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         .accept_ra_rt_table     = 0,
 202         .proxy_ndp              = 0,
 203         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 204         .disable_ipv6           = 0,
 205         .accept_dad             = 1,
 206 };
 207
 208 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 209         .forwarding             = 0,
 210         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 211         .mtu6                   = IPV6_MIN_MTU,
 212         .accept_ra              = 1,
 213         .accept_redirects       = 1,
 214         .autoconf               = 1,
 215         .dad_transmits          = 1,
 216         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 217         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 218         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 219 #ifdef CONFIG_IPV6_PRIVACY
 220         .use_tempaddr           = 0,
 221         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 222         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 223         .regen_max_retry        = REGEN_MAX_RETRY,
 224         .max_desync_factor      = MAX_DESYNC_FACTOR,
 225 #endif
 226         .max_addresses          = IPV6_MAX_ADDRESSES,
 227         .accept_ra_defrtr       = 1,
 228         .accept_ra_pinfo        = 1,
 229 #ifdef CONFIG_IPV6_ROUTER_PREF
 230         .accept_ra_rtr_pref     = 1,
 231         .rtr_probe_interval     = 60 * HZ,
 232 #ifdef CONFIG_IPV6_ROUTE_INFO
 233         .accept_ra_rt_info_max_plen = 0,
 234 #endif
 235 #endif
 236         .accept_ra_rt_table     = 0,
 237         .proxy_ndp              = 0,
 238         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 239         .disable_ipv6           = 0,
 240         .accept_dad             = 1,
 241 };
 242
 243 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 244 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 245 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 246 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 247 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 248 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 249 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 250 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 251
 252 /* Check if a valid qdisc is available */
 253 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 254 {
 255         return !qdisc_tx_is_noop(dev);
 256 }
 257
 258 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 259 {
 260         if (del_timer(&ifp->timer))
 261                 __in6_ifa_put(ifp);
 262 }
 263
 264 enum addrconf_timer_t {
 265         AC_NONE,
 266         AC_DAD,
 267         AC_RS,
 268 };
 269
 270 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 271                                enum addrconf_timer_t what,
 272                                unsigned long when)
 273 {
 274         if (!del_timer(&ifp->timer))
 275                 in6_ifa_hold(ifp);
 276
 277         switch (what) {
 278         case AC_DAD:
 279                 ifp->timer.function = addrconf_dad_timer;
 280                 break;
 281         case AC_RS:
 282                 ifp->timer.function = addrconf_rs_timer;
 283                 break;
 284         default:
 285                 break;
 286         }
 287         ifp->timer.expires = jiffies + when;
 288         add_timer(&ifp->timer);
 289 }
 290
 291 static int snmp6_alloc_dev(struct inet6_dev *idev)
 292 {
 293         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 294                           sizeof(struct ipstats_mib),
 295                           __alignof__(struct ipstats_mib)) < 0)
 296                 goto err_ip;
 297         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 298                                         GFP_KERNEL);
 299         if (!idev->stats.icmpv6dev)
 300                 goto err_icmp;
 301         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 302                                            GFP_KERNEL);
 303         if (!idev->stats.icmpv6msgdev)
 304                 goto err_icmpmsg;
 305
 306         return 0;
 307
 308 err_icmpmsg:
 309         kfree(idev->stats.icmpv6dev);
 310 err_icmp:
 311         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 312 err_ip:
 313         return -ENOMEM;
 314 }
 315
 316 static void snmp6_free_dev(struct inet6_dev *idev)
 317 {
 318         kfree(idev->stats.icmpv6msgdev);
 319         kfree(idev->stats.icmpv6dev);
 320         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 321 }
 322
 323 /* Nobody refers to this device, we may destroy it. */
 324
 325 void in6_dev_finish_destroy(struct inet6_dev *idev)
 326 {
 327         struct net_device *dev = idev->dev;
 328
 329         WARN_ON(!list_empty(&idev->addr_list));
 330         WARN_ON(idev->mc_list != NULL);
 331
 332 #ifdef NET_REFCNT_DEBUG
 333         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 334 #endif
 335         dev_put(dev);
 336         if (!idev->dead) {
 337                 pr_warn("Freeing alive inet6 device %p\n", idev);
 338                 return;
 339         }
 340         snmp6_free_dev(idev);
 341         kfree_rcu(idev, rcu);
 342 }
 343 EXPORT_SYMBOL(in6_dev_finish_destroy);
 344
 345 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 346 {
 347         struct inet6_dev *ndev;
 348
 349         ASSERT_RTNL();
 350
 351         if (dev->mtu < IPV6_MIN_MTU)
 352                 return NULL;
 353
 354         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 355
 356         if (ndev == NULL)
 357                 return NULL;
 358
 359         rwlock_init(&ndev->lock);
 360         ndev->dev = dev;
 361         INIT_LIST_HEAD(&ndev->addr_list);
 362
 363         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 364         ndev->cnf.mtu6 = dev->mtu;
 365         ndev->cnf.sysctl = NULL;
 366         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 367         if (ndev->nd_parms == NULL) {
 368                 kfree(ndev);
 369                 return NULL;
 370         }
 371         if (ndev->cnf.forwarding)
 372                 dev_disable_lro(dev);
 373         /* We refer to the device */
 374         dev_hold(dev);
 375
 376         if (snmp6_alloc_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 dev_put(dev);
 382                 kfree(ndev);
 383                 return NULL;
 384         }
 385
 386         if (snmp6_register_dev(ndev) < 0) {
 387                 ADBG((KERN_WARNING
 388                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 389                         __func__, dev->name));
 390                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 391                 ndev->dead = 1;
 392                 in6_dev_finish_destroy(ndev);
 393                 return NULL;
 394         }
 395
 396         /* One reference from device.  We must do this before
 397          * we invoke __ipv6_regen_rndid().
 398          */
 399         in6_dev_hold(ndev);
 400
 401         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 402                 ndev->cnf.accept_dad = -1;
 403
 404 #if IS_ENABLED(CONFIG_IPV6_SIT)
 405         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 406                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 407                 ndev->cnf.rtr_solicits = 0;
 408         }
 409 #endif
 410
 411 #ifdef CONFIG_IPV6_PRIVACY
 412         INIT_LIST_HEAD(&ndev->tempaddr_list);
 413         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 414         if ((dev->flags&IFF_LOOPBACK) ||
 415             dev->type == ARPHRD_TUNNEL ||
 416             dev->type == ARPHRD_TUNNEL6 ||
 417             dev->type == ARPHRD_SIT ||
 418             dev->type == ARPHRD_NONE) {
 419                 ndev->cnf.use_tempaddr = -1;
 420         } else {
 421                 in6_dev_hold(ndev);
 422                 ipv6_regen_rndid((unsigned long) ndev);
 423         }
 424 #endif
 425         ndev->token = in6addr_any;
 426
 427         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 428                 ndev->if_flags |= IF_READY;
 429
 430         ipv6_mc_init_dev(ndev);
 431         ndev->tstamp = jiffies;
 432         addrconf_sysctl_register(ndev);
 433         /* protected by rtnl_lock */
 434         rcu_assign_pointer(dev->ip6_ptr, ndev);
 435
 436         /* Join interface-local all-node multicast group */
 437         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 438
 439         /* Join all-node multicast group */
 440         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 441
 442         /* Join all-router multicast group if forwarding is set */
 443         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 444                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 445
 446         return ndev;
 447 }
 448
 449 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 450 {
 451         struct inet6_dev *idev;
 452
 453         ASSERT_RTNL();
 454
 455         idev = __in6_dev_get(dev);
 456         if (!idev) {
 457                 idev = ipv6_add_dev(dev);
 458                 if (!idev)
 459                         return NULL;
 460         }
 461
 462         if (dev->flags&IFF_UP)
 463                 ipv6_mc_up(idev);
 464         return idev;
 465 }
 466
 467 static int inet6_netconf_msgsize_devconf(int type)
 468 {
 469         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 470                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 471
 472         /* type -1 is used for ALL */
 473         if (type == -1 || type == NETCONFA_FORWARDING)
 474                 size += nla_total_size(4);
 475 #ifdef CONFIG_IPV6_MROUTE
 476         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 477                 size += nla_total_size(4);
 478 #endif
 479
 480         return size;
 481 }
 482
 483 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 484                                       struct ipv6_devconf *devconf, u32 portid,
 485                                       u32 seq, int event, unsigned int flags,
 486                                       int type)
 487 {
 488         struct nlmsghdr  *nlh;
 489         struct netconfmsg *ncm;
 490
 491         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 492                         flags);
 493         if (nlh == NULL)
 494                 return -EMSGSIZE;
 495
 496         ncm = nlmsg_data(nlh);
 497         ncm->ncm_family = AF_INET6;
 498
 499         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 500                 goto nla_put_failure;
 501
 502         /* type -1 is used for ALL */
 503         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 504             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 505                 goto nla_put_failure;
 506 #ifdef CONFIG_IPV6_MROUTE
 507         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 508             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 509                         devconf->mc_forwarding) < 0)
 510                 goto nla_put_failure;
 511 #endif
 512         return nlmsg_end(skb, nlh);
 513
 514 nla_put_failure:
 515         nlmsg_cancel(skb, nlh);
 516         return -EMSGSIZE;
 517 }
 518
 519 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 520                                   struct ipv6_devconf *devconf)
 521 {
 522         struct sk_buff *skb;
 523         int err = -ENOBUFS;
 524
 525         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 526         if (skb == NULL)
 527                 goto errout;
 528
 529         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 530                                          RTM_NEWNETCONF, 0, type);
 531         if (err < 0) {
 532                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 533                 WARN_ON(err == -EMSGSIZE);
 534                 kfree_skb(skb);
 535                 goto errout;
 536         }
 537         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 538         return;
 539 errout:
 540         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 541 }
 542
 543 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 544         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 545         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 546 };
 547
 548 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 549                                      struct nlmsghdr *nlh)
 550 {
 551         struct net *net = sock_net(in_skb->sk);
 552         struct nlattr *tb[NETCONFA_MAX+1];
 553         struct netconfmsg *ncm;
 554         struct sk_buff *skb;
 555         struct ipv6_devconf *devconf;
 556         struct inet6_dev *in6_dev;
 557         struct net_device *dev;
 558         int ifindex;
 559         int err;
 560
 561         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 562                           devconf_ipv6_policy);
 563         if (err < 0)
 564                 goto errout;
 565
 566         err = EINVAL;
 567         if (!tb[NETCONFA_IFINDEX])
 568                 goto errout;
 569
 570         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 571         switch (ifindex) {
 572         case NETCONFA_IFINDEX_ALL:
 573                 devconf = net->ipv6.devconf_all;
 574                 break;
 575         case NETCONFA_IFINDEX_DEFAULT:
 576                 devconf = net->ipv6.devconf_dflt;
 577                 break;
 578         default:
 579                 dev = __dev_get_by_index(net, ifindex);
 580                 if (dev == NULL)
 581                         goto errout;
 582                 in6_dev = __in6_dev_get(dev);
 583                 if (in6_dev == NULL)
 584                         goto errout;
 585                 devconf = &in6_dev->cnf;
 586                 break;
 587         }
 588
 589         err = -ENOBUFS;
 590         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 591         if (skb == NULL)
 592                 goto errout;
 593
 594         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 595                                          NETLINK_CB(in_skb).portid,
 596                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 597                                          -1);
 598         if (err < 0) {
 599                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 600                 WARN_ON(err == -EMSGSIZE);
 601                 kfree_skb(skb);
 602                 goto errout;
 603         }
 604         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 605 errout:
 606         return err;
 607 }
 608
 609 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 610                                       struct netlink_callback *cb)
 611 {
 612         struct net *net = sock_net(skb->sk);
 613         int h, s_h;
 614         int idx, s_idx;
 615         struct net_device *dev;
 616         struct inet6_dev *idev;
 617         struct hlist_head *head;
 618
 619         s_h = cb->args[0];
 620         s_idx = idx = cb->args[1];
 621
 622         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 623                 idx = 0;
 624                 head = &net->dev_index_head[h];
 625                 rcu_read_lock();
 626                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 627                           net->dev_base_seq;
 628                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 629                         if (idx < s_idx)
 630                                 goto cont;
 631                         idev = __in6_dev_get(dev);
 632                         if (!idev)
 633                                 goto cont;
 634
 635                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 636                                                        &idev->cnf,
 637                                                        NETLINK_CB(cb->skb).portid,
 638                                                        cb->nlh->nlmsg_seq,
 639                                                        RTM_NEWNETCONF,
 640                                                        NLM_F_MULTI,
 641                                                        -1) <= 0) {
 642                                 rcu_read_unlock();
 643                                 goto done;
 644                         }
 645                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 646 cont:
 647                         idx++;
 648                 }
 649                 rcu_read_unlock();
 650         }
 651         if (h == NETDEV_HASHENTRIES) {
 652                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 653                                                net->ipv6.devconf_all,
 654                                                NETLINK_CB(cb->skb).portid,
 655                                                cb->nlh->nlmsg_seq,
 656                                                RTM_NEWNETCONF, NLM_F_MULTI,
 657                                                -1) <= 0)
 658                         goto done;
 659                 else
 660                         h++;
 661         }
 662         if (h == NETDEV_HASHENTRIES + 1) {
 663                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 664                                                net->ipv6.devconf_dflt,
 665                                                NETLINK_CB(cb->skb).portid,
 666                                                cb->nlh->nlmsg_seq,
 667                                                RTM_NEWNETCONF, NLM_F_MULTI,
 668                                                -1) <= 0)
 669                         goto done;
 670                 else
 671                         h++;
 672         }
 673 done:
 674         cb->args[0] = h;
 675         cb->args[1] = idx;
 676
 677         return skb->len;
 678 }
 679
 680 #ifdef CONFIG_SYSCTL
 681 static void dev_forward_change(struct inet6_dev *idev)
 682 {
 683         struct net_device *dev;
 684         struct inet6_ifaddr *ifa;
 685
 686         if (!idev)
 687                 return;
 688         dev = idev->dev;
 689         if (idev->cnf.forwarding)
 690                 dev_disable_lro(dev);
 691         if (dev->flags & IFF_MULTICAST) {
 692                 if (idev->cnf.forwarding) {
 693                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 694                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 695                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 696                 } else {
 697                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 698                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 699                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 700                 }
 701         }
 702
 703         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 704                 if (ifa->flags&IFA_F_TENTATIVE)
 705                         continue;
 706                 if (idev->cnf.forwarding)
 707                         addrconf_join_anycast(ifa);
 708                 else
 709                         addrconf_leave_anycast(ifa);
 710         }
 711         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 712                                      dev->ifindex, &idev->cnf);
 713 }
 714
 715
 716 static void addrconf_forward_change(struct net *net, __s32 newf)
 717 {
 718         struct net_device *dev;
 719         struct inet6_dev *idev;
 720
 721         for_each_netdev(net, dev) {
 722                 idev = __in6_dev_get(dev);
 723                 if (idev) {
 724                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 725                         idev->cnf.forwarding = newf;
 726                         if (changed)
 727                                 dev_forward_change(idev);
 728                 }
 729         }
 730 }
 731
 732 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 733 {
 734         struct net *net;
 735         int old;
 736
 737         if (!rtnl_trylock())
 738                 return restart_syscall();
 739
 740         net = (struct net *)table->extra2;
 741         old = *p;
 742         *p = newf;
 743
 744         if (p == &net->ipv6.devconf_dflt->forwarding) {
 745                 if ((!newf) ^ (!old))
 746                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 747                                                      NETCONFA_IFINDEX_DEFAULT,
 748                                                      net->ipv6.devconf_dflt);
 749                 rtnl_unlock();
 750                 return 0;
 751         }
 752
 753         if (p == &net->ipv6.devconf_all->forwarding) {
 754                 net->ipv6.devconf_dflt->forwarding = newf;
 755                 addrconf_forward_change(net, newf);
 756                 if ((!newf) ^ (!old))
 757                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 758                                                      NETCONFA_IFINDEX_ALL,
 759                                                      net->ipv6.devconf_all);
 760         } else if ((!newf) ^ (!old))
 761                 dev_forward_change((struct inet6_dev *)table->extra1);
 762         rtnl_unlock();
 763
 764         if (newf)
 765                 rt6_purge_dflt_routers(net);
 766         return 1;
 767 }
 768 #endif
 769
 770 /* Nobody refers to this ifaddr, destroy it */
 771 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 772 {
 773         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 774
 775 #ifdef NET_REFCNT_DEBUG
 776         pr_debug("%s\n", __func__);
 777 #endif
 778
 779         in6_dev_put(ifp->idev);
 780
 781         if (del_timer(&ifp->timer))
 782                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 783
 784         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 785                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 786                 return;
 787         }
 788         ip6_rt_put(ifp->rt);
 789
 790         kfree_rcu(ifp, rcu);
 791 }
 792
 793 static void
 794 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 795 {
 796         struct list_head *p;
 797         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 798
 799         /*
 800          * Each device address list is sorted in order of scope -
 801          * global before linklocal.
