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 ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 173
 174 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 175         .forwarding             = 0,
 176         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 177         .mtu6                   = IPV6_MIN_MTU,
 178         .accept_ra              = 1,
 179         .accept_redirects       = 1,
 180         .autoconf               = 1,
 181         .force_mld_version      = 0,
 182         .dad_transmits          = 1,
 183         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 184         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 185         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 186 #ifdef CONFIG_IPV6_PRIVACY
 187         .use_tempaddr           = 0,
 188         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 189         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 190         .regen_max_retry        = REGEN_MAX_RETRY,
 191         .max_desync_factor      = MAX_DESYNC_FACTOR,
 192 #endif
 193         .max_addresses          = IPV6_MAX_ADDRESSES,
 194         .accept_ra_defrtr       = 1,
 195         .accept_ra_pinfo        = 1,
 196 #ifdef CONFIG_IPV6_ROUTER_PREF
 197         .accept_ra_rtr_pref     = 1,
 198         .rtr_probe_interval     = 60 * HZ,
 199 #ifdef CONFIG_IPV6_ROUTE_INFO
 200         .accept_ra_rt_info_max_plen = 0,
 201 #endif
 202 #endif
 203         .proxy_ndp              = 0,
 204         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 205         .disable_ipv6           = 0,
 206         .accept_dad             = 1,
 207 };
 208
 209 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 210         .forwarding             = 0,
 211         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 212         .mtu6                   = IPV6_MIN_MTU,
 213         .accept_ra              = 1,
 214         .accept_redirects       = 1,
 215         .autoconf               = 1,
 216         .dad_transmits          = 1,
 217         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 218         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 219         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 220 #ifdef CONFIG_IPV6_PRIVACY
 221         .use_tempaddr           = 0,
 222         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 223         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 224         .regen_max_retry        = REGEN_MAX_RETRY,
 225         .max_desync_factor      = MAX_DESYNC_FACTOR,
 226 #endif
 227         .max_addresses          = IPV6_MAX_ADDRESSES,
 228         .accept_ra_defrtr       = 1,
 229         .accept_ra_pinfo        = 1,
 230 #ifdef CONFIG_IPV6_ROUTER_PREF
 231         .accept_ra_rtr_pref     = 1,
 232         .rtr_probe_interval     = 60 * HZ,
 233 #ifdef CONFIG_IPV6_ROUTE_INFO
 234         .accept_ra_rt_info_max_plen = 0,
 235 #endif
 236 #endif
 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
 426         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 427                 ndev->if_flags |= IF_READY;
 428
 429         ipv6_mc_init_dev(ndev);
 430         ndev->tstamp = jiffies;
 431         addrconf_sysctl_register(ndev);
 432         /* protected by rtnl_lock */
 433         rcu_assign_pointer(dev->ip6_ptr, ndev);
 434
 435         /* Join interface-local all-node multicast group */
 436         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 437
 438         /* Join all-node multicast group */
 439         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 440
 441         /* Join all-router multicast group if forwarding is set */
 442         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 443                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 444
 445         return ndev;
 446 }
 447
 448 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 449 {
 450         struct inet6_dev *idev;
 451
 452         ASSERT_RTNL();
 453
 454         idev = __in6_dev_get(dev);
 455         if (!idev) {
 456                 idev = ipv6_add_dev(dev);
 457                 if (!idev)
 458                         return NULL;
 459         }
 460
 461         if (dev->flags&IFF_UP)
 462                 ipv6_mc_up(idev);
 463         return idev;
 464 }
 465
 466 static int inet6_netconf_msgsize_devconf(int type)
 467 {
 468         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 469                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 470
 471         /* type -1 is used for ALL */
 472         if (type == -1 || type == NETCONFA_FORWARDING)
 473                 size += nla_total_size(4);
 474 #ifdef CONFIG_IPV6_MROUTE
 475         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 476                 size += nla_total_size(4);
 477 #endif
 478
 479         return size;
 480 }
 481
 482 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 483                                       struct ipv6_devconf *devconf, u32 portid,
 484                                       u32 seq, int event, unsigned int flags,
 485                                       int type)
 486 {
 487         struct nlmsghdr  *nlh;
 488         struct netconfmsg *ncm;
 489
 490         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 491                         flags);
 492         if (nlh == NULL)
 493                 return -EMSGSIZE;
 494
 495         ncm = nlmsg_data(nlh);
 496         ncm->ncm_family = AF_INET6;
 497
 498         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 499                 goto nla_put_failure;
 500
 501         /* type -1 is used for ALL */
 502         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 503             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 504                 goto nla_put_failure;
 505 #ifdef CONFIG_IPV6_MROUTE
 506         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 507             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 508                         devconf->mc_forwarding) < 0)
 509                 goto nla_put_failure;
 510 #endif
 511         return nlmsg_end(skb, nlh);
 512
 513 nla_put_failure:
 514         nlmsg_cancel(skb, nlh);
 515         return -EMSGSIZE;
 516 }
 517
 518 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 519                                   struct ipv6_devconf *devconf)
 520 {
 521         struct sk_buff *skb;
 522         int err = -ENOBUFS;
 523
 524         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 525         if (skb == NULL)
 526                 goto errout;
 527
 528         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 529                                          RTM_NEWNETCONF, 0, type);
 530         if (err < 0) {
 531                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 532                 WARN_ON(err == -EMSGSIZE);
 533                 kfree_skb(skb);
 534                 goto errout;
 535         }
 536         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 537         return;
 538 errout:
 539         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 540 }
 541
 542 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 543         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 544         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 545 };
 546
