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          * Use age calculation as in addrconf_verify to avoid unnecessary
1117          * temporary addresses being generated.
1118          */
1119         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1120         if (tmp_prefered_lft <= regen_advance + age) {
1121                 in6_ifa_put(ifp);
1122                 in6_dev_put(idev);
1123                 ret = -1;
1124                 goto out;
1125         }
1126
1127         addr_flags = IFA_F_TEMPORARY;
1128         /* set in addrconf_prefix_rcv() */
1129         if (ifp->flags & IFA_F_OPTIMISTIC)
1130                 addr_flags |= IFA_F_OPTIMISTIC;
1131
1132         ift = ipv6_add_addr(idev, &addr, tmp_plen,
1133                             ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1134                             addr_flags);
1135         if (IS_ERR(ift)) {
1136                 in6_ifa_put(ifp);
1137                 in6_dev_put(idev);
1138                 pr_info("%s: retry temporary address regeneration\n", __func__);
1139                 tmpaddr = &addr;
1140                 write_lock(&idev->lock);
1141                 goto retry;
1142         }
1143
1144         spin_lock_bh(&ift->lock);
1145         ift->ifpub = ifp;
1146         ift->valid_lft = tmp_valid_lft;
1147         ift->prefered_lft = tmp_prefered_lft;
1148         ift->cstamp = now;
1149         ift->tstamp = tmp_tstamp;
1150         spin_unlock_bh(&ift->lock);
1151
1152         addrconf_dad_start(ift);
1153         in6_ifa_put(ift);
1154         in6_dev_put(idev);
1155 out:
1156         return ret;
1157 }
1158 #endif
1159
1160 /*
1161  *      Choose an appropriate source address (RFC3484)
1162  */
1163 enum {
1164         IPV6_SADDR_RULE_INIT = 0,
1165         IPV6_SADDR_RULE_LOCAL,
1166         IPV6_SADDR_RULE_SCOPE,
1167         IPV6_SADDR_RULE_PREFERRED,
1168 #ifdef CONFIG_IPV6_MIP6
1169         IPV6_SADDR_RULE_HOA,
1170 #endif
1171         IPV6_SADDR_RULE_OIF,
1172         IPV6_SADDR_RULE_LABEL,
1173 #ifdef CONFIG_IPV6_PRIVACY
1174         IPV6_SADDR_RULE_PRIVACY,
1175 #endif
1176         IPV6_SADDR_RULE_ORCHID,
1177         IPV6_SADDR_RULE_PREFIX,
1178 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1179         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1180 #endif
1181         IPV6_SADDR_RULE_MAX
1182 };
1183
1184 struct ipv6_saddr_score {
1185         int                     rule;
1186         int                     addr_type;
1187         struct inet6_ifaddr     *ifa;
1188         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1189         int                     scopedist;
1190         int                     matchlen;
1191 };
1192
1193 struct ipv6_saddr_dst {
1194         const struct in6_addr *addr;
1195         int ifindex;
1196         int scope;
1197         int label;
1198         unsigned int prefs;
1199 };
1200
1201 static inline int ipv6_saddr_preferred(int type)
1202 {
1203         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1204                 return 1;
1205         return 0;
1206 }
1207
1208 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1209 {
1210 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1211         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1212 #else
1213         return false;
1214 #endif
1215 }
1216
1217 static int ipv6_get_saddr_eval(struct net *net,
1218                                struct ipv6_saddr_score *score,
1219                                struct ipv6_saddr_dst *dst,
1220                                int i)
1221 {
1222         int ret;
1223
1224         if (i <= score->rule) {
1225                 switch (i) {
1226                 case IPV6_SADDR_RULE_SCOPE:
1227                         ret = score->scopedist;
1228                         break;
1229                 case IPV6_SADDR_RULE_PREFIX:
1230                         ret = score->matchlen;
1231                         break;
1232                 default:
1233                         ret = !!test_bit(i, score->scorebits);
1234                 }
1235                 goto out;
1236         }
1237
1238         switch (i) {
1239         case IPV6_SADDR_RULE_INIT:
1240                 /* Rule 0: remember if hiscore is not ready yet */
1241                 ret = !!score->ifa;
1242                 break;
1243         case IPV6_SADDR_RULE_LOCAL:
1244                 /* Rule 1: Prefer same address */
1245                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1246                 break;
1247         case IPV6_SADDR_RULE_SCOPE:
1248                 /* Rule 2: Prefer appropriate scope
1249                  *
1250                  *      ret
1251                  *       ^
1252                  *    -1 |  d 15
1253                  *    ---+--+-+---> scope
1254                  *       |
1255                  *       |             d is scope of the destination.
1256                  *  B-d  |  \
1257                  *       |   \      <- smaller scope is better if
1258                  *  B-15 |    \        if scope is enough for destinaion.
1259                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1260                  * d-C-1 | /
1261                  *       |/         <- greater is better
1262                  *   -C  /             if scope is not enough for destination.
1263                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1264                  *
1265                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1266                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1267                  * Assume B = 0 and we get C > 29.
1268                  */
1269                 ret = __ipv6_addr_src_scope(score->addr_type);
1270                 if (ret >= dst->scope)
1271                         ret = -ret;
1272                 else
1273                         ret -= 128;     /* 30 is enough */
1274                 score->scopedist = ret;
1275                 break;
1276         case IPV6_SADDR_RULE_PREFERRED:
1277             {
1278                 /* Rule 3: Avoid deprecated and optimistic addresses */
1279                 u8 avoid = IFA_F_DEPRECATED;
1280
1281                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1282                         avoid |= IFA_F_OPTIMISTIC;
1283                 ret = ipv6_saddr_preferred(score->addr_type) ||
1284                       !(score->ifa->flags & avoid);
1285                 break;
1286             }
1287 #ifdef CONFIG_IPV6_MIP6
1288         case IPV6_SADDR_RULE_HOA:
1289             {
1290                 /* Rule 4: Prefer home address */
1291                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1292                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1293                 break;
1294             }
1295 #endif
1296         case IPV6_SADDR_RULE_OIF:
1297                 /* Rule 5: Prefer outgoing interface */
1298                 ret = (!dst->ifindex ||
1299                        dst->ifindex == score->ifa->idev->dev->ifindex);
1300                 break;
1301         case IPV6_SADDR_RULE_LABEL:
1302                 /* Rule 6: Prefer matching label */
1303                 ret = ipv6_addr_label(net,
1304                                       &score->ifa->addr, score->addr_type,
1305                                       score->ifa->idev->dev->ifindex) == dst->label;
1306                 break;
1307 #ifdef CONFIG_IPV6_PRIVACY
1308         case IPV6_SADDR_RULE_PRIVACY:
1309             {
1310                 /* Rule 7: Prefer public address
1311                  * Note: prefer temporary address if use_tempaddr >= 2
1312                  */
1313                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1314                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1315                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1316                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1317                 break;
1318             }
1319 #endif
1320         case IPV6_SADDR_RULE_ORCHID:
1321                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1322                  *          non-ORCHID vs non-ORCHID
1323                  */
1324                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1325                         ipv6_addr_orchid(dst->addr));
1326                 break;
1327         case IPV6_SADDR_RULE_PREFIX:
1328                 /* Rule 8: Use longest matching prefix */
1329                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1330                 if (ret > score->ifa->prefix_len)
1331                         ret = score->ifa->prefix_len;
1332                 score->matchlen = ret;
1333                 break;
1334 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1335         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1336                 /* Optimistic addresses still have lower precedence than other
1337                  * preferred addresses.
1338                  */
1339                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1340                 break;
1341 #endif
1342         default:
1343                 ret = 0;
1344         }
1345
1346         if (ret)
1347                 __set_bit(i, score->scorebits);
1348         score->rule = i;
1349 out:
1350         return ret;
1351 }
1352
1353 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1354                        const struct in6_addr *daddr, unsigned int prefs,
1355                        struct in6_addr *saddr)
1356 {
1357         struct ipv6_saddr_score scores[2],
1358                                 *score = &scores[0], *hiscore = &scores[1];
1359         struct ipv6_saddr_dst dst;
1360         struct net_device *dev;
1361         int dst_type;
1362
1363         dst_type = __ipv6_addr_type(daddr);
1364         dst.addr = daddr;
1365         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1366         dst.scope = __ipv6_addr_src_scope(dst_type);
1367         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1368         dst.prefs = prefs;
1369
1370         hiscore->rule = -1;
1371         hiscore->ifa = NULL;
1372
1373         rcu_read_lock();
1374
1375         for_each_netdev_rcu(net, dev) {
1376                 struct inet6_dev *idev;
1377
1378                 /* Candidate Source Address (section 4)
1379                  *  - multicast and link-local destination address,
1380                  *    the set of candidate source address MUST only
1381                  *    include addresses assigned to interfaces
1382                  *    belonging to the same link as the outgoing
1383                  *    interface.
1384                  * (- For site-local destination addresses, the
1385                  *    set of candidate source addresses MUST only
1386                  *    include addresses assigned to interfaces
1387                  *    belonging to the same site as the outgoing
1388                  *    interface.)
1389                  */
1390                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1391                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1392                     dst.ifindex && dev->ifindex != dst.ifindex)
1393                         continue;
1394
1395                 idev = __in6_dev_get(dev);
1396                 if (!idev)
1397                         continue;
1398
1399                 read_lock_bh(&idev->lock);
1400                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1401                         int i;
1402
1403                         /*
1404                          * - Tentative Address (RFC2462 section 5.4)
1405                          *  - A tentative address is not considered
1406                          *    "assigned to an interface" in the traditional
1407                          *    sense, unless it is also flagged as optimistic.
1408                          * - Candidate Source Address (section 4)
1409                          *  - In any case, anycast addresses, multicast
1410                          *    addresses, and the unspecified address MUST
1411                          *    NOT be included in a candidate set.
1412                          */
1413                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1414                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1415                                 continue;
1416
1417                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1418
1419                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1420                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1421                                 LIMIT_NETDEBUG(KERN_DEBUG
1422                                                "ADDRCONF: unspecified / multicast address "
1423                                                "assigned as unicast address on %s",
1424                                                dev->name);
1425                                 continue;
1426                         }
1427
1428                         score->rule = -1;
1429                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1430
1431                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1432                                 int minihiscore, miniscore;
1433
1434                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1435                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1436
1437                                 if (minihiscore > miniscore) {
1438                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1439                                             score->scopedist > 0) {
1440                                                 /*
1441                                                  * special case:
1442                                                  * each remaining entry
1443                                                  * has too small (not enough)
1444                                                  * scope, because ifa entries
1445                                                  * are sorted by their scope
1446                                                  * values.
1447                                                  */
1448                                                 goto try_nextdev;
1449                                         }
1450                                         break;
1451                                 } else if (minihiscore < miniscore) {
1452                                         if (hiscore->ifa)
1453                                                 in6_ifa_put(hiscore->ifa);
1454
1455                                         in6_ifa_hold(score->ifa);
1456
1457                                         swap(hiscore, score);
1458
1459                                         /* restore our iterator */
1460                                         score->ifa = hiscore->ifa;
1461
1462                                         break;
1463                                 }
1464                         }
1465                 }
1466 try_nextdev:
1467                 read_unlock_bh(&idev->lock);
1468         }
1469         rcu_read_unlock();
1470
1471         if (!hiscore->ifa)
1472                 return -EADDRNOTAVAIL;
1473
1474         *saddr = hiscore->ifa->addr;
1475         in6_ifa_put(hiscore->ifa);
1476         return 0;
1477 }
1478 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1479
1480 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1481                       unsigned char banned_flags)
1482 {
1483         struct inet6_ifaddr *ifp;
1484         int err = -EADDRNOTAVAIL;
1485
1486         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1487                 if (ifp->scope == IFA_LINK &&
1488                     !(ifp->flags & banned_flags)) {
1489                         *addr = ifp->addr;
1490                         err = 0;
1491                         break;
1492                 }
1493         }
1494         return err;
1495 }
1496
1497 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1498                     unsigned char banned_flags)
1499 {
1500         struct inet6_dev *idev;
1501         int err = -EADDRNOTAVAIL;
1502
1503         rcu_read_lock();
1504         idev = __in6_dev_get(dev);
1505         if (idev) {
1506                 read_lock_bh(&idev->lock);
1507                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1508                 read_unlock_bh(&idev->lock);
1509         }
1510         rcu_read_unlock();
1511         return err;
1512 }
1513
1514 static int ipv6_count_addresses(struct inet6_dev *idev)
1515 {
1516         int cnt = 0;
1517         struct inet6_ifaddr *ifp;
1518
1519         read_lock_bh(&idev->lock);
1520         list_for_each_entry(ifp, &idev->addr_list, if_list)
1521                 cnt++;
1522         read_unlock_bh(&idev->lock);
1523         return cnt;
1524 }
1525
1526 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1527                   const struct net_device *dev, int strict)
1528 {
1529         struct inet6_ifaddr *ifp;
1530         unsigned int hash = inet6_addr_hash(addr);
1531
1532         rcu_read_lock_bh();
1533         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1534                 if (!net_eq(dev_net(ifp->idev->dev), net))
1535                         continue;
1536                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1537                     !(ifp->flags&IFA_F_TENTATIVE) &&
1538                     (dev == NULL || ifp->idev->dev == dev ||
1539                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1540                         rcu_read_unlock_bh();
1541                         return 1;
1542                 }
1543         }
1544
1545         rcu_read_unlock_bh();
1546         return 0;
1547 }
1548 EXPORT_SYMBOL(ipv6_chk_addr);
1549
1550 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1551                                struct net_device *dev)
1552 {
1553         unsigned int hash = inet6_addr_hash(addr);
1554         struct inet6_ifaddr *ifp;
1555
1556         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1557                 if (!net_eq(dev_net(ifp->idev->dev), net))
1558                         continue;
1559                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1560                         if (dev == NULL || ifp->idev->dev == dev)
1561                                 return true;
1562                 }
1563         }
1564         return false;
1565 }
1566
1567 /* Compares an address/prefix_len with addresses on device @dev.
1568  * If one is found it returns true.