 802          */
 803         list_for_each(p, &idev->addr_list) {
 804                 struct inet6_ifaddr *ifa
 805                         = list_entry(p, struct inet6_ifaddr, if_list);
 806                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 807                         break;
 808         }
 809
 810         list_add_tail(&ifp->if_list, p);
 811 }
 812
 813 static u32 inet6_addr_hash(const struct in6_addr *addr)
 814 {
 815         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 816 }
 817
 818 /* On success it returns ifp with increased reference count */
 819
 820 static struct inet6_ifaddr *
 821 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 822               int scope, u32 flags)
 823 {
 824         struct inet6_ifaddr *ifa = NULL;
 825         struct rt6_info *rt;
 826         unsigned int hash;
 827         int err = 0;
 828         int addr_type = ipv6_addr_type(addr);
 829
 830         if (addr_type == IPV6_ADDR_ANY ||
 831             addr_type & IPV6_ADDR_MULTICAST ||
 832             (!(idev->dev->flags & IFF_LOOPBACK) &&
 833              addr_type & IPV6_ADDR_LOOPBACK))
 834                 return ERR_PTR(-EADDRNOTAVAIL);
 835
 836         rcu_read_lock_bh();
 837         if (idev->dead) {
 838                 err = -ENODEV;                  /*XXX*/
 839                 goto out2;
 840         }
 841
 842         if (idev->cnf.disable_ipv6) {
 843                 err = -EACCES;
 844                 goto out2;
 845         }
 846
 847         spin_lock(&addrconf_hash_lock);
 848
 849         /* Ignore adding duplicate addresses on an interface */
 850         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 851                 ADBG(("ipv6_add_addr: already assigned\n"));
 852                 err = -EEXIST;
 853                 goto out;
 854         }
 855
 856         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 857
 858         if (ifa == NULL) {
 859                 ADBG(("ipv6_add_addr: malloc failed\n"));
 860                 err = -ENOBUFS;
 861                 goto out;
 862         }
 863
 864         rt = addrconf_dst_alloc(idev, addr, false);
 865         if (IS_ERR(rt)) {
 866                 err = PTR_ERR(rt);
 867                 goto out;
 868         }
 869
 870         ifa->addr = *addr;
 871
 872         spin_lock_init(&ifa->lock);
 873         spin_lock_init(&ifa->state_lock);
 874         init_timer(&ifa->timer);
 875         INIT_HLIST_NODE(&ifa->addr_lst);
 876         ifa->timer.data = (unsigned long) ifa;
 877         ifa->scope = scope;
 878         ifa->prefix_len = pfxlen;
 879         ifa->flags = flags | IFA_F_TENTATIVE;
 880         ifa->cstamp = ifa->tstamp = jiffies;
 881         ifa->tokenized = false;
 882
 883         ifa->rt = rt;
 884
 885         ifa->idev = idev;
 886         in6_dev_hold(idev);
 887         /* For caller */
 888         in6_ifa_hold(ifa);
 889
 890         /* Add to big hash table */
 891         hash = inet6_addr_hash(addr);
 892
 893         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 894         spin_unlock(&addrconf_hash_lock);
 895
 896         write_lock(&idev->lock);
 897         /* Add to inet6_dev unicast addr list. */
 898         ipv6_link_dev_addr(idev, ifa);
 899
 900 #ifdef CONFIG_IPV6_PRIVACY
 901         if (ifa->flags&IFA_F_TEMPORARY) {
 902                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 903                 in6_ifa_hold(ifa);
 904         }
 905 #endif
 906
 907         in6_ifa_hold(ifa);
 908         write_unlock(&idev->lock);
 909 out2:
 910         rcu_read_unlock_bh();
 911
 912         if (likely(err == 0))
 913                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 914         else {
 915                 kfree(ifa);
 916                 ifa = ERR_PTR(err);
 917         }
 918
 919         return ifa;
 920 out:
 921         spin_unlock(&addrconf_hash_lock);
 922         goto out2;
 923 }
 924
 925 /* This function wants to get referenced ifp and releases it before return */
 926
 927 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 928 {
 929         struct inet6_ifaddr *ifa, *ifn;
 930         struct inet6_dev *idev = ifp->idev;
 931         int state;
 932         int deleted = 0, onlink = 0;
 933         unsigned long expires = jiffies;
 934
 935         spin_lock_bh(&ifp->state_lock);
 936         state = ifp->state;
 937         ifp->state = INET6_IFADDR_STATE_DEAD;
 938         spin_unlock_bh(&ifp->state_lock);
 939
 940         if (state == INET6_IFADDR_STATE_DEAD)
 941                 goto out;
 942
 943         spin_lock_bh(&addrconf_hash_lock);
 944         hlist_del_init_rcu(&ifp->addr_lst);
 945         spin_unlock_bh(&addrconf_hash_lock);
 946
 947         write_lock_bh(&idev->lock);
 948 #ifdef CONFIG_IPV6_PRIVACY
 949         if (ifp->flags&IFA_F_TEMPORARY) {
 950                 list_del(&ifp->tmp_list);
 951                 if (ifp->ifpub) {
 952                         in6_ifa_put(ifp->ifpub);
 953                         ifp->ifpub = NULL;
 954                 }
 955                 __in6_ifa_put(ifp);
 956         }
 957 #endif
 958
 959         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 960                 if (ifa == ifp) {
 961                         list_del_init(&ifp->if_list);
 962                         __in6_ifa_put(ifp);
 963
 964                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 965                                 break;
 966                         deleted = 1;
 967                         continue;
 968                 } else if (ifp->flags & IFA_F_PERMANENT) {
 969                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 970                                               ifp->prefix_len)) {
 971                                 if (ifa->flags & IFA_F_PERMANENT) {
 972                                         onlink = 1;
 973                                         if (deleted)
 974                                                 break;
 975                                 } else {
 976                                         unsigned long lifetime;
 977
 978                                         if (!onlink)
 979                                                 onlink = -1;
 980
 981                                         spin_lock(&ifa->lock);
 982
 983                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 984                                         /*
 985                                          * Note: Because this address is
 986                                          * not permanent, lifetime <
 987                                          * LONG_MAX / HZ here.
 988                                          */
 989                                         if (time_before(expires,
 990                                                         ifa->tstamp + lifetime * HZ))
 991                                                 expires = ifa->tstamp + lifetime * HZ;
 992                                         spin_unlock(&ifa->lock);
 993                                 }
 994                         }
 995                 }
 996         }
 997         write_unlock_bh(&idev->lock);
 998
 999         addrconf_del_timer(ifp);
1000
1001         ipv6_ifa_notify(RTM_DELADDR, ifp);
1002
1003         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1004
1005         /*
1006          * Purge or update corresponding prefix
1007          *
1008          * 1) we don't purge prefix here if address was not permanent.
1009          *    prefix is managed by its own lifetime.
1010          * 2) if there're no addresses, delete prefix.
1011          * 3) if there're still other permanent address(es),
1012          *    corresponding prefix is still permanent.
1013          * 4) otherwise, update prefix lifetime to the
1014          *    longest valid lifetime among the corresponding
1015          *    addresses on the device.
1016          *    Note: subsequent RA will update lifetime.
1017          *
1018          * --yoshfuji
1019          */
1020         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1021                 struct in6_addr prefix;
1022                 struct rt6_info *rt;
1023
1024                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1025
1026                 rt = addrconf_get_prefix_route(&prefix,
1027                                                ifp->prefix_len,
1028                                                ifp->idev->dev,
1029                                                0, RTF_GATEWAY | RTF_DEFAULT);
1030
1031                 if (rt) {
1032                         if (onlink == 0) {
1033                                 ip6_del_rt(rt);
1034                                 rt = NULL;
1035                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1036                                 rt6_set_expires(rt, expires);
1037                         }
1038                 }
1039                 ip6_rt_put(rt);
1040         }
1041
1042         /* clean up prefsrc entries */
1043         rt6_remove_prefsrc(ifp);
1044 out:
1045         in6_ifa_put(ifp);
1046 }
1047
1048 #ifdef CONFIG_IPV6_PRIVACY
1049 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1050 {
1051         struct inet6_dev *idev = ifp->idev;
1052         struct in6_addr addr, *tmpaddr;
1053         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1054         unsigned long regen_advance;
1055         int tmp_plen;
1056         int ret = 0;
1057         int max_addresses;
1058         u32 addr_flags;
1059         unsigned long now = jiffies;
1060
1061         write_lock(&idev->lock);
1062         if (ift) {
1063                 spin_lock_bh(&ift->lock);
1064                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1065                 spin_unlock_bh(&ift->lock);
1066                 tmpaddr = &addr;
1067         } else {
1068                 tmpaddr = NULL;
1069         }
1070 retry:
1071         in6_dev_hold(idev);
1072         if (idev->cnf.use_tempaddr <= 0) {
1073                 write_unlock(&idev->lock);
1074                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1075                 in6_dev_put(idev);
1076                 ret = -1;
1077                 goto out;
1078         }
1079         spin_lock_bh(&ifp->lock);
1080         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1081                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1082                 spin_unlock_bh(&ifp->lock);
1083                 write_unlock(&idev->lock);
1084                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1085                         __func__);
1086                 in6_dev_put(idev);
1087                 ret = -1;
1088                 goto out;
1089         }
1090         in6_ifa_hold(ifp);
1091         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1092         __ipv6_try_regen_rndid(idev, tmpaddr);
1093         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1094         age = (now - ifp->tstamp) / HZ;
1095         tmp_valid_lft = min_t(__u32,
1096                               ifp->valid_lft,
1097                               idev->cnf.temp_valid_lft + age);
1098         tmp_prefered_lft = min_t(__u32,
1099                                  ifp->prefered_lft,
1100                                  idev->cnf.temp_prefered_lft + age -
1101                                  idev->cnf.max_desync_factor);
1102         tmp_plen = ifp->prefix_len;
1103         max_addresses = idev->cnf.max_addresses;
1104         tmp_tstamp = ifp->tstamp;
1105         spin_unlock_bh(&ifp->lock);
1106
1107         regen_advance = idev->cnf.regen_max_retry *
1108                         idev->cnf.dad_transmits *
1109                         idev->nd_parms->retrans_time / HZ;
1110         write_unlock(&idev->lock);
1111
1112         /* A temporary address is created only if this calculated Preferred
1113          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1114          * an implementation must not create a temporary address with a zero
1115          * Preferred Lifetime.
1116          */
1117         if (tmp_prefered_lft <= regen_advance) {
1118                 in6_ifa_put(ifp);
1119                 in6_dev_put(idev);
1120                 ret = -1;
1121                 goto out;
1122         }
1123
1124         addr_flags = IFA_F_TEMPORARY;
1125         /* set in addrconf_prefix_rcv() */
1126         if (ifp->flags & IFA_F_OPTIMISTIC)
1127                 addr_flags |= IFA_F_OPTIMISTIC;
1128
1129         ift = !max_addresses ||
1130               ipv6_count_addresses(idev) < max_addresses ?
1131                 ipv6_add_addr(idev, &addr, tmp_plen,
1132                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1133                               addr_flags) : NULL;
1134         if (IS_ERR_OR_NULL(ift)) {
1135                 in6_ifa_put(ifp);
1136                 in6_dev_put(idev);
1137                 pr_info("%s: retry temporary address regeneration\n", __func__);
1138                 tmpaddr = &addr;
1139                 write_lock(&idev->lock);
1140                 goto retry;
1141         }
1142
1143         spin_lock_bh(&ift->lock);
1144         ift->ifpub = ifp;
1145         ift->valid_lft = tmp_valid_lft;
1146         ift->prefered_lft = tmp_prefered_lft;
1147         ift->cstamp = now;
1148         ift->tstamp = tmp_tstamp;
1149         spin_unlock_bh(&ift->lock);
1150
1151         addrconf_dad_start(ift);
1152         in6_ifa_put(ift);
1153         in6_dev_put(idev);
1154 out:
1155         return ret;
1156 }
1157 #endif
1158
1159 /*
1160  *      Choose an appropriate source address (RFC3484)
1161  */
1162 enum {
1163         IPV6_SADDR_RULE_INIT = 0,
1164         IPV6_SADDR_RULE_LOCAL,
1165         IPV6_SADDR_RULE_SCOPE,
1166         IPV6_SADDR_RULE_PREFERRED,
1167 #ifdef CONFIG_IPV6_MIP6
1168         IPV6_SADDR_RULE_HOA,
1169 #endif
1170         IPV6_SADDR_RULE_OIF,
1171         IPV6_SADDR_RULE_LABEL,
1172 #ifdef CONFIG_IPV6_PRIVACY
1173         IPV6_SADDR_RULE_PRIVACY,
1174 #endif
1175         IPV6_SADDR_RULE_ORCHID,
1176         IPV6_SADDR_RULE_PREFIX,
1177 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1178         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1179 #endif
1180         IPV6_SADDR_RULE_MAX
1181 };
1182
1183 struct ipv6_saddr_score {
1184         int                     rule;
1185         int                     addr_type;
1186         struct inet6_ifaddr     *ifa;
1187         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1188         int                     scopedist;
1189         int                     matchlen;
1190 };
1191
1192 struct ipv6_saddr_dst {
1193         const struct in6_addr *addr;
1194         int ifindex;
1195         int scope;
1196         int label;
1197         unsigned int prefs;
1198 };
1199
1200 static inline int ipv6_saddr_preferred(int type)
1201 {
1202         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1203                 return 1;
1204         return 0;
1205 }
1206
1207 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1208 {
1209 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1210         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1211 #else
1212         return false;
1213 #endif
1214 }
1215
1216 static int ipv6_get_saddr_eval(struct net *net,
1217                                struct ipv6_saddr_score *score,
1218                                struct ipv6_saddr_dst *dst,
1219                                int i)
1220 {
1221         int ret;
1222
1223         if (i <= score->rule) {
1224                 switch (i) {
1225                 case IPV6_SADDR_RULE_SCOPE:
1226                         ret = score->scopedist;
1227                         break;
1228                 case IPV6_SADDR_RULE_PREFIX:
1229                         ret = score->matchlen;
1230                         break;
1231                 default:
1232                         ret = !!test_bit(i, score->scorebits);
1233                 }
1234                 goto out;
1235         }
1236
1237         switch (i) {
1238         case IPV6_SADDR_RULE_INIT:
1239                 /* Rule 0: remember if hiscore is not ready yet */
1240                 ret = !!score->ifa;
1241                 break;
1242         case IPV6_SADDR_RULE_LOCAL:
1243                 /* Rule 1: Prefer same address */
1244                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1245                 break;
1246         case IPV6_SADDR_RULE_SCOPE:
1247                 /* Rule 2: Prefer appropriate scope
1248                  *
1249                  *      ret
1250                  *       ^
1251                  *    -1 |  d 15
1252                  *    ---+--+-+---> scope
1253                  *       |
1254                  *       |             d is scope of the destination.
1255                  *  B-d  |  \
1256                  *       |   \      <- smaller scope is better if
1257                  *  B-15 |    \        if scope is enough for destinaion.
1258                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1259                  * d-C-1 | /
1260                  *       |/         <- greater is better
1261                  *   -C  /             if scope is not enough for destination.
1262                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1263                  *
1264                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1265                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1266                  * Assume B = 0 and we get C > 29.
1267                  */
1268                 ret = __ipv6_addr_src_scope(score->addr_type);
1269                 if (ret >= dst->scope)
1270                         ret = -ret;
1271                 else
1272                         ret -= 128;     /* 30 is enough */
1273                 score->scopedist = ret;
1274                 break;
1275         case IPV6_SADDR_RULE_PREFERRED:
1276             {
1277                 /* Rule 3: Avoid deprecated and optimistic addresses */
1278                 u8 avoid = IFA_F_DEPRECATED;
1279
1280                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1281                         avoid |= IFA_F_OPTIMISTIC;
1282                 ret = ipv6_saddr_preferred(score->addr_type) ||
1283                       !(score->ifa->flags & avoid);
1284                 break;
1285             }
1286 #ifdef CONFIG_IPV6_MIP6
1287         case IPV6_SADDR_RULE_HOA:
1288             {
1289                 /* Rule 4: Prefer home address */
1290                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1291                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1292                 break;
1293             }
1294 #endif
1295         case IPV6_SADDR_RULE_OIF:
1296                 /* Rule 5: Prefer outgoing interface */
1297                 ret = (!dst->ifindex ||
1298                        dst->ifindex == score->ifa->idev->dev->ifindex);
1299                 break;
1300         case IPV6_SADDR_RULE_LABEL:
1301                 /* Rule 6: Prefer matching label */
1302                 ret = ipv6_addr_label(net,
1303                                       &score->ifa->addr, score->addr_type,
1304                                       score->ifa->idev->dev->ifindex) == dst->label;
1305                 break;
1306 #ifdef CONFIG_IPV6_PRIVACY
1307         case IPV6_SADDR_RULE_PRIVACY:
1308             {
1309                 /* Rule 7: Prefer public address
1310                  * Note: prefer temporary address if use_tempaddr >= 2
1311                  */
1312                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1313                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1314                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1315                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1316                 break;
1317             }
1318 #endif
1319         case IPV6_SADDR_RULE_ORCHID:
1320                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1321                  *          non-ORCHID vs non-ORCHID
1322                  */
1323                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1324                         ipv6_addr_orchid(dst->addr));
1325                 break;
1326         case IPV6_SADDR_RULE_PREFIX:
1327                 /* Rule 8: Use longest matching prefix */
1328                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1329                 if (ret > score->ifa->prefix_len)
1330                         ret = score->ifa->prefix_len;
1331                 score->matchlen = ret;
1332                 break;
1333 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1334         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1335                 /* Optimistic addresses still have lower precedence than other
1336                  * preferred addresses.
1337                  */
1338                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1339                 break;
1340 #endif
1341         default:
1342                 ret = 0;
1343         }
1344
1345         if (ret)
1346                 __set_bit(i, score->scorebits);
1347         score->rule = i;
1348 out:
1349         return ret;
1350 }
1351
1352 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1353                        const struct in6_addr *daddr, unsigned int prefs,
1354                        struct in6_addr *saddr)
1355 {
1356         struct ipv6_saddr_score scores[2],
1357                                 *score = &scores[0], *hiscore = &scores[1];
1358         struct ipv6_saddr_dst dst;
1359         struct net_device *dev;
1360         int dst_type;
1361
1362         dst_type = __ipv6_addr_type(daddr);
1363         dst.addr = daddr;
1364         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1365         dst.scope = __ipv6_addr_src_scope(dst_type);
1366         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1367         dst.prefs = prefs;
1368
1369         hiscore->rule = -1;
1370         hiscore->ifa = NULL;
1371
1372         rcu_read_lock();
1373
1374         for_each_netdev_rcu(net, dev) {
1375                 struct inet6_dev *idev;
1376
1377                 /* Candidate Source Address (section 4)
1378                  *  - multicast and link-local destination address,
1379                  *    the set of candidate source address MUST only
1380                  *    include addresses assigned to interfaces
1381                  *    belonging to the same link as the outgoing
1382                  *    interface.
1383                  * (- For site-local destination addresses, the
1384                  *    set of candidate source addresses MUST only
1385                  *    include addresses assigned to interfaces
1386                  *    belonging to the same site as the outgoing
1387                  *    interface.)
1388                  */
1389                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1390                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1391                     dst.ifindex && dev->ifindex != dst.ifindex)
1392                         continue;
1393
1394                 idev = __in6_dev_get(dev);
1395                 if (!idev)
1396                         continue;
1397
1398                 read_lock_bh(&idev->lock);
1399                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1400                         int i;
1401
1402                         /*
1403                          * - Tentative Address (RFC2462 section 5.4)
1404                          *  - A tentative address is not considered
1405                          *    "assigned to an interface" in the traditional
1406                          *    sense, unless it is also flagged as optimistic.
1407                          * - Candidate Source Address (section 4)
1408                          *  - In any case, anycast addresses, multicast
1409                          *    addresses, and the unspecified address MUST
1410                          *    NOT be included in a candidate set.
1411                          */
1412                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1413                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1414                                 continue;
1415
1416                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1417
1418                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1419                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1420                                 LIMIT_NETDEBUG(KERN_DEBUG
1421                                                "ADDRCONF: unspecified / multicast address "
1422                                                "assigned as unicast address on %s",
1423                                                dev->name);
1424                                 continue;
1425                         }
1426
1427                         score->rule = -1;
1428                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1429
1430                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1431                                 int minihiscore, miniscore;
1432
1433                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1434                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1435
1436                                 if (minihiscore > miniscore) {
1437                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1438                                             score->scopedist > 0) {
1439                                                 /*
1440                                                  * special case:
1441                                                  * each remaining entry
1442                                                  * has too small (not enough)
1443                                                  * scope, because ifa entries
1444                                                  * are sorted by their scope
1445                                                  * values.