 547 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 548                                      struct nlmsghdr *nlh)
 549 {
 550         struct net *net = sock_net(in_skb->sk);
 551         struct nlattr *tb[NETCONFA_MAX+1];
 552         struct netconfmsg *ncm;
 553         struct sk_buff *skb;
 554         struct ipv6_devconf *devconf;
 555         struct inet6_dev *in6_dev;
 556         struct net_device *dev;
 557         int ifindex;
 558         int err;
 559
 560         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 561                           devconf_ipv6_policy);
 562         if (err < 0)
 563                 goto errout;
 564
 565         err = EINVAL;
 566         if (!tb[NETCONFA_IFINDEX])
 567                 goto errout;
 568
 569         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 570         switch (ifindex) {
 571         case NETCONFA_IFINDEX_ALL:
 572                 devconf = net->ipv6.devconf_all;
 573                 break;
 574         case NETCONFA_IFINDEX_DEFAULT:
 575                 devconf = net->ipv6.devconf_dflt;
 576                 break;
 577         default:
 578                 dev = __dev_get_by_index(net, ifindex);
 579                 if (dev == NULL)
 580                         goto errout;
 581                 in6_dev = __in6_dev_get(dev);
 582                 if (in6_dev == NULL)
 583                         goto errout;
 584                 devconf = &in6_dev->cnf;
 585                 break;
 586         }
 587
 588         err = -ENOBUFS;
 589         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 590         if (skb == NULL)
 591                 goto errout;
 592
 593         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 594                                          NETLINK_CB(in_skb).portid,
 595                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 596                                          -1);
 597         if (err < 0) {
 598                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 599                 WARN_ON(err == -EMSGSIZE);
 600                 kfree_skb(skb);
 601                 goto errout;
 602         }
 603         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 604 errout:
 605         return err;
 606 }
 607
 608 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 609                                       struct netlink_callback *cb)
 610 {
 611         struct net *net = sock_net(skb->sk);
 612         int h, s_h;
 613         int idx, s_idx;
 614         struct net_device *dev;
 615         struct inet6_dev *idev;
 616         struct hlist_head *head;
 617
 618         s_h = cb->args[0];
 619         s_idx = idx = cb->args[1];
 620
 621         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 622                 idx = 0;
 623                 head = &net->dev_index_head[h];
 624                 rcu_read_lock();
 625                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 626                           net->dev_base_seq;
 627                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 628                         if (idx < s_idx)
 629                                 goto cont;
 630                         idev = __in6_dev_get(dev);
 631                         if (!idev)
 632                                 goto cont;
 633
 634                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 635                                                        &idev->cnf,
 636                                                        NETLINK_CB(cb->skb).portid,
 637                                                        cb->nlh->nlmsg_seq,
 638                                                        RTM_NEWNETCONF,
 639                                                        NLM_F_MULTI,
 640                                                        -1) <= 0) {
 641                                 rcu_read_unlock();
 642                                 goto done;
 643                         }
 644                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 645 cont:
 646                         idx++;
 647                 }
 648                 rcu_read_unlock();
 649         }
 650         if (h == NETDEV_HASHENTRIES) {
 651                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 652                                                net->ipv6.devconf_all,
 653                                                NETLINK_CB(cb->skb).portid,
 654                                                cb->nlh->nlmsg_seq,
 655                                                RTM_NEWNETCONF, NLM_F_MULTI,
 656                                                -1) <= 0)
 657                         goto done;
 658                 else
 659                         h++;
 660         }
 661         if (h == NETDEV_HASHENTRIES + 1) {
 662                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 663                                                net->ipv6.devconf_dflt,
 664                                                NETLINK_CB(cb->skb).portid,
 665                                                cb->nlh->nlmsg_seq,
 666                                                RTM_NEWNETCONF, NLM_F_MULTI,
 667                                                -1) <= 0)
 668                         goto done;
 669                 else
 670                         h++;
 671         }
 672 done:
 673         cb->args[0] = h;
 674         cb->args[1] = idx;
 675
 676         return skb->len;
 677 }
 678
 679 #ifdef CONFIG_SYSCTL
 680 static void dev_forward_change(struct inet6_dev *idev)
 681 {
 682         struct net_device *dev;
 683         struct inet6_ifaddr *ifa;
 684
 685         if (!idev)
 686                 return;
 687         dev = idev->dev;
 688         if (idev->cnf.forwarding)
 689                 dev_disable_lro(dev);
 690         if (dev->flags & IFF_MULTICAST) {
 691                 if (idev->cnf.forwarding) {
 692                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 693                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 694                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 695                 } else {
 696                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 697                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 698                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 699                 }
 700         }
 701
 702         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 703                 if (ifa->flags&IFA_F_TENTATIVE)
 704                         continue;
 705                 if (idev->cnf.forwarding)
 706                         addrconf_join_anycast(ifa);
 707                 else
 708                         addrconf_leave_anycast(ifa);
 709         }
 710         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 711                                      dev->ifindex, &idev->cnf);
 712 }
 713
 714
 715 static void addrconf_forward_change(struct net *net, __s32 newf)
 716 {
 717         struct net_device *dev;
 718         struct inet6_dev *idev;
 719
 720         for_each_netdev(net, dev) {
 721                 idev = __in6_dev_get(dev);
 722                 if (idev) {
 723                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 724                         idev->cnf.forwarding = newf;