1569  */
1570 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1571         const unsigned int prefix_len, struct net_device *dev)
1572 {
1573         struct inet6_dev *idev;
1574         struct inet6_ifaddr *ifa;
1575         bool ret = false;
1576
1577         rcu_read_lock();
1578         idev = __in6_dev_get(dev);
1579         if (idev) {
1580                 read_lock_bh(&idev->lock);
1581                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1582                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1583                         if (ret)
1584                                 break;
1585                 }
1586                 read_unlock_bh(&idev->lock);
1587         }
1588         rcu_read_unlock();
1589
1590         return ret;
1591 }
1592 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1593
1594 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1595 {
1596         struct inet6_dev *idev;
1597         struct inet6_ifaddr *ifa;
1598         int     onlink;
1599
1600         onlink = 0;
1601         rcu_read_lock();
1602         idev = __in6_dev_get(dev);
1603         if (idev) {
1604                 read_lock_bh(&idev->lock);
1605                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1606                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1607                                                    ifa->prefix_len);
1608                         if (onlink)
1609                                 break;
1610                 }
1611                 read_unlock_bh(&idev->lock);
1612         }
1613         rcu_read_unlock();
1614         return onlink;
1615 }
1616 EXPORT_SYMBOL(ipv6_chk_prefix);
1617
1618 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1619                                      struct net_device *dev, int strict)
1620 {
1621         struct inet6_ifaddr *ifp, *result = NULL;
1622         unsigned int hash = inet6_addr_hash(addr);
1623
1624         rcu_read_lock_bh();
1625         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1626                 if (!net_eq(dev_net(ifp->idev->dev), net))
1627                         continue;
1628                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1629                         if (dev == NULL || ifp->idev->dev == dev ||
1630                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1631                                 result = ifp;
1632                                 in6_ifa_hold(ifp);
1633                                 break;
1634                         }
1635                 }
1636         }
1637         rcu_read_unlock_bh();
1638
1639         return result;
1640 }
1641
1642 /* Gets referenced address, destroys ifaddr */
1643
1644 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1645 {
1646         if (ifp->flags&IFA_F_PERMANENT) {
1647                 spin_lock_bh(&ifp->lock);
1648                 addrconf_del_timer(ifp);
1649                 ifp->flags |= IFA_F_TENTATIVE;
1650                 if (dad_failed)
1651                         ifp->flags |= IFA_F_DADFAILED;
1652                 spin_unlock_bh(&ifp->lock);
1653                 if (dad_failed)
1654                         ipv6_ifa_notify(0, ifp);
1655                 in6_ifa_put(ifp);
1656 #ifdef CONFIG_IPV6_PRIVACY
1657         } else if (ifp->flags&IFA_F_TEMPORARY) {
1658                 struct inet6_ifaddr *ifpub;
1659                 spin_lock_bh(&ifp->lock);
1660                 ifpub = ifp->ifpub;
1661                 if (ifpub) {
1662                         in6_ifa_hold(ifpub);
1663                         spin_unlock_bh(&ifp->lock);
1664                         ipv6_create_tempaddr(ifpub, ifp);
1665                         in6_ifa_put(ifpub);
1666                 } else {
1667                         spin_unlock_bh(&ifp->lock);
1668                 }
1669                 ipv6_del_addr(ifp);
1670 #endif
1671         } else
1672                 ipv6_del_addr(ifp);
1673 }
1674
1675 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1676 {
1677         int err = -ENOENT;
1678
1679         spin_lock(&ifp->state_lock);
1680         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1681                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1682                 err = 0;
1683         }
1684         spin_unlock(&ifp->state_lock);
1685
1686         return err;
1687 }
1688
1689 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1690 {
1691         struct inet6_dev *idev = ifp->idev;
1692
1693         if (addrconf_dad_end(ifp)) {
1694                 in6_ifa_put(ifp);
1695                 return;
1696         }
1697
1698         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1699                              ifp->idev->dev->name, &ifp->addr);
1700
1701         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1702                 struct in6_addr addr;
1703
1704                 addr.s6_addr32[0] = htonl(0xfe800000);
1705                 addr.s6_addr32[1] = 0;
1706
1707                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1708                     ipv6_addr_equal(&ifp->addr, &addr)) {
1709                         /* DAD failed for link-local based on MAC address */
1710                         idev->cnf.disable_ipv6 = 1;
1711
1712                         pr_info("%s: IPv6 being disabled!\n",
1713                                 ifp->idev->dev->name);
1714                 }
1715         }
1716
1717         addrconf_dad_stop(ifp, 1);
1718 }
1719
1720 /* Join to solicited addr multicast group. */
1721
1722 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1723 {
1724         struct in6_addr maddr;
1725
1726         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1727                 return;
1728
1729         addrconf_addr_solict_mult(addr, &maddr);
1730         ipv6_dev_mc_inc(dev, &maddr);
1731 }
1732
1733 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1734 {
1735         struct in6_addr maddr;
1736
1737         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1738                 return;
1739
1740         addrconf_addr_solict_mult(addr, &maddr);
1741         __ipv6_dev_mc_dec(idev, &maddr);
1742 }
1743
1744 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1745 {
1746         struct in6_addr addr;
1747         if (ifp->prefix_len == 127) /* RFC 6164 */
1748                 return;
1749         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1750         if (ipv6_addr_any(&addr))
1751                 return;
1752         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1753 }
1754
1755 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1756 {
1757         struct in6_addr addr;
1758         if (ifp->prefix_len == 127) /* RFC 6164 */
1759                 return;
1760         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1761         if (ipv6_addr_any(&addr))
1762                 return;
1763         __ipv6_dev_ac_dec(ifp->idev, &addr);
1764 }
1765
1766 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1767 {
1768         if (dev->addr_len != ETH_ALEN)
1769                 return -1;
1770         memcpy(eui, dev->dev_addr, 3);
1771         memcpy(eui + 5, dev->dev_addr + 3, 3);
1772
1773         /*
1774          * The zSeries OSA network cards can be shared among various
1775          * OS instances, but the OSA cards have only one MAC address.
1776          * This leads to duplicate address conflicts in conjunction
1777          * with IPv6 if more than one instance uses the same card.
1778          *
1779          * The driver for these cards can deliver a unique 16-bit
1780          * identifier for each instance sharing the same card.  It is
1781          * placed instead of 0xFFFE in the interface identifier.  The
1782          * "u" bit of the interface identifier is not inverted in this
1783          * case.  Hence the resulting interface identifier has local
1784          * scope according to RFC2373.
1785          */
1786         if (dev->dev_id) {
1787                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1788                 eui[4] = dev->dev_id & 0xFF;
1789         } else {
1790                 eui[3] = 0xFF;
1791                 eui[4] = 0xFE;
1792                 eui[0] ^= 2;
1793         }
1794         return 0;
1795 }
1796
1797 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1798 {
1799         if (dev->addr_len != IEEE802154_ADDR_LEN)
1800                 return -1;
1801         memcpy(eui, dev->dev_addr, 8);
1802         eui[0] ^= 2;
1803         return 0;
1804 }
1805
1806 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1807 {
1808         union fwnet_hwaddr *ha;
1809
1810         if (dev->addr_len != FWNET_ALEN)
1811                 return -1;
1812
1813         ha = (union fwnet_hwaddr *)dev->dev_addr;
1814
1815         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1816         eui[0] ^= 2;
1817         return 0;
1818 }
1819
1820 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1821 {
1822         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1823         if (dev->addr_len != ARCNET_ALEN)
1824                 return -1;
1825         memset(eui, 0, 7);
1826         eui[7] = *(u8 *)dev->dev_addr;
1827         return 0;
1828 }
1829
1830 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1831 {
1832         if (dev->addr_len != INFINIBAND_ALEN)
1833                 return -1;
1834         memcpy(eui, dev->dev_addr + 12, 8);
1835         eui[0] |= 2;
1836         return 0;
1837 }
1838
1839 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1840 {
1841         if (addr == 0)
1842                 return -1;
1843         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1844                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1845                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1846                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1847                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1848                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1849         eui[1] = 0;
1850         eui[2] = 0x5E;
1851         eui[3] = 0xFE;
1852         memcpy(eui + 4, &addr, 4);
1853         return 0;
1854 }
1855
1856 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1857 {
1858         if (dev->priv_flags & IFF_ISATAP)
1859                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1860         return -1;
1861 }
1862
1863 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1864 {
1865         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1866 }
1867
1868 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1869 {
1870         switch (dev->type) {
1871         case ARPHRD_ETHER:
1872         case ARPHRD_FDDI:
1873                 return addrconf_ifid_eui48(eui, dev);
1874         case ARPHRD_ARCNET:
1875                 return addrconf_ifid_arcnet(eui, dev);
1876         case ARPHRD_INFINIBAND:
1877                 return addrconf_ifid_infiniband(eui, dev);
1878         case ARPHRD_SIT:
1879                 return addrconf_ifid_sit(eui, dev);
1880         case ARPHRD_IPGRE:
1881                 return addrconf_ifid_gre(eui, dev);
1882         case ARPHRD_IEEE802154:
1883                 return addrconf_ifid_eui64(eui, dev);
1884         case ARPHRD_IEEE1394:
1885                 return addrconf_ifid_ieee1394(eui, dev);
1886         }
1887         return -1;
1888 }
1889
1890 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1891 {
1892         int err = -1;
1893         struct inet6_ifaddr *ifp;
1894
1895         read_lock_bh(&idev->lock);
1896         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1897                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1898                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1899                         err = 0;
1900                         break;
1901                 }
1902         }
1903         read_unlock_bh(&idev->lock);
1904         return err;
1905 }
1906
1907 #ifdef CONFIG_IPV6_PRIVACY
1908 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1909 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1910 {
1911 regen:
1912         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1913         idev->rndid[0] &= ~0x02;
1914
1915         /*
1916          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1917          * check if generated address is not inappropriate
1918          *
1919          *  - Reserved subnet anycast (RFC 2526)
1920          *      11111101 11....11 1xxxxxxx
1921          *  - ISATAP (RFC4214) 6.1
1922          *      00-00-5E-FE-xx-xx-xx-xx
1923          *  - value 0
1924          *  - XXX: already assigned to an address on the device
1925          */
1926         if (idev->rndid[0] == 0xfd &&
1927             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1928             (idev->rndid[7]&0x80))
1929                 goto regen;
1930         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1931                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1932                         goto regen;
1933                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1934                         goto regen;
1935         }
1936 }
1937
1938 static void ipv6_regen_rndid(unsigned long data)
1939 {
1940         struct inet6_dev *idev = (struct inet6_dev *) data;
1941         unsigned long expires;
1942
1943         rcu_read_lock_bh();
1944         write_lock_bh(&idev->lock);
1945
1946         if (idev->dead)
1947                 goto out;
1948
1949         __ipv6_regen_rndid(idev);
1950
1951         expires = jiffies +
1952                 idev->cnf.temp_prefered_lft * HZ -
1953                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1954                 idev->cnf.max_desync_factor * HZ;
1955         if (time_before(expires, jiffies)) {
1956                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1957                         __func__, idev->dev->name);
1958                 goto out;
1959         }
1960
1961         if (!mod_timer(&idev->regen_timer, expires))
1962                 in6_dev_hold(idev);
1963
1964 out:
1965         write_unlock_bh(&idev->lock);
1966         rcu_read_unlock_bh();
1967         in6_dev_put(idev);
1968 }
1969
1970 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1971 {
1972         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1973                 __ipv6_regen_rndid(idev);
1974 }
1975 #endif
1976
1977 u32 addrconf_rt_table(const struct net_device *dev, u32 default_table) {
1978         /* Determines into what table to put autoconf PIO/RIO/default routes
1979          * learned on this device.
1980          *
1981          * - If 0, use the same table for every device. This puts routes into
1982          *   one of RT_TABLE_{PREFIX,INFO,DFLT} depending on the type of route
1983          *   (but note that these three are currently all equal to
1984          *   RT6_TABLE_MAIN).
1985          * - If > 0, use the specified table.
1986          * - If < 0, put routes into table dev->ifindex + (-rt_table).
1987          */
1988         struct inet6_dev *idev = in6_dev_get(dev);
1989         u32 table;
1990         int sysctl = idev->cnf.accept_ra_rt_table;
1991         if (sysctl == 0) {
1992                 table = default_table;
1993         } else if (sysctl > 0) {
1994                 table = (u32) sysctl;
1995         } else {
1996                 table = (unsigned) dev->ifindex + (-sysctl);
1997         }
1998         in6_dev_put(idev);
1999         return table;
2000 }
2001
2002 /*
2003  *      Add prefix route.
2004  */
2005
2006 static void
2007 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2008                       unsigned long expires, u32 flags)
2009 {
2010         struct fib6_config cfg = {
2011                 .fc_table = addrconf_rt_table(dev, RT6_TABLE_PREFIX),
2012                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2013                 .fc_ifindex = dev->ifindex,
2014                 .fc_expires = expires,
2015                 .fc_dst_len = plen,
2016                 .fc_flags = RTF_UP | flags,
2017                 .fc_nlinfo.nl_net = dev_net(dev),
2018                 .fc_protocol = RTPROT_KERNEL,
2019         };
2020
2021         cfg.fc_dst = *pfx;
2022
2023         /* Prevent useless cloning on PtP SIT.
2024            This thing is done here expecting that the whole
2025            class of non-broadcast devices need not cloning.
2026          */
2027 #if IS_ENABLED(CONFIG_IPV6_SIT)
2028         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2029                 cfg.fc_flags |= RTF_NONEXTHOP;
2030 #endif
2031
2032         ip6_route_add(&cfg);
2033 }
2034
2035
2036 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2037                                                   int plen,
2038                                                   const struct net_device *dev,
2039                                                   u32 flags, u32 noflags)
2040 {
2041         struct fib6_node *fn;
2042         struct rt6_info *rt = NULL;
2043         struct fib6_table *table;
2044
2045         table = fib6_get_table(dev_net(dev),
2046                                addrconf_rt_table(dev, RT6_TABLE_PREFIX));
2047         if (table == NULL)
2048                 return NULL;
2049
2050         read_lock_bh(&table->tb6_lock);
2051         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2052         if (!fn)
2053                 goto out;
2054         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2055                 if (rt->dst.dev->ifindex != dev->ifindex)
2056                         continue;
2057                 if ((rt->rt6i_flags & flags) != flags)
2058                         continue;
2059                 if ((rt->rt6i_flags & noflags) != 0)
2060                         continue;
2061                 dst_hold(&rt->dst);
2062                 break;
2063         }
2064 out:
2065         read_unlock_bh(&table->tb6_lock);
2066         return rt;
2067 }
2068
2069
2070 /* Create "default" multicast route to the interface */
2071
2072 static void addrconf_add_mroute(struct net_device *dev)
2073 {
2074         struct fib6_config cfg = {
2075                 .fc_table = RT6_TABLE_LOCAL,
2076                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2077                 .fc_ifindex = dev->ifindex,
2078                 .fc_dst_len = 8,
2079                 .fc_flags = RTF_UP,
2080                 .fc_nlinfo.nl_net = dev_net(dev),
2081         };
2082
2083         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2084
2085         ip6_route_add(&cfg);
2086 }
2087
2088 #if IS_ENABLED(CONFIG_IPV6_SIT)
2089 static void sit_route_add(struct net_device *dev)
2090 {
2091         struct fib6_config cfg = {
2092                 .fc_table = RT6_TABLE_MAIN,
2093                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2094                 .fc_ifindex = dev->ifindex,
2095                 .fc_dst_len = 96,
2096                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2097                 .fc_nlinfo.nl_net = dev_net(dev),
2098         };
2099
2100         /* prefix length - 96 bits "::d.d.d.d" */
2101         ip6_route_add(&cfg);
2102 }
2103 #endif
2104
2105 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2106 {
2107         struct inet6_dev *idev;
2108
2109         ASSERT_RTNL();
2110
2111         idev = ipv6_find_idev(dev);
2112         if (!idev)
2113                 return ERR_PTR(-ENOBUFS);
2114
2115         if (idev->cnf.disable_ipv6)
2116                 return ERR_PTR(-EACCES);
2117
2118         /* Add default multicast route */
2119         if (!(dev->flags & IFF_LOOPBACK))
2120                 addrconf_add_mroute(dev);
2121
2122         return idev;
2123 }
2124
2125 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2126 {
2127         struct prefix_info *pinfo;
2128         __u32 valid_lft;
2129         __u32 prefered_lft;
2130         int addr_type;
2131         struct inet6_dev *in6_dev;
2132         struct net *net = dev_net(dev);
2133
2134         pinfo = (struct prefix_info *) opt;
2135
2136         if (len < sizeof(struct prefix_info)) {
2137                 ADBG(("addrconf: prefix option too short\n"));
2138                 return;
2139         }
2140
2141         /*
2142          *      Validation checks ([ADDRCONF], page 19)
2143          */
2144
2145         addr_type = ipv6_addr_type(&pinfo->prefix);
2146
2147         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2148                 return;
2149
2150         valid_lft = ntohl(pinfo->valid);
2151         prefered_lft = ntohl(pinfo->prefered);
2152
2153         if (prefered_lft > valid_lft) {
2154                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2155                 return;
2156         }
2157
2158         in6_dev = in6_dev_get(dev);
2159
2160         if (in6_dev == NULL) {
2161                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2162                                     dev->name);
2163                 return;
2164         }
2165
2166         /*
2167          *      Two things going on here:
2168          *      1) Add routes for on-link prefixes
2169          *      2) Configure prefixes with the auto flag set
2170          */
2171
2172         if (pinfo->onlink) {
2173                 struct rt6_info *rt;
2174                 unsigned long rt_expires;
2175
2176                 /* Avoid arithmetic overflow. Really, we could
2177                  * save rt_expires in seconds, likely valid_lft,
2178                  * but it would require division in fib gc, that it
2179                  * not good.