1446                                                  */
1447                                                 goto try_nextdev;
1448                                         }
1449                                         break;
1450                                 } else if (minihiscore < miniscore) {
1451                                         if (hiscore->ifa)
1452                                                 in6_ifa_put(hiscore->ifa);
1453
1454                                         in6_ifa_hold(score->ifa);
1455
1456                                         swap(hiscore, score);
1457
1458                                         /* restore our iterator */
1459                                         score->ifa = hiscore->ifa;
1460
1461                                         break;
1462                                 }
1463                         }
1464                 }
1465 try_nextdev:
1466                 read_unlock_bh(&idev->lock);
1467         }
1468         rcu_read_unlock();
1469
1470         if (!hiscore->ifa)
1471                 return -EADDRNOTAVAIL;
1472
1473         *saddr = hiscore->ifa->addr;
1474         in6_ifa_put(hiscore->ifa);
1475         return 0;
1476 }
1477 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1478
1479 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1480                     unsigned char banned_flags)
1481 {
1482         struct inet6_dev *idev;
1483         int err = -EADDRNOTAVAIL;
1484
1485         rcu_read_lock();
1486         idev = __in6_dev_get(dev);
1487         if (idev) {
1488                 struct inet6_ifaddr *ifp;
1489
1490                 read_lock_bh(&idev->lock);
1491                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1492                         if (ifp->scope == IFA_LINK &&
1493                             !(ifp->flags & banned_flags)) {
1494                                 *addr = ifp->addr;
1495                                 err = 0;
1496                                 break;
1497                         }
1498                 }
1499                 read_unlock_bh(&idev->lock);
1500         }
1501         rcu_read_unlock();
1502         return err;
1503 }
1504
1505 static int ipv6_count_addresses(struct inet6_dev *idev)
1506 {
1507         int cnt = 0;
1508         struct inet6_ifaddr *ifp;
1509
1510         read_lock_bh(&idev->lock);
1511         list_for_each_entry(ifp, &idev->addr_list, if_list)
1512                 cnt++;
1513         read_unlock_bh(&idev->lock);
1514         return cnt;
1515 }
1516
1517 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1518                   const struct net_device *dev, int strict)
1519 {
1520         struct inet6_ifaddr *ifp;
1521         unsigned int hash = inet6_addr_hash(addr);
1522
1523         rcu_read_lock_bh();
1524         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1525                 if (!net_eq(dev_net(ifp->idev->dev), net))
1526                         continue;
1527                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1528                     (!(ifp->flags&IFA_F_TENTATIVE) ||
1529                      (ipv6_use_optimistic_addr(ifp->idev) &&
1530                       ifp->flags&IFA_F_OPTIMISTIC)) &&
1531                     (dev == NULL || ifp->idev->dev == dev ||
1532                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1533                         rcu_read_unlock_bh();
1534                         return 1;
1535                 }
1536         }
1537
1538         rcu_read_unlock_bh();
1539         return 0;
1540 }
1541 EXPORT_SYMBOL(ipv6_chk_addr);
1542
1543 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1544                                struct net_device *dev)
1545 {
1546         unsigned int hash = inet6_addr_hash(addr);
1547         struct inet6_ifaddr *ifp;
1548
1549         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1550                 if (!net_eq(dev_net(ifp->idev->dev), net))
1551                         continue;
1552                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1553                         if (dev == NULL || ifp->idev->dev == dev)
1554                                 return true;
1555                 }
1556         }
1557         return false;
1558 }
1559
1560 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1561 {
1562         struct inet6_dev *idev;
1563         struct inet6_ifaddr *ifa;
1564         int     onlink;
1565
1566         onlink = 0;
1567         rcu_read_lock();
1568         idev = __in6_dev_get(dev);
1569         if (idev) {
1570                 read_lock_bh(&idev->lock);
1571                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1572                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1573                                                    ifa->prefix_len);
1574                         if (onlink)
1575                                 break;
1576                 }
1577                 read_unlock_bh(&idev->lock);
1578         }
1579         rcu_read_unlock();
1580         return onlink;
1581 }
1582 EXPORT_SYMBOL(ipv6_chk_prefix);
1583
1584 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1585                                      struct net_device *dev, int strict)
1586 {
1587         struct inet6_ifaddr *ifp, *result = NULL;
1588         unsigned int hash = inet6_addr_hash(addr);
1589
1590         rcu_read_lock_bh();
1591         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1592                 if (!net_eq(dev_net(ifp->idev->dev), net))
1593                         continue;
1594                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1595                         if (dev == NULL || ifp->idev->dev == dev ||
1596                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1597                                 result = ifp;
1598                                 in6_ifa_hold(ifp);
1599                                 break;
1600                         }
1601                 }
1602         }
1603         rcu_read_unlock_bh();
1604
1605         return result;
1606 }
1607
1608 /* Gets referenced address, destroys ifaddr */
1609
1610 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1611 {
1612         if (ifp->flags&IFA_F_PERMANENT) {
1613                 spin_lock_bh(&ifp->lock);
1614                 addrconf_del_timer(ifp);
1615                 ifp->flags |= IFA_F_TENTATIVE;
1616                 if (dad_failed)
1617                         ifp->flags |= IFA_F_DADFAILED;
1618                 spin_unlock_bh(&ifp->lock);
1619                 if (dad_failed)
1620                         ipv6_ifa_notify(0, ifp);
1621                 in6_ifa_put(ifp);
1622 #ifdef CONFIG_IPV6_PRIVACY
1623         } else if (ifp->flags&IFA_F_TEMPORARY) {
1624                 struct inet6_ifaddr *ifpub;
1625                 spin_lock_bh(&ifp->lock);
1626                 ifpub = ifp->ifpub;
1627                 if (ifpub) {
1628                         in6_ifa_hold(ifpub);
1629                         spin_unlock_bh(&ifp->lock);
1630                         ipv6_create_tempaddr(ifpub, ifp);
1631                         in6_ifa_put(ifpub);
1632                 } else {
1633                         spin_unlock_bh(&ifp->lock);
1634                 }
1635                 ipv6_del_addr(ifp);
1636 #endif
1637         } else
1638                 ipv6_del_addr(ifp);
1639 }
1640
1641 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1642 {
1643         int err = -ENOENT;
1644
1645         spin_lock(&ifp->state_lock);
1646         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1647                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1648                 err = 0;
1649         }
1650         spin_unlock(&ifp->state_lock);
1651
1652         return err;
1653 }
1654
1655 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1656 {
1657         struct inet6_dev *idev = ifp->idev;
1658
1659         if (addrconf_dad_end(ifp)) {
1660                 in6_ifa_put(ifp);
1661                 return;
1662         }
1663
1664         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1665                              ifp->idev->dev->name, &ifp->addr);
1666
1667         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1668                 struct in6_addr addr;
1669
1670                 addr.s6_addr32[0] = htonl(0xfe800000);
1671                 addr.s6_addr32[1] = 0;
1672
1673                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1674                     ipv6_addr_equal(&ifp->addr, &addr)) {
1675                         /* DAD failed for link-local based on MAC address */
1676                         idev->cnf.disable_ipv6 = 1;
1677
1678                         pr_info("%s: IPv6 being disabled!\n",
1679                                 ifp->idev->dev->name);
1680                 }
1681         }
1682
1683         addrconf_dad_stop(ifp, 1);
1684 }
1685
1686 /* Join to solicited addr multicast group. */
1687
1688 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1689 {
1690         struct in6_addr maddr;
1691
1692         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1693                 return;
1694
1695         addrconf_addr_solict_mult(addr, &maddr);
1696         ipv6_dev_mc_inc(dev, &maddr);
1697 }
1698
1699 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1700 {
1701         struct in6_addr maddr;
1702
1703         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1704                 return;
1705
1706         addrconf_addr_solict_mult(addr, &maddr);
1707         __ipv6_dev_mc_dec(idev, &maddr);
1708 }
1709
1710 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1711 {
1712         struct in6_addr addr;
1713         if (ifp->prefix_len == 127) /* RFC 6164 */
1714                 return;
1715         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1716         if (ipv6_addr_any(&addr))
1717                 return;
1718         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1719 }
1720
1721 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1722 {
1723         struct in6_addr addr;
1724         if (ifp->prefix_len == 127) /* RFC 6164 */
1725                 return;
1726         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1727         if (ipv6_addr_any(&addr))
1728                 return;
1729         __ipv6_dev_ac_dec(ifp->idev, &addr);
1730 }
1731
1732 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1733 {
1734         if (dev->addr_len != ETH_ALEN)
1735                 return -1;
1736         memcpy(eui, dev->dev_addr, 3);
1737         memcpy(eui + 5, dev->dev_addr + 3, 3);
1738
1739         /*
1740          * The zSeries OSA network cards can be shared among various
1741          * OS instances, but the OSA cards have only one MAC address.
1742          * This leads to duplicate address conflicts in conjunction
1743          * with IPv6 if more than one instance uses the same card.
1744          *
1745          * The driver for these cards can deliver a unique 16-bit
1746          * identifier for each instance sharing the same card.  It is
1747          * placed instead of 0xFFFE in the interface identifier.  The
1748          * "u" bit of the interface identifier is not inverted in this
1749          * case.  Hence the resulting interface identifier has local
1750          * scope according to RFC2373.
1751          */
1752         if (dev->dev_id) {
1753                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1754                 eui[4] = dev->dev_id & 0xFF;
1755         } else {
1756                 eui[3] = 0xFF;
1757                 eui[4] = 0xFE;
1758                 eui[0] ^= 2;
1759         }
1760         return 0;
1761 }
1762
1763 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1764 {
1765         if (dev->addr_len != IEEE802154_ADDR_LEN)
1766                 return -1;
1767         memcpy(eui, dev->dev_addr, 8);
1768         eui[0] ^= 2;
1769         return 0;
1770 }
1771
1772 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1773 {
1774         union fwnet_hwaddr *ha;
1775
1776         if (dev->addr_len != FWNET_ALEN)
1777                 return -1;
1778
1779         ha = (union fwnet_hwaddr *)dev->dev_addr;
1780
1781         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1782         eui[0] ^= 2;
1783         return 0;
1784 }
1785
1786 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1787 {
1788         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1789         if (dev->addr_len != ARCNET_ALEN)
1790                 return -1;
1791         memset(eui, 0, 7);
1792         eui[7] = *(u8 *)dev->dev_addr;
1793         return 0;
1794 }
1795
1796 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1797 {
1798         if (dev->addr_len != INFINIBAND_ALEN)
1799                 return -1;
1800         memcpy(eui, dev->dev_addr + 12, 8);
1801         eui[0] |= 2;
1802         return 0;
1803 }
1804
1805 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1806 {
1807         if (addr == 0)
1808                 return -1;
1809         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1810                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1811                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1812                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1813                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1814                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1815         eui[1] = 0;
1816         eui[2] = 0x5E;
1817         eui[3] = 0xFE;
1818         memcpy(eui + 4, &addr, 4);
1819         return 0;
1820 }
1821
1822 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1823 {
1824         if (dev->priv_flags & IFF_ISATAP)
1825                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1826         return -1;
1827 }
1828
1829 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1830 {
1831         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1832 }
1833
1834 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1835 {
1836         switch (dev->type) {
1837         case ARPHRD_ETHER:
1838         case ARPHRD_FDDI:
1839                 return addrconf_ifid_eui48(eui, dev);
1840         case ARPHRD_ARCNET:
1841                 return addrconf_ifid_arcnet(eui, dev);
1842         case ARPHRD_INFINIBAND:
1843                 return addrconf_ifid_infiniband(eui, dev);
1844         case ARPHRD_SIT:
1845                 return addrconf_ifid_sit(eui, dev);
1846         case ARPHRD_IPGRE:
1847                 return addrconf_ifid_gre(eui, dev);
1848         case ARPHRD_IEEE802154:
1849                 return addrconf_ifid_eui64(eui, dev);
1850         case ARPHRD_IEEE1394:
1851                 return addrconf_ifid_ieee1394(eui, dev);
1852         }
1853         return -1;
1854 }
1855
1856 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1857 {
1858         int err = -1;
1859         struct inet6_ifaddr *ifp;
1860
1861         read_lock_bh(&idev->lock);
1862         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1863                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1864                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1865                         err = 0;
1866                         break;
1867                 }
1868         }
1869         read_unlock_bh(&idev->lock);
1870         return err;
1871 }
1872
1873 #ifdef CONFIG_IPV6_PRIVACY
1874 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1875 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1876 {
1877 regen:
1878         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1879         idev->rndid[0] &= ~0x02;
1880
1881         /*
1882          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1883          * check if generated address is not inappropriate
1884          *
1885          *  - Reserved subnet anycast (RFC 2526)
1886          *      11111101 11....11 1xxxxxxx
1887          *  - ISATAP (RFC4214) 6.1
1888          *      00-00-5E-FE-xx-xx-xx-xx
1889          *  - value 0
1890          *  - XXX: already assigned to an address on the device
1891          */
1892         if (idev->rndid[0] == 0xfd &&
1893             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1894             (idev->rndid[7]&0x80))
1895                 goto regen;
1896         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1897                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1898                         goto regen;
1899                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1900                         goto regen;
1901         }
1902 }
1903
1904 static void ipv6_regen_rndid(unsigned long data)
1905 {
1906         struct inet6_dev *idev = (struct inet6_dev *) data;
1907         unsigned long expires;
1908
1909         rcu_read_lock_bh();
1910         write_lock_bh(&idev->lock);
1911
1912         if (idev->dead)
1913                 goto out;
1914
1915         __ipv6_regen_rndid(idev);
1916
1917         expires = jiffies +
1918                 idev->cnf.temp_prefered_lft * HZ -
1919                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1920                 idev->cnf.max_desync_factor * HZ;
1921         if (time_before(expires, jiffies)) {
1922                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1923                         __func__, idev->dev->name);
1924                 goto out;
1925         }
1926
1927         if (!mod_timer(&idev->regen_timer, expires))
1928                 in6_dev_hold(idev);
1929
1930 out:
1931         write_unlock_bh(&idev->lock);
1932         rcu_read_unlock_bh();
1933         in6_dev_put(idev);
1934 }
1935
1936 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1937 {
1938         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1939                 __ipv6_regen_rndid(idev);
1940 }
1941 #endif
1942
1943 u32 addrconf_rt_table(const struct net_device *dev, u32 default_table) {
1944         /* Determines into what table to put autoconf PIO/RIO/default routes
1945          * learned on this device.
1946          *
1947          * - If 0, use the same table for every device. This puts routes into
1948          *   one of RT_TABLE_{PREFIX,INFO,DFLT} depending on the type of route
1949          *   (but note that these three are currently all equal to
1950          *   RT6_TABLE_MAIN).
1951          * - If > 0, use the specified table.
1952          * - If < 0, put routes into table dev->ifindex + (-rt_table).
1953          */
1954         struct inet6_dev *idev = in6_dev_get(dev);
1955         u32 table;
1956         int sysctl = idev->cnf.accept_ra_rt_table;
1957         if (sysctl == 0) {
1958                 table = default_table;
1959         } else if (sysctl > 0) {
1960                 table = (u32) sysctl;
1961         } else {
1962                 table = (unsigned) dev->ifindex + (-sysctl);
1963         }
1964         in6_dev_put(idev);
1965         return table;
1966 }
1967
1968 /*
1969  *      Add prefix route.
1970  */
1971
1972 static void
1973 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1974                       unsigned long expires, u32 flags)
1975 {
1976         struct fib6_config cfg = {
1977                 .fc_table = addrconf_rt_table(dev, RT6_TABLE_PREFIX),
1978                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1979                 .fc_ifindex = dev->ifindex,
1980                 .fc_expires = expires,
1981                 .fc_dst_len = plen,
1982                 .fc_flags = RTF_UP | flags,
1983                 .fc_nlinfo.nl_net = dev_net(dev),
1984                 .fc_protocol = RTPROT_KERNEL,
1985         };
1986
1987         cfg.fc_dst = *pfx;
1988
1989         /* Prevent useless cloning on PtP SIT.
1990            This thing is done here expecting that the whole
1991            class of non-broadcast devices need not cloning.
1992          */
1993 #if IS_ENABLED(CONFIG_IPV6_SIT)
1994         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1995                 cfg.fc_flags |= RTF_NONEXTHOP;
1996 #endif
1997
1998         ip6_route_add(&cfg);
1999 }
2000
2001
2002 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2003                                                   int plen,
2004                                                   const struct net_device *dev,
2005                                                   u32 flags, u32 noflags)
2006 {
2007         struct fib6_node *fn;
2008         struct rt6_info *rt = NULL;
2009         struct fib6_table *table;
2010
2011         table = fib6_get_table(dev_net(dev),
2012                                addrconf_rt_table(dev, RT6_TABLE_PREFIX));
2013         if (table == NULL)
2014                 return NULL;
2015
2016         read_lock_bh(&table->tb6_lock);
2017         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2018         if (!fn)
2019                 goto out;
2020         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2021                 if (rt->dst.dev->ifindex != dev->ifindex)
2022                         continue;
2023                 if ((rt->rt6i_flags & flags) != flags)
2024                         continue;
2025                 if ((rt->rt6i_flags & noflags) != 0)
2026                         continue;
2027                 dst_hold(&rt->dst);
2028                 break;
2029         }
2030 out:
2031         read_unlock_bh(&table->tb6_lock);
2032         return rt;
2033 }
2034
2035
2036 /* Create "default" multicast route to the interface */
2037
2038 static void addrconf_add_mroute(struct net_device *dev)
2039 {
2040         struct fib6_config cfg = {
2041                 .fc_table = RT6_TABLE_LOCAL,
2042                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2043                 .fc_ifindex = dev->ifindex,
2044                 .fc_dst_len = 8,
2045                 .fc_flags = RTF_UP,
2046                 .fc_nlinfo.nl_net = dev_net(dev),
2047         };
2048
2049         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2050
2051         ip6_route_add(&cfg);
2052 }
2053
2054 #if IS_ENABLED(CONFIG_IPV6_SIT)
2055 static void sit_route_add(struct net_device *dev)
2056 {
2057         struct fib6_config cfg = {
2058                 .fc_table = RT6_TABLE_MAIN,
2059                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2060                 .fc_ifindex = dev->ifindex,
2061                 .fc_dst_len = 96,
2062                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2063                 .fc_nlinfo.nl_net = dev_net(dev),
2064         };
2065
2066         /* prefix length - 96 bits "::d.d.d.d" */
2067         ip6_route_add(&cfg);
2068 }
2069 #endif
2070
2071 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2072 {
2073         struct inet6_dev *idev;
2074
2075         ASSERT_RTNL();
2076
2077         idev = ipv6_find_idev(dev);
2078         if (!idev)
2079                 return ERR_PTR(-ENOBUFS);
2080
2081         if (idev->cnf.disable_ipv6)
2082                 return ERR_PTR(-EACCES);
2083
2084         /* Add default multicast route */
2085         if (!(dev->flags & IFF_LOOPBACK))
2086                 addrconf_add_mroute(dev);
2087
2088         return idev;
2089 }
2090
2091 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2092 {
2093         struct prefix_info *pinfo;
2094         __u32 valid_lft;
2095         __u32 prefered_lft;
2096         int addr_type;
2097         struct inet6_dev *in6_dev;
2098         struct net *net = dev_net(dev);
2099
2100         pinfo = (struct prefix_info *) opt;
2101
2102         if (len < sizeof(struct prefix_info)) {
2103                 ADBG(("addrconf: prefix option too short\n"));
2104                 return;
2105         }
2106
2107         /*
2108          *      Validation checks ([ADDRCONF], page 19)
2109          */
2110
2111         addr_type = ipv6_addr_type(&pinfo->prefix);
2112
2113         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2114                 return;
2115
2116         valid_lft = ntohl(pinfo->valid);
2117         prefered_lft = ntohl(pinfo->prefered);
2118
2119         if (prefered_lft > valid_lft) {
2120                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2121                 return;
2122         }
2123
2124         in6_dev = in6_dev_get(dev);
2125
2126         if (in6_dev == NULL) {
2127                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2128                                     dev->name);
2129                 return;
2130         }
2131
2132         /*
2133          *      Two things going on here:
2134          *      1) Add routes for on-link prefixes
2135          *      2) Configure prefixes with the auto flag set
2136          */
2137
2138         if (pinfo->onlink) {
2139                 struct rt6_info *rt;
2140                 unsigned long rt_expires;
2141
2142                 /* Avoid arithmetic overflow. Really, we could
2143                  * save rt_expires in seconds, likely valid_lft,
2144                  * but it would require division in fib gc, that it
2145                  * not good.
