_netdev(&init_net, dev) {
5201                 if (__in6_dev_get(dev) == NULL)
5202                         continue;
5203                 addrconf_ifdown(dev, 1);
5204         }
5205         addrconf_ifdown(init_net.loopback_dev, 2);
5206
5207         /*
5208          *      Check hash table.
5209          */
5210         spin_lock_bh(&addrconf_hash_lock);
5211         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5212                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5213         spin_unlock_bh(&addrconf_hash_lock);
5214
5215         del_timer(&addr_chk_timer);
5216         rtnl_unlock();
5217 }