2180                  */
2181                 if (HZ > USER_HZ)
2182                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2183                 else
2184                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2185
2186                 if (addrconf_finite_timeout(rt_expires))
2187                         rt_expires *= HZ;
2188
2189                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2190                                                pinfo->prefix_len,
2191                                                dev,
2192                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2193                                                RTF_GATEWAY | RTF_DEFAULT);
2194
2195                 if (rt) {
2196                         /* Autoconf prefix route */
2197                         if (valid_lft == 0) {
2198                                 ip6_del_rt(rt);
2199                                 rt = NULL;
2200                         } else if (addrconf_finite_timeout(rt_expires)) {
2201                                 /* not infinity */
2202                                 rt6_set_expires(rt, jiffies + rt_expires);
2203                         } else {
2204                                 rt6_clean_expires(rt);
2205                         }
2206                 } else if (valid_lft) {
2207                         clock_t expires = 0;
2208                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2209                         if (addrconf_finite_timeout(rt_expires)) {
2210                                 /* not infinity */
2211                                 flags |= RTF_EXPIRES;
2212                                 expires = jiffies_to_clock_t(rt_expires);
2213                         }
2214                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2215                                               dev, expires, flags);
2216                 }
2217                 ip6_rt_put(rt);
2218         }
2219
2220         /* Try to figure out our local address for this prefix */
2221
2222         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2223                 struct inet6_ifaddr *ifp;
2224                 struct in6_addr addr;
2225                 int create = 0, update_lft = 0;
2226                 bool tokenized = false;
2227
2228                 if (pinfo->prefix_len == 64) {
2229                         memcpy(&addr, &pinfo->prefix, 8);
2230
2231                         if (!ipv6_addr_any(&in6_dev->token)) {
2232                                 read_lock_bh(&in6_dev->lock);
2233                                 memcpy(addr.s6_addr + 8,
2234                                        in6_dev->token.s6_addr + 8, 8);
2235                                 read_unlock_bh(&in6_dev->lock);
2236                                 tokenized = true;
2237                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2238                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2239                                 in6_dev_put(in6_dev);
2240                                 return;
2241                         }
2242                         goto ok;
2243                 }
2244                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2245                                     pinfo->prefix_len);
2246                 in6_dev_put(in6_dev);
2247                 return;
2248
2249 ok:
2250
2251                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2252
2253                 if (ifp == NULL && valid_lft) {
2254                         int max_addresses = in6_dev->cnf.max_addresses;
2255                         u32 addr_flags = 0;
2256
2257 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2258                         if (in6_dev->cnf.optimistic_dad &&
2259                             !net->ipv6.devconf_all->forwarding && sllao)
2260                                 addr_flags = IFA_F_OPTIMISTIC;
2261 #endif
2262
2263                         /* Do not allow to create too much of autoconfigured
2264                          * addresses; this would be too easy way to crash kernel.
2265                          */
2266                         if (!max_addresses ||
2267                             ipv6_count_addresses(in6_dev) < max_addresses)
2268                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2269                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2270                                                     addr_flags);
2271
2272                         if (IS_ERR_OR_NULL(ifp)) {
2273                                 in6_dev_put(in6_dev);
2274                                 return;
2275                         }
2276
2277                         update_lft = create = 1;
2278                         ifp->cstamp = jiffies;
2279                         ifp->tokenized = tokenized;
2280                         addrconf_dad_start(ifp);
2281                 }
2282
2283                 if (ifp) {
2284                         int flags;
2285                         unsigned long now;
2286 #ifdef CONFIG_IPV6_PRIVACY
2287                         struct inet6_ifaddr *ift;
2288 #endif
2289                         u32 stored_lft;
2290
2291                         /* update lifetime (RFC2462 5.5.3 e) */
2292                         spin_lock(&ifp->lock);
2293                         now = jiffies;
2294                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2295                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2296                         else
2297                                 stored_lft = 0;
2298                         if (!update_lft && stored_lft) {
2299                                 if (valid_lft > MIN_VALID_LIFETIME ||
2300                                     valid_lft > stored_lft)
2301                                         update_lft = 1;
2302                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2303                                         /* valid_lft <= stored_lft is always true */
2304                                         /*
2305                                          * RFC 4862 Section 5.5.3e:
2306                                          * "Note that the preferred lifetime of
2307                                          *  the corresponding address is always
2308                                          *  reset to the Preferred Lifetime in
2309                                          *  the received Prefix Information
2310                                          *  option, regardless of whether the
2311                                          *  valid lifetime is also reset or
2312                                          *  ignored."
2313                                          *
2314                                          *  So if the preferred lifetime in
2315                                          *  this advertisement is different
2316                                          *  than what we have stored, but the
2317                                          *  valid lifetime is invalid, just
2318                                          *  reset prefered_lft.
2319                                          *
2320                                          *  We must set the valid lifetime
2321                                          *  to the stored lifetime since we'll
2322                                          *  be updating the timestamp below,
2323                                          *  else we'll set it back to the
2324                                          *  minimum.
2325                                          */
2326                                         if (prefered_lft != ifp->prefered_lft) {
2327                                                 valid_lft = stored_lft;
2328                                                 update_lft = 1;
2329                                         }
2330                                 } else {
2331                                         valid_lft = MIN_VALID_LIFETIME;
2332                                         if (valid_lft < prefered_lft)
2333                                                 prefered_lft = valid_lft;
2334                                         update_lft = 1;
2335                                 }
2336                         }
2337
2338                         if (update_lft) {
2339                                 ifp->valid_lft = valid_lft;
2340                                 ifp->prefered_lft = prefered_lft;
2341                                 ifp->tstamp = now;
2342                                 flags = ifp->flags;
2343                                 ifp->flags &= ~IFA_F_DEPRECATED;
2344                                 spin_unlock(&ifp->lock);
2345
2346                                 if (!(flags&IFA_F_TENTATIVE))
2347                                         ipv6_ifa_notify(0, ifp);
2348                         } else
2349                                 spin_unlock(&ifp->lock);
2350
2351 #ifdef CONFIG_IPV6_PRIVACY
2352                         read_lock_bh(&in6_dev->lock);
2353                         /* update all temporary addresses in the list */
2354                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2355                                             tmp_list) {
2356                                 int age, max_valid, max_prefered;
2357
2358                                 if (ifp != ift->ifpub)
2359                                         continue;
2360
2361                                 /*
2362                                  * RFC 4941 section 3.3:
2363                                  * If a received option will extend the lifetime
2364                                  * of a public address, the lifetimes of
2365                                  * temporary addresses should be extended,
2366                                  * subject to the overall constraint that no
2367                                  * temporary addresses should ever remain
2368                                  * "valid" or "preferred" for a time longer than
2369                                  * (TEMP_VALID_LIFETIME) or
2370                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2371                                  * respectively.
2372                                  */
2373                                 age = (now - ift->cstamp) / HZ;
2374                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2375                                 if (max_valid < 0)
2376                                         max_valid = 0;
2377
2378                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2379                                                in6_dev->cnf.max_desync_factor -
2380                                                age;
2381                                 if (max_prefered < 0)
2382                                         max_prefered = 0;
2383
2384                                 if (valid_lft > max_valid)
2385                                         valid_lft = max_valid;
2386
2387                                 if (prefered_lft > max_prefered)
2388                                         prefered_lft = max_prefered;
2389
2390                                 spin_lock(&ift->lock);
2391                                 flags = ift->flags;
2392                                 ift->valid_lft = valid_lft;
2393                                 ift->prefered_lft = prefered_lft;
2394                                 ift->tstamp = now;
2395                                 if (prefered_lft > 0)
2396                                         ift->flags &= ~IFA_F_DEPRECATED;
2397
2398                                 spin_unlock(&ift->lock);
2399                                 if (!(flags&IFA_F_TENTATIVE))
2400                                         ipv6_ifa_notify(0, ift);
2401                         }
2402
2403                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2404                                 /*
2405                                  * When a new public address is created as
2406                                  * described in [ADDRCONF], also create a new
2407                                  * temporary address. Also create a temporary
2408                                  * address if it's enabled but no temporary
2409                                  * address currently exists.
2410                                  */
2411                                 read_unlock_bh(&in6_dev->lock);
2412                                 ipv6_create_tempaddr(ifp, NULL);
2413                         } else {
2414                                 read_unlock_bh(&in6_dev->lock);
2415                         }
2416 #endif
2417                         in6_ifa_put(ifp);
2418                         addrconf_verify(0);
2419                 }
2420         }
2421         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2422         in6_dev_put(in6_dev);
2423 }
2424
2425 /*
2426  *      Set destination address.
2427  *      Special case for SIT interfaces where we create a new "virtual"
2428  *      device.
2429  */
2430 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2431 {
2432         struct in6_ifreq ireq;
2433         struct net_device *dev;
2434         int err = -EINVAL;
2435
2436         rtnl_lock();
2437
2438         err = -EFAULT;
2439         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2440                 goto err_exit;
2441
2442         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2443
2444         err = -ENODEV;
2445         if (dev == NULL)
2446                 goto err_exit;
2447
2448 #if IS_ENABLED(CONFIG_IPV6_SIT)
2449         if (dev->type == ARPHRD_SIT) {
2450                 const struct net_device_ops *ops = dev->netdev_ops;
2451                 struct ifreq ifr;
2452                 struct ip_tunnel_parm p;
2453
2454                 err = -EADDRNOTAVAIL;
2455                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2456                         goto err_exit;
2457
2458                 memset(&p, 0, sizeof(p));
2459                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2460                 p.iph.saddr = 0;
2461                 p.iph.version = 4;
2462                 p.iph.ihl = 5;
2463                 p.iph.protocol = IPPROTO_IPV6;
2464                 p.iph.ttl = 64;
2465                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2466
2467                 if (ops->ndo_do_ioctl) {
2468                         mm_segment_t oldfs = get_fs();
2469
2470                         set_fs(KERNEL_DS);
2471                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2472                         set_fs(oldfs);
2473                 } else
2474                         err = -EOPNOTSUPP;
2475
2476                 if (err == 0) {
2477                         err = -ENOBUFS;
2478                         dev = __dev_get_by_name(net, p.name);
2479                         if (!dev)
2480                                 goto err_exit;
2481                         err = dev_open(dev);
2482                 }
2483         }
2484 #endif
2485
2486 err_exit:
2487         rtnl_unlock();
2488         return err;
2489 }
2490
2491 /*
2492  *      Manual configuration of address on an interface
2493  */
2494 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2495                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2496                           __u32 valid_lft)
2497 {
2498         struct inet6_ifaddr *ifp;
2499         struct inet6_dev *idev;
2500         struct net_device *dev;
2501         int scope;
2502         u32 flags;
2503         clock_t expires;
2504         unsigned long timeout;
2505
2506         ASSERT_RTNL();
2507
2508         if (plen > 128)
2509                 return -EINVAL;
2510
2511         /* check the lifetime */
2512         if (!valid_lft || prefered_lft > valid_lft)
2513                 return -EINVAL;
2514
2515         dev = __dev_get_by_index(net, ifindex);
2516         if (!dev)
2517                 return -ENODEV;
2518
2519         idev = addrconf_add_dev(dev);
2520         if (IS_ERR(idev))
2521                 return PTR_ERR(idev);
2522
2523         scope = ipv6_addr_scope(pfx);
2524
2525         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2526         if (addrconf_finite_timeout(timeout)) {
2527                 expires = jiffies_to_clock_t(timeout * HZ);
2528                 valid_lft = timeout;
2529                 flags = RTF_EXPIRES;
2530         } else {
2531                 expires = 0;
2532                 flags = 0;
2533                 ifa_flags |= IFA_F_PERMANENT;
2534         }
2535
2536         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2537         if (addrconf_finite_timeout(timeout)) {
2538                 if (timeout == 0)
2539                         ifa_flags |= IFA_F_DEPRECATED;
2540                 prefered_lft = timeout;
2541         }
2542
2543         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2544
2545         if (!IS_ERR(ifp)) {
2546                 spin_lock_bh(&ifp->lock);
2547                 ifp->valid_lft = valid_lft;
2548                 ifp->prefered_lft = prefered_lft;
2549                 ifp->tstamp = jiffies;
2550                 spin_unlock_bh(&ifp->lock);
2551
2552                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2553                                       expires, flags);
2554                 /*
2555                  * Note that section 3.1 of RFC 4429 indicates
2556                  * that the Optimistic flag should not be set for
2557                  * manually configured addresses
2558                  */
2559                 addrconf_dad_start(ifp);
2560                 in6_ifa_put(ifp);
2561                 addrconf_verify(0);
2562                 return 0;
2563         }
2564
2565         return PTR_ERR(ifp);
2566 }
2567
2568 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2569                           unsigned int plen)
2570 {
2571         struct inet6_ifaddr *ifp;
2572         struct inet6_dev *idev;
2573         struct net_device *dev;
2574
2575         if (plen > 128)
2576                 return -EINVAL;
2577
2578         dev = __dev_get_by_index(net, ifindex);
2579         if (!dev)
2580                 return -ENODEV;
2581
2582         if ((idev = __in6_dev_get(dev)) == NULL)
2583                 return -ENXIO;
2584
2585         read_lock_bh(&idev->lock);
2586         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2587                 if (ifp->prefix_len == plen &&
2588                     ipv6_addr_equal(pfx, &ifp->addr)) {
2589                         in6_ifa_hold(ifp);
2590                         read_unlock_bh(&idev->lock);
2591
2592                         ipv6_del_addr(ifp);
2593
2594                         /* If the last address is deleted administratively,
2595                            disable IPv6 on this interface.