2146                  */
2147                 if (HZ > USER_HZ)
2148                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2149                 else
2150                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2151
2152                 if (addrconf_finite_timeout(rt_expires))
2153                         rt_expires *= HZ;
2154
2155                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2156                                                pinfo->prefix_len,
2157                                                dev,
2158                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2159                                                RTF_GATEWAY | RTF_DEFAULT);
2160
2161                 if (rt) {
2162                         /* Autoconf prefix route */
2163                         if (valid_lft == 0) {
2164                                 ip6_del_rt(rt);
2165                                 rt = NULL;
2166                         } else if (addrconf_finite_timeout(rt_expires)) {
2167                                 /* not infinity */
2168                                 rt6_set_expires(rt, jiffies + rt_expires);
2169                         } else {
2170                                 rt6_clean_expires(rt);
2171                         }
2172                 } else if (valid_lft) {
2173                         clock_t expires = 0;
2174                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2175                         if (addrconf_finite_timeout(rt_expires)) {
2176                                 /* not infinity */
2177                                 flags |= RTF_EXPIRES;
2178                                 expires = jiffies_to_clock_t(rt_expires);
2179                         }
2180                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2181                                               dev, expires, flags);
2182                 }
2183                 ip6_rt_put(rt);
2184         }
2185
2186         /* Try to figure out our local address for this prefix */
2187
2188         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2189                 struct inet6_ifaddr *ifp;
2190                 struct in6_addr addr;
2191                 int create = 0, update_lft = 0;
2192                 bool tokenized = false;
2193
2194                 if (pinfo->prefix_len == 64) {
2195                         memcpy(&addr, &pinfo->prefix, 8);
2196
2197                         if (!ipv6_addr_any(&in6_dev->token)) {
2198                                 read_lock_bh(&in6_dev->lock);
2199                                 memcpy(addr.s6_addr + 8,
2200                                        in6_dev->token.s6_addr + 8, 8);
2201                                 read_unlock_bh(&in6_dev->lock);
2202                                 tokenized = true;
2203                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2204                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2205                                 in6_dev_put(in6_dev);
2206                                 return;
2207                         }
2208                         goto ok;
2209                 }
2210                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2211                                     pinfo->prefix_len);
2212                 in6_dev_put(in6_dev);
2213                 return;
2214
2215 ok:
2216
2217                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2218
2219                 if (ifp == NULL && valid_lft) {
2220                         int max_addresses = in6_dev->cnf.max_addresses;
2221                         u32 addr_flags = 0;
2222
2223 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2224                         if (in6_dev->cnf.optimistic_dad &&
2225                             !net->ipv6.devconf_all->forwarding && sllao)
2226                                 addr_flags = IFA_F_OPTIMISTIC;
2227 #endif
2228
2229                         /* Do not allow to create too much of autoconfigured
2230                          * addresses; this would be too easy way to crash kernel.
2231                          */
2232                         if (!max_addresses ||
2233                             ipv6_count_addresses(in6_dev) < max_addresses)
2234                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2235                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2236                                                     addr_flags);
2237
2238                         if (IS_ERR_OR_NULL(ifp)) {
2239                                 in6_dev_put(in6_dev);
2240                                 return;
2241                         }
2242
2243                         update_lft = create = 1;
2244                         ifp->cstamp = jiffies;
2245                         ifp->tokenized = tokenized;
2246                         addrconf_dad_start(ifp);
2247                 }
2248
2249                 if (ifp) {
2250                         int flags;
2251                         unsigned long now;
2252 #ifdef CONFIG_IPV6_PRIVACY
2253                         struct inet6_ifaddr *ift;
2254 #endif
2255                         u32 stored_lft;
2256
2257                         /* update lifetime (RFC2462 5.5.3 e) */
2258                         spin_lock(&ifp->lock);
2259                         now = jiffies;
2260                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2261                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2262                         else
2263                                 stored_lft = 0;
2264                         if (!update_lft && stored_lft) {
2265                                 if (valid_lft > MIN_VALID_LIFETIME ||
2266                                     valid_lft > stored_lft)
2267                                         update_lft = 1;
2268                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2269                                         /* valid_lft <= stored_lft is always true */
2270                                         /*
2271                                          * RFC 4862 Section 5.5.3e:
2272                                          * "Note that the preferred lifetime of
2273                                          *  the corresponding address is always
2274                                          *  reset to the Preferred Lifetime in
2275                                          *  the received Prefix Information
2276                                          *  option, regardless of whether the
2277                                          *  valid lifetime is also reset or
2278                                          *  ignored."
2279                                          *
2280                                          *  So if the preferred lifetime in
2281                                          *  this advertisement is different
2282                                          *  than what we have stored, but the
2283                                          *  valid lifetime is invalid, just
2284                                          *  reset prefered_lft.
2285                                          *
2286                                          *  We must set the valid lifetime
2287                                          *  to the stored lifetime since we'll
2288                                          *  be updating the timestamp below,
2289                                          *  else we'll set it back to the
2290                                          *  minimum.
2291                                          */
2292                                         if (prefered_lft != ifp->prefered_lft) {
2293                                                 valid_lft = stored_lft;
2294                                                 update_lft = 1;
2295                                         }
2296                                 } else {
2297                                         valid_lft = MIN_VALID_LIFETIME;
2298                                         if (valid_lft < prefered_lft)
2299                                                 prefered_lft = valid_lft;
2300                                         update_lft = 1;
2301                                 }
2302                         }
2303
2304                         if (update_lft) {
2305                                 ifp->valid_lft = valid_lft;
2306                                 ifp->prefered_lft = prefered_lft;
2307                                 ifp->tstamp = now;
2308                                 flags = ifp->flags;
2309                                 ifp->flags &= ~IFA_F_DEPRECATED;
2310                                 spin_unlock(&ifp->lock);
2311
2312                                 if (!(flags&IFA_F_TENTATIVE))
2313                                         ipv6_ifa_notify(0, ifp);
2314                         } else
2315                                 spin_unlock(&ifp->lock);
2316
2317 #ifdef CONFIG_IPV6_PRIVACY
2318                         read_lock_bh(&in6_dev->lock);
2319                         /* update all temporary addresses in the list */
2320                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2321                                             tmp_list) {
2322                                 int age, max_valid, max_prefered;
2323
2324                                 if (ifp != ift->ifpub)
2325                                         continue;
2326
2327                                 /*
2328                                  * RFC 4941 section 3.3:
2329                                  * If a received option will extend the lifetime
2330                                  * of a public address, the lifetimes of
2331                                  * temporary addresses should be extended,
2332                                  * subject to the overall constraint that no
2333                                  * temporary addresses should ever remain
2334                                  * "valid" or "preferred" for a time longer than
2335                                  * (TEMP_VALID_LIFETIME) or
2336                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2337                                  * respectively.
2338                                  */
2339                                 age = (now - ift->cstamp) / HZ;
2340                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2341                                 if (max_valid < 0)
2342                                         max_valid = 0;
2343
2344                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2345                                                in6_dev->cnf.max_desync_factor -
2346                                                age;
2347                                 if (max_prefered < 0)
2348                                         max_prefered = 0;
2349
2350                                 if (valid_lft > max_valid)
2351                                         valid_lft = max_valid;
2352
2353                                 if (prefered_lft > max_prefered)
2354                                         prefered_lft = max_prefered;
2355
2356                                 spin_lock(&ift->lock);
2357                                 flags = ift->flags;
2358                                 ift->valid_lft = valid_lft;
2359                                 ift->prefered_lft = prefered_lft;
2360                                 ift->tstamp = now;
2361                                 if (prefered_lft > 0)
2362                                         ift->flags &= ~IFA_F_DEPRECATED;
2363
2364                                 spin_unlock(&ift->lock);
2365                                 if (!(flags&IFA_F_TENTATIVE))
2366                                         ipv6_ifa_notify(0, ift);
2367                         }
2368
2369                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2370                                 /*
2371                                  * When a new public address is created as
2372                                  * described in [ADDRCONF], also create a new
2373                                  * temporary address. Also create a temporary
2374                                  * address if it's enabled but no temporary
2375                                  * address currently exists.
2376                                  */
2377                                 read_unlock_bh(&in6_dev->lock);
2378                                 ipv6_create_tempaddr(ifp, NULL);
2379                         } else {
2380                                 read_unlock_bh(&in6_dev->lock);
2381                         }
2382 #endif
2383                         in6_ifa_put(ifp);
2384                         addrconf_verify(0);
2385                 }
2386         }
2387         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2388         in6_dev_put(in6_dev);
2389 }
2390
2391 /*
2392  *      Set destination address.
2393  *      Special case for SIT interfaces where we create a new "virtual"
2394  *      device.
2395  */
2396 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2397 {
2398         struct in6_ifreq ireq;
2399         struct net_device *dev;
2400         int err = -EINVAL;
2401
2402         rtnl_lock();
2403
2404         err = -EFAULT;
2405         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2406                 goto err_exit;
2407
2408         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2409
2410         err = -ENODEV;
2411         if (dev == NULL)
2412                 goto err_exit;
2413
2414 #if IS_ENABLED(CONFIG_IPV6_SIT)
2415         if (dev->type == ARPHRD_SIT) {
2416                 const struct net_device_ops *ops = dev->netdev_ops;
2417                 struct ifreq ifr;
2418                 struct ip_tunnel_parm p;
2419
2420                 err = -EADDRNOTAVAIL;
2421                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2422                         goto err_exit;
2423
2424                 memset(&p, 0, sizeof(p));
2425                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2426                 p.iph.saddr = 0;
2427                 p.iph.version = 4;
2428                 p.iph.ihl = 5;
2429                 p.iph.protocol = IPPROTO_IPV6;
2430                 p.iph.ttl = 64;
2431                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2432
2433                 if (ops->ndo_do_ioctl) {
2434                         mm_segment_t oldfs = get_fs();
2435
2436                         set_fs(KERNEL_DS);
2437                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2438                         set_fs(oldfs);
2439                 } else
2440                         err = -EOPNOTSUPP;
2441
2442                 if (err == 0) {
2443                         err = -ENOBUFS;
2444                         dev = __dev_get_by_name(net, p.name);
2445                         if (!dev)
2446                                 goto err_exit;
2447                         err = dev_open(dev);
2448                 }
2449         }
2450 #endif
2451
2452 err_exit:
2453         rtnl_unlock();
2454         return err;
2455 }
2456
2457 /*
2458  *      Manual configuration of address on an interface
2459  */
2460 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2461                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2462                           __u32 valid_lft)
2463 {
2464         struct inet6_ifaddr *ifp;
2465         struct inet6_dev *idev;
2466         struct net_device *dev;
2467         int scope;
2468         u32 flags;
2469         clock_t expires;
2470         unsigned long timeout;
2471
2472         ASSERT_RTNL();
2473
2474         if (plen > 128)
2475                 return -EINVAL;
2476
2477         /* check the lifetime */
2478         if (!valid_lft || prefered_lft > valid_lft)
2479                 return -EINVAL;
2480
2481         dev = __dev_get_by_index(net, ifindex);
2482         if (!dev)
2483                 return -ENODEV;
2484
2485         idev = addrconf_add_dev(dev);
2486         if (IS_ERR(idev))
2487                 return PTR_ERR(idev);
2488
2489         scope = ipv6_addr_scope(pfx);
2490
2491         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2492         if (addrconf_finite_timeout(timeout)) {
2493                 expires = jiffies_to_clock_t(timeout * HZ);
2494                 valid_lft = timeout;
2495                 flags = RTF_EXPIRES;
2496         } else {
2497                 expires = 0;
2498                 flags = 0;
2499                 ifa_flags |= IFA_F_PERMANENT;
2500         }
2501
2502         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2503         if (addrconf_finite_timeout(timeout)) {
2504                 if (timeout == 0)
2505                         ifa_flags |= IFA_F_DEPRECATED;
2506                 prefered_lft = timeout;
2507         }
2508
2509         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2510
2511         if (!IS_ERR(ifp)) {
2512                 spin_lock_bh(&ifp->lock);
2513                 ifp->valid_lft = valid_lft;
2514                 ifp->prefered_lft = prefered_lft;
2515                 ifp->tstamp = jiffies;
2516                 spin_unlock_bh(&ifp->lock);
2517
2518                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2519                                       expires, flags);
2520                 /*
2521                  * Note that section 3.1 of RFC 4429 indicates
2522                  * that the Optimistic flag should not be set for
2523                  * manually configured addresses
2524                  */
2525                 addrconf_dad_start(ifp);
2526                 in6_ifa_put(ifp);
2527                 addrconf_verify(0);
2528                 return 0;
2529         }
2530
2531         return PTR_ERR(ifp);
2532 }
2533
2534 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2535                           unsigned int plen)
2536 {
2537         struct inet6_ifaddr *ifp;
2538         struct inet6_dev *idev;
2539         struct net_device *dev;
2540
2541         if (plen > 128)
2542                 return -EINVAL;
2543
2544         dev = __dev_get_by_index(net, ifindex);
2545         if (!dev)
2546                 return -ENODEV;
2547
2548         if ((idev = __in6_dev_get(dev)) == NULL)
2549                 return -ENXIO;
2550
2551         read_lock_bh(&idev->lock);
2552         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2553                 if (ifp->prefix_len == plen &&
2554                     ipv6_addr_equal(pfx, &ifp->addr)) {
2555                         in6_ifa_hold(ifp);
2556                         read_unlock_bh(&idev->lock);
2557
2558                         ipv6_del_addr(ifp);
2559
2560                         /* If the last address is deleted administratively,
2561                            disable IPv6 on this interface.