 725                         if (changed)
 726                                 dev_forward_change(idev);
 727                 }
 728         }
 729 }
 730
 731 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 732 {
 733         struct net *net;
 734         int old;
 735
 736         if (!rtnl_trylock())
 737                 return restart_syscall();
 738
 739         net = (struct net *)table->extra2;
 740         old = *p;
 741         *p = newf;
 742
 743         if (p == &net->ipv6.devconf_dflt->forwarding) {
 744                 if ((!newf) ^ (!old))
 745                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 746                                                      NETCONFA_IFINDEX_DEFAULT,
 747                                                      net->ipv6.devconf_dflt);
 748                 rtnl_unlock();
 749                 return 0;
 750         }
 751
 752         if (p == &net->ipv6.devconf_all->forwarding) {
 753                 net->ipv6.devconf_dflt->forwarding = newf;
 754                 addrconf_forward_change(net, newf);
 755                 if ((!newf) ^ (!old))
 756                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 757                                                      NETCONFA_IFINDEX_ALL,
 758                                                      net->ipv6.devconf_all);
 759         } else if ((!newf) ^ (!old))
 760                 dev_forward_change((struct inet6_dev *)table->extra1);
 761         rtnl_unlock();
 762
 763         if (newf)
 764                 rt6_purge_dflt_routers(net);
 765         return 1;
 766 }
 767 #endif
 768
 769 /* Nobody refers to this ifaddr, destroy it */
 770 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 771 {
 772         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 773
 774 #ifdef NET_REFCNT_DEBUG
 775         pr_debug("%s\n", __func__);
 776 #endif
 777
 778         in6_dev_put(ifp->idev);
 779
 780         if (del_timer(&ifp->timer))
 781                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 782
 783         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 784                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 785                 return;
 786         }
 787         ip6_rt_put(ifp->rt);
 788
 789         kfree_rcu(ifp, rcu);
 790 }
 791
 792 static void
 793 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 794 {
 795         struct list_head *p;
 796         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 797
 798         /*
 799          * Each device address list is sorted in order of scope -
 800          * global before linklocal.
 801          */
 802         list_for_each(p, &idev->addr_list) {
 803                 struct inet6_ifaddr *ifa
 804                         = list_entry(p, struct inet6_ifaddr, if_list);
 805                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 806                         break;
 807         }
 808
 809         list_add_tail(&ifp->if_list, p);
 810 }
 811
 812 static u32 inet6_addr_hash(const struct in6_addr *addr)
 813 {
 814         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 815 }
 816
 817 /* On success it returns ifp with increased reference count */
 818
 819 static struct inet6_ifaddr *
 820 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 821               int scope, u32 flags)
 822 {
 823         struct inet6_ifaddr *ifa = NULL;
 824         struct rt6_info *rt;
 825         unsigned int hash;
 826         int err = 0;
 827         int addr_type = ipv6_addr_type(addr);
 828
 829         if (addr_type == IPV6_ADDR_ANY ||
 830             addr_type & IPV6_ADDR_MULTICAST ||
 831             (!(idev->dev->flags & IFF_LOOPBACK) &&
 832              addr_type & IPV6_ADDR_LOOPBACK))
 833                 return ERR_PTR(-EADDRNOTAVAIL);
 834
 835         rcu_read_lock_bh();
 836         if (idev->dead) {
 837                 err = -ENODEV;                  /*XXX*/
 838                 goto out2;
 839         }
 840
 841         if (idev->cnf.disable_ipv6) {
 842                 err = -EACCES;
 843                 goto out2;
 844         }
 845
 846         spin_lock(&addrconf_hash_lock);
 847
 848         /* Ignore adding duplicate addresses on an interface */
 849         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 850                 ADBG(("ipv6_add_addr: already assigned\n"));
 851                 err = -EEXIST;
 852                 goto out;
 853         }
 854
 855         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 856
 857         if (ifa == NULL) {
 858                 ADBG(("ipv6_add_addr: malloc failed\n"));
 859                 err = -ENOBUFS;
 860                 goto out;
 861         }
 862
 863         rt = addrconf_dst_alloc(idev, addr, false);
 864         if (IS_ERR(rt)) {
 865                 err = PTR_ERR(rt);
 866                 goto out;
 867         }
 868
 869         ifa->addr = *addr;
 870
 871         spin_lock_init(&ifa->lock);
 872         spin_lock_init(&ifa->state_lock);
 873         init_timer(&ifa->timer);
 874         INIT_HLIST_NODE(&ifa->addr_lst);
 875         ifa->timer.data = (unsigned long) ifa;
 876         ifa->scope = scope;
 877         ifa->prefix_len = pfxlen;
 878         ifa->flags = flags | IFA_F_TENTATIVE;
 879         ifa->cstamp = ifa->tstamp = jiffies;
 880
 881         ifa->rt = rt;
 882
 883         ifa->idev = idev;
 884         in6_dev_hold(idev);
 885         /* For caller */
 886         in6_ifa_hold(ifa);
 887
 888         /* Add to big hash table */
 889         hash = inet6_addr_hash(addr);
 890
 891         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 892         spin_unlock(&addrconf_hash_lock);
 893
 894         write_lock(&idev->lock);
 895         /* Add to inet6_dev unicast addr list. */
 896         ipv6_link_dev_addr(idev, ifa);
 897
 898 #ifdef CONFIG_IPV6_PRIVACY
 899         if (ifa->flags&IFA_F_TEMPORARY) {
 900                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 901                 in6_ifa_hold(ifa);
 902         }
 903 #endif
 904
 905         in6_ifa_hold(ifa);
 906         write_unlock(&idev->lock);
 907 out2:
 908         rcu_read_unlock_bh();
 909
 910         if (likely(err == 0))
 911                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 912         else {
 913                 kfree(ifa);
 914                 ifa = ERR_PTR(err);
 915         }
 916
 917         return ifa;
 918 out:
 919         spin_unlock(&addrconf_hash_lock);
 920         goto out2;
 921 }
 922
 923 /* This function wants to get referenced ifp and releases it before return */
 924
 925 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 926 {
 927         struct inet6_ifaddr *ifa, *ifn;
 928         struct inet6_dev *idev = ifp->idev;
 929         int state;
 930         int deleted = 0, onlink = 0;
 931         unsigned long expires = jiffies;
 932
 933         spin_lock_bh(&ifp->state_lock);
 934         state = ifp->state;
 935         ifp->state = INET6_IFADDR_STATE_DEAD;
 936         spin_unlock_bh(&ifp->state_lock);
 937
 938         if (state == INET6_IFADDR_STATE_DEAD)
 939                 goto out;
 940
 941         spin_lock_bh(&addrconf_hash_lock);
 942         hlist_del_init_rcu(&ifp->addr_lst);
 943         spin_unlock_bh(&addrconf_hash_lock);
 944
 945         write_lock_bh(&idev->lock);