2596                          */
2597                         if (list_empty(&idev->addr_list))
2598                                 addrconf_ifdown(idev->dev, 1);
2599                         return 0;
2600                 }
2601         }
2602         read_unlock_bh(&idev->lock);
2603         return -EADDRNOTAVAIL;
2604 }
2605
2606
2607 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2608 {
2609         struct in6_ifreq ireq;
2610         int err;
2611
2612         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2613                 return -EPERM;
2614
2615         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2616                 return -EFAULT;
2617
2618         rtnl_lock();
2619         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2620                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2621                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2622         rtnl_unlock();
2623         return err;
2624 }
2625
2626 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2627 {
2628         struct in6_ifreq ireq;
2629         int err;
2630
2631         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2632                 return -EPERM;
2633
2634         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2635                 return -EFAULT;
2636
2637         rtnl_lock();
2638         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2639                              ireq.ifr6_prefixlen);
2640         rtnl_unlock();
2641         return err;
2642 }
2643
2644 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2645                      int plen, int scope)
2646 {
2647         struct inet6_ifaddr *ifp;
2648
2649         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2650         if (!IS_ERR(ifp)) {
2651                 spin_lock_bh(&ifp->lock);
2652                 ifp->flags &= ~IFA_F_TENTATIVE;
2653                 spin_unlock_bh(&ifp->lock);
2654                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2655                 in6_ifa_put(ifp);
2656         }
2657 }
2658
2659 #if IS_ENABLED(CONFIG_IPV6_SIT)
2660 static void sit_add_v4_addrs(struct inet6_dev *idev)
2661 {
2662         struct in6_addr addr;
2663         struct net_device *dev;
2664         struct net *net = dev_net(idev->dev);
2665         int scope;
2666
2667         ASSERT_RTNL();
2668
2669         memset(&addr, 0, sizeof(struct in6_addr));
2670         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2671
2672         if (idev->dev->flags&IFF_POINTOPOINT) {
2673                 addr.s6_addr32[0] = htonl(0xfe800000);
2674                 scope = IFA_LINK;
2675         } else {
2676                 scope = IPV6_ADDR_COMPATv4;
2677         }
2678
2679         if (addr.s6_addr32[3]) {
2680                 add_addr(idev, &addr, 128, scope);
2681                 return;
2682         }
2683
2684         for_each_netdev(net, dev) {
2685                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2686                 if (in_dev && (dev->flags & IFF_UP)) {
2687                         struct in_ifaddr *ifa;
2688
2689                         int flag = scope;
2690
2691                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2692                                 int plen;
2693
2694                                 addr.s6_addr32[3] = ifa->ifa_local;
2695
2696                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2697                                         continue;
2698                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2699                                         if (idev->dev->flags&IFF_POINTOPOINT)
2700                                                 continue;
2701                                         flag |= IFA_HOST;
2702                                 }
2703                                 if (idev->dev->flags&IFF_POINTOPOINT)
2704                                         plen = 64;
2705                                 else
2706                                         plen = 96;
2707
2708                                 add_addr(idev, &addr, plen, flag);
2709                         }
2710                 }
2711         }
2712 }
2713 #endif
2714
2715 static void init_loopback(struct net_device *dev)
2716 {
2717         struct inet6_dev  *idev;
2718         struct net_device *sp_dev;
2719         struct inet6_ifaddr *sp_ifa;
2720         struct rt6_info *sp_rt;
2721
2722         /* ::1 */
2723
2724         ASSERT_RTNL();
2725
2726         if ((idev = ipv6_find_idev(dev)) == NULL) {
2727                 pr_debug("%s: add_dev failed\n", __func__);
2728                 return;
2729         }
2730
2731         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2732
2733         /* Add routes to other interface's IPv6 addresses */
2734         for_each_netdev(dev_net(dev), sp_dev) {
2735                 if (!strcmp(sp_dev->name, dev->name))
2736                         continue;
2737
2738                 idev = __in6_dev_get(sp_dev);
2739                 if (!idev)
2740                         continue;
2741
2742                 read_lock_bh(&idev->lock);
2743                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2744
2745                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2746                                 continue;
2747
2748                         if (sp_ifa->rt) {
2749                                 /* This dst has been added to garbage list when
2750                                  * lo device down, release this obsolete dst and
2751                                  * reallocate a new router for ifa.
2752                                  */
2753                                 if (sp_ifa->rt->dst.obsolete > 0) {
2754                                         ip6_rt_put(sp_ifa->rt);
2755                                         sp_ifa->rt = NULL;
2756                                 } else {
2757                                         continue;
2758                                 }
2759                         }
2760
2761                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2762
2763                         /* Failure cases are ignored */
2764                         if (!IS_ERR(sp_rt)) {
2765                                 sp_ifa->rt = sp_rt;
2766                                 ip6_ins_rt(sp_rt);
2767                         }
2768                 }
2769                 read_unlock_bh(&idev->lock);
2770         }
2771 }
2772
2773 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2774 {
2775         struct inet6_ifaddr *ifp;
2776         u32 addr_flags = IFA_F_PERMANENT;
2777
2778 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2779         if (idev->cnf.optimistic_dad &&
2780             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2781                 addr_flags |= IFA_F_OPTIMISTIC;
2782 #endif
2783
2784
2785         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2786         if (!IS_ERR(ifp)) {
2787                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2788                 addrconf_dad_start(ifp);
2789                 in6_ifa_put(ifp);
2790         }
2791 }
2792
2793 static void addrconf_dev_config(struct net_device *dev)
2794 {
2795         struct in6_addr addr;
2796         struct inet6_dev *idev;
2797
2798         ASSERT_RTNL();
2799
2800         if ((dev->type != ARPHRD_ETHER) &&
2801             (dev->type != ARPHRD_FDDI) &&
2802             (dev->type != ARPHRD_ARCNET) &&
2803             (dev->type != ARPHRD_INFINIBAND) &&
2804             (dev->type != ARPHRD_IEEE802154) &&
2805             (dev->type != ARPHRD_IEEE1394)) {
2806                 /* Alas, we support only Ethernet autoconfiguration. */
2807                 return;
2808         }
2809
2810         idev = addrconf_add_dev(dev);
2811         if (IS_ERR(idev))
2812                 return;
2813
2814         memset(&addr, 0, sizeof(struct in6_addr));
2815         addr.s6_addr32[0] = htonl(0xFE800000);
2816
2817         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2818                 addrconf_add_linklocal(idev, &addr);
2819 }
2820
2821 #if IS_ENABLED(CONFIG_IPV6_SIT)
2822 static void addrconf_sit_config(struct net_device *dev)
2823 {
2824         struct inet6_dev *idev;
2825
2826         ASSERT_RTNL();
2827
2828         /*
2829          * Configure the tunnel with one of our IPv4
2830          * addresses... we should configure all of
2831          * our v4 addrs in the tunnel
2832          */
2833
2834         if ((idev = ipv6_find_idev(dev)) == NULL) {
2835                 pr_debug("%s: add_dev failed\n", __func__);
2836                 return;
2837         }
2838
2839         if (dev->priv_flags & IFF_ISATAP) {
2840                 struct in6_addr addr;
2841
2842                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2843                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2844                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2845                         addrconf_add_linklocal(idev, &addr);
2846                 return;
2847         }
2848
2849         sit_add_v4_addrs(idev);
2850
2851         if (dev->flags&IFF_POINTOPOINT)
2852                 addrconf_add_mroute(dev);
2853         else
2854                 sit_route_add(dev);
2855 }
2856 #endif
2857
2858 #if IS_ENABLED(CONFIG_NET_IPGRE)
2859 static void addrconf_gre_config(struct net_device *dev)
2860 {
2861         struct inet6_dev *idev;
2862         struct in6_addr addr;
2863
2864         pr_info("%s(%s)\n", __func__, dev->name);
2865
2866         ASSERT_RTNL();
2867
2868         if ((idev = ipv6_find_idev(dev)) == NULL) {
2869                 pr_debug("%s: add_dev failed\n", __func__);
2870                 return;
2871         }
2872
2873         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2874         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2875
2876         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2877                 addrconf_add_linklocal(idev, &addr);
2878 }
2879 #endif
2880
2881 static inline int
2882 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2883 {
2884         struct in6_addr lladdr;
2885
2886         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2887                 addrconf_add_linklocal(idev, &lladdr);
2888                 return 0;
2889         }
2890         return -1;
2891 }
2892
2893 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2894 {
2895         struct net_device *link_dev;
2896         struct net *net = dev_net(idev->dev);
2897
2898         /* first try to inherit the link-local address from the link device */
2899         if (idev->dev->iflink &&
2900             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2901                 if (!ipv6_inherit_linklocal(idev, link_dev))
2902                         return;
2903         }
2904         /* then try to inherit it from any device */
2905         for_each_netdev(net, link_dev) {
2906                 if (!ipv6_inherit_linklocal(idev, link_dev))
2907                         return;
2908         }
2909         pr_debug("init ip6-ip6: add_linklocal failed\n");
2910 }
2911
2912 /*
2913  * Autoconfigure tunnel with a link-local address so routing protocols,
2914  * DHCPv6, MLD etc. can be run over the virtual link
2915  */
2916
2917 static void addrconf_ip6_tnl_config(struct net_device *dev)
2918 {
2919         struct inet6_dev *idev;
2920
2921         ASSERT_RTNL();
2922
2923         idev = addrconf_add_dev(dev);
2924         if (IS_ERR(idev)) {
2925                 pr_debug("init ip6-ip6: add_dev failed\n");
2926                 return;
2927         }
2928         ip6_tnl_add_linklocal(idev);
2929 }
2930
2931 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2932                            void *data)
2933 {
2934         struct net_device *dev = (struct net_device *) data;
2935         struct inet6_dev *idev = __in6_dev_get(dev);
2936         int run_pending = 0;
2937         int err;
2938
2939         switch (event) {
2940         case NETDEV_REGISTER:
2941                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2942                         idev = ipv6_add_dev(dev);
2943                         if (!idev)
2944                                 return notifier_from_errno(-ENOMEM);
2945                 }
2946                 break;
2947
2948         case NETDEV_UP:
2949         case NETDEV_CHANGE:
2950                 if (dev->flags & IFF_SLAVE)
2951                         break;
2952
2953                 if (event == NETDEV_UP) {
2954                         if (!addrconf_qdisc_ok(dev)) {
2955                                 /* device is not ready yet. */
2956                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2957                                         dev->name);
2958                                 break;
2959                         }
2960
2961                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2962                                 idev = ipv6_add_dev(dev);
2963
2964                         if (idev) {
2965                                 idev->if_flags |= IF_READY;
2966                                 run_pending = 1;
2967                         }
2968                 } else {
2969                         if (!addrconf_qdisc_ok(dev)) {
2970                                 /* device is still not ready. */
2971                                 break;
2972                         }
2973
2974                         if (idev) {
2975                                 if (idev->if_flags & IF_READY)
2976                                         /* device is already configured. */
2977                                         break;
2978                                 idev->if_flags |= IF_READY;
2979                         }
2980
2981                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2982                                 dev->name);
2983
2984                         run_pending = 1;
2985                 }
2986
2987                 switch (dev->type) {
2988 #if IS_ENABLED(CONFIG_IPV6_SIT)
2989                 case ARPHRD_SIT:
2990                         addrconf_sit_config(dev);
2991                         break;
2992 #endif
2993 #if IS_ENABLED(CONFIG_NET_IPGRE)
2994                 case ARPHRD_IPGRE:
2995                         addrconf_gre_config(dev);
2996                         break;
2997 #endif
2998                 case ARPHRD_TUNNEL6:
2999                         addrconf_ip6_tnl_config(dev);
3000                         break;
3001                 case ARPHRD_LOOPBACK:
3002                         init_loopback(dev);
3003                         break;
3004
3005                 default:
3006                         addrconf_dev_config(dev);
3007                         break;
3008                 }
3009
3010                 if (idev) {
3011                         if (run_pending)
3012                                 addrconf_dad_run(idev);
3013
3014                         /*
3015                          * If the MTU changed during the interface down,
3016                          * when the interface up, the changed MTU must be
3017                          * reflected in the idev as well as routers.
3018                          */
3019                         if (idev->cnf.mtu6 != dev->mtu &&
3020                             dev->mtu >= IPV6_MIN_MTU) {
3021                                 rt6_mtu_change(dev, dev->mtu);
3022                                 idev->cnf.mtu6 = dev->mtu;
3023                         }
3024                         idev->tstamp = jiffies;
3025                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3026
3027                         /*
3028                          * If the changed mtu during down is lower than
3029                          * IPV6_MIN_MTU stop IPv6 on this interface.
3030                          */
3031                         if (dev->mtu < IPV6_MIN_MTU)
3032                                 addrconf_ifdown(dev, 1);
3033                 }
3034                 break;
3035
3036         case NETDEV_CHANGEMTU:
3037                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3038                         rt6_mtu_change(dev, dev->mtu);
3039                         idev->cnf.mtu6 = dev->mtu;
3040                         break;
3041                 }
3042
3043                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3044                         idev = ipv6_add_dev(dev);
3045                         if (idev)
3046                                 break;
3047                 }
3048
3049                 /*
3050                  * MTU falled under IPV6_MIN_MTU.
3051                  * Stop IPv6 on this interface.
3052                  */
3053
3054         case NETDEV_DOWN:
3055         case NETDEV_UNREGISTER:
3056                 /*
3057                  *      Remove all addresses from this interface.
3058                  */
3059                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3060                 break;
3061
3062         case NETDEV_CHANGENAME:
3063                 if (idev) {
3064                         snmp6_unregister_dev(idev);
3065                         addrconf_sysctl_unregister(idev);
3066                         addrconf_sysctl_register(idev);
3067                         err = snmp6_register_dev(idev);
3068                         if (err)
3069                                 return notifier_from_errno(err);
3070                 }
3071                 break;
3072
3073         case NETDEV_PRE_TYPE_CHANGE:
3074         case NETDEV_POST_TYPE_CHANGE:
3075                 addrconf_type_change(dev, event);
3076                 break;
3077         }
3078
3079         return NOTIFY_OK;
3080 }
3081
3082 /*
3083  *      addrconf module should be notified of a device going up
3084  */
3085 static struct notifier_block ipv6_dev_notf = {
3086         .notifier_call = addrconf_notify,
3087 };
3088
3089 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3090 {
3091         struct inet6_dev *idev;
3092         ASSERT_RTNL();
3093
3094         idev = __in6_dev_get(dev);
3095
3096         if (event == NETDEV_POST_TYPE_CHANGE)
3097                 ipv6_mc_remap(idev);
3098         else if (event == NETDEV_PRE_TYPE_CHANGE)
3099                 ipv6_mc_unmap(idev);
3100 }
3101
3102 static int addrconf_ifdown(struct net_device *dev, int how)
3103 {
3104         struct net *net = dev_net(dev);
3105         struct inet6_dev *idev;
3106         struct inet6_ifaddr *ifa;
3107         int state, i;
3108
3109         ASSERT_RTNL();
3110
3111         rt6_ifdown(net, dev);
3112         neigh_ifdown(&nd_tbl, dev);
3113
3114         idev = __in6_dev_get(dev);
3115         if (idev == NULL)
3116                 return -ENODEV;
3117
3118         /*
3119          * Step 1: remove reference to ipv6 device from parent device.
3120          *         Do not dev_put!