2562                          */
2563                         if (list_empty(&idev->addr_list))
2564                                 addrconf_ifdown(idev->dev, 1);
2565                         return 0;
2566                 }
2567         }
2568         read_unlock_bh(&idev->lock);
2569         return -EADDRNOTAVAIL;
2570 }
2571
2572
2573 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2574 {
2575         struct in6_ifreq ireq;
2576         int err;
2577
2578         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2579                 return -EPERM;
2580
2581         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2582                 return -EFAULT;
2583
2584         rtnl_lock();
2585         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2586                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2587                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2588         rtnl_unlock();
2589         return err;
2590 }
2591
2592 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2593 {
2594         struct in6_ifreq ireq;
2595         int err;
2596
2597         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2598                 return -EPERM;
2599
2600         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2601                 return -EFAULT;
2602
2603         rtnl_lock();
2604         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2605                              ireq.ifr6_prefixlen);
2606         rtnl_unlock();
2607         return err;
2608 }
2609
2610 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2611                      int plen, int scope)
2612 {
2613         struct inet6_ifaddr *ifp;
2614
2615         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2616         if (!IS_ERR(ifp)) {
2617                 spin_lock_bh(&ifp->lock);
2618                 ifp->flags &= ~IFA_F_TENTATIVE;
2619                 spin_unlock_bh(&ifp->lock);
2620                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2621                 in6_ifa_put(ifp);
2622         }
2623 }
2624
2625 #if IS_ENABLED(CONFIG_IPV6_SIT)
2626 static void sit_add_v4_addrs(struct inet6_dev *idev)
2627 {
2628         struct in6_addr addr;
2629         struct net_device *dev;
2630         struct net *net = dev_net(idev->dev);
2631         int scope;
2632
2633         ASSERT_RTNL();
2634
2635         memset(&addr, 0, sizeof(struct in6_addr));
2636         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2637
2638         if (idev->dev->flags&IFF_POINTOPOINT) {
2639                 addr.s6_addr32[0] = htonl(0xfe800000);
2640                 scope = IFA_LINK;
2641         } else {
2642                 scope = IPV6_ADDR_COMPATv4;
2643         }
2644
2645         if (addr.s6_addr32[3]) {
2646                 add_addr(idev, &addr, 128, scope);
2647                 return;
2648         }
2649
2650         for_each_netdev(net, dev) {
2651                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2652                 if (in_dev && (dev->flags & IFF_UP)) {
2653                         struct in_ifaddr *ifa;
2654
2655                         int flag = scope;
2656
2657                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2658                                 int plen;
2659
2660                                 addr.s6_addr32[3] = ifa->ifa_local;
2661
2662                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2663                                         continue;
2664                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2665                                         if (idev->dev->flags&IFF_POINTOPOINT)
2666                                                 continue;
2667                                         flag |= IFA_HOST;
2668                                 }
2669                                 if (idev->dev->flags&IFF_POINTOPOINT)
2670                                         plen = 64;
2671                                 else
2672                                         plen = 96;
2673
2674                                 add_addr(idev, &addr, plen, flag);
2675                         }
2676                 }
2677         }
2678 }
2679 #endif
2680
2681 static void init_loopback(struct net_device *dev)
2682 {
2683         struct inet6_dev  *idev;
2684         struct net_device *sp_dev;
2685         struct inet6_ifaddr *sp_ifa;
2686         struct rt6_info *sp_rt;
2687
2688         /* ::1 */
2689
2690         ASSERT_RTNL();
2691
2692         if ((idev = ipv6_find_idev(dev)) == NULL) {
2693                 pr_debug("%s: add_dev failed\n", __func__);
2694                 return;
2695         }
2696
2697         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2698
2699         /* Add routes to other interface's IPv6 addresses */
2700         for_each_netdev(dev_net(dev), sp_dev) {
2701                 if (!strcmp(sp_dev->name, dev->name))
2702                         continue;
2703
2704                 idev = __in6_dev_get(sp_dev);
2705                 if (!idev)
2706                         continue;
2707
2708                 read_lock_bh(&idev->lock);
2709                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2710
2711                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2712                                 continue;
2713
2714                         if (sp_ifa->rt)
2715                                 continue;
2716
2717                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2718
2719                         /* Failure cases are ignored */
2720                         if (!IS_ERR(sp_rt)) {
2721                                 sp_ifa->rt = sp_rt;
2722                                 ip6_ins_rt(sp_rt);
2723                         }
2724                 }
2725                 read_unlock_bh(&idev->lock);
2726         }
2727 }
2728
2729 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2730 {
2731         struct inet6_ifaddr *ifp;
2732         u32 addr_flags = IFA_F_PERMANENT;
2733
2734 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2735         if (idev->cnf.optimistic_dad &&
2736             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2737                 addr_flags |= IFA_F_OPTIMISTIC;
2738 #endif
2739
2740
2741         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2742         if (!IS_ERR(ifp)) {
2743                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2744                 addrconf_dad_start(ifp);
2745                 in6_ifa_put(ifp);
2746         }
2747 }
2748
2749 static void addrconf_dev_config(struct net_device *dev)
2750 {
2751         struct in6_addr addr;
2752         struct inet6_dev *idev;
2753
2754         ASSERT_RTNL();
2755
2756         if ((dev->type != ARPHRD_ETHER) &&
2757             (dev->type != ARPHRD_FDDI) &&
2758             (dev->type != ARPHRD_ARCNET) &&
2759             (dev->type != ARPHRD_INFINIBAND) &&
2760             (dev->type != ARPHRD_IEEE802154) &&
2761             (dev->type != ARPHRD_IEEE1394)) {
2762                 /* Alas, we support only Ethernet autoconfiguration. */
2763                 return;
2764         }
2765
2766         idev = addrconf_add_dev(dev);
2767         if (IS_ERR(idev))
2768                 return;
2769
2770         memset(&addr, 0, sizeof(struct in6_addr));
2771         addr.s6_addr32[0] = htonl(0xFE800000);
2772
2773         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2774                 addrconf_add_linklocal(idev, &addr);
2775 }
2776
2777 #if IS_ENABLED(CONFIG_IPV6_SIT)
2778 static void addrconf_sit_config(struct net_device *dev)
2779 {
2780         struct inet6_dev *idev;
2781
2782         ASSERT_RTNL();
2783
2784         /*
2785          * Configure the tunnel with one of our IPv4
2786          * addresses... we should configure all of
2787          * our v4 addrs in the tunnel
2788          */
2789
2790         if ((idev = ipv6_find_idev(dev)) == NULL) {
2791                 pr_debug("%s: add_dev failed\n", __func__);
2792                 return;
2793         }
2794
2795         if (dev->priv_flags & IFF_ISATAP) {
2796                 struct in6_addr addr;
2797
2798                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2799                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2800                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2801                         addrconf_add_linklocal(idev, &addr);
2802                 return;
2803         }
2804
2805         sit_add_v4_addrs(idev);
2806
2807         if (dev->flags&IFF_POINTOPOINT)
2808                 addrconf_add_mroute(dev);
2809         else
2810                 sit_route_add(dev);
2811 }
2812 #endif
2813
2814 #if IS_ENABLED(CONFIG_NET_IPGRE)
2815 static void addrconf_gre_config(struct net_device *dev)
2816 {
2817         struct inet6_dev *idev;
2818         struct in6_addr addr;
2819
2820         pr_info("%s(%s)\n", __func__, dev->name);
2821
2822         ASSERT_RTNL();
2823
2824         if ((idev = ipv6_find_idev(dev)) == NULL) {
2825                 pr_debug("%s: add_dev failed\n", __func__);
2826                 return;
2827         }
2828
2829         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2830         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2831
2832         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2833                 addrconf_add_linklocal(idev, &addr);
2834 }
2835 #endif
2836
2837 static inline int
2838 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2839 {
2840         struct in6_addr lladdr;
2841
2842         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2843                 addrconf_add_linklocal(idev, &lladdr);
2844                 return 0;
2845         }
2846         return -1;
2847 }
2848
2849 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2850 {
2851         struct net_device *link_dev;
2852         struct net *net = dev_net(idev->dev);
2853
2854         /* first try to inherit the link-local address from the link device */
2855         if (idev->dev->iflink &&
2856             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2857                 if (!ipv6_inherit_linklocal(idev, link_dev))
2858                         return;
2859         }
2860         /* then try to inherit it from any device */
2861         for_each_netdev(net, link_dev) {
2862                 if (!ipv6_inherit_linklocal(idev, link_dev))
2863                         return;
2864         }
2865         pr_debug("init ip6-ip6: add_linklocal failed\n");
2866 }
2867
2868 /*
2869  * Autoconfigure tunnel with a link-local address so routing protocols,
2870  * DHCPv6, MLD etc. can be run over the virtual link
2871  */
2872
2873 static void addrconf_ip6_tnl_config(struct net_device *dev)
2874 {
2875         struct inet6_dev *idev;
2876
2877         ASSERT_RTNL();
2878
2879         idev = addrconf_add_dev(dev);
2880         if (IS_ERR(idev)) {
2881                 pr_debug("init ip6-ip6: add_dev failed\n");
2882                 return;
2883         }
2884         ip6_tnl_add_linklocal(idev);
2885 }
2886
2887 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2888                            void *data)
2889 {
2890         struct net_device *dev = (struct net_device *) data;
2891         struct inet6_dev *idev = __in6_dev_get(dev);
2892         int run_pending = 0;
2893         int err;
2894
2895         switch (event) {
2896         case NETDEV_REGISTER:
2897                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2898                         idev = ipv6_add_dev(dev);
2899                         if (!idev)
2900                                 return notifier_from_errno(-ENOMEM);
2901                 }
2902                 break;
2903
2904         case NETDEV_UP:
2905         case NETDEV_CHANGE:
2906                 if (dev->flags & IFF_SLAVE)
2907                         break;
2908
2909                 if (event == NETDEV_UP) {
2910                         if (!addrconf_qdisc_ok(dev)) {
2911                                 /* device is not ready yet. */
2912                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2913                                         dev->name);
2914                                 break;
2915                         }
2916
2917                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2918                                 idev = ipv6_add_dev(dev);
2919
2920                         if (idev) {
2921                                 idev->if_flags |= IF_READY;
2922                                 run_pending = 1;
2923                         }
2924                 } else {
2925                         if (!addrconf_qdisc_ok(dev)) {
2926                                 /* device is still not ready. */
2927                                 break;
2928                         }
2929
2930                         if (idev) {
2931                                 if (idev->if_flags & IF_READY)
2932                                         /* device is already configured. */
2933                                         break;
2934                                 idev->if_flags |= IF_READY;
2935                         }
2936
2937                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2938                                 dev->name);
2939
2940                         run_pending = 1;
2941                 }
2942
2943                 switch (dev->type) {
2944 #if IS_ENABLED(CONFIG_IPV6_SIT)
2945                 case ARPHRD_SIT:
2946                         addrconf_sit_config(dev);
2947                         break;
2948 #endif
2949 #if IS_ENABLED(CONFIG_NET_IPGRE)
2950                 case ARPHRD_IPGRE:
2951                         addrconf_gre_config(dev);
2952                         break;
2953 #endif
2954                 case ARPHRD_TUNNEL6:
2955                         addrconf_ip6_tnl_config(dev);
2956                         break;
2957                 case ARPHRD_LOOPBACK:
2958                         init_loopback(dev);
2959                         break;
2960
2961                 default:
2962                         addrconf_dev_config(dev);
2963                         break;
2964                 }
2965
2966                 if (idev) {
2967                         if (run_pending)
2968                                 addrconf_dad_run(idev);
2969
2970                         /*
2971                          * If the MTU changed during the interface down,
2972                          * when the interface up, the changed MTU must be
2973                          * reflected in the idev as well as routers.
2974                          */
2975                         if (idev->cnf.mtu6 != dev->mtu &&
2976                             dev->mtu >= IPV6_MIN_MTU) {
2977                                 rt6_mtu_change(dev, dev->mtu);
2978                                 idev->cnf.mtu6 = dev->mtu;
2979                         }
2980                         idev->tstamp = jiffies;
2981                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2982
2983                         /*
2984                          * If the changed mtu during down is lower than
2985                          * IPV6_MIN_MTU stop IPv6 on this interface.
2986                          */
2987                         if (dev->mtu < IPV6_MIN_MTU)
2988                                 addrconf_ifdown(dev, 1);
2989                 }
2990                 break;
2991
2992         case NETDEV_CHANGEMTU:
2993                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2994                         rt6_mtu_change(dev, dev->mtu);
2995                         idev->cnf.mtu6 = dev->mtu;
2996                         break;
2997                 }
2998
2999                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3000                         idev = ipv6_add_dev(dev);
3001                         if (idev)
3002                                 break;
3003                 }
3004
3005                 /*
3006                  * MTU falled under IPV6_MIN_MTU.
3007                  * Stop IPv6 on this interface.
3008                  */
3009
3010         case NETDEV_DOWN:
3011         case NETDEV_UNREGISTER:
3012                 /*
3013                  *      Remove all addresses from this interface.
3014                  */
3015                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3016                 break;
3017
3018         case NETDEV_CHANGENAME:
3019                 if (idev) {
3020                         snmp6_unregister_dev(idev);
3021                         addrconf_sysctl_unregister(idev);
3022                         addrconf_sysctl_register(idev);
3023                         err = snmp6_register_dev(idev);
3024                         if (err)
3025                                 return notifier_from_errno(err);
3026                 }
3027                 break;
3028
3029         case NETDEV_PRE_TYPE_CHANGE:
3030         case NETDEV_POST_TYPE_CHANGE:
3031                 addrconf_type_change(dev, event);
3032                 break;
3033         }
3034
3035         return NOTIFY_OK;
3036 }
3037
3038 /*
3039  *      addrconf module should be notified of a device going up
3040  */
3041 static struct notifier_block ipv6_dev_notf = {
3042         .notifier_call = addrconf_notify,
3043 };
3044
3045 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3046 {
3047         struct inet6_dev *idev;
3048         ASSERT_RTNL();
3049
3050         idev = __in6_dev_get(dev);
3051
3052         if (event == NETDEV_POST_TYPE_CHANGE)
3053                 ipv6_mc_remap(idev);
3054         else if (event == NETDEV_PRE_TYPE_CHANGE)
3055                 ipv6_mc_unmap(idev);
3056 }
3057
3058 static int addrconf_ifdown(struct net_device *dev, int how)
3059 {
3060         struct net *net = dev_net(dev);
3061         struct inet6_dev *idev;
3062         struct inet6_ifaddr *ifa;
3063         int state, i;
3064
3065         ASSERT_RTNL();
3066
3067         rt6_ifdown(net, dev);
3068         neigh_ifdown(&nd_tbl, dev);
3069
3070         idev = __in6_dev_get(dev);
3071         if (idev == NULL)
3072                 return -ENODEV;
3073
3074         /*
3075          * Step 1: remove reference to ipv6 device from parent device.
3076          *         Do not dev_put!
3077          */
3078         if (how) {
3079                 idev->dead = 1;
3080
3081                 /* protected by rtnl_lock */
3082                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3083
3084                 /* Step 1.5: remove snmp6 entry */
3085                 snmp6_unregister_dev(idev);
3086
3087         }
3088
3089         /* Step 2: clear hash table */
3090         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3091                 struct hlist_head *h = &inet6_addr_lst[i];
3092
3093                 spin_lock_bh(&addrconf_hash_lock);
3094         restart:
3095                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3096                         if (ifa->idev == idev) {
3097                                 hlist_del_init_rcu(&ifa->addr_lst);
3098                                 addrconf_del_timer(ifa);
3099                                 goto restart;
3100                         }
3101                 }
3102                 spin_unlock_bh(&addrconf_hash_lock);
3103         }
3104
3105         write_lock_bh(&idev->lock);
3106
3107         /* Step 2: clear flags for stateless addrconf */
3108         if (!how)
3109                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3110
3111 #ifdef CONFIG_IPV6_PRIVACY
3112         if (how && del_timer(&idev->regen_timer))
3113                 in6_dev_put(idev);
3114
3115         /* Step 3: clear tempaddr list */
3116         while (!list_empty(&idev->tempaddr_list)) {
3117                 ifa = list_first_entry(&idev->tempaddr_list,
3118                                        struct inet6_ifaddr, tmp_list);
3119                 list_del(&ifa->tmp_list);
3120                 write_unlock_bh(&idev->lock);
3121                 spin_lock_bh(&ifa->lock);
3122
3123                 if (ifa->ifpub) {
3124                         in6_ifa_put(ifa->ifpub);
3125                         ifa->ifpub = NULL;
3126                 }
3127                 spin_unlock_bh(&ifa->lock);
3128                 in6_ifa_put(ifa);
3129                 write_lock_bh(&idev->lock);
3130         }
3131 #endif
3132
3133         while (!list_empty(&idev->addr_list)) {
3134                 ifa = list_first_entry(&idev->addr_list,
3135                                        struct inet6_ifaddr, if_list);
3136                 addrconf_del_timer(ifa);
3137
3138                 list_del(&ifa->if_list);
3139
3140                 write_unlock_bh(&idev->lock);
3141
3142                 spin_lock_bh(&ifa->state_lock);
3143                 state = ifa->state;
3144                 ifa->state = INET6_IFADDR_STATE_DEAD;
3145                 spin_unlock_bh(&ifa->state_lock);
3146
3147                 if (state != INET6_IFADDR_STATE_DEAD) {
3148                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3149                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3150                 }
3151                 in6_ifa_put(ifa);
3152
3153                 write_lock_bh(&idev->lock);
3154         }
3155
3156         write_unlock_bh(&idev->lock);
3157
3158         /* Step 5: Discard anycast and multicast list */
3159         if (how) {
3160                 ipv6_ac_destroy_dev(idev);
3161                 ipv6_mc_destroy_dev(idev);
3162         } else {
3163                 ipv6_mc_down(idev);
3164         }
3165
3166         idev->tstamp = jiffies;
3167
3168         /* Last: Shot the device (if unregistered) */
3169         if (how) {
3170                 addrconf_sysctl_unregister(idev);
3171                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3172                 neigh_ifdown(&nd_tbl, dev);
3173                 in6_dev_put(idev);
3174         }
3175         return 0;
3176 }
3177
3178 static void addrconf_rs_timer(unsigned long data)
3179 {
3180         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3181         struct inet6_dev *idev = ifp->idev;
3182
3183         read_lock(&idev->lock);
3184         if (idev->dead || !(idev->if_flags & IF_READY))
3185                 goto out;
3186
3187         if (!ipv6_accept_ra(idev))
3188                 goto out;
3189
3190         /* Announcement received after solicitation was sent */
3191         if (idev->if_flags & IF_RA_RCVD)
3192                 goto out;
3193
3194         spin_lock(&ifp->lock);
3195         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3196                 /* The wait after the last probe can be shorter */
3197                 addrconf_mod_timer(ifp, AC_RS,
3198                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3199                                    idev->cnf.rtr_solicit_delay :
3200                                    idev->cnf.rtr_solicit_interval);
3201                 spin_unlock(&ifp->lock);
3202
3203                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3204         } else {
3205                 spin_unlock(&ifp->lock);
3206                 /*
3207                  * Note: we do not support deprecated "all on-link"
3208                  * assumption any longer.
3209                  */
3210                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3211         }
3212
3213 out:
3214         read_unlock(&idev->lock);
3215         in6_ifa_put(ifp);
3216 }
3217
3218 /*
3219  *      Duplicate Address Detection
3220  */
3221 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3222 {
3223         unsigned long rand_num;
3224         struct inet6_dev *idev = ifp->idev;
3225
3226         if (ifp->flags & IFA_F_OPTIMISTIC)
3227                 rand_num = 0;
3228         else
3229                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3230
3231         ifp->probes = idev->cnf.dad_transmits;
3232         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3233 }
3234
3235 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3236 {
3237         struct inet6_dev *idev = ifp->idev;
3238         struct net_device *dev = idev->dev;
3239
3240         addrconf_join_solict(dev, &ifp->addr);
3241
3242         net_srandom(ifp->addr.s6_addr32[3]);
3243
3244         read_lock_bh(&idev->lock);
3245         spin_lock(&ifp->lock);
3246         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3247                 goto out;
3248
3249         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3250             idev->cnf.accept_dad < 1 ||
3251             !(ifp->flags&IFA_F_TENTATIVE) ||
3252             ifp->flags & IFA_F_NODAD) {
3253                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3254                 spin_unlock(&ifp->lock);
3255                 read_unlock_bh(&idev->lock);
3256
3257                 addrconf_dad_completed(ifp);
3258                 return;
3259         }
3260
3261         if (!(idev->if_flags & IF_READY)) {
3262                 spin_unlock(&ifp->lock);
3263                 read_unlock_bh(&idev->lock);
3264                 /*
3265                  * If the device is not ready:
3266                  * - keep it tentative if it is a permanent address.
3267                  * - otherwise, kill it.
3268                  */
3269                 in6_ifa_hold(ifp);
3270                 addrconf_dad_stop(ifp, 0);
3271                 return;
3272         }
3273
3274         /*
3275          * Optimistic nodes can start receiving
3276          * Frames right away
3277          */
3278         if (ifp->flags & IFA_F_OPTIMISTIC) {
3279                 ip6_ins_rt(ifp->rt);
3280                 if (ipv6_use_optimistic_addr(idev)) {
3281                         /* Because optimistic nodes can use this address,
3282                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3283                          */
3284                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3285                 }
3286         }
3287
3288         addrconf_dad_kick(ifp);
3289 out:
3290         spin_unlock(&ifp->lock);
3291         read_unlock_bh(&idev->lock);
3292 }
3293
3294 static void addrconf_dad_timer(unsigned long data)
3295 {
3296         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3297         struct inet6_dev *idev = ifp->idev;
3298         struct in6_addr mcaddr;
3299
3300         if (!ifp->probes && addrconf_dad_end(ifp))
3301                 goto out;
3302
3303         read_lock(&idev->lock);
3304         if (idev->dead || !(idev->if_flags & IF_READY)) {
3305                 read_unlock(&idev->lock);
3306                 goto out;
3307         }
3308
3309         spin_lock(&ifp->lock);
3310         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3311                 spin_unlock(&ifp->lock);
3312                 read_unlock(&idev->lock);
3313                 goto out;
3314         }
3315
3316         if (ifp->probes == 0) {
3317                 /*
3318                  * DAD was successful
3319                  */
3320
3321                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3322                 spin_unlock(&ifp->lock);
3323                 read_unlock(&idev->lock);
3324
3325                 addrconf_dad_completed(ifp);
3326
3327                 goto out;
3328         }
3329
3330         ifp->probes--;
3331         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3332         spin_unlock(&ifp->lock);
3333         read_unlock(&idev->lock);
3334
3335         /* send a neighbour solicitation for our addr */
3336         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3337         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3338 out:
3339         in6_ifa_put(ifp);
3340 }
3341
3342 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3343 {
3344         struct net_device *dev = ifp->idev->dev;
3345
3346         /*
3347          *      Configure the address for reception. Now it is valid.