 946 #ifdef CONFIG_IPV6_PRIVACY
 947         if (ifp->flags&IFA_F_TEMPORARY) {
 948                 list_del(&ifp->tmp_list);
 949                 if (ifp->ifpub) {
 950                         in6_ifa_put(ifp->ifpub);
 951                         ifp->ifpub = NULL;
 952                 }
 953                 __in6_ifa_put(ifp);
 954         }
 955 #endif
 956
 957         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 958                 if (ifa == ifp) {
 959                         list_del_init(&ifp->if_list);
 960                         __in6_ifa_put(ifp);
 961
 962                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 963                                 break;
 964                         deleted = 1;
 965                         continue;
 966                 } else if (ifp->flags & IFA_F_PERMANENT) {
 967                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 968                                               ifp->prefix_len)) {
 969                                 if (ifa->flags & IFA_F_PERMANENT) {
 970                                         onlink = 1;
 971                                         if (deleted)
 972                                                 break;
 973                                 } else {
 974                                         unsigned long lifetime;
 975
 976                                         if (!onlink)
 977                                                 onlink = -1;
 978
 979                                         spin_lock(&ifa->lock);
 980
 981                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 982                                         /*
 983                                          * Note: Because this address is
 984                                          * not permanent, lifetime <
 985                                          * LONG_MAX / HZ here.
 986                                          */
 987                                         if (time_before(expires,
 988                                                         ifa->tstamp + lifetime * HZ))
 989                                                 expires = ifa->tstamp + lifetime * HZ;
 990                                         spin_unlock(&ifa->lock);
 991                                 }
 992                         }
 993                 }
 994         }
 995         write_unlock_bh(&idev->lock);
 996
 997         addrconf_del_timer(ifp);
 998
 999         ipv6_ifa_notify(RTM_DELADDR, ifp);
1000
1001         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
1002
1003         /*
1004          * Purge or update corresponding prefix
1005          *
1006          * 1) we don't purge prefix here if address was not permanent.
1007          *    prefix is managed by its own lifetime.
1008          * 2) if there're no addresses, delete prefix.
1009          * 3) if there're still other permanent address(es),
1010          *    corresponding prefix is still permanent.
1011          * 4) otherwise, update prefix lifetime to the
1012          *    longest valid lifetime among the corresponding
1013          *    addresses on the device.
1014          *    Note: subsequent RA will update lifetime.
1015          *
1016          * --yoshfuji
1017          */
1018         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1019                 struct in6_addr prefix;
1020                 struct rt6_info *rt;
1021
1022                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1023
1024                 rt = addrconf_get_prefix_route(&prefix,
1025                                                ifp->prefix_len,
1026                                                ifp->idev->dev,
1027                                                0, RTF_GATEWAY | RTF_DEFAULT);
1028
1029                 if (rt) {
1030                         if (onlink == 0) {
1031                                 ip6_del_rt(rt);
1032                                 rt = NULL;
1033                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1034                                 rt6_set_expires(rt, expires);
1035                         }
1036                 }
1037                 ip6_rt_put(rt);
1038         }
1039
1040         /* clean up prefsrc entries */
1041         rt6_remove_prefsrc(ifp);
1042 out:
1043         in6_ifa_put(ifp);
1044 }
1045
1046 #ifdef CONFIG_IPV6_PRIVACY
1047 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1048 {
1049         struct inet6_dev *idev = ifp->idev;
1050         struct in6_addr addr, *tmpaddr;
1051         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1052         unsigned long regen_advance;
1053         int tmp_plen;
1054         int ret = 0;
1055         int max_addresses;
1056         u32 addr_flags;
1057         unsigned long now = jiffies;
1058
1059         write_lock(&idev->lock);
1060         if (ift) {
1061                 spin_lock_bh(&ift->lock);
1062                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1063                 spin_unlock_bh(&ift->lock);
1064                 tmpaddr = &addr;
1065         } else {
1066                 tmpaddr = NULL;
1067         }
1068 retry:
1069         in6_dev_hold(idev);
1070         if (idev->cnf.use_tempaddr <= 0) {
1071                 write_unlock(&idev->lock);
1072                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1073                 in6_dev_put(idev);
1074                 ret = -1;
1075                 goto out;
1076         }
1077         spin_lock_bh(&ifp->lock);
1078         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1079                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1080                 spin_unlock_bh(&ifp->lock);
1081                 write_unlock(&idev->lock);
1082                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1083                         __func__);
1084                 in6_dev_put(idev);
1085                 ret = -1;
1086                 goto out;
1087         }
1088         in6_ifa_hold(ifp);
1089         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1090         __ipv6_try_regen_rndid(idev, tmpaddr);
1091         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1092         age = (now - ifp->tstamp) / HZ;
1093         tmp_valid_lft = min_t(__u32,
1094                               ifp->valid_lft,
1095                               idev->cnf.temp_valid_lft + age);
1096         tmp_prefered_lft = min_t(__u32,
1097                                  ifp->prefered_lft,
1098                                  idev->cnf.temp_prefered_lft + age -
1099                                  idev->cnf.max_desync_factor);
1100         tmp_plen = ifp->prefix_len;
1101         max_addresses = idev->cnf.max_addresses;
1102         tmp_tstamp = ifp->tstamp;
1103         spin_unlock_bh(&ifp->lock);
1104
1105         regen_advance = idev->cnf.regen_max_retry *
1106                         idev->cnf.dad_transmits *
1107                         idev->nd_parms->retrans_time / HZ;
1108         write_unlock(&idev->lock);
1109
1110         /* A temporary address is created only if this calculated Preferred
1111          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1112          * an implementation must not create a temporary address with a zero
1113          * Preferred Lifetime.