3121          */
3122         if (how) {
3123                 idev->dead = 1;
3124
3125                 /* protected by rtnl_lock */
3126                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3127
3128                 /* Step 1.5: remove snmp6 entry */
3129                 snmp6_unregister_dev(idev);
3130
3131         }
3132
3133         /* Step 2: clear hash table */
3134         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3135                 struct hlist_head *h = &inet6_addr_lst[i];
3136
3137                 spin_lock_bh(&addrconf_hash_lock);
3138         restart:
3139                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3140                         if (ifa->idev == idev) {
3141                                 hlist_del_init_rcu(&ifa->addr_lst);
3142                                 addrconf_del_timer(ifa);
3143                                 goto restart;
3144                         }
3145                 }
3146                 spin_unlock_bh(&addrconf_hash_lock);
3147         }
3148
3149         write_lock_bh(&idev->lock);
3150
3151         /* Step 2: clear flags for stateless addrconf */
3152         if (!how)
3153                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3154
3155 #ifdef CONFIG_IPV6_PRIVACY
3156         if (how && del_timer(&idev->regen_timer))
3157                 in6_dev_put(idev);
3158
3159         /* Step 3: clear tempaddr list */
3160         while (!list_empty(&idev->tempaddr_list)) {
3161                 ifa = list_first_entry(&idev->tempaddr_list,
3162                                        struct inet6_ifaddr, tmp_list);
3163                 list_del(&ifa->tmp_list);
3164                 write_unlock_bh(&idev->lock);
3165                 spin_lock_bh(&ifa->lock);
3166
3167                 if (ifa->ifpub) {
3168                         in6_ifa_put(ifa->ifpub);
3169                         ifa->ifpub = NULL;
3170                 }
3171                 spin_unlock_bh(&ifa->lock);
3172                 in6_ifa_put(ifa);
3173                 write_lock_bh(&idev->lock);
3174         }
3175 #endif
3176
3177         while (!list_empty(&idev->addr_list)) {
3178                 ifa = list_first_entry(&idev->addr_list,
3179                                        struct inet6_ifaddr, if_list);
3180                 addrconf_del_timer(ifa);
3181
3182                 list_del(&ifa->if_list);
3183
3184                 write_unlock_bh(&idev->lock);
3185
3186                 spin_lock_bh(&ifa->state_lock);
3187                 state = ifa->state;
3188                 ifa->state = INET6_IFADDR_STATE_DEAD;
3189                 spin_unlock_bh(&ifa->state_lock);
3190
3191                 if (state != INET6_IFADDR_STATE_DEAD) {
3192                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3193                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3194                 }
3195                 in6_ifa_put(ifa);
3196
3197                 write_lock_bh(&idev->lock);
3198         }
3199
3200         write_unlock_bh(&idev->lock);
3201
3202         /* Step 5: Discard multicast list */
3203         if (how)
3204                 ipv6_mc_destroy_dev(idev);
3205         else
3206                 ipv6_mc_down(idev);
3207
3208         idev->tstamp = jiffies;
3209
3210         /* Last: Shot the device (if unregistered) */
3211         if (how) {
3212                 addrconf_sysctl_unregister(idev);
3213                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3214                 neigh_ifdown(&nd_tbl, dev);
3215                 in6_dev_put(idev);
3216         }
3217         return 0;
3218 }
3219
3220 static void addrconf_rs_timer(unsigned long data)
3221 {
3222         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3223         struct inet6_dev *idev = ifp->idev;
3224
3225         read_lock(&idev->lock);
3226         if (idev->dead || !(idev->if_flags & IF_READY))
3227                 goto out;
3228
3229         if (!ipv6_accept_ra(idev))
3230                 goto out;
3231
3232         /* Announcement received after solicitation was sent */
3233         if (idev->if_flags & IF_RA_RCVD)
3234                 goto out;
3235
3236         spin_lock(&ifp->lock);
3237         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3238                 /* The wait after the last probe can be shorter */
3239                 addrconf_mod_timer(ifp, AC_RS,
3240                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3241                                    idev->cnf.rtr_solicit_delay :
3242                                    idev->cnf.rtr_solicit_interval);
3243                 spin_unlock(&ifp->lock);
3244
3245                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3246         } else {
3247                 spin_unlock(&ifp->lock);
3248                 /*
3249                  * Note: we do not support deprecated "all on-link"
3250                  * assumption any longer.
3251                  */
3252                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3253         }
3254
3255 out:
3256         read_unlock(&idev->lock);
3257         in6_ifa_put(ifp);
3258 }
3259
3260 /*
3261  *      Duplicate Address Detection
3262  */
3263 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3264 {
3265         unsigned long rand_num;
3266         struct inet6_dev *idev = ifp->idev;
3267
3268         if (ifp->flags & IFA_F_OPTIMISTIC)
3269                 rand_num = 0;
3270         else
3271                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3272
3273         ifp->probes = idev->cnf.dad_transmits;
3274         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3275 }
3276
3277 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3278 {
3279         struct inet6_dev *idev = ifp->idev;
3280         struct net_device *dev = idev->dev;
3281
3282         addrconf_join_solict(dev, &ifp->addr);
3283
3284         net_srandom(ifp->addr.s6_addr32[3]);
3285
3286         read_lock_bh(&idev->lock);
3287         spin_lock(&ifp->lock);
3288         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3289                 goto out;
3290
3291         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3292             idev->cnf.accept_dad < 1 ||
3293             !(ifp->flags&IFA_F_TENTATIVE) ||
3294             ifp->flags & IFA_F_NODAD) {
3295                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3296                 spin_unlock(&ifp->lock);
3297                 read_unlock_bh(&idev->lock);
3298
3299                 addrconf_dad_completed(ifp);
3300                 return;
3301         }
3302
3303         if (!(idev->if_flags & IF_READY)) {
3304                 spin_unlock(&ifp->lock);
3305                 read_unlock_bh(&idev->lock);
3306                 /*
3307                  * If the device is not ready:
3308                  * - keep it tentative if it is a permanent address.
3309                  * - otherwise, kill it.
3310                  */
3311                 in6_ifa_hold(ifp);
3312                 addrconf_dad_stop(ifp, 0);
3313                 return;
3314         }
3315
3316         /*
3317          * Optimistic nodes can start receiving
3318          * Frames right away
3319          */
3320         if (ifp->flags & IFA_F_OPTIMISTIC) {
3321                 ip6_ins_rt(ifp->rt);
3322                 if (ipv6_use_optimistic_addr(idev)) {
3323                         /* Because optimistic nodes can use this address,
3324                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3325                          */
3326                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3327                 }
3328         }
3329
3330         addrconf_dad_kick(ifp);
3331 out:
3332         spin_unlock(&ifp->lock);
3333         read_unlock_bh(&idev->lock);
3334 }
3335
3336 static void addrconf_dad_timer(unsigned long data)
3337 {
3338         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3339         struct inet6_dev *idev = ifp->idev;
3340         struct in6_addr mcaddr;
3341
3342         if (!ifp->probes && addrconf_dad_end(ifp))
3343                 goto out;
3344
3345         read_lock(&idev->lock);
3346         if (idev->dead || !(idev->if_flags & IF_READY)) {
3347                 read_unlock(&idev->lock);
3348                 goto out;
3349         }
3350
3351         spin_lock(&ifp->lock);
3352         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3353                 spin_unlock(&ifp->lock);
3354                 read_unlock(&idev->lock);
3355                 goto out;
3356         }
3357
3358         if (ifp->probes == 0) {
3359                 /*
3360                  * DAD was successful
3361                  */
3362
3363                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3364                 spin_unlock(&ifp->lock);
3365                 read_unlock(&idev->lock);
3366
3367                 addrconf_dad_completed(ifp);
3368
3369                 goto out;
3370         }
3371
3372         ifp->probes--;
3373         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3374         spin_unlock(&ifp->lock);
3375         read_unlock(&idev->lock);
3376
3377         /* send a neighbour solicitation for our addr */
3378         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3379         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3380 out:
3381         in6_ifa_put(ifp);
3382 }
3383
3384 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3385 {
3386         struct net_device *dev = ifp->idev->dev;
3387
3388         /*
3389          *      Configure the address for reception. Now it is valid.
3390          */
3391
3392         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3393
3394         /* If added prefix is link local and we are prepared to process
3395            router advertisements, start sending router solicitations.
3396          */
3397
3398         if (ipv6_accept_ra(ifp->idev) &&
3399             ifp->idev->cnf.rtr_solicits > 0 &&
3400             (dev->flags&IFF_LOOPBACK) == 0 &&
3401             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3402                 /*
3403                  *      If a host as already performed a random delay
3404                  *      [...] as part of DAD [...] there is no need
3405                  *      to delay again before sending the first RS
3406                  */
3407                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3408
3409                 spin_lock_bh(&ifp->lock);
3410                 ifp->probes = 1;
3411                 ifp->idev->if_flags |= IF_RS_SENT;
3412                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3413                 spin_unlock_bh(&ifp->lock);
3414         }
3415 }
3416
3417 static void addrconf_dad_run(struct inet6_dev *idev)
3418 {
3419         struct inet6_ifaddr *ifp;
3420
3421         read_lock_bh(&idev->lock);
3422         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3423                 spin_lock(&ifp->lock);
3424                 if (ifp->flags & IFA_F_TENTATIVE &&
3425                     ifp->state == INET6_IFADDR_STATE_DAD)
3426                         addrconf_dad_kick(ifp);
3427                 spin_unlock(&ifp->lock);
3428         }
3429         read_unlock_bh(&idev->lock);
3430 }
3431
3432 #ifdef CONFIG_PROC_FS
3433 struct if6_iter_state {
3434         struct seq_net_private p;
3435         int bucket;
3436         int offset;
3437 };
3438
3439 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3440 {
3441         struct inet6_ifaddr *ifa = NULL;
3442         struct if6_iter_state *state = seq->private;
3443         struct net *net = seq_file_net(seq);
3444         int p = 0;
3445
3446         /* initial bucket if pos is 0 */
3447         if (pos == 0) {
3448                 state->bucket = 0;
3449                 state->offset = 0;
3450         }
3451
3452         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3453                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3454                                          addr_lst) {
3455                         if (!net_eq(dev_net(ifa->idev->dev), net))
3456                                 continue;
3457                         /* sync with offset */
3458                         if (p < state->offset) {
3459                                 p++;
3460                                 continue;
3461                         }
3462                         state->offset++;
3463                         return ifa;
3464                 }
3465
3466                 /* prepare for next bucket */
3467                 state->offset = 0;
3468                 p = 0;
3469         }
3470         return NULL;
3471 }
3472
3473 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3474                                          struct inet6_ifaddr *ifa)
3475 {
3476         struct if6_iter_state *state = seq->private;
3477         struct net *net = seq_file_net(seq);
3478
3479         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3480                 if (!net_eq(dev_net(ifa->idev->dev), net))
3481                         continue;
3482                 state->offset++;
3483                 return ifa;
3484         }
3485
3486         while (++state->bucket < IN6_ADDR_HSIZE) {
3487                 state->offset = 0;
3488                 hlist_for_each_entry_rcu_bh(ifa,
3489                                      &inet6_addr_lst[state->bucket], addr_lst) {
3490                         if (!net_eq(dev_net(ifa->idev->dev), net))
3491                                 continue;
3492                         state->offset++;
3493                         return ifa;
3494                 }
3495         }
3496
3497         return NULL;
3498 }
3499
3500 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3501         __acquires(rcu_bh)
3502 {
3503         rcu_read_lock_bh();
3504         return if6_get_first(seq, *pos);
3505 }
3506
3507 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3508 {
3509         struct inet6_ifaddr *ifa;
3510
3511         ifa = if6_get_next(seq, v);
3512         ++*pos;
3513         return ifa;
3514 }
3515
3516 static void if6_seq_stop(struct seq_file *seq, void *v)
3517         __releases(rcu_bh)
3518 {
3519         rcu_read_unlock_bh();
3520 }
3521
3522 static int if6_seq_show(struct seq_file *seq, void *v)
3523 {
3524         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3525         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3526                    &ifp->addr,
3527                    ifp->idev->dev->ifindex,
3528                    ifp->prefix_len,
3529                    ifp->scope,
3530                    ifp->flags,
3531                    ifp->idev->dev->name);
3532         return 0;
3533 }
3534
3535 static const struct seq_operations if6_seq_ops = {
3536         .start  = if6_seq_start,
3537         .next   = if6_seq_next,
3538         .show   = if6_seq_show,
3539         .stop   = if6_seq_stop,
3540 };
3541
3542 static int if6_seq_open(struct inode *inode, struct file *file)
3543 {
3544         return seq_open_net(inode, file, &if6_seq_ops,
3545                             sizeof(struct if6_iter_state));
3546 }
3547
3548 static const struct file_operations if6_fops = {
3549         .owner          = THIS_MODULE,
3550         .open           = if6_seq_open,
3551         .read           = seq_read,
3552         .llseek         = seq_lseek,
3553         .release        = seq_release_net,
3554 };
3555
3556 static int __net_init if6_proc_net_init(struct net *net)
3557 {
3558         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3559                 return -ENOMEM;
3560         return 0;
3561 }
3562
3563 static void __net_exit if6_proc_net_exit(struct net *net)
3564 {
3565         remove_proc_entry("if_inet6", net->proc_net);
3566 }
3567
3568 static struct pernet_operations if6_proc_net_ops = {
3569        .init = if6_proc_net_init,
3570        .exit = if6_proc_net_exit,
3571 };
3572
3573 int __init if6_proc_init(void)
3574 {
3575         return register_pernet_subsys(&if6_proc_net_ops);
3576 }
3577
3578 void if6_proc_exit(void)
3579 {
3580         unregister_pernet_subsys(&if6_proc_net_ops);
3581 }
3582 #endif  /* CONFIG_PROC_FS */
3583
3584 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3585 /* Check if address is a home address configured on any interface. */
3586 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3587 {
3588         int ret = 0;
3589         struct inet6_ifaddr *ifp = NULL;
3590         unsigned int hash = inet6_addr_hash(addr);
3591
3592         rcu_read_lock_bh();
3593         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3594                 if (!net_eq(dev_net(ifp->idev->dev), net))
3595                         continue;
3596                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3597                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3598                         ret = 1;
3599                         break;
3600                 }
3601         }
3602         rcu_read_unlock_bh();
3603         return ret;
3604 }
3605 #endif
3606
3607 /*
3608  *      Periodic address status verification
3609  */
3610
3611 static void addrconf_verify(unsigned long foo)
3612 {
3613         unsigned long now, next, next_sec, next_sched;
3614         struct inet6_ifaddr *ifp;
3615         int i;
3616
3617         rcu_read_lock_bh();
3618         spin_lock(&addrconf_verify_lock);
3619         now = jiffies;
3620         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3621
3622         del_timer(&addr_chk_timer);
3623
3624         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3625 restart:
3626                 hlist_for_each_entry_rcu_bh(ifp,
3627                                          &inet6_addr_lst[i], addr_lst) {
3628                         unsigned long age;
3629
3630                         if (ifp->flags & IFA_F_PERMANENT)
3631                                 continue;
3632
3633                         spin_lock(&ifp->lock);
3634                         /* We try to batch several events at once. */
3635                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3636
3637                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3638                             age >= ifp->valid_lft) {
3639                                 spin_unlock(&ifp->lock);
3640                                 in6_ifa_hold(ifp);
3641                                 ipv6_del_addr(ifp);
3642                                 goto restart;
3643                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3644                                 spin_unlock(&ifp->lock);
3645                                 continue;
3646                         } else if (age >= ifp->prefered_lft) {
3647                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3648                                 int deprecate = 0;
3649
3650                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3651                                         deprecate = 1;
3652                                         ifp->flags |= IFA_F_DEPRECATED;
3653                                 }
3654
3655                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3656                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3657
3658                                 spin_unlock(&ifp->lock);
3659
3660                                 if (deprecate) {
3661                                         in6_ifa_hold(ifp);
3662
3663                                         ipv6_ifa_notify(0, ifp);
3664                                         in6_ifa_put(ifp);
3665                                         goto restart;
3666                                 }
3667 #ifdef CONFIG_IPV6_PRIVACY
3668                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3669                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3670                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3671                                         ifp->idev->cnf.dad_transmits *
3672                                         ifp->idev->nd_parms->retrans_time / HZ;
3673
3674                                 if (age >= ifp->prefered_lft - regen_advance) {
3675                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3676                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3677                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3678                                         if (!ifp->regen_count && ifpub) {
3679                                                 ifp->regen_count++;
3680                                                 in6_ifa_hold(ifp);
3681                                                 in6_ifa_hold(ifpub);
3682                                                 spin_unlock(&ifp->lock);
3683
3684                                                 spin_lock(&ifpub->lock);
3685                                                 ifpub->regen_count = 0;
3686                                                 spin_unlock(&ifpub->lock);
3687                                                 ipv6_create_tempaddr(ifpub, ifp);
3688                                                 in6_ifa_put(ifpub);
3689                                                 in6_ifa_put(ifp);
3690                                                 goto restart;
3691                                         }
3692                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3693                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3694                                 spin_unlock(&ifp->lock);
3695 #endif
3696                         } else {
3697                                 /* ifp->prefered_lft <= ifp->valid_lft */
3698                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3699                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3700                                 spin_unlock(&ifp->lock);
3701                         }
3702                 }
3703         }
3704
3705         next_sec = round_jiffies_up(next);
3706         next_sched = next;
3707
3708         /* If rounded timeout is accurate enough, accept it. */
3709         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3710                 next_sched = next_sec;
3711
3712         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3713         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3714                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3715
3716         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3717               now, next, next_sec, next_sched));
3718
3719         addr_chk_timer.expires = next_sched;
3720         add_timer(&addr_chk_timer);
3721         spin_unlock(&addrconf_verify_lock);
3722         rcu_read_unlock_bh();
3723 }
3724
3725 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3726 {
3727         struct in6_addr *pfx = NULL;
3728
3729         if (addr)
3730                 pfx = nla_data(addr);
3731
3732         if (local) {
3733                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3734                         pfx = NULL;
3735                 else
3736                         pfx = nla_data(local);
3737         }
3738
3739         return pfx;
3740 }
3741
3742 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3743         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3744         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3745         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3746 };
3747
3748 static int
3749 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3750 {
3751         struct net *net = sock_net(skb->sk);
3752         struct ifaddrmsg *ifm;
3753         struct nlattr *tb[IFA_MAX+1];
3754         struct in6_addr *pfx;
3755         int err;
3756
3757         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3758         if (err < 0)
3759                 return err;
3760
3761         ifm = nlmsg_data(nlh);
3762         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3763         if (pfx == NULL)
3764                 return -EINVAL;
3765
3766         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3767 }
3768
3769 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3770                              u32 prefered_lft, u32 valid_lft)
3771 {
3772         u32 flags;
3773         clock_t expires;
3774         unsigned long timeout;
3775
3776         if (!valid_lft || (prefered_lft > valid_lft))
3777                 return -EINVAL;
3778
3779         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3780         if (addrconf_finite_timeout(timeout)) {
3781                 expires = jiffies_to_clock_t(timeout * HZ);
3782                 valid_lft = timeout;
3783                 flags = RTF_EXPIRES;
3784         } else {
3785                 expires = 0;
3786                 flags = 0;
3787                 ifa_flags |= IFA_F_PERMANENT;
3788         }
3789
3790         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3791         if (addrconf_finite_timeout(timeout)) {
3792                 if (timeout == 0)
3793                         ifa_flags |= IFA_F_DEPRECATED;
3794                 prefered_lft = timeout;
3795         }
3796
3797         spin_lock_bh(&ifp->lock);
3798         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3799         ifp->tstamp = jiffies;
3800         ifp->valid_lft = valid_lft;
3801         ifp->prefered_lft = prefered_lft;
3802
3803         spin_unlock_bh(&ifp->lock);
3804         if (!(ifp->flags&IFA_F_TENTATIVE))
3805                 ipv6_ifa_notify(0, ifp);
3806
3807         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3808                               expires, flags);
3809         addrconf_verify(0);
3810
3811         return 0;
3812 }
3813
3814 static int
3815 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3816 {
3817         struct net *net = sock_net(skb->sk);
3818         struct ifaddrmsg *ifm;
3819         struct nlattr *tb[IFA_MAX+1];
3820         struct in6_addr *pfx;
3821         struct inet6_ifaddr *ifa;
3822         struct net_device *dev;
3823         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3824         u8 ifa_flags;
3825         int err;
3826
3827         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3828         if (err < 0)
3829                 return err;
3830
3831         ifm = nlmsg_data(nlh);
3832         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3833         if (pfx == NULL)
3834                 return -EINVAL;
3835
3836         if (tb[IFA_CACHEINFO]) {
3837                 struct ifa_cacheinfo *ci;
3838
3839                 ci = nla_data(tb[IFA_CACHEINFO]);
3840                 valid_lft = ci->ifa_valid;
3841                 preferred_lft = ci->ifa_prefered;
3842         } else {
3843                 preferred_lft = INFINITY_LIFE_TIME;
3844                 valid_lft = INFINITY_LIFE_TIME;
3845         }
3846
3847         dev =  __dev_get_by_index(net, ifm->ifa_index);
3848         if (dev == NULL)
3849                 return -ENODEV;
3850
3851         /* We ignore other flags so far. */
3852         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3853
3854         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3855         if (ifa == NULL) {
3856                 /*
3857                  * It would be best to check for !NLM_F_CREATE here but
3858                  * userspace alreay relies on not having to provide this.