3348          */
3349
3350         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3351
3352         /* If added prefix is link local and we are prepared to process
3353            router advertisements, start sending router solicitations.
3354          */
3355
3356         if (ipv6_accept_ra(ifp->idev) &&
3357             ifp->idev->cnf.rtr_solicits > 0 &&
3358             (dev->flags&IFF_LOOPBACK) == 0 &&
3359             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3360                 /*
3361                  *      If a host as already performed a random delay
3362                  *      [...] as part of DAD [...] there is no need
3363                  *      to delay again before sending the first RS
3364                  */
3365                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3366
3367                 spin_lock_bh(&ifp->lock);
3368                 ifp->probes = 1;
3369                 ifp->idev->if_flags |= IF_RS_SENT;
3370                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3371                 spin_unlock_bh(&ifp->lock);
3372         }
3373 }
3374
3375 static void addrconf_dad_run(struct inet6_dev *idev)
3376 {
3377         struct inet6_ifaddr *ifp;
3378
3379         read_lock_bh(&idev->lock);
3380         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3381                 spin_lock(&ifp->lock);
3382                 if (ifp->flags & IFA_F_TENTATIVE &&
3383                     ifp->state == INET6_IFADDR_STATE_DAD)
3384                         addrconf_dad_kick(ifp);
3385                 spin_unlock(&ifp->lock);
3386         }
3387         read_unlock_bh(&idev->lock);
3388 }
3389
3390 #ifdef CONFIG_PROC_FS
3391 struct if6_iter_state {
3392         struct seq_net_private p;
3393         int bucket;
3394         int offset;
3395 };
3396
3397 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3398 {
3399         struct inet6_ifaddr *ifa = NULL;
3400         struct if6_iter_state *state = seq->private;
3401         struct net *net = seq_file_net(seq);
3402         int p = 0;
3403
3404         /* initial bucket if pos is 0 */
3405         if (pos == 0) {
3406                 state->bucket = 0;
3407                 state->offset = 0;
3408         }
3409
3410         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3411                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3412                                          addr_lst) {
3413                         if (!net_eq(dev_net(ifa->idev->dev), net))
3414                                 continue;
3415                         /* sync with offset */
3416                         if (p < state->offset) {
3417                                 p++;
3418                                 continue;
3419                         }
3420                         state->offset++;
3421                         return ifa;
3422                 }
3423
3424                 /* prepare for next bucket */
3425                 state->offset = 0;
3426                 p = 0;
3427         }
3428         return NULL;
3429 }
3430
3431 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3432                                          struct inet6_ifaddr *ifa)
3433 {
3434         struct if6_iter_state *state = seq->private;
3435         struct net *net = seq_file_net(seq);
3436
3437         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3438                 if (!net_eq(dev_net(ifa->idev->dev), net))
3439                         continue;
3440                 state->offset++;
3441                 return ifa;
3442         }
3443
3444         while (++state->bucket < IN6_ADDR_HSIZE) {
3445                 state->offset = 0;
3446                 hlist_for_each_entry_rcu_bh(ifa,
3447                                      &inet6_addr_lst[state->bucket], addr_lst) {
3448                         if (!net_eq(dev_net(ifa->idev->dev), net))
3449                                 continue;
3450                         state->offset++;
3451                         return ifa;
3452                 }
3453         }
3454
3455         return NULL;
3456 }
3457
3458 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3459         __acquires(rcu_bh)
3460 {
3461         rcu_read_lock_bh();
3462         return if6_get_first(seq, *pos);
3463 }
3464
3465 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3466 {
3467         struct inet6_ifaddr *ifa;
3468
3469         ifa = if6_get_next(seq, v);
3470         ++*pos;
3471         return ifa;
3472 }
3473
3474 static void if6_seq_stop(struct seq_file *seq, void *v)
3475         __releases(rcu_bh)
3476 {
3477         rcu_read_unlock_bh();
3478 }
3479
3480 static int if6_seq_show(struct seq_file *seq, void *v)
3481 {
3482         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3483         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3484                    &ifp->addr,
3485                    ifp->idev->dev->ifindex,
3486                    ifp->prefix_len,
3487                    ifp->scope,
3488                    ifp->flags,
3489                    ifp->idev->dev->name);
3490         return 0;
3491 }
3492
3493 static const struct seq_operations if6_seq_ops = {
3494         .start  = if6_seq_start,
3495         .next   = if6_seq_next,
3496         .show   = if6_seq_show,
3497         .stop   = if6_seq_stop,
3498 };
3499
3500 static int if6_seq_open(struct inode *inode, struct file *file)
3501 {
3502         return seq_open_net(inode, file, &if6_seq_ops,
3503                             sizeof(struct if6_iter_state));
3504 }
3505
3506 static const struct file_operations if6_fops = {
3507         .owner          = THIS_MODULE,
3508         .open           = if6_seq_open,
3509         .read           = seq_read,
3510         .llseek         = seq_lseek,
3511         .release        = seq_release_net,
3512 };
3513
3514 static int __net_init if6_proc_net_init(struct net *net)
3515 {
3516         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3517                 return -ENOMEM;
3518         return 0;
3519 }
3520
3521 static void __net_exit if6_proc_net_exit(struct net *net)
3522 {
3523         remove_proc_entry("if_inet6", net->proc_net);
3524 }
3525
3526 static struct pernet_operations if6_proc_net_ops = {
3527        .init = if6_proc_net_init,
3528        .exit = if6_proc_net_exit,
3529 };
3530
3531 int __init if6_proc_init(void)
3532 {
3533         return register_pernet_subsys(&if6_proc_net_ops);
3534 }
3535
3536 void if6_proc_exit(void)
3537 {
3538         unregister_pernet_subsys(&if6_proc_net_ops);
3539 }
3540 #endif  /* CONFIG_PROC_FS */
3541
3542 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3543 /* Check if address is a home address configured on any interface. */
3544 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3545 {
3546         int ret = 0;
3547         struct inet6_ifaddr *ifp = NULL;
3548         unsigned int hash = inet6_addr_hash(addr);
3549
3550         rcu_read_lock_bh();
3551         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3552                 if (!net_eq(dev_net(ifp->idev->dev), net))
3553                         continue;
3554                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3555                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3556                         ret = 1;
3557                         break;
3558                 }
3559         }
3560         rcu_read_unlock_bh();
3561         return ret;
3562 }
3563 #endif
3564
3565 /*
3566  *      Periodic address status verification
3567  */
3568
3569 static void addrconf_verify(unsigned long foo)
3570 {
3571         unsigned long now, next, next_sec, next_sched;
3572         struct inet6_ifaddr *ifp;
3573         int i;
3574
3575         rcu_read_lock_bh();
3576         spin_lock(&addrconf_verify_lock);
3577         now = jiffies;
3578         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3579
3580         del_timer(&addr_chk_timer);
3581
3582         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3583 restart:
3584                 hlist_for_each_entry_rcu_bh(ifp,
3585                                          &inet6_addr_lst[i], addr_lst) {
3586                         unsigned long age;
3587
3588                         if (ifp->flags & IFA_F_PERMANENT)
3589                                 continue;
3590
3591                         spin_lock(&ifp->lock);
3592                         /* We try to batch several events at once. */
3593                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3594
3595                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3596                             age >= ifp->valid_lft) {
3597                                 spin_unlock(&ifp->lock);
3598                                 in6_ifa_hold(ifp);
3599                                 ipv6_del_addr(ifp);
3600                                 goto restart;
3601                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3602                                 spin_unlock(&ifp->lock);
3603                                 continue;
3604                         } else if (age >= ifp->prefered_lft) {
3605                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3606                                 int deprecate = 0;
3607
3608                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3609                                         deprecate = 1;
3610                                         ifp->flags |= IFA_F_DEPRECATED;
3611                                 }
3612
3613                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3614                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3615
3616                                 spin_unlock(&ifp->lock);
3617
3618                                 if (deprecate) {
3619                                         in6_ifa_hold(ifp);
3620
3621                                         ipv6_ifa_notify(0, ifp);
3622                                         in6_ifa_put(ifp);
3623                                         goto restart;
3624                                 }
3625 #ifdef CONFIG_IPV6_PRIVACY
3626                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3627                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3628                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3629                                         ifp->idev->cnf.dad_transmits *
3630                                         ifp->idev->nd_parms->retrans_time / HZ;
3631
3632                                 if (age >= ifp->prefered_lft - regen_advance) {
3633                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3634                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3635                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3636                                         if (!ifp->regen_count && ifpub) {
3637                                                 ifp->regen_count++;
3638                                                 in6_ifa_hold(ifp);
3639                                                 in6_ifa_hold(ifpub);
3640                                                 spin_unlock(&ifp->lock);
3641
3642                                                 spin_lock(&ifpub->lock);
3643                                                 ifpub->regen_count = 0;
3644                                                 spin_unlock(&ifpub->lock);
3645                                                 ipv6_create_tempaddr(ifpub, ifp);
3646                                                 in6_ifa_put(ifpub);
3647                                                 in6_ifa_put(ifp);
3648                                                 goto restart;
3649                                         }
3650                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3651                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3652                                 spin_unlock(&ifp->lock);
3653 #endif
3654                         } else {
3655                                 /* ifp->prefered_lft <= ifp->valid_lft */
3656                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3657                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3658                                 spin_unlock(&ifp->lock);
3659                         }
3660                 }
3661         }
3662
3663         next_sec = round_jiffies_up(next);
3664         next_sched = next;
3665
3666         /* If rounded timeout is accurate enough, accept it. */
3667         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3668                 next_sched = next_sec;
3669
3670         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3671         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3672                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3673
3674         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3675               now, next, next_sec, next_sched));
3676
3677         addr_chk_timer.expires = next_sched;
3678         add_timer(&addr_chk_timer);
3679         spin_unlock(&addrconf_verify_lock);
3680         rcu_read_unlock_bh();
3681 }
3682
3683 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3684 {
3685         struct in6_addr *pfx = NULL;
3686
3687         if (addr)
3688                 pfx = nla_data(addr);
3689
3690         if (local) {
3691                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3692                         pfx = NULL;
3693                 else
3694                         pfx = nla_data(local);
3695         }
3696
3697         return pfx;
3698 }
3699
3700 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3701         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3702         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3703         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3704 };
3705
3706 static int
3707 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3708 {
3709         struct net *net = sock_net(skb->sk);
3710         struct ifaddrmsg *ifm;
3711         struct nlattr *tb[IFA_MAX+1];
3712         struct in6_addr *pfx;
3713         int err;
3714
3715         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3716         if (err < 0)
3717                 return err;
3718
3719         ifm = nlmsg_data(nlh);
3720         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3721         if (pfx == NULL)
3722                 return -EINVAL;
3723
3724         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3725 }
3726
3727 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3728                              u32 prefered_lft, u32 valid_lft)
3729 {
3730         u32 flags;
3731         clock_t expires;
3732         unsigned long timeout;
3733
3734         if (!valid_lft || (prefered_lft > valid_lft))
3735                 return -EINVAL;
3736
3737         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3738         if (addrconf_finite_timeout(timeout)) {
3739                 expires = jiffies_to_clock_t(timeout * HZ);
3740                 valid_lft = timeout;
3741                 flags = RTF_EXPIRES;
3742         } else {
3743                 expires = 0;
3744                 flags = 0;
3745                 ifa_flags |= IFA_F_PERMANENT;
3746         }
3747
3748         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3749         if (addrconf_finite_timeout(timeout)) {
3750                 if (timeout == 0)
3751                         ifa_flags |= IFA_F_DEPRECATED;
3752                 prefered_lft = timeout;
3753         }
3754
3755         spin_lock_bh(&ifp->lock);
3756         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3757         ifp->tstamp = jiffies;
3758         ifp->valid_lft = valid_lft;
3759         ifp->prefered_lft = prefered_lft;
3760
3761         spin_unlock_bh(&ifp->lock);
3762         if (!(ifp->flags&IFA_F_TENTATIVE))
3763                 ipv6_ifa_notify(0, ifp);
3764
3765         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3766                               expires, flags);
3767         addrconf_verify(0);
3768
3769         return 0;
3770 }
3771
3772 static int
3773 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3774 {
3775         struct net *net = sock_net(skb->sk);
3776         struct ifaddrmsg *ifm;
3777         struct nlattr *tb[IFA_MAX+1];
3778         struct in6_addr *pfx;
3779         struct inet6_ifaddr *ifa;
3780         struct net_device *dev;
3781         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3782         u8 ifa_flags;
3783         int err;
3784
3785         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3786         if (err < 0)
3787                 return err;
3788
3789         ifm = nlmsg_data(nlh);
3790         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3791         if (pfx == NULL)
3792                 return -EINVAL;
3793
3794         if (tb[IFA_CACHEINFO]) {
3795                 struct ifa_cacheinfo *ci;
3796
3797                 ci = nla_data(tb[IFA_CACHEINFO]);
3798                 valid_lft = ci->ifa_valid;
3799                 preferred_lft = ci->ifa_prefered;
3800         } else {
3801                 preferred_lft = INFINITY_LIFE_TIME;
3802                 valid_lft = INFINITY_LIFE_TIME;
3803         }
3804
3805         dev =  __dev_get_by_index(net, ifm->ifa_index);
3806         if (dev == NULL)
3807                 return -ENODEV;
3808
3809         /* We ignore other flags so far. */
3810         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3811
3812         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3813         if (ifa == NULL) {
3814                 /*
3815                  * It would be best to check for !NLM_F_CREATE here but
3816                  * userspace alreay relies on not having to provide this.