1114          */
1115         if (tmp_prefered_lft <= regen_advance) {
1116                 in6_ifa_put(ifp);
1117                 in6_dev_put(idev);
1118                 ret = -1;
1119                 goto out;
1120         }
1121
1122         addr_flags = IFA_F_TEMPORARY;
1123         /* set in addrconf_prefix_rcv() */
1124         if (ifp->flags & IFA_F_OPTIMISTIC)
1125                 addr_flags |= IFA_F_OPTIMISTIC;
1126
1127         ift = !max_addresses ||
1128               ipv6_count_addresses(idev) < max_addresses ?
1129                 ipv6_add_addr(idev, &addr, tmp_plen,
1130                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1131                               addr_flags) : NULL;
1132         if (IS_ERR_OR_NULL(ift)) {
1133                 in6_ifa_put(ifp);
1134                 in6_dev_put(idev);
1135                 pr_info("%s: retry temporary address regeneration\n", __func__);
1136                 tmpaddr = &addr;
1137                 write_lock(&idev->lock);
1138                 goto retry;
1139         }
1140
1141         spin_lock_bh(&ift->lock);
1142         ift->ifpub = ifp;
1143         ift->valid_lft = tmp_valid_lft;
1144         ift->prefered_lft = tmp_prefered_lft;
1145         ift->cstamp = now;
1146         ift->tstamp = tmp_tstamp;
1147         spin_unlock_bh(&ift->lock);
1148
1149         addrconf_dad_start(ift);
1150         in6_ifa_put(ift);
1151         in6_dev_put(idev);
1152 out:
1153         return ret;
1154 }
1155 #endif
1156
1157 /*
1158  *      Choose an appropriate source address (RFC3484)
1159  */
1160 enum {
1161         IPV6_SADDR_RULE_INIT = 0,
1162         IPV6_SADDR_RULE_LOCAL,
1163         IPV6_SADDR_RULE_SCOPE,
1164         IPV6_SADDR_RULE_PREFERRED,
1165 #ifdef CONFIG_IPV6_MIP6
1166         IPV6_SADDR_RULE_HOA,
1167 #endif
1168         IPV6_SADDR_RULE_OIF,
1169         IPV6_SADDR_RULE_LABEL,
1170 #ifdef CONFIG_IPV6_PRIVACY
1171         IPV6_SADDR_RULE_PRIVACY,
1172 #endif
1173         IPV6_SADDR_RULE_ORCHID,
1174         IPV6_SADDR_RULE_PREFIX,
1175         IPV6_SADDR_RULE_MAX
1176 };
1177
1178 struct ipv6_saddr_score {
1179         int                     rule;
1180         int                     addr_type;
1181         struct inet6_ifaddr     *ifa;
1182         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1183         int                     scopedist;
1184         int                     matchlen;
1185 };
1186
1187 struct ipv6_saddr_dst {
1188         const struct in6_addr *addr;
1189         int ifindex;
1190         int scope;
1191         int label;
1192         unsigned int prefs;
1193 };
1194
1195 static inline int ipv6_saddr_preferred(int type)
1196 {
1197         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1198                 return 1;
1199         return 0;
1200 }
1201
1202 static int ipv6_get_saddr_eval(struct net *net,
1203                                struct ipv6_saddr_score *score,
1204                                struct ipv6_saddr_dst *dst,
1205                                int i)
1206 {
1207         int ret;
1208
1209         if (i <= score->rule) {
1210                 switch (i) {
1211                 case IPV6_SADDR_RULE_SCOPE:
1212                         ret = score->scopedist;
1213                         break;
1214                 case IPV6_SADDR_RULE_PREFIX:
1215                         ret = score->matchlen;
1216                         break;
1217                 default:
1218                         ret = !!test_bit(i, score->scorebits);
1219                 }
1220                 goto out;
1221         }
1222
1223         switch (i) {
1224         case IPV6_SADDR_RULE_INIT:
1225                 /* Rule 0: remember if hiscore is not ready yet */
1226                 ret = !!score->ifa;
1227                 break;
1228         case IPV6_SADDR_RULE_LOCAL:
1229                 /* Rule 1: Prefer same address */
1230                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1231                 break;
1232         case IPV6_SADDR_RULE_SCOPE:
1233                 /* Rule 2: Prefer appropriate scope
1234                  *
1235                  *      ret
1236                  *       ^
1237                  *    -1 |  d 15
1238                  *    ---+--+-+---> scope
1239                  *       |
1240                  *       |             d is scope of the destination.