3859                  */
3860                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3861                                       ifm->ifa_prefixlen, ifa_flags,
3862                                       preferred_lft, valid_lft);
3863         }
3864
3865         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3866             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3867                 err = -EEXIST;
3868         else
3869                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3870
3871         in6_ifa_put(ifa);
3872
3873         return err;
3874 }
3875
3876 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3877                           u8 scope, int ifindex)
3878 {
3879         struct ifaddrmsg *ifm;
3880
3881         ifm = nlmsg_data(nlh);
3882         ifm->ifa_family = AF_INET6;
3883         ifm->ifa_prefixlen = prefixlen;
3884         ifm->ifa_flags = flags;
3885         ifm->ifa_scope = scope;
3886         ifm->ifa_index = ifindex;
3887 }
3888
3889 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3890                          unsigned long tstamp, u32 preferred, u32 valid)
3891 {
3892         struct ifa_cacheinfo ci;
3893
3894         ci.cstamp = cstamp_delta(cstamp);
3895         ci.tstamp = cstamp_delta(tstamp);
3896         ci.ifa_prefered = preferred;
3897         ci.ifa_valid = valid;
3898
3899         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3900 }
3901
3902 static inline int rt_scope(int ifa_scope)
3903 {
3904         if (ifa_scope & IFA_HOST)
3905                 return RT_SCOPE_HOST;
3906         else if (ifa_scope & IFA_LINK)
3907                 return RT_SCOPE_LINK;
3908         else if (ifa_scope & IFA_SITE)
3909                 return RT_SCOPE_SITE;
3910         else
3911                 return RT_SCOPE_UNIVERSE;
3912 }
3913
3914 static inline int inet6_ifaddr_msgsize(void)
3915 {
3916         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3917                + nla_total_size(16) /* IFA_ADDRESS */
3918                + nla_total_size(sizeof(struct ifa_cacheinfo));
3919 }
3920
3921 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3922                              u32 portid, u32 seq, int event, unsigned int flags)
3923 {
3924         struct nlmsghdr  *nlh;
3925         u32 preferred, valid;
3926
3927         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3928         if (nlh == NULL)
3929                 return -EMSGSIZE;
3930
3931         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3932                       ifa->idev->dev->ifindex);
3933
3934         if (!(ifa->flags&IFA_F_PERMANENT)) {
3935                 preferred = ifa->prefered_lft;
3936                 valid = ifa->valid_lft;
3937                 if (preferred != INFINITY_LIFE_TIME) {
3938                         long tval = (jiffies - ifa->tstamp)/HZ;
3939                         if (preferred > tval)
3940                                 preferred -= tval;
3941                         else
3942                                 preferred = 0;
3943                         if (valid != INFINITY_LIFE_TIME) {
3944                                 if (valid > tval)
3945                                         valid -= tval;
3946                                 else
3947                                         valid = 0;
3948                         }
3949                 }
3950         } else {
3951                 preferred = INFINITY_LIFE_TIME;
3952                 valid = INFINITY_LIFE_TIME;
3953         }
3954
3955         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3956             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3957                 nlmsg_cancel(skb, nlh);
3958                 return -EMSGSIZE;
3959         }
3960
3961         return nlmsg_end(skb, nlh);
3962 }
3963
3964 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3965                                 u32 portid, u32 seq, int event, u16 flags)
3966 {
3967         struct nlmsghdr  *nlh;
3968         u8 scope = RT_SCOPE_UNIVERSE;
3969         int ifindex = ifmca->idev->dev->ifindex;
3970
3971         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3972                 scope = RT_SCOPE_SITE;
3973
3974         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3975         if (nlh == NULL)
3976                 return -EMSGSIZE;
3977
3978         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3979         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3980             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3981                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3982                 nlmsg_cancel(skb, nlh);
3983                 return -EMSGSIZE;
3984         }
3985
3986         return nlmsg_end(skb, nlh);
3987 }
3988
3989 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3990                                 u32 portid, u32 seq, int event, unsigned int flags)
3991 {
3992         struct nlmsghdr  *nlh;
3993         u8 scope = RT_SCOPE_UNIVERSE;
3994         int ifindex = ifaca->aca_idev->dev->ifindex;
3995
3996         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3997                 scope = RT_SCOPE_SITE;
3998
3999         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4000         if (nlh == NULL)
4001                 return -EMSGSIZE;
4002
4003         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4004         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4005             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4006                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4007                 nlmsg_cancel(skb, nlh);
4008                 return -EMSGSIZE;
4009         }
4010
4011         return nlmsg_end(skb, nlh);
4012 }
4013
4014 enum addr_type_t {
4015         UNICAST_ADDR,
4016         MULTICAST_ADDR,
4017         ANYCAST_ADDR,
4018 };
4019
4020 /* called with rcu_read_lock() */
4021 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4022                           struct netlink_callback *cb, enum addr_type_t type,
4023                           int s_ip_idx, int *p_ip_idx)
4024 {
4025         struct ifmcaddr6 *ifmca;
4026         struct ifacaddr6 *ifaca;
4027         int err = 1;
4028         int ip_idx = *p_ip_idx;
4029
4030         read_lock_bh(&idev->lock);
4031         switch (type) {
4032         case UNICAST_ADDR: {
4033                 struct inet6_ifaddr *ifa;
4034
4035                 /* unicast address incl. temp addr */
4036                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4037                         if (++ip_idx < s_ip_idx)
4038                                 continue;
4039                         err = inet6_fill_ifaddr(skb, ifa,
4040                                                 NETLINK_CB(cb->skb).portid,
4041                                                 cb->nlh->nlmsg_seq,
4042                                                 RTM_NEWADDR,
4043                                                 NLM_F_MULTI);
4044                         if (err <= 0)
4045                                 break;
4046                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4047                 }
4048                 break;
4049         }
4050         case MULTICAST_ADDR:
4051                 /* multicast address */
4052                 for (ifmca = idev->mc_list; ifmca;
4053                      ifmca = ifmca->next, ip_idx++) {
4054                         if (ip_idx < s_ip_idx)
4055                                 continue;
4056                         err = inet6_fill_ifmcaddr(skb, ifmca,
4057                                                   NETLINK_CB(cb->skb).portid,
4058                                                   cb->nlh->nlmsg_seq,
4059                                                   RTM_GETMULTICAST,
4060                                                   NLM_F_MULTI);
4061                         if (err <= 0)
4062                                 break;
4063                 }
4064                 break;
4065         case ANYCAST_ADDR:
4066                 /* anycast address */
4067                 for (ifaca = idev->ac_list; ifaca;
4068                      ifaca = ifaca->aca_next, ip_idx++) {
4069                         if (ip_idx < s_ip_idx)
4070                                 continue;
4071                         err = inet6_fill_ifacaddr(skb, ifaca,
4072                                                   NETLINK_CB(cb->skb).portid,
4073                                                   cb->nlh->nlmsg_seq,
4074                                                   RTM_GETANYCAST,
4075                                                   NLM_F_MULTI);
4076                         if (err <= 0)
4077                                 break;
4078                 }
4079                 break;
4080         default:
4081                 break;
4082         }
4083         read_unlock_bh(&idev->lock);
4084         *p_ip_idx = ip_idx;
4085         return err;
4086 }
4087
4088 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4089                            enum addr_type_t type)
4090 {
4091         struct net *net = sock_net(skb->sk);
4092         int h, s_h;
4093         int idx, ip_idx;
4094         int s_idx, s_ip_idx;
4095         struct net_device *dev;
4096         struct inet6_dev *idev;
4097         struct hlist_head *head;
4098
4099         s_h = cb->args[0];
4100         s_idx = idx = cb->args[1];
4101         s_ip_idx = ip_idx = cb->args[2];
4102
4103         rcu_read_lock();
4104         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4105         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4106                 idx = 0;
4107                 head = &net->dev_index_head[h];
4108                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4109                         if (idx < s_idx)
4110                                 goto cont;
4111                         if (h > s_h || idx > s_idx)
4112                                 s_ip_idx = 0;
4113                         ip_idx = 0;
4114                         idev = __in6_dev_get(dev);
4115                         if (!idev)
4116                                 goto cont;
4117
4118                         if (in6_dump_addrs(idev, skb, cb, type,
4119                                            s_ip_idx, &ip_idx) <= 0)
4120                                 goto done;
4121 cont:
4122                         idx++;
4123                 }
4124         }
4125 done:
4126         rcu_read_unlock();
4127         cb->args[0] = h;
4128         cb->args[1] = idx;
4129         cb->args[2] = ip_idx;
4130
4131         return skb->len;
4132 }
4133
4134 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4135 {
4136         enum addr_type_t type = UNICAST_ADDR;
4137
4138         return inet6_dump_addr(skb, cb, type);
4139 }
4140
4141 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4142 {
4143         enum addr_type_t type = MULTICAST_ADDR;
4144
4145         return inet6_dump_addr(skb, cb, type);
4146 }
4147
4148
4149 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4150 {
4151         enum addr_type_t type = ANYCAST_ADDR;
4152
4153         return inet6_dump_addr(skb, cb, type);
4154 }
4155
4156 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4157 {
4158         struct net *net = sock_net(in_skb->sk);
4159         struct ifaddrmsg *ifm;
4160         struct nlattr *tb[IFA_MAX+1];
4161         struct in6_addr *addr = NULL;
4162         struct net_device *dev = NULL;
4163         struct inet6_ifaddr *ifa;
4164         struct sk_buff *skb;
4165         int err;
4166
4167         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4168         if (err < 0)
4169                 goto errout;
4170
4171         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4172         if (addr == NULL) {
4173                 err = -EINVAL;
4174                 goto errout;
4175         }
4176
4177         ifm = nlmsg_data(nlh);
4178         if (ifm->ifa_index)
4179                 dev = __dev_get_by_index(net, ifm->ifa_index);
4180
4181         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4182         if (!ifa) {
4183                 err = -EADDRNOTAVAIL;
4184                 goto errout;
4185         }
4186
4187         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4188         if (!skb) {
4189                 err = -ENOBUFS;
4190                 goto errout_ifa;
4191         }
4192
4193         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4194                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4195         if (err < 0) {
4196                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4197                 WARN_ON(err == -EMSGSIZE);
4198                 kfree_skb(skb);
4199                 goto errout_ifa;
4200         }
4201         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4202 errout_ifa:
4203         in6_ifa_put(ifa);
4204 errout:
4205         return err;
4206 }
4207
4208 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4209 {
4210         struct sk_buff *skb;
4211         struct net *net = dev_net(ifa->idev->dev);
4212         int err = -ENOBUFS;
4213
4214         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4215         if (skb == NULL)
4216                 goto errout;
4217
4218         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4219         if (err < 0) {
4220                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4221                 WARN_ON(err == -EMSGSIZE);
4222                 kfree_skb(skb);
4223                 goto errout;
4224         }
4225         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4226         return;
4227 errout:
4228         if (err < 0)
4229                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4230 }
4231
4232 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4233                                 __s32 *array, int bytes)
4234 {
4235         BUG_ON(bytes < (DEVCONF_MAX * 4));
4236
4237         memset(array, 0, bytes);
4238         array[DEVCONF_FORWARDING] = cnf->forwarding;
4239         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4240         array[DEVCONF_MTU6] = cnf->mtu6;
4241         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4242         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4243         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4244         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4245         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4246         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4247                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4248         array[DEVCONF_RTR_SOLICIT_DELAY] =
4249                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4250         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4251 #ifdef CONFIG_IPV6_PRIVACY
4252         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4253         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4254         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4255         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4256         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4257 #endif
4258         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4259         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4260         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4261 #ifdef CONFIG_IPV6_ROUTER_PREF
4262         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4263         array[DEVCONF_RTR_PROBE_INTERVAL] =
4264                 jiffies_to_msecs(cnf->rtr_probe_interval);
4265 #ifdef CONFIG_IPV6_ROUTE_INFO
4266         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4267 #endif
4268 #endif
4269         array[DEVCONF_ACCEPT_RA_RT_TABLE] = cnf->accept_ra_rt_table;
4270         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4271         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4272 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4273         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4274         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4275 #endif
4276 #ifdef CONFIG_IPV6_MROUTE
4277         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4278 #endif
4279         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4280         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4281         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4282         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4283 }
4284
4285 static inline size_t inet6_ifla6_size(void)
4286 {
4287         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4288              + nla_total_size(sizeof(struct ifla_cacheinfo))
4289              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4290              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4291              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4292              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4293 }
4294
4295 static inline size_t inet6_if_nlmsg_size(void)
4296 {
4297         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4298                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4299                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4300                + nla_total_size(4) /* IFLA_MTU */
4301                + nla_total_size(4) /* IFLA_LINK */
4302                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4303 }
4304
4305 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4306                                       int items, int bytes)
4307 {
4308         int i;
4309         int pad = bytes - sizeof(u64) * items;
4310         BUG_ON(pad < 0);
4311
4312         /* Use put_unaligned() because stats may not be aligned for u64. */
4313         put_unaligned(items, &stats[0]);
4314         for (i = 1; i < items; i++)
4315                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4316
4317         memset(&stats[items], 0, pad);
4318 }
4319
4320 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4321                                       int items, int bytes, size_t syncpoff)
4322 {
4323         int i;
4324         int pad = bytes - sizeof(u64) * items;
4325         BUG_ON(pad < 0);
4326
4327         /* Use put_unaligned() because stats may not be aligned for u64. */
4328         put_unaligned(items, &stats[0]);
4329         for (i = 1; i < items; i++)
4330                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4331
4332         memset(&stats[items], 0, pad);
4333 }
4334
4335 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4336                              int bytes)
4337 {
4338         switch (attrtype) {
4339         case IFLA_INET6_STATS:
4340                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4341                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4342                 break;
4343         case IFLA_INET6_ICMP6STATS:
4344                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4345                 break;
4346         }
4347 }
4348
4349 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4350 {
4351         struct nlattr *nla;
4352         struct ifla_cacheinfo ci;
4353
4354         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4355                 goto nla_put_failure;
4356         ci.max_reasm_len = IPV6_MAXPLEN;
4357         ci.tstamp = cstamp_delta(idev->tstamp);
4358         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4359         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4360         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4361                 goto nla_put_failure;
4362         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4363         if (nla == NULL)
4364                 goto nla_put_failure;
4365         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4366
4367         /* XXX - MC not implemented */
4368
4369         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4370         if (nla == NULL)
4371                 goto nla_put_failure;
4372         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4373
4374         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4375         if (nla == NULL)
4376                 goto nla_put_failure;
4377         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4378
4379         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4380         if (nla == NULL)
4381                 goto nla_put_failure;
4382         read_lock_bh(&idev->lock);
4383         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4384         read_unlock_bh(&idev->lock);
4385
4386         return 0;
4387
4388 nla_put_failure:
4389         return -EMSGSIZE;
4390 }
4391
4392 static size_t inet6_get_link_af_size(const struct net_device *dev)
4393 {
4394         if (!__in6_dev_get(dev))
4395                 return 0;
4396
4397         return inet6_ifla6_size();
4398 }
4399
4400 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4401 {
4402         struct inet6_dev *idev = __in6_dev_get(dev);
4403
4404         if (!idev)
4405                 return -ENODATA;
4406
4407         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4408                 return -EMSGSIZE;
4409
4410         return 0;
4411 }
4412
4413 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4414 {
4415         struct inet6_ifaddr *ifp;
4416         struct net_device *dev = idev->dev;
4417         bool update_rs = false;
4418         struct in6_addr ll_addr;
4419
4420         if (token == NULL)
4421                 return -EINVAL;
4422         if (ipv6_addr_any(token))
4423                 return -EINVAL;
4424         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4425                 return -EINVAL;
4426         if (!ipv6_accept_ra(idev))
4427                 return -EINVAL;
4428         if (idev->cnf.rtr_solicits <= 0)
4429                 return -EINVAL;
4430
4431         write_lock_bh(&idev->lock);
4432
4433         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4434         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4435
4436         write_unlock_bh(&idev->lock);
4437
4438         if (!idev->dead && (idev->if_flags & IF_READY) &&
4439             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4440                              IFA_F_OPTIMISTIC)) {
4441
4442                 /* If we're not ready, then normal ifup will take care
4443                  * of this. Otherwise, we need to request our rs here.