3817                  */
3818                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3819                                       ifm->ifa_prefixlen, ifa_flags,
3820                                       preferred_lft, valid_lft);
3821         }
3822
3823         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3824             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3825                 err = -EEXIST;
3826         else
3827                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3828
3829         in6_ifa_put(ifa);
3830
3831         return err;
3832 }
3833
3834 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3835                           u8 scope, int ifindex)
3836 {
3837         struct ifaddrmsg *ifm;
3838
3839         ifm = nlmsg_data(nlh);
3840         ifm->ifa_family = AF_INET6;
3841         ifm->ifa_prefixlen = prefixlen;
3842         ifm->ifa_flags = flags;
3843         ifm->ifa_scope = scope;
3844         ifm->ifa_index = ifindex;
3845 }
3846
3847 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3848                          unsigned long tstamp, u32 preferred, u32 valid)
3849 {
3850         struct ifa_cacheinfo ci;
3851
3852         ci.cstamp = cstamp_delta(cstamp);
3853         ci.tstamp = cstamp_delta(tstamp);
3854         ci.ifa_prefered = preferred;
3855         ci.ifa_valid = valid;
3856
3857         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3858 }
3859
3860 static inline int rt_scope(int ifa_scope)
3861 {
3862         if (ifa_scope & IFA_HOST)
3863                 return RT_SCOPE_HOST;
3864         else if (ifa_scope & IFA_LINK)
3865                 return RT_SCOPE_LINK;
3866         else if (ifa_scope & IFA_SITE)
3867                 return RT_SCOPE_SITE;
3868         else
3869                 return RT_SCOPE_UNIVERSE;
3870 }
3871
3872 static inline int inet6_ifaddr_msgsize(void)
3873 {
3874         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3875                + nla_total_size(16) /* IFA_ADDRESS */
3876                + nla_total_size(sizeof(struct ifa_cacheinfo));
3877 }
3878
3879 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3880                              u32 portid, u32 seq, int event, unsigned int flags)
3881 {
3882         struct nlmsghdr  *nlh;
3883         u32 preferred, valid;
3884
3885         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3886         if (nlh == NULL)
3887                 return -EMSGSIZE;
3888
3889         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3890                       ifa->idev->dev->ifindex);
3891
3892         if (!(ifa->flags&IFA_F_PERMANENT)) {
3893                 preferred = ifa->prefered_lft;
3894                 valid = ifa->valid_lft;
3895                 if (preferred != INFINITY_LIFE_TIME) {
3896                         long tval = (jiffies - ifa->tstamp)/HZ;
3897                         if (preferred > tval)
3898                                 preferred -= tval;
3899                         else
3900                                 preferred = 0;
3901                         if (valid != INFINITY_LIFE_TIME) {
3902                                 if (valid > tval)
3903                                         valid -= tval;
3904                                 else
3905                                         valid = 0;
3906                         }
3907                 }
3908         } else {
3909                 preferred = INFINITY_LIFE_TIME;
3910                 valid = INFINITY_LIFE_TIME;
3911         }
3912
3913         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3914             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3915                 nlmsg_cancel(skb, nlh);
3916                 return -EMSGSIZE;
3917         }
3918
3919         return nlmsg_end(skb, nlh);
3920 }
3921
3922 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3923                                 u32 portid, u32 seq, int event, u16 flags)
3924 {
3925         struct nlmsghdr  *nlh;
3926         u8 scope = RT_SCOPE_UNIVERSE;
3927         int ifindex = ifmca->idev->dev->ifindex;
3928
3929         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3930                 scope = RT_SCOPE_SITE;
3931
3932         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3933         if (nlh == NULL)
3934                 return -EMSGSIZE;
3935
3936         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3937         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3938             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3939                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3940                 nlmsg_cancel(skb, nlh);
3941                 return -EMSGSIZE;
3942         }
3943
3944         return nlmsg_end(skb, nlh);
3945 }
3946
3947 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3948                                 u32 portid, u32 seq, int event, unsigned int flags)
3949 {
3950         struct nlmsghdr  *nlh;
3951         u8 scope = RT_SCOPE_UNIVERSE;
3952         int ifindex = ifaca->aca_idev->dev->ifindex;
3953
3954         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3955                 scope = RT_SCOPE_SITE;
3956
3957         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3958         if (nlh == NULL)
3959                 return -EMSGSIZE;
3960
3961         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3962         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3963             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3964                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3965                 nlmsg_cancel(skb, nlh);
3966                 return -EMSGSIZE;
3967         }
3968
3969         return nlmsg_end(skb, nlh);
3970 }
3971
3972 enum addr_type_t {
3973         UNICAST_ADDR,
3974         MULTICAST_ADDR,
3975         ANYCAST_ADDR,
3976 };
3977
3978 /* called with rcu_read_lock() */
3979 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3980                           struct netlink_callback *cb, enum addr_type_t type,
3981                           int s_ip_idx, int *p_ip_idx)
3982 {
3983         struct ifmcaddr6 *ifmca;
3984         struct ifacaddr6 *ifaca;
3985         int err = 1;
3986         int ip_idx = *p_ip_idx;
3987
3988         read_lock_bh(&idev->lock);
3989         switch (type) {
3990         case UNICAST_ADDR: {
3991                 struct inet6_ifaddr *ifa;
3992
3993                 /* unicast address incl. temp addr */
3994                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3995                         if (++ip_idx < s_ip_idx)
3996                                 continue;
3997                         err = inet6_fill_ifaddr(skb, ifa,
3998                                                 NETLINK_CB(cb->skb).portid,
3999                                                 cb->nlh->nlmsg_seq,
4000                                                 RTM_NEWADDR,
4001                                                 NLM_F_MULTI);
4002                         if (err <= 0)
4003                                 break;
4004                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4005                 }
4006                 break;
4007         }
4008         case MULTICAST_ADDR:
4009                 /* multicast address */
4010                 for (ifmca = idev->mc_list; ifmca;
4011                      ifmca = ifmca->next, ip_idx++) {
4012                         if (ip_idx < s_ip_idx)
4013                                 continue;
4014                         err = inet6_fill_ifmcaddr(skb, ifmca,
4015                                                   NETLINK_CB(cb->skb).portid,
4016                                                   cb->nlh->nlmsg_seq,
4017                                                   RTM_GETMULTICAST,
4018                                                   NLM_F_MULTI);
4019                         if (err <= 0)
4020                                 break;
4021                 }
4022                 break;
4023         case ANYCAST_ADDR:
4024                 /* anycast address */
4025                 for (ifaca = idev->ac_list; ifaca;
4026                      ifaca = ifaca->aca_next, ip_idx++) {
4027                         if (ip_idx < s_ip_idx)
4028                                 continue;
4029                         err = inet6_fill_ifacaddr(skb, ifaca,
4030                                                   NETLINK_CB(cb->skb).portid,
4031                                                   cb->nlh->nlmsg_seq,
4032                                                   RTM_GETANYCAST,
4033                                                   NLM_F_MULTI);
4034                         if (err <= 0)
4035                                 break;
4036                 }
4037                 break;
4038         default:
4039                 break;
4040         }
4041         read_unlock_bh(&idev->lock);
4042         *p_ip_idx = ip_idx;
4043         return err;
4044 }
4045
4046 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4047                            enum addr_type_t type)
4048 {
4049         struct net *net = sock_net(skb->sk);
4050         int h, s_h;
4051         int idx, ip_idx;
4052         int s_idx, s_ip_idx;
4053         struct net_device *dev;
4054         struct inet6_dev *idev;
4055         struct hlist_head *head;
4056
4057         s_h = cb->args[0];
4058         s_idx = idx = cb->args[1];
4059         s_ip_idx = ip_idx = cb->args[2];
4060
4061         rcu_read_lock();
4062         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4063         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4064                 idx = 0;
4065                 head = &net->dev_index_head[h];
4066                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4067                         if (idx < s_idx)
4068                                 goto cont;
4069                         if (h > s_h || idx > s_idx)
4070                                 s_ip_idx = 0;
4071                         ip_idx = 0;
4072                         idev = __in6_dev_get(dev);
4073                         if (!idev)
4074                                 goto cont;
4075
4076                         if (in6_dump_addrs(idev, skb, cb, type,
4077                                            s_ip_idx, &ip_idx) <= 0)
4078                                 goto done;
4079 cont:
4080                         idx++;
4081                 }
4082         }
4083 done:
4084         rcu_read_unlock();
4085         cb->args[0] = h;
4086         cb->args[1] = idx;
4087         cb->args[2] = ip_idx;
4088
4089         return skb->len;
4090 }
4091
4092 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4093 {
4094         enum addr_type_t type = UNICAST_ADDR;
4095
4096         return inet6_dump_addr(skb, cb, type);
4097 }
4098
4099 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4100 {
4101         enum addr_type_t type = MULTICAST_ADDR;
4102
4103         return inet6_dump_addr(skb, cb, type);
4104 }
4105
4106
4107 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4108 {
4109         enum addr_type_t type = ANYCAST_ADDR;
4110
4111         return inet6_dump_addr(skb, cb, type);
4112 }
4113
4114 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4115 {
4116         struct net *net = sock_net(in_skb->sk);
4117         struct ifaddrmsg *ifm;
4118         struct nlattr *tb[IFA_MAX+1];
4119         struct in6_addr *addr = NULL;
4120         struct net_device *dev = NULL;
4121         struct inet6_ifaddr *ifa;
4122         struct sk_buff *skb;
4123         int err;
4124
4125         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4126         if (err < 0)
4127                 goto errout;
4128
4129         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4130         if (addr == NULL) {
4131                 err = -EINVAL;
4132                 goto errout;
4133         }
4134
4135         ifm = nlmsg_data(nlh);
4136         if (ifm->ifa_index)
4137                 dev = __dev_get_by_index(net, ifm->ifa_index);
4138
4139         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4140         if (!ifa) {
4141                 err = -EADDRNOTAVAIL;
4142                 goto errout;
4143         }
4144
4145         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4146         if (!skb) {
4147                 err = -ENOBUFS;
4148                 goto errout_ifa;
4149         }
4150
4151         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4152                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4153         if (err < 0) {
4154                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4155                 WARN_ON(err == -EMSGSIZE);
4156                 kfree_skb(skb);
4157                 goto errout_ifa;
4158         }
4159         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4160 errout_ifa:
4161         in6_ifa_put(ifa);
4162 errout:
4163         return err;
4164 }
4165
4166 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4167 {
4168         struct sk_buff *skb;
4169         struct net *net = dev_net(ifa->idev->dev);
4170         int err = -ENOBUFS;
4171
4172         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4173         if (skb == NULL)
4174                 goto errout;
4175
4176         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4177         if (err < 0) {
4178                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4179                 WARN_ON(err == -EMSGSIZE);
4180                 kfree_skb(skb);
4181                 goto errout;
4182         }
4183         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4184         return;
4185 errout:
4186         if (err < 0)
4187                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4188 }
4189
4190 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4191                                 __s32 *array, int bytes)
4192 {
4193         BUG_ON(bytes < (DEVCONF_MAX * 4));
4194
4195         memset(array, 0, bytes);
4196         array[DEVCONF_FORWARDING] = cnf->forwarding;
4197         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4198         array[DEVCONF_MTU6] = cnf->mtu6;
4199         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4200         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4201         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4202         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4203         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4204         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4205                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4206         array[DEVCONF_RTR_SOLICIT_DELAY] =
4207                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4208         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4209 #ifdef CONFIG_IPV6_PRIVACY
4210         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4211         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4212         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4213         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4214         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4215 #endif
4216         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4217         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4218         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4219 #ifdef CONFIG_IPV6_ROUTER_PREF
4220         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4221         array[DEVCONF_RTR_PROBE_INTERVAL] =
4222                 jiffies_to_msecs(cnf->rtr_probe_interval);
4223 #ifdef CONFIG_IPV6_ROUTE_INFO
4224         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4225 #endif
4226 #endif
4227         array[DEVCONF_ACCEPT_RA_RT_TABLE] = cnf->accept_ra_rt_table;
4228         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4229         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4230 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4231         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4232         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4233 #endif
4234 #ifdef CONFIG_IPV6_MROUTE
4235         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4236 #endif
4237         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4238         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4239         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4240         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4241 }
4242
4243 static inline size_t inet6_ifla6_size(void)
4244 {
4245         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4246              + nla_total_size(sizeof(struct ifla_cacheinfo))
4247              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4248              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4249              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4250              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4251 }
4252
4253 static inline size_t inet6_if_nlmsg_size(void)
4254 {
4255         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4256                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4257                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4258                + nla_total_size(4) /* IFLA_MTU */
4259                + nla_total_size(4) /* IFLA_LINK */
4260                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4261 }
4262
4263 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4264                                       int items, int bytes)
4265 {
4266         int i;
4267         int pad = bytes - sizeof(u64) * items;
4268         BUG_ON(pad < 0);
4269
4270         /* Use put_unaligned() because stats may not be aligned for u64. */
4271         put_unaligned(items, &stats[0]);
4272         for (i = 1; i < items; i++)
4273                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4274
4275         memset(&stats[items], 0, pad);
4276 }
4277
4278 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4279                                       int items, int bytes, size_t syncpoff)
4280 {
4281         int i;
4282         int pad = bytes - sizeof(u64) * items;
4283         BUG_ON(pad < 0);
4284
4285         /* Use put_unaligned() because stats may not be aligned for u64. */
4286         put_unaligned(items, &stats[0]);
4287         for (i = 1; i < items; i++)
4288                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4289
4290         memset(&stats[items], 0, pad);
4291 }
4292
4293 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4294                              int bytes)
4295 {
4296         switch (attrtype) {
4297         case IFLA_INET6_STATS:
4298                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4299                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4300                 break;
4301         case IFLA_INET6_ICMP6STATS:
4302                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4303                 break;
4304         }
4305 }
4306
4307 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4308 {
4309         struct nlattr *nla;
4310         struct ifla_cacheinfo ci;
4311
4312         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4313                 goto nla_put_failure;
4314         ci.max_reasm_len = IPV6_MAXPLEN;
4315         ci.tstamp = cstamp_delta(idev->tstamp);
4316         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4317         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4318         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4319                 goto nla_put_failure;
4320         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4321         if (nla == NULL)
4322                 goto nla_put_failure;
4323         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4324
4325         /* XXX - MC not implemented */
4326
4327         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4328         if (nla == NULL)
4329                 goto nla_put_failure;
4330         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4331
4332         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4333         if (nla == NULL)
4334                 goto nla_put_failure;
4335         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4336
4337         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4338         if (nla == NULL)
4339                 goto nla_put_failure;
4340         read_lock_bh(&idev->lock);
4341         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4342         read_unlock_bh(&idev->lock);
4343
4344         return 0;
4345
4346 nla_put_failure:
4347         return -EMSGSIZE;
4348 }
4349
4350 static size_t inet6_get_link_af_size(const struct net_device *dev)
4351 {
4352         if (!__in6_dev_get(dev))
4353                 return 0;
4354
4355         return inet6_ifla6_size();
4356 }
4357
4358 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4359 {
4360         struct inet6_dev *idev = __in6_dev_get(dev);
4361
4362         if (!idev)
4363                 return -ENODATA;
4364
4365         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4366                 return -EMSGSIZE;
4367
4368         return 0;
4369 }
4370
4371 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4372 {
4373         struct inet6_ifaddr *ifp;
4374         struct net_device *dev = idev->dev;
4375         bool update_rs = false;
4376         struct in6_addr ll_addr;
4377
4378         if (token == NULL)
4379                 return -EINVAL;
4380         if (ipv6_addr_any(token))
4381                 return -EINVAL;
4382         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4383                 return -EINVAL;
4384         if (!ipv6_accept_ra(idev))
4385                 return -EINVAL;
4386         if (idev->cnf.rtr_solicits <= 0)
4387                 return -EINVAL;
4388
4389         write_lock_bh(&idev->lock);
4390
4391         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4392         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4393
4394         write_unlock_bh(&idev->lock);
4395
4396         if (!idev->dead && (idev->if_flags & IF_READY) &&
4397             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4398                              IFA_F_OPTIMISTIC)) {
4399
4400                 /* If we're not ready, then normal ifup will take care
4401                  * of this. Otherwise, we need to request our rs here.
4402                  */
4403                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4404                 update_rs = true;
4405         }
4406
4407         write_lock_bh(&idev->lock);
4408
4409         if (update_rs)
4410                 idev->if_flags |= IF_RS_SENT;
4411
4412         /* Well, that's kinda nasty ... */
4413         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4414                 spin_lock(&ifp->lock);
4415                 if (ifp->tokenized) {
4416                         ifp->valid_lft = 0;
4417                         ifp->prefered_lft = 0;
4418                 }
4419                 spin_unlock(&ifp->lock);
4420         }
4421
4422         write_unlock_bh(&idev->lock);
4423         return 0;
4424 }
4425
4426 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4427 {
4428         int err = -EINVAL;
4429         struct inet6_dev *idev = __in6_dev_get(dev);
4430         struct nlattr *tb[IFLA_INET6_MAX + 1];
4431
4432         if (!idev)
4433                 return -EAFNOSUPPORT;
4434
4435         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4436                 BUG();
4437
4438         if (tb[IFLA_INET6_TOKEN])
4439                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4440
4441         return err;
4442 }
4443
4444 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4445                              u32 portid, u32 seq, int event, unsigned int flags)
4446 {
4447         struct net_device *dev = idev->dev;
4448         struct ifinfomsg *hdr;
4449         struct nlmsghdr *nlh;
4450         void *protoinfo;
4451
4452         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4453         if (nlh == NULL)
4454                 return -EMSGSIZE;
4455
4456         hdr = nlmsg_data(nlh);
4457         hdr->ifi_family = AF_INET6;
4458         hdr->__ifi_pad = 0;
4459         hdr->ifi_type = dev->type;
4460         hdr->ifi_index = dev->ifindex;
4461         hdr->ifi_flags = dev_get_flags(dev);
4462         hdr->ifi_change = 0;
4463
4464         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4465             (dev->addr_len &&
4466              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4467             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4468             (dev->ifindex != dev->iflink &&
4469              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4470                 goto nla_put_failure;
4471         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4472         if (protoinfo == NULL)
4473                 goto nla_put_failure;
4474
4475         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4476                 goto nla_put_failure;
4477
4478         nla_nest_end(skb, protoinfo);
4479         return nlmsg_end(skb, nlh);
4480
4481 nla_put_failure:
4482         nlmsg_cancel(skb, nlh);
4483         return -EMSGSIZE;
4484 }
4485
4486 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4487 {
4488         struct net *net = sock_net(skb->sk);
4489         int h, s_h;
4490         int idx = 0, s_idx;
4491         struct net_device *dev;
4492         struct inet6_dev *idev;
4493         struct hlist_head *head;
4494
4495         s_h = cb->args[0];
4496         s_idx = cb->args[1];
4497
4498         rcu_read_lock();
4499         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4500                 idx = 0;
4501                 head = &net->dev_index_head[h];
4502                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4503                         if (idx < s_idx)
4504                                 goto cont;
4505                         idev = __in6_dev_get(dev);
4506                         if (!idev)
4507                                 goto cont;
4508                         if (inet6_fill_ifinfo(skb, idev,
4509                                               NETLINK_CB(cb->skb).portid,
4510                                               cb->nlh->nlmsg_seq,
4511                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4512                                 goto out;
4513 cont:
4514                         idx++;
4515                 }
4516         }
4517 out:
4518         rcu_read_unlock();
4519         cb->args[1] = idx;
4520         cb->args[0] = h;
4521
4522         return skb->len;
4523 }
4524
4525 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4526 {
4527         struct sk_buff *skb;
4528         struct net *net = dev_net(idev->dev);
4529         int err = -ENOBUFS;
4530
4531         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4532         if (skb == NULL)
4533                 goto errout;
4534
4535         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4536         if (err < 0) {
4537                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4538                 WARN_ON(err == -EMSGSIZE);
4539                 kfree_skb(skb);
4540                 goto errout;
4541         }
4542         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4543         return;
4544 errout:
4545         if (err < 0)
4546                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4547 }
4548
4549 static inline size_t inet6_prefix_nlmsg_size(void)
4550 {
4551         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4552                + nla_total_size(sizeof(struct in6_addr))
4553                + nla_total_size(sizeof(struct prefix_cacheinfo));
4554 }
4555
4556 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4557                              struct prefix_info *pinfo, u32 portid, u32 seq,
4558                              int event, unsigned int flags)
4559 {
4560         struct prefixmsg *pmsg;
4561         struct nlmsghdr *nlh;
4562         struct prefix_cacheinfo ci;
4563
4564         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4565         if (nlh == NULL)
4566                 return -EMSGSIZE;
4567
4568         pmsg = nlmsg_data(nlh);
4569         pmsg->prefix_family = AF_INET6;
4570         pmsg->prefix_pad1 = 0;
4571         pmsg->prefix_pad2 = 0;
4572         pmsg->prefix_ifindex = idev->dev->ifindex;
4573         pmsg->prefix_len = pinfo->prefix_len;
4574         pmsg->prefix_type = pinfo->type;
4575         pmsg->prefix_pad3 = 0;
4576         pmsg->prefix_flags = 0;
4577         if (pinfo->onlink)
4578                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4579         if (pinfo->autoconf)
4580                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4581
4582         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4583                 goto nla_put_failure;
4584         ci.preferred_time = ntohl(pinfo->prefered);
4585         ci.valid_time = ntohl(pinfo->valid);
4586         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4587                 goto nla_put_failure;
4588         return nlmsg_end(skb, nlh);
4589
4590 nla_put_failure:
4591         nlmsg_cancel(skb, nlh);
4592         return -EMSGSIZE;
4593 }
4594
4595 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4596                          struct prefix_info *pinfo)
4597 {
4598         struct sk_buff *skb;
4599         struct net *net = dev_net(idev->dev);
4600         int err = -ENOBUFS;
4601
4602         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4603         if (skb == NULL)
4604                 goto errout;
4605
4606         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4607         if (err < 0) {
4608                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4609                 WARN_ON(err == -EMSGSIZE);
4610                 kfree_skb(skb);
4611                 goto errout;
4612         }
4613         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4614         return;
4615 errout:
4616         if (err < 0)
4617                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4618 }
4619
4620 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4621 {
4622         struct net *net = dev_net(ifp->idev->dev);
4623
4624         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4625
4626         switch (event) {
4627         case RTM_NEWADDR:
4628                 /*
4629                  * If the address was optimistic
4630                  * we inserted the route at the start of
4631                  * our DAD process, so we don't need
4632                  * to do it again
4633                  */
4634                 if (!(ifp->rt->rt6i_node))
4635                         ip6_ins_rt(ifp->rt);
4636                 if (ifp->idev->cnf.forwarding)
4637                         addrconf_join_anycast(ifp);
4638                 break;
4639         case RTM_DELADDR:
4640                 if (ifp->idev->cnf.forwarding)
4641                         addrconf_leave_anycast(ifp);
4642                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4643                 dst_hold(&ifp->rt->dst);
4644
4645                 if (ip6_del_rt(ifp->rt))
4646                         dst_free(&ifp->rt->dst);
4647                 break;
4648         }
4649         atomic_inc(&net->ipv6.dev_addr_genid);
4650 }
4651
4652 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4653 {
4654         rcu_read_lock_bh();
4655         if (likely(ifp->idev->dead == 0))
4656                 __ipv6_ifa_notify(event, ifp);
4657         rcu_read_unlock_bh();
4658 }
4659
4660 #ifdef CONFIG_SYSCTL
4661
4662 static
4663 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4664                            void __user *buffer, size_t *lenp, loff_t *ppos)
4665 {
4666         int *valp = ctl->data;
4667         int val = *valp;
4668         loff_t pos = *ppos;
4669         ctl_table lctl;
4670         int ret;
4671
4672         /*
4673          * ctl->data points to idev->cnf.forwarding, we should
4674          * not modify it until we get the rtnl lock.