1241                  *  B-d  |  \
1242                  *       |   \      <- smaller scope is better if
1243                  *  B-15 |    \        if scope is enough for destinaion.
1244                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1245                  * d-C-1 | /
1246                  *       |/         <- greater is better
1247                  *   -C  /             if scope is not enough for destination.
1248                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1249                  *
1250                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1251                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1252                  * Assume B = 0 and we get C > 29.
1253                  */
1254                 ret = __ipv6_addr_src_scope(score->addr_type);
1255                 if (ret >= dst->scope)
1256                         ret = -ret;
1257                 else
1258                         ret -= 128;     /* 30 is enough */
1259                 score->scopedist = ret;
1260                 break;
1261         case IPV6_SADDR_RULE_PREFERRED:
1262                 /* Rule 3: Avoid deprecated and optimistic addresses */
1263                 ret = ipv6_saddr_preferred(score->addr_type) ||
1264                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1265                 break;
1266 #ifdef CONFIG_IPV6_MIP6
1267         case IPV6_SADDR_RULE_HOA:
1268             {
1269                 /* Rule 4: Prefer home address */
1270                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1271                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1272                 break;
1273             }
1274 #endif
1275         case IPV6_SADDR_RULE_OIF:
1276                 /* Rule 5: Prefer outgoing interface */
1277                 ret = (!dst->ifindex ||
1278                        dst->ifindex == score->ifa->idev->dev->ifindex);
1279                 break;
1280         case IPV6_SADDR_RULE_LABEL:
1281                 /* Rule 6: Prefer matching label */
1282                 ret = ipv6_addr_label(net,
1283                                       &score->ifa->addr, score->addr_type,
1284                                       score->ifa->idev->dev->ifindex) == dst->label;
1285                 break;
1286 #ifdef CONFIG_IPV6_PRIVACY
1287         case IPV6_SADDR_RULE_PRIVACY:
1288             {
1289                 /* Rule 7: Prefer public address
1290                  * Note: prefer temporary address if use_tempaddr >= 2
1291                  */
1292                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1293                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1294                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1295                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1296                 break;
1297             }
1298 #endif
1299         case IPV6_SADDR_RULE_ORCHID:
1300                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1301                  *          non-ORCHID vs non-ORCHID
1302                  */
1303                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1304                         ipv6_addr_orchid(dst->addr));
1305                 break;
1306         case IPV6_SADDR_RULE_PREFIX:
1307                 /* Rule 8: Use longest matching prefix */
1308                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1309                 if (ret > score->ifa->prefix_len)
1310                         ret = score->ifa->prefix_len;
1311                 score->matchlen = ret;
1312                 break;
1313         default:
1314                 ret = 0;
1315         }
1316
1317         if (ret)
1318                 __set_bit(i, score->scorebits);
1319         score->rule = i;
1320 out:
1321         return ret;
1322 }
1323
1324 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1325                        const struct in6_addr *daddr, unsigned int prefs,
1326                        struct in6_addr *saddr)
1327 {
1328         struct ipv6_saddr_score scores[2],
1329                                 *score = &scores[0], *hiscore = &scores[1];
1330         struct ipv6_saddr_dst dst;
1331         struct net_device *dev;
1332         int dst_type;
1333
1334         dst_type = __ipv6_addr_type(daddr);
1335         dst.addr = daddr;
1336         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1337         dst.scope = __ipv6_addr_src_scope(dst_type);
1338         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1339         dst.prefs = prefs;
1340
1341         hiscore->rule = -1;
1342         hiscore->ifa = NULL;
1343
1344         rcu_read_lock();
1345
1346         for_each_netdev_rcu(net, dev) {
1347                 struct inet6_dev *idev;
1348
1349                 /* Candidate Source Address (section 4)
1350                  *  - multicast and link-local destination address,
1351                  *    the set of candidate source address MUST only
1352                  *    include addresses assigned to interfaces
1353                  *    belonging to the same link as the outgoing
1354                  *    interface.
1355                  * (- For site-local destination addresses, the
1356                  *    set of candidate source addresses MUST only
1357                  *    include addresses assigned to interfaces
1358                  *    belonging to the same site as the outgoing
1359                  *    interface.)
1360                  */
1361                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1362                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1363                     dst.ifindex && dev->ifindex != dst.ifindex)
1364                         continue;
1365
1366                 idev = __in6_dev_get(dev);
1367                 if (!idev)
1368                         continue;
1369
1370                 read_lock_bh(&idev->lock);
1371                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1372                         int i;
1373
1374                         /*
1375                          * - Tentative Address (RFC2462 section 5.4)
1376                          *  - A tentative address is not considered
1377                          *    "assigned to an interface" in the traditional
1378                          *    sense, unless it is also flagged as optimistic.
1379                          * - Candidate Source Address (section 4)
1380                          *  - In any case, anycast addresses, multicast
1381                          *    addresses, and the unspecified address MUST
1382                          *    NOT be included in a candidate set.