4444                  */
4445                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4446                 update_rs = true;
4447         }
4448
4449         write_lock_bh(&idev->lock);
4450
4451         if (update_rs)
4452                 idev->if_flags |= IF_RS_SENT;
4453
4454         /* Well, that's kinda nasty ... */
4455         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4456                 spin_lock(&ifp->lock);
4457                 if (ifp->tokenized) {
4458                         ifp->valid_lft = 0;
4459                         ifp->prefered_lft = 0;
4460                 }
4461                 spin_unlock(&ifp->lock);
4462         }
4463
4464         write_unlock_bh(&idev->lock);
4465         return 0;
4466 }
4467
4468 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4469 {
4470         int err = -EINVAL;
4471         struct inet6_dev *idev = __in6_dev_get(dev);
4472         struct nlattr *tb[IFLA_INET6_MAX + 1];
4473
4474         if (!idev)
4475                 return -EAFNOSUPPORT;
4476
4477         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4478                 BUG();
4479
4480         if (tb[IFLA_INET6_TOKEN])
4481                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4482
4483         return err;
4484 }
4485
4486 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4487                              u32 portid, u32 seq, int event, unsigned int flags)
4488 {
4489         struct net_device *dev = idev->dev;
4490         struct ifinfomsg *hdr;
4491         struct nlmsghdr *nlh;
4492         void *protoinfo;
4493
4494         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4495         if (nlh == NULL)
4496                 return -EMSGSIZE;
4497
4498         hdr = nlmsg_data(nlh);
4499         hdr->ifi_family = AF_INET6;
4500         hdr->__ifi_pad = 0;
4501         hdr->ifi_type = dev->type;
4502         hdr->ifi_index = dev->ifindex;
4503         hdr->ifi_flags = dev_get_flags(dev);
4504         hdr->ifi_change = 0;
4505
4506         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4507             (dev->addr_len &&
4508              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4509             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4510             (dev->ifindex != dev->iflink &&
4511              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4512                 goto nla_put_failure;
4513         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4514         if (protoinfo == NULL)
4515                 goto nla_put_failure;
4516
4517         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4518                 goto nla_put_failure;
4519
4520         nla_nest_end(skb, protoinfo);
4521         return nlmsg_end(skb, nlh);
4522
4523 nla_put_failure:
4524         nlmsg_cancel(skb, nlh);
4525         return -EMSGSIZE;
4526 }
4527
4528 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4529 {
4530         struct net *net = sock_net(skb->sk);
4531         int h, s_h;
4532         int idx = 0, s_idx;
4533         struct net_device *dev;
4534         struct inet6_dev *idev;
4535         struct hlist_head *head;
4536
4537         s_h = cb->args[0];
4538         s_idx = cb->args[1];
4539
4540         rcu_read_lock();
4541         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4542                 idx = 0;
4543                 head = &net->dev_index_head[h];
4544                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4545                         if (idx < s_idx)
4546                                 goto cont;
4547                         idev = __in6_dev_get(dev);
4548                         if (!idev)
4549                                 goto cont;
4550                         if (inet6_fill_ifinfo(skb, idev,
4551                                               NETLINK_CB(cb->skb).portid,
4552                                               cb->nlh->nlmsg_seq,
4553                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4554                                 goto out;
4555 cont:
4556                         idx++;
4557                 }
4558         }
4559 out:
4560         rcu_read_unlock();
4561         cb->args[1] = idx;
4562         cb->args[0] = h;
4563
4564         return skb->len;
4565 }
4566
4567 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4568 {
4569         struct sk_buff *skb;
4570         struct net *net = dev_net(idev->dev);
4571         int err = -ENOBUFS;
4572
4573         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4574         if (skb == NULL)
4575                 goto errout;
4576
4577         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4578         if (err < 0) {
4579                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4580                 WARN_ON(err == -EMSGSIZE);
4581                 kfree_skb(skb);
4582                 goto errout;
4583         }
4584         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4585         return;
4586 errout:
4587         if (err < 0)
4588                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4589 }
4590
4591 static inline size_t inet6_prefix_nlmsg_size(void)
4592 {
4593         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4594                + nla_total_size(sizeof(struct in6_addr))
4595                + nla_total_size(sizeof(struct prefix_cacheinfo));
4596 }
4597
4598 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4599                              struct prefix_info *pinfo, u32 portid, u32 seq,
4600                              int event, unsigned int flags)
4601 {
4602         struct prefixmsg *pmsg;
4603         struct nlmsghdr *nlh;
4604         struct prefix_cacheinfo ci;
4605
4606         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4607         if (nlh == NULL)
4608                 return -EMSGSIZE;
4609
4610         pmsg = nlmsg_data(nlh);
4611         pmsg->prefix_family = AF_INET6;
4612         pmsg->prefix_pad1 = 0;
4613         pmsg->prefix_pad2 = 0;
4614         pmsg->prefix_ifindex = idev->dev->ifindex;
4615         pmsg->prefix_len = pinfo->prefix_len;
4616         pmsg->prefix_type = pinfo->type;
4617         pmsg->prefix_pad3 = 0;
4618         pmsg->prefix_flags = 0;
4619         if (pinfo->onlink)
4620                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4621         if (pinfo->autoconf)
4622                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4623
4624         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4625                 goto nla_put_failure;
4626         ci.preferred_time = ntohl(pinfo->prefered);
4627         ci.valid_time = ntohl(pinfo->valid);
4628         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4629                 goto nla_put_failure;
4630         return nlmsg_end(skb, nlh);
4631
4632 nla_put_failure:
4633         nlmsg_cancel(skb, nlh);
4634         return -EMSGSIZE;
4635 }
4636
4637 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4638                          struct prefix_info *pinfo)
4639 {
4640         struct sk_buff *skb;
4641         struct net *net = dev_net(idev->dev);
4642         int err = -ENOBUFS;
4643
4644         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4645         if (skb == NULL)
4646                 goto errout;
4647
4648         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4649         if (err < 0) {
4650                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4651                 WARN_ON(err == -EMSGSIZE);
4652                 kfree_skb(skb);
4653                 goto errout;
4654         }
4655         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4656         return;
4657 errout:
4658         if (err < 0)
4659                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4660 }
4661
4662 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4663 {
4664         struct net *net = dev_net(ifp->idev->dev);
4665
4666         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4667
4668         switch (event) {
4669         case RTM_NEWADDR:
4670                 /*
4671                  * If the address was optimistic
4672                  * we inserted the route at the start of
4673                  * our DAD process, so we don't need
4674                  * to do it again
4675                  */
4676                 if (!(ifp->rt->rt6i_node))
4677                         ip6_ins_rt(ifp->rt);
4678                 if (ifp->idev->cnf.forwarding)
4679                         addrconf_join_anycast(ifp);
4680                 break;
4681         case RTM_DELADDR:
4682                 if (ifp->idev->cnf.forwarding)
4683                         addrconf_leave_anycast(ifp);
4684                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4685                 dst_hold(&ifp->rt->dst);
4686
4687                 if (ip6_del_rt(ifp->rt))
4688                         dst_free(&ifp->rt->dst);
4689                 break;
4690         }
4691         atomic_inc(&net->ipv6.dev_addr_genid);
4692 }
4693
4694 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4695 {
4696         rcu_read_lock_bh();
4697         if (likely(ifp->idev->dead == 0))
4698                 __ipv6_ifa_notify(event, ifp);
4699         rcu_read_unlock_bh();
4700 }
4701
4702 #ifdef CONFIG_SYSCTL
4703
4704 static
4705 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4706                            void __user *buffer, size_t *lenp, loff_t *ppos)
4707 {
4708         int *valp = ctl->data;
4709         int val = *valp;
4710         loff_t pos = *ppos;
4711         ctl_table lctl;
4712         int ret;
4713
4714         /*
4715          * ctl->data points to idev->cnf.forwarding, we should
4716          * not modify it until we get the rtnl lock.
4717          */
4718         lctl = *ctl;
4719         lctl.data = &val;
4720
4721         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4722
4723         if (write)
4724                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4725         if (ret)
4726                 *ppos = pos;
4727         return ret;
4728 }
4729
4730 static void dev_disable_change(struct inet6_dev *idev)
4731 {
4732         if (!idev || !idev->dev)
4733                 return;
4734
4735         if (idev->cnf.disable_ipv6)
4736                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4737         else
4738                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4739 }
4740
4741 static void addrconf_disable_change(struct net *net, __s32 newf)
4742 {
4743         struct net_device *dev;
4744         struct inet6_dev *idev;
4745
4746         rcu_read_lock();
4747         for_each_netdev_rcu(net, dev) {
4748                 idev = __in6_dev_get(dev);
4749                 if (idev) {
4750                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4751                         idev->cnf.disable_ipv6 = newf;
4752                         if (changed)
4753                                 dev_disable_change(idev);
4754                 }
4755         }
4756         rcu_read_unlock();
4757 }
4758
4759 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4760 {
4761         struct net *net;
4762         int old;
4763
4764         if (!rtnl_trylock())
4765                 return restart_syscall();
4766
4767         net = (struct net *)table->extra2;
4768         old = *p;
4769         *p = newf;
4770
4771         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4772                 rtnl_unlock();
4773                 return 0;
4774         }
4775
4776         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4777                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4778                 addrconf_disable_change(net, newf);
4779         } else if ((!newf) ^ (!old))
4780                 dev_disable_change((struct inet6_dev *)table->extra1);
4781
4782         rtnl_unlock();
4783         return 0;
4784 }
4785
4786 static
4787 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4788                             void __user *buffer, size_t *lenp, loff_t *ppos)
4789 {
4790         int *valp = ctl->data;
4791         int val = *valp;
4792         loff_t pos = *ppos;
4793         ctl_table lctl;
4794         int ret;
4795
4796         /*
4797          * ctl->data points to idev->cnf.disable_ipv6, we should
4798          * not modify it until we get the rtnl lock.