4675          */
4676         lctl = *ctl;
4677         lctl.data = &val;
4678
4679         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4680
4681         if (write)
4682                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4683         if (ret)
4684                 *ppos = pos;
4685         return ret;
4686 }
4687
4688 static void dev_disable_change(struct inet6_dev *idev)
4689 {
4690         if (!idev || !idev->dev)
4691                 return;
4692
4693         if (idev->cnf.disable_ipv6)
4694                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4695         else
4696                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4697 }
4698
4699 static void addrconf_disable_change(struct net *net, __s32 newf)
4700 {
4701         struct net_device *dev;
4702         struct inet6_dev *idev;
4703
4704         rcu_read_lock();
4705         for_each_netdev_rcu(net, dev) {
4706                 idev = __in6_dev_get(dev);
4707                 if (idev) {
4708                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4709                         idev->cnf.disable_ipv6 = newf;
4710                         if (changed)
4711                                 dev_disable_change(idev);
4712                 }
4713         }
4714         rcu_read_unlock();
4715 }
4716
4717 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4718 {
4719         struct net *net;
4720         int old;
4721
4722         if (!rtnl_trylock())
4723                 return restart_syscall();
4724
4725         net = (struct net *)table->extra2;
4726         old = *p;
4727         *p = newf;
4728
4729         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4730                 rtnl_unlock();
4731                 return 0;
4732         }
4733
4734         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4735                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4736                 addrconf_disable_change(net, newf);
4737         } else if ((!newf) ^ (!old))
4738                 dev_disable_change((struct inet6_dev *)table->extra1);
4739
4740         rtnl_unlock();
4741         return 0;
4742 }
4743
4744 static
4745 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4746                             void __user *buffer, size_t *lenp, loff_t *ppos)
4747 {
4748         int *valp = ctl->data;
4749         int val = *valp;
4750         loff_t pos = *ppos;
4751         ctl_table lctl;
4752         int ret;
4753
4754         /*
4755          * ctl->data points to idev->cnf.disable_ipv6, we should
4756          * not modify it until we get the rtnl lock.
4757          */
4758         lctl = *ctl;
4759         lctl.data = &val;
4760
4761         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4762
4763         if (write)
4764                 ret = addrconf_disable_ipv6(ctl, valp, val);
4765         if (ret)
4766                 *ppos = pos;
4767         return ret;
4768 }
4769
4770 static struct addrconf_sysctl_table
4771 {
4772         struct ctl_table_header *sysctl_header;
4773         ctl_table addrconf_vars[DEVCONF_MAX+1];
4774 } addrconf_sysctl __read_mostly = {
4775         .sysctl_header = NULL,
4776         .addrconf_vars = {
4777                 {
4778                         .procname       = "forwarding",
4779                         .data           = &ipv6_devconf.forwarding,
4780                         .maxlen         = sizeof(int),
4781                         .mode           = 0644,
4782                         .proc_handler   = addrconf_sysctl_forward,
4783                 },
4784                 {
4785                         .procname       = "hop_limit",
4786                         .data           = &ipv6_devconf.hop_limit,
4787                         .maxlen         = sizeof(int),
4788                         .mode           = 0644,
4789                         .proc_handler   = proc_dointvec,
4790                 },
4791                 {
4792                         .procname       = "mtu",
4793                         .data           = &ipv6_devconf.mtu6,
4794                         .maxlen         = sizeof(int),
4795                         .mode           = 0644,
4796                         .proc_handler   = proc_dointvec,
4797                 },
4798                 {
4799                         .procname       = "accept_ra",
4800                         .data           = &ipv6_devconf.accept_ra,
4801                         .maxlen         = sizeof(int),
4802                         .mode           = 0644,
4803                         .proc_handler   = proc_dointvec,
4804                 },
4805                 {
4806                         .procname       = "accept_redirects",
4807                         .data           = &ipv6_devconf.accept_redirects,
4808                         .maxlen         = sizeof(int),
4809                         .mode           = 0644,
4810                         .proc_handler   = proc_dointvec,
4811                 },
4812                 {
4813                         .procname       = "autoconf",
4814                         .data           = &ipv6_devconf.autoconf,
4815                         .maxlen         = sizeof(int),
4816                         .mode           = 0644,
4817                         .proc_handler   = proc_dointvec,
4818                 },
4819                 {
4820                         .procname       = "dad_transmits",
4821                         .data           = &ipv6_devconf.dad_transmits,
4822                         .maxlen         = sizeof(int),
4823                         .mode           = 0644,
4824                         .proc_handler   = proc_dointvec,
4825                 },
4826                 {
4827                         .procname       = "router_solicitations",
4828                         .data           = &ipv6_devconf.rtr_solicits,
4829                         .maxlen         = sizeof(int),
4830                         .mode           = 0644,
4831                         .proc_handler   = proc_dointvec,
4832                 },
4833                 {
4834                         .procname       = "router_solicitation_interval",
4835                         .data           = &ipv6_devconf.rtr_solicit_interval,
4836                         .maxlen         = sizeof(int),
4837                         .mode           = 0644,
4838                         .proc_handler   = proc_dointvec_jiffies,
4839                 },
4840                 {
4841                         .procname       = "router_solicitation_delay",
4842                         .data           = &ipv6_devconf.rtr_solicit_delay,
4843                         .maxlen         = sizeof(int),
4844                         .mode           = 0644,
4845                         .proc_handler   = proc_dointvec_jiffies,
4846                 },
4847                 {
4848                         .procname       = "force_mld_version",
4849                         .data           = &ipv6_devconf.force_mld_version,
4850                         .maxlen         = sizeof(int),
4851                         .mode           = 0644,
4852                         .proc_handler   = proc_dointvec,
4853                 },
4854 #ifdef CONFIG_IPV6_PRIVACY
4855                 {
4856                         .procname       = "use_tempaddr",
4857                         .data           = &ipv6_devconf.use_tempaddr,
4858                         .maxlen         = sizeof(int),
4859                         .mode           = 0644,
4860                         .proc_handler   = proc_dointvec,
4861                 },
4862                 {
4863                         .procname       = "temp_valid_lft",
4864                         .data           = &ipv6_devconf.temp_valid_lft,
4865                         .maxlen         = sizeof(int),
4866                         .mode           = 0644,
4867                         .proc_handler   = proc_dointvec,
4868                 },
4869                 {
4870                         .procname       = "temp_prefered_lft",
4871                         .data           = &ipv6_devconf.temp_prefered_lft,
4872                         .maxlen         = sizeof(int),
4873                         .mode           = 0644,
4874                         .proc_handler   = proc_dointvec,
4875                 },
4876                 {
4877                         .procname       = "regen_max_retry",
4878                         .data           = &ipv6_devconf.regen_max_retry,
4879                         .maxlen         = sizeof(int),
4880                         .mode           = 0644,
4881                         .proc_handler   = proc_dointvec,
4882                 },
4883                 {
4884                         .procname       = "max_desync_factor",
4885                         .data           = &ipv6_devconf.max_desync_factor,
4886                         .maxlen         = sizeof(int),
4887                         .mode           = 0644,
4888                         .proc_handler   = proc_dointvec,
4889                 },
4890 #endif
4891                 {
4892                         .procname       = "max_addresses",
4893                         .data           = &ipv6_devconf.max_addresses,
4894                         .maxlen         = sizeof(int),
4895                         .mode           = 0644,
4896                         .proc_handler   = proc_dointvec,
4897                 },
4898                 {
4899                         .procname       = "accept_ra_defrtr",
4900                         .data           = &ipv6_devconf.accept_ra_defrtr,
4901                         .maxlen         = sizeof(int),
4902                         .mode           = 0644,
4903                         .proc_handler   = proc_dointvec,
4904                 },
4905                 {
4906                         .procname       = "accept_ra_pinfo",
4907                         .data           = &ipv6_devconf.accept_ra_pinfo,
4908                         .maxlen         = sizeof(int),
4909                         .mode           = 0644,
4910                         .proc_handler   = proc_dointvec,
4911                 },
4912 #ifdef CONFIG_IPV6_ROUTER_PREF
4913                 {
4914                         .procname       = "accept_ra_rtr_pref",
4915                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4916                         .maxlen         = sizeof(int),
4917                         .mode           = 0644,
4918                         .proc_handler   = proc_dointvec,
4919                 },
4920                 {
4921                         .procname       = "router_probe_interval",
4922                         .data           = &ipv6_devconf.rtr_probe_interval,
4923                         .maxlen         = sizeof(int),
4924                         .mode           = 0644,
4925                         .proc_handler   = proc_dointvec_jiffies,
4926                 },
4927 #ifdef CONFIG_IPV6_ROUTE_INFO
4928                 {
4929                         .procname       = "accept_ra_rt_info_max_plen",
4930                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4931                         .maxlen         = sizeof(int),
4932                         .mode           = 0644,
4933                         .proc_handler   = proc_dointvec,
4934                 },
4935 #endif
4936 #endif
4937                 {
4938                         .procname       = "accept_ra_rt_table",
4939                         .data           = &ipv6_devconf.accept_ra_rt_table,
4940                         .maxlen         = sizeof(int),
4941                         .mode           = 0644,
4942                         .proc_handler   = proc_dointvec,
4943                 },
4944                 {
4945                         .procname       = "proxy_ndp",
4946                         .data           = &ipv6_devconf.proxy_ndp,
4947                         .maxlen         = sizeof(int),
4948                         .mode           = 0644,
4949                         .proc_handler   = proc_dointvec,
4950                 },
4951                 {
4952                         .procname       = "accept_source_route",
4953                         .data           = &ipv6_devconf.accept_source_route,
4954                         .maxlen         = sizeof(int),
4955                         .mode           = 0644,
4956                         .proc_handler   = proc_dointvec,
4957                 },
4958 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4959                 {
4960                         .procname       = "optimistic_dad",
4961                         .data           = &ipv6_devconf.optimistic_dad,
4962                         .maxlen         = sizeof(int),
4963                         .mode           = 0644,
4964                         .proc_handler   = proc_dointvec,
4965
4966                 },
4967                 {
4968                         .procname       = "use_optimistic",
4969                         .data           = &ipv6_devconf.use_optimistic,
4970                         .maxlen         = sizeof(int),
4971                         .mode           = 0644,
4972                         .proc_handler   = proc_dointvec,
4973
4974                 },
4975 #endif
4976 #ifdef CONFIG_IPV6_MROUTE
4977                 {
4978                         .procname       = "mc_forwarding",
4979                         .data           = &ipv6_devconf.mc_forwarding,
4980                         .maxlen         = sizeof(int),
4981                         .mode           = 0444,
4982                         .proc_handler   = proc_dointvec,
4983                 },
4984 #endif
4985                 {
4986                         .procname       = "disable_ipv6",
4987                         .data           = &ipv6_devconf.disable_ipv6,
4988                         .maxlen         = sizeof(int),
4989                         .mode           = 0644,
4990                         .proc_handler   = addrconf_sysctl_disable,
4991                 },
4992                 {
4993                         .procname       = "accept_dad",
4994                         .data           = &ipv6_devconf.accept_dad,
4995                         .maxlen         = sizeof(int),
4996                         .mode           = 0644,
4997                         .proc_handler   = proc_dointvec,
4998                 },
4999                 {
5000                         .procname       = "force_tllao",
5001                         .data           = &ipv6_devconf.force_tllao,
5002                         .maxlen         = sizeof(int),
5003                         .mode           = 0644,
5004                         .proc_handler   = proc_dointvec
5005                 },
5006                 {
5007                         .procname       = "ndisc_notify",
5008                         .data           = &ipv6_devconf.ndisc_notify,
5009                         .maxlen         = sizeof(int),
5010                         .mode           = 0644,
5011                         .proc_handler   = proc_dointvec
5012                 },
5013                 {
5014                         /* sentinel */
5015                 }
5016         },
5017 };
5018
5019 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5020                 struct inet6_dev *idev, struct ipv6_devconf *p)
5021 {
5022         int i;
5023         struct addrconf_sysctl_table *t;
5024         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5025
5026         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5027         if (t == NULL)
5028                 goto out;
5029
5030         for (i = 0; t->addrconf_vars[i].data; i++) {
5031                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5032                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5033                 t->addrconf_vars[i].extra2 = net;
5034         }
5035
5036         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5037
5038         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5039         if (t->sysctl_header == NULL)
5040                 goto free;
5041
5042         p->sysctl = t;
5043         return 0;
5044
5045 free:
5046         kfree(t);
5047 out:
5048         return -ENOBUFS;
5049 }
5050
5051 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5052 {
5053         struct addrconf_sysctl_table *t;
5054
5055         if (p->sysctl == NULL)
5056                 return;
5057
5058         t = p->sysctl;
5059         p->sysctl = NULL;
5060         unregister_net_sysctl_table(t->sysctl_header);
5061         kfree(t);
5062 }
5063
5064 static void addrconf_sysctl_register(struct inet6_dev *idev)
5065 {
5066         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5067                               &ndisc_ifinfo_sysctl_change);
5068         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5069                                         idev, &idev->cnf);
5070 }
5071
5072 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5073 {
5074         __addrconf_sysctl_unregister(&idev->cnf);
5075         neigh_sysctl_unregister(idev->nd_parms);
5076 }
5077
5078
5079 #endif
5080
5081 static int __net_init addrconf_init_net(struct net *net)
5082 {
5083         int err = -ENOMEM;
5084         struct ipv6_devconf *all, *dflt;
5085
5086         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5087         if (all == NULL)
5088                 goto err_alloc_all;
5089
5090         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5091         if (dflt == NULL)
5092                 goto err_alloc_dflt;
5093
5094         /* these will be inherited by all namespaces */
5095         dflt->autoconf = ipv6_defaults.autoconf;
5096         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5097
5098         net->ipv6.devconf_all = all;
5099         net->ipv6.devconf_dflt = dflt;
5100
5101 #ifdef CONFIG_SYSCTL
5102         err = __addrconf_sysctl_register(net, "all", NULL, all);
5103         if (err < 0)
5104                 goto err_reg_all;
5105
5106         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5107         if (err < 0)
5108                 goto err_reg_dflt;
5109 #endif
5110         return 0;
5111
5112 #ifdef CONFIG_SYSCTL
5113 err_reg_dflt:
5114         __addrconf_sysctl_unregister(all);
5115 err_reg_all:
5116         kfree(dflt);
5117 #endif
5118 err_alloc_dflt:
5119         kfree(all);
5120 err_alloc_all:
5121         return err;
5122 }
5123
5124 static void __net_exit addrconf_exit_net(struct net *net)
5125 {
5126 #ifdef CONFIG_SYSCTL
5127         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5128         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5129 #endif
5130         if (!net_eq(net, &init_net)) {
5131                 kfree(net->ipv6.devconf_dflt);
5132                 kfree(net->ipv6.devconf_all);
5133         }
5134 }
5135
5136 static struct pernet_operations addrconf_ops = {
5137         .init = addrconf_init_net,
5138         .exit = addrconf_exit_net,
5139 };
5140
5141 static struct rtnl_af_ops inet6_ops = {
5142         .family           = AF_INET6,
5143         .fill_link_af     = inet6_fill_link_af,
5144         .get_link_af_size = inet6_get_link_af_size,
5145         .set_link_af      = inet6_set_link_af,
5146 };
5147
5148 /*
5149  *      Init / cleanup code
5150  */
5151
5152 int __init addrconf_init(void)
5153 {
5154         int i, err;
5155
5156         err = ipv6_addr_label_init();
5157         if (err < 0) {
5158                 pr_crit("%s: cannot initialize default policy table: %d\n",
5159                         __func__, err);
5160                 goto out;
5161         }
5162
5163         err = register_pernet_subsys(&addrconf_ops);
5164         if (err < 0)
5165                 goto out_addrlabel;
5166
5167         /* The addrconf netdev notifier requires that loopback_dev
5168          * has it's ipv6 private information allocated and setup
5169          * before it can bring up and give link-local addresses
5170          * to other devices which are up.
5171          *
5172          * Unfortunately, loopback_dev is not necessarily the first
5173          * entry in the global dev_base list of net devices.  In fact,
5174          * it is likely to be the very last entry on that list.
5175          * So this causes the notifier registry below to try and
5176          * give link-local addresses to all devices besides loopback_dev
5177          * first, then loopback_dev, which cases all the non-loopback_dev
5178          * devices to fail to get a link-local address.
5179          *
5180          * So, as a temporary fix, allocate the ipv6 structure for
5181          * loopback_dev first by hand.
5182          * Longer term, all of the dependencies ipv6 has upon the loopback
5183          * device and it being up should be removed.
5184          */
5185         rtnl_lock();
5186         if (!ipv6_add_dev(init_net.loopback_dev))
5187                 err = -ENOMEM;
5188         rtnl_unlock();
5189         if (err)
5190                 goto errlo;
5191
5192         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5193                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5194
5195         register_netdevice_notifier(&ipv6_dev_notf);
5196
5197         addrconf_verify(0);
5198
5199         err = rtnl_af_register(&inet6_ops);
5200         if (err < 0)
5201                 goto errout_af;
5202
5203         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5204                               NULL);
5205         if (err < 0)
5206                 goto errout;
5207
5208         /* Only the first call to __rtnl_register can fail */
5209         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5210         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5211         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5212                         inet6_dump_ifaddr, NULL);
5213         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5214                         inet6_dump_ifmcaddr, NULL);
5215         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5216                         inet6_dump_ifacaddr, NULL);
5217         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5218                         inet6_netconf_dump_devconf, NULL);
5219
5220         ipv6_addr_label_rtnl_register();
5221
5222         return 0;
5223 errout:
5224         rtnl_af_unregister(&inet6_ops);
5225 errout_af:
5226         unregister_netdevice_notifier(&ipv6_dev_notf);
5227 errlo:
5228         unregister_pernet_subsys(&addrconf_ops);
5229 out_addrlabel:
5230         ipv6_addr_label_cleanup();
5231 out:
5232         return err;
5233 }
5234
5235 void addrconf_cleanup(void)
5236 {
5237         struct net_device *dev;
5238         int i;
5239
5240         unregister_netdevice_notifier(&ipv6_dev_notf);
5241         unregister_pernet_subsys(&addrconf_ops);
5242         ipv6_addr_label_cleanup();
5243
5244         rtnl_lock();
5245
5246         __rtnl_af_unregister(&inet6_ops);
5247
5248         /* clean dev list */
5249         for_each_netdev(&init_net, dev) {
5250                 if (__in6_dev_get(dev) == NULL)
5251                         continue;
5252                 addrconf_ifdown(dev, 1);
5253         }
5254         addrconf_ifdown(init_net.loopback_dev, 2);
5255
5256         /*
5257          *      Check hash table.
5258          */
5259         spin_lock_bh(&addrconf_hash_lock);
5260         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5261                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5262         spin_unlock_bh(&addrconf_hash_lock);
5263
5264         del_timer(&addr_chk_timer);
5265         rtnl_unlock();
5266 }