1383                          */
1384                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1385                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1386                                 continue;
1387
1388                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1389
1390                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1391                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1392                                 LIMIT_NETDEBUG(KERN_DEBUG
1393                                                "ADDRCONF: unspecified / multicast address "
1394                                                "assigned as unicast address on %s",
1395                                                dev->name);
1396                                 continue;
1397                         }
1398
1399                         score->rule = -1;
1400                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1401
1402                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1403                                 int minihiscore, miniscore;
1404
1405                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1406                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1407
1408                                 if (minihiscore > miniscore) {
1409                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1410                                             score->scopedist > 0) {
1411                                                 /*
1412                                                  * special case:
1413                                                  * each remaining entry
1414                                                  * has too small (not enough)
1415                                                  * scope, because ifa entries
1416                                                  * are sorted by their scope
1417                                                  * values.
1418                                                  */
1419                                                 goto try_nextdev;
1420                                         }
1421                                         break;
1422                                 } else if (minihiscore < miniscore) {
1423                                         if (hiscore->ifa)
1424                                                 in6_ifa_put(hiscore->ifa);
1425
1426                                         in6_ifa_hold(score->ifa);
1427
1428                                         swap(hiscore, score);
1429
1430                                         /* restore our iterator */
1431                                         score->ifa = hiscore->ifa;
1432
1433                                         break;
1434                                 }
1435                         }
1436                 }
1437 try_nextdev:
1438                 read_unlock_bh(&idev->lock);
1439         }
1440         rcu_read_unlock();
1441
1442         if (!hiscore->ifa)
1443                 return -EADDRNOTAVAIL;
1444
1445         *saddr = hiscore->ifa->addr;
1446         in6_ifa_put(hiscore->ifa);
1447         return 0;
1448 }
1449 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1450
1451 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1452                     unsigned char banned_flags)
1453 {
1454         struct inet6_dev *idev;
1455         int err = -EADDRNOTAVAIL;
1456
1457         rcu_read_lock();
1458         idev = __in6_dev_get(dev);
1459         if (idev) {
1460                 struct inet6_ifaddr *ifp;
1461
1462                 read_lock_bh(&idev->lock);
1463                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1464                         if (ifp->scope == IFA_LINK &&
1465                             !(ifp->flags & banned_flags)) {
1466                                 *addr = ifp->addr;
1467                                 err = 0;
1468                                 break;
1469                         }
1470                 }
1471                 read_unlock_bh(&idev->lock);
1472         }
1473         rcu_read_unlock();
1474         return err;
1475 }
1476
1477 static int ipv6_count_addresses(struct inet6_dev *idev)
1478 {
1479         int cnt = 0;
1480         struct inet6_ifaddr *ifp;
1481
1482         read_lock_bh(&idev->lock);
1483         list_for_each_entry(ifp, &idev->addr_list, if_list)
1484                 cnt++;
1485         read_unlock_bh(&idev->lock);
1486         return cnt;
1487 }
1488
1489 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1490                   struct net_device *dev, int strict)
1491 {
1492         struct inet6_ifaddr *ifp;
1493         unsigned int hash = inet6_addr_hash(addr);
1494
1495         rcu_read_lock_bh();
1496         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1497                 if (!net_eq(dev_net(ifp->idev->dev), net))
1498                         continue;
1499                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1500                     !(ifp->flags&IFA_F_TENTATIVE) &&
1501                     (dev == NULL || ifp->idev->dev == dev ||
1502                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1503                         rcu_read_unlock_bh();
1504                         return 1;
1505                 }
1506         }
1507
1508         rcu_read_unlock_bh();
1509         return 0;
1510 }
1511 EXPORT_SYMBOL(ipv6_chk_addr);
1512
1513 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1514                                struct net_device *dev)
1515 {
1516         unsigned int hash = inet6_addr_hash(addr);
1517         struct inet6_ifaddr *ifp;
1518
1519         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1520                 if (!net_eq(dev_net(ifp->idev->dev), net))
1521                         continue;
1522                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1523                         if (dev == NULL || ifp->idev->dev == dev)
1524                                 return true;
1525                 }
1526         }
1527         return false;
1528 }
1529
1530 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1531 {
1532         struct inet6_dev *idev;
1533         struct inet6_ifaddr *ifa;
1534         int     onlink;
1535
1536         onlink = 0;
1537         rcu_read_lock();
1538         idev = __in6_dev_get(dev);
1539         if (idev) {
1540                 read_lock_bh(&idev->lock);
1541                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1542                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1543                                                    ifa->prefix_len);
1544                         if (onlink)
1545                                 break;
1546                 }
1547                 read_unlock_bh(&idev->lock);
1548         }
1549         rcu_read_unlock();
1550         return onlink;
1551 }
1552 EXPORT_SYMBOL(ipv6_chk_prefix);
1553
1554 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1555                                      struct net_device *dev, int strict)
1556 {
1557         struct inet6_ifaddr *ifp, *result = NULL;
1558         unsigned int hash = inet6_addr_hash(addr);
1559