4799          */
4800         lctl = *ctl;
4801         lctl.data = &val;
4802
4803         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4804
4805         if (write)
4806                 ret = addrconf_disable_ipv6(ctl, valp, val);
4807         if (ret)
4808                 *ppos = pos;
4809         return ret;
4810 }
4811
4812 static struct addrconf_sysctl_table
4813 {
4814         struct ctl_table_header *sysctl_header;
4815         ctl_table addrconf_vars[DEVCONF_MAX+1];
4816 } addrconf_sysctl __read_mostly = {
4817         .sysctl_header = NULL,
4818         .addrconf_vars = {
4819                 {
4820                         .procname       = "forwarding",
4821                         .data           = &ipv6_devconf.forwarding,
4822                         .maxlen         = sizeof(int),
4823                         .mode           = 0644,
4824                         .proc_handler   = addrconf_sysctl_forward,
4825                 },
4826                 {
4827                         .procname       = "hop_limit",
4828                         .data           = &ipv6_devconf.hop_limit,
4829                         .maxlen         = sizeof(int),
4830                         .mode           = 0644,
4831                         .proc_handler   = proc_dointvec,
4832                 },
4833                 {
4834                         .procname       = "mtu",
4835                         .data           = &ipv6_devconf.mtu6,
4836                         .maxlen         = sizeof(int),
4837                         .mode           = 0644,
4838                         .proc_handler   = proc_dointvec,
4839                 },
4840                 {
4841                         .procname       = "accept_ra",
4842                         .data           = &ipv6_devconf.accept_ra,
4843                         .maxlen         = sizeof(int),
4844                         .mode           = 0644,
4845                         .proc_handler   = proc_dointvec,
4846                 },
4847                 {
4848                         .procname       = "accept_redirects",
4849                         .data           = &ipv6_devconf.accept_redirects,
4850                         .maxlen         = sizeof(int),
4851                         .mode           = 0644,
4852                         .proc_handler   = proc_dointvec,
4853                 },
4854                 {
4855                         .procname       = "autoconf",
4856                         .data           = &ipv6_devconf.autoconf,
4857                         .maxlen         = sizeof(int),
4858                         .mode           = 0644,
4859                         .proc_handler   = proc_dointvec,
4860                 },
4861                 {
4862                         .procname       = "dad_transmits",
4863                         .data           = &ipv6_devconf.dad_transmits,
4864                         .maxlen         = sizeof(int),
4865                         .mode           = 0644,
4866                         .proc_handler   = proc_dointvec,
4867                 },
4868                 {
4869                         .procname       = "router_solicitations",
4870                         .data           = &ipv6_devconf.rtr_solicits,
4871                         .maxlen         = sizeof(int),
4872                         .mode           = 0644,
4873                         .proc_handler   = proc_dointvec,
4874                 },
4875                 {
4876                         .procname       = "router_solicitation_interval",
4877                         .data           = &ipv6_devconf.rtr_solicit_interval,
4878                         .maxlen         = sizeof(int),
4879                         .mode           = 0644,
4880                         .proc_handler   = proc_dointvec_jiffies,
4881                 },
4882                 {
4883                         .procname       = "router_solicitation_delay",
4884                         .data           = &ipv6_devconf.rtr_solicit_delay,
4885                         .maxlen         = sizeof(int),
4886                         .mode           = 0644,
4887                         .proc_handler   = proc_dointvec_jiffies,
4888                 },
4889                 {
4890                         .procname       = "force_mld_version",
4891                         .data           = &ipv6_devconf.force_mld_version,
4892                         .maxlen         = sizeof(int),
4893                         .mode           = 0644,
4894                         .proc_handler   = proc_dointvec,
4895                 },
4896 #ifdef CONFIG_IPV6_PRIVACY
4897                 {
4898                         .procname       = "use_tempaddr",
4899                         .data           = &ipv6_devconf.use_tempaddr,
4900                         .maxlen         = sizeof(int),
4901                         .mode           = 0644,
4902                         .proc_handler   = proc_dointvec,
4903                 },
4904                 {
4905                         .procname       = "temp_valid_lft",
4906                         .data           = &ipv6_devconf.temp_valid_lft,
4907                         .maxlen         = sizeof(int),
4908                         .mode           = 0644,
4909                         .proc_handler   = proc_dointvec,
4910                 },
4911                 {
4912                         .procname       = "temp_prefered_lft",
4913                         .data           = &ipv6_devconf.temp_prefered_lft,
4914                         .maxlen         = sizeof(int),
4915                         .mode           = 0644,
4916                         .proc_handler   = proc_dointvec,
4917                 },
4918                 {
4919                         .procname       = "regen_max_retry",
4920                         .data           = &ipv6_devconf.regen_max_retry,
4921                         .maxlen         = sizeof(int),
4922                         .mode           = 0644,
4923                         .proc_handler   = proc_dointvec,
4924                 },
4925                 {
4926                         .procname       = "max_desync_factor",
4927                         .data           = &ipv6_devconf.max_desync_factor,
4928                         .maxlen         = sizeof(int),
4929                         .mode           = 0644,
4930                         .proc_handler   = proc_dointvec,
4931                 },
4932 #endif
4933                 {
4934                         .procname       = "max_addresses",
4935                         .data           = &ipv6_devconf.max_addresses,
4936                         .maxlen         = sizeof(int),
4937                         .mode           = 0644,
4938                         .proc_handler   = proc_dointvec,
4939                 },
4940                 {
4941                         .procname       = "accept_ra_defrtr",
4942                         .data           = &ipv6_devconf.accept_ra_defrtr,
4943                         .maxlen         = sizeof(int),
4944                         .mode           = 0644,
4945                         .proc_handler   = proc_dointvec,
4946                 },
4947                 {
4948                         .procname       = "accept_ra_pinfo",
4949                         .data           = &ipv6_devconf.accept_ra_pinfo,
4950                         .maxlen         = sizeof(int),
4951                         .mode           = 0644,
4952                         .proc_handler   = proc_dointvec,
4953                 },
4954 #ifdef CONFIG_IPV6_ROUTER_PREF
4955                 {
4956                         .procname       = "accept_ra_rtr_pref",
4957                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4958                         .maxlen         = sizeof(int),
4959                         .mode           = 0644,
4960                         .proc_handler   = proc_dointvec,
4961                 },
4962                 {
4963                         .procname       = "router_probe_interval",
4964                         .data           = &ipv6_devconf.rtr_probe_interval,
4965                         .maxlen         = sizeof(int),
4966                         .mode           = 0644,
4967                         .proc_handler   = proc_dointvec_jiffies,
4968                 },
4969 #ifdef CONFIG_IPV6_ROUTE_INFO
4970                 {
4971                         .procname       = "accept_ra_rt_info_max_plen",
4972                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4973                         .maxlen         = sizeof(int),
4974                         .mode           = 0644,
4975                         .proc_handler   = proc_dointvec,
4976                 },
4977 #endif
4978 #endif
4979                 {
4980                         .procname       = "accept_ra_rt_table",
4981                         .data           = &ipv6_devconf.accept_ra_rt_table,
4982                         .maxlen         = sizeof(int),
4983                         .mode           = 0644,
4984                         .proc_handler   = proc_dointvec,
4985                 },
4986                 {
4987                         .procname       = "proxy_ndp",
4988                         .data           = &ipv6_devconf.proxy_ndp,
4989                         .maxlen         = sizeof(int),
4990                         .mode           = 0644,
4991                         .proc_handler   = proc_dointvec,
4992                 },
4993                 {
4994                         .procname       = "accept_source_route",
4995                         .data           = &ipv6_devconf.accept_source_route,
4996                         .maxlen         = sizeof(int),
4997                         .mode           = 0644,
4998                         .proc_handler   = proc_dointvec,
4999                 },
5000 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5001                 {
5002                         .procname       = "optimistic_dad",
5003                         .data           = &ipv6_devconf.optimistic_dad,
5004                         .maxlen         = sizeof(int),
5005                         .mode           = 0644,
5006                         .proc_handler   = proc_dointvec,
5007
5008                 },
5009                 {
5010                         .procname       = "use_optimistic",
5011                         .data           = &ipv6_devconf.use_optimistic,
5012                         .maxlen         = sizeof(int),
5013                         .mode           = 0644,
5014                         .proc_handler   = proc_dointvec,
5015
5016                 },
5017 #endif
5018 #ifdef CONFIG_IPV6_MROUTE
5019                 {
5020                         .procname       = "mc_forwarding",
5021                         .data           = &ipv6_devconf.mc_forwarding,
5022                         .maxlen         = sizeof(int),
5023                         .mode           = 0444,
5024                         .proc_handler   = proc_dointvec,
5025                 },
5026 #endif
5027                 {
5028                         .procname       = "disable_ipv6",
5029                         .data           = &ipv6_devconf.disable_ipv6,
5030                         .maxlen         = sizeof(int),
5031                         .mode           = 0644,
5032                         .proc_handler   = addrconf_sysctl_disable,
5033                 },
5034                 {
5035                         .procname       = "accept_dad",
5036                         .data           = &ipv6_devconf.accept_dad,
5037                         .maxlen         = sizeof(int),
5038                         .mode           = 0644,
5039                         .proc_handler   = proc_dointvec,
5040                 },
5041                 {
5042                         .procname       = "force_tllao",
5043                         .data           = &ipv6_devconf.force_tllao,
5044                         .maxlen         = sizeof(int),
5045                         .mode           = 0644,
5046                         .proc_handler   = proc_dointvec
5047                 },
5048                 {
5049                         .procname       = "ndisc_notify",
5050                         .data           = &ipv6_devconf.ndisc_notify,
5051                         .maxlen         = sizeof(int),
5052                         .mode           = 0644,
5053                         .proc_handler   = proc_dointvec
5054                 },
5055                 {
5056                         /* sentinel */
5057                 }
5058         },
5059 };
5060
5061 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5062                 struct inet6_dev *idev, struct ipv6_devconf *p)
5063 {
5064         int i;
5065         struct addrconf_sysctl_table *t;
5066         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5067
5068         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5069         if (t == NULL)
5070                 goto out;
5071
5072         for (i = 0; t->addrconf_vars[i].data; i++) {
5073                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5074                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5075                 t->addrconf_vars[i].extra2 = net;
5076         }
5077
5078         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5079
5080         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5081         if (t->sysctl_header == NULL)
5082                 goto free;
5083
5084         p->sysctl = t;
5085         return 0;
5086
5087 free:
5088         kfree(t);
5089 out:
5090         return -ENOBUFS;
5091 }
5092
5093 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5094 {
5095         struct addrconf_sysctl_table *t;
5096
5097         if (p->sysctl == NULL)
5098                 return;
5099
5100         t = p->sysctl;
5101         p->sysctl = NULL;
5102         unregister_net_sysctl_table(t->sysctl_header);
5103         kfree(t);
5104 }
5105
5106 static void addrconf_sysctl_register(struct inet6_dev *idev)
5107 {
5108         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5109                               &ndisc_ifinfo_sysctl_change);
5110         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5111                                         idev, &idev->cnf);
5112 }
5113
5114 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5115 {
5116         __addrconf_sysctl_unregister(&idev->cnf);
5117         neigh_sysctl_unregister(idev->nd_parms);
5118 }
5119
5120
5121 #endif
5122
5123 static int __net_init addrconf_init_net(struct net *net)
5124 {
5125         int err = -ENOMEM;
5126         struct ipv6_devconf *all, *dflt;
5127
5128         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5129         if (all == NULL)
5130                 goto err_alloc_all;
5131
5132         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5133         if (dflt == NULL)
5134                 goto err_alloc_dflt;
5135
5136         /* these will be inherited by all namespaces */
5137         dflt->autoconf = ipv6_defaults.autoconf;
5138         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5139
5140         net->ipv6.devconf_all = all;
5141         net->ipv6.devconf_dflt = dflt;
5142
5143 #ifdef CONFIG_SYSCTL
5144         err = __addrconf_sysctl_register(net, "all", NULL, all);
5145         if (err < 0)
5146                 goto err_reg_all;
5147
5148         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5149         if (err < 0)
5150                 goto err_reg_dflt;
5151 #endif
5152         return 0;
5153
5154 #ifdef CONFIG_SYSCTL
5155 err_reg_dflt:
5156         __addrconf_sysctl_unregister(all);
5157 err_reg_all:
5158         kfree(dflt);
5159 #endif
5160 err_alloc_dflt:
5161         kfree(all);
5162 err_alloc_all:
5163         return err;
5164 }
5165
5166 static void __net_exit addrconf_exit_net(struct net *net)
5167 {
5168 #ifdef CONFIG_SYSCTL
5169         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5170         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5171 #endif
5172         if (!net_eq(net, &init_net)) {
5173                 kfree(net->ipv6.devconf_dflt);
5174                 kfree(net->ipv6.devconf_all);
5175         }
5176 }
5177
5178 static struct pernet_operations addrconf_ops = {
5179         .init = addrconf_init_net,
5180         .exit = addrconf_exit_net,
5181 };
5182
5183 static struct rtnl_af_ops inet6_ops = {
5184         .family           = AF_INET6,
5185         .fill_link_af     = inet6_fill_link_af,
5186         .get_link_af_size = inet6_get_link_af_size,
5187         .set_link_af      = inet6_set_link_af,
5188 };
5189
5190 /*
5191  *      Init / cleanup code
5192  */
5193
5194 int __init addrconf_init(void)
5195 {
5196         int i, err;
5197
5198         err = ipv6_addr_label_init();
5199         if (err < 0) {
5200                 pr_crit("%s: cannot initialize default policy table: %d\n",
5201                         __func__, err);
5202                 goto out;
5203         }
5204
5205         err = register_pernet_subsys(&addrconf_ops);
5206         if (err < 0)
5207                 goto out_addrlabel;
5208
5209         /* The addrconf netdev notifier requires that loopback_dev
5210          * has it's ipv6 private information allocated and setup
5211          * before it can bring up and give link-local addresses
5212          * to other devices which are up.
5213          *
5214          * Unfortunately, loopback_dev is not necessarily the first
5215          * entry in the global dev_base list of net devices.  In fact,
5216          * it is likely to be the very last entry on that list.
5217          * So this causes the notifier registry below to try and
5218          * give link-local addresses to all devices besides loopback_dev
5219          * first, then loopback_dev, which cases all the non-loopback_dev
5220          * devices to fail to get a link-local address.
5221          *
5222          * So, as a temporary fix, allocate the ipv6 structure for
5223          * loopback_dev first by hand.
5224          * Longer term, all of the dependencies ipv6 has upon the loopback
5225          * device and it being up should be removed.
5226          */
5227         rtnl_lock();
5228         if (!ipv6_add_dev(init_net.loopback_dev))
5229                 err = -ENOMEM;
5230         rtnl_unlock();
5231         if (err)
5232                 goto errlo;
5233
5234         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5235                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5236
5237         register_netdevice_notifier(&ipv6_dev_notf);
5238
5239         addrconf_verify(0);
5240
5241         err = rtnl_af_register(&inet6_ops);
5242         if (err < 0)
5243                 goto errout_af;
5244
5245         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5246                               NULL);
5247         if (err < 0)
5248                 goto errout;
5249
5250         /* Only the first call to __rtnl_register can fail */
5251         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5252         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5253         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5254                         inet6_dump_ifaddr, NULL);
5255         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5256                         inet6_dump_ifmcaddr, NULL);
5257         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5258                         inet6_dump_ifacaddr, NULL);
5259         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5260                         inet6_netconf_dump_devconf, NULL);
5261
5262         ipv6_addr_label_rtnl_register();
5263
5264         return 0;
5265 errout:
5266         rtnl_af_unregister(&inet6_ops);
5267 errout_af:
5268         unregister_netdevice_notifier(&ipv6_dev_notf);
5269 errlo:
5270         unregister_pernet_subsys(&addrconf_ops);
5271 out_addrlabel:
5272         ipv6_addr_label_cleanup();
5273 out:
5274         return err;
5275 }
5276
5277 void addrconf_cleanup(void)
5278 {
5279         struct net_device *dev;
5280         int i;
5281
5282         unregister_netdevice_notifier(&ipv6_dev_notf);
5283         unregister_pernet_subsys(&addrconf_ops);
5284         ipv6_addr_label_cleanup();
5285
5286         rtnl_lock();
5287
5288         __rtnl_af_unregister(&inet6_ops);
5289
5290         /* clean dev list */
5291         for_each_netdev(&init_net, dev) {
5292                 if (__in6_dev_get(dev) == NULL)
5293                         continue;
5294                 addrconf_ifdown(dev, 1);
5295         }
5296         addrconf_ifdown(init_net.loopback_dev, 2);
5297
5298         /*
5299          *      Check hash table.
5300          */
5301         spin_lock_bh(&addrconf_hash_lock);
5302         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5303                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5304         spin_unlock_bh(&addrconf_hash_lock);
5305
5306         del_timer(&addr_chk_timer);
5307         rtnl_unlock();
5308 }