1560         rcu_read_lock_bh();
1561         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1562                 if (!net_eq(dev_net(ifp->idev->dev), net))
1563                         continue;
1564                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1565                         if (dev == NULL || ifp->idev->dev == dev ||
1566                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1567                                 result = ifp;
1568                                 in6_ifa_hold(ifp);
1569                                 break;
1570                         }
1571                 }
1572         }
1573         rcu_read_unlock_bh();
1574
1575         return result;
1576 }
1577
1578 /* Gets referenced address, destroys ifaddr */
1579
1580 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1581 {
1582         if (ifp->flags&IFA_F_PERMANENT) {
1583                 spin_lock_bh(&ifp->lock);
1584                 addrconf_del_timer(ifp);
1585                 ifp->flags |= IFA_F_TENTATIVE;
1586                 if (dad_failed)
1587                         ifp->flags |= IFA_F_DADFAILED;
1588                 spin_unlock_bh(&ifp->lock);
1589                 if (dad_failed)
1590                         ipv6_ifa_notify(0, ifp);
1591                 in6_ifa_put(ifp);
1592 #ifdef CONFIG_IPV6_PRIVACY
1593         } else if (ifp->flags&IFA_F_TEMPORARY) {
1594                 struct inet6_ifaddr *ifpub;
1595                 spin_lock_bh(&ifp->lock);
1596                 ifpub = ifp->ifpub;
1597                 if (ifpub) {
1598                         in6_ifa_hold(ifpub);
1599                         spin_unlock_bh(&ifp->lock);
1600                         ipv6_create_tempaddr(ifpub, ifp);
1601                         in6_ifa_put(ifpub);
1602                 } else {
1603                         spin_unlock_bh(&ifp->lock);
1604                 }
1605                 ipv6_del_addr(ifp);
1606 #endif
1607         } else
1608                 ipv6_del_addr(ifp);
1609 }
1610
1611 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1612 {
1613         int err = -ENOENT;
1614
1615         spin_lock(&ifp->state_lock);
1616         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1617                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1618                 err = 0;
1619         }
1620         spin_unlock(&ifp->state_lock);
1621
1622         return err;
1623 }
1624
1625 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1626 {
1627         struct inet6_dev *idev = ifp->idev;
1628
1629         if (addrconf_dad_end(ifp)) {
1630                 in6_ifa_put(ifp);
1631                 return;
1632         }
1633
1634         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1635                              ifp->idev->dev->name, &ifp->addr);
1636
1637         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1638                 struct in6_addr addr;
1639
1640                 addr.s6_addr32[0] = htonl(0xfe800000);
1641                 addr.s6_addr32[1] = 0;
1642
1643                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1644                     ipv6_addr_equal(&ifp->addr, &addr)) {
1645                         /* DAD failed for link-local based on MAC address */
1646                         idev->cnf.disable_ipv6 = 1;
1647
1648                         pr_info("%s: IPv6 being disabled!\n",
1649                                 ifp->idev->dev->name);
1650                 }
1651         }
1652
1653         addrconf_dad_stop(ifp, 1);
1654 }
1655
1656 /* Join to solicited addr multicast group. */
1657
1658 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1659 {
1660         struct in6_addr maddr;
1661
1662         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1663                 return;
1664
1665         addrconf_addr_solict_mult(addr, &maddr);
1666         ipv6_dev_mc_inc(dev, &maddr);
1667 }
1668
1669 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1670 {
1671         struct in6_addr maddr;
1672
1673         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1674                 return;
1675
1676         addrconf_addr_solict_mult(addr, &maddr);
1677         __ipv6_dev_mc_dec(idev, &maddr);
1678 }
1679
1680 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1681 {
1682         struct in6_addr addr;
1683         if (ifp->prefix_len == 127) /* RFC 6164 */
1684                 return;
1685         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1686         if (ipv6_addr_any(&addr))
1687                 return;
1688         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1689 }
1690
1691 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1692 {
1693         struct in6_addr addr;
1694         if (ifp->prefix_len == 127) /* RFC 6164 */
1695                 return;
1696         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1697         if (ipv6_addr_any(&addr))
1698                 return;
1699         __ipv6_dev_ac_dec(ifp->idev, &addr);
1700 }
1701
1702 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1703 {
1704         if (dev->addr_len != ETH_ALEN)
1705                 return -1;
1706         memcpy(eui, dev->dev_addr, 3);
1707         memcpy(eui + 5, dev->dev_addr + 3, 3);
1708
1709         /*
1710          * The zSeries OSA network cards can be shared among various
1711          * OS instances, but the OSA cards have only one MAC address.
1712          * This leads to duplicate address conflicts in conjunction
1713          * with IPv6 if more than one instance uses the same card.
1714          *
1715          * The driver for these cards can deliver a unique 16-bit
1716          * identifier for each instance sharing the same card.  It is
1717          * placed instead of 0xFFFE in the interface identifier.  The
1718          * "u" bit of the interface identifier is not inverted in this
1719          * case.  Hence the resulting interface identifier has local
1720          * scope according to RFC2373.
1721          */
1722         if (dev->dev_id) {
1723                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1724                 eui[4] = dev->dev_id & 0xFF;
1725         } else {
1726                 eui[3] = 0xFF;
1727                 eui[4] = 0xFE;
1728                 eui[0] ^= 2;
1729         }
1730         return 0;
1731 }
1732
1733 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1734 {
1735         if (dev->addr_len != IEEE802154_ADDR_LEN)
1736                 return -1;
1737         memcpy(eui, dev->dev_addr, 8);
1738         eui[0] ^= 2;
1739         return 0;
1740 }
1741
1742 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1743 {
1744         union fwnet_hwaddr *ha;
1745
1746         if (dev->addr_len != FWNET_ALEN)
1747                 return -1;
1748
1749         ha = (union fwnet_hwaddr *)dev->dev_addr;
1750
1751         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1752         eui[0] ^= 2;
1753         return 0;
1754 }
1755
1756 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1757 {
1758         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1759         if (dev->addr_len != ARCNET_ALEN)
1760                 return -1;
1761         memset(eui, 0, 7);
1762         eui[7] = *(u8 *)dev->dev_addr;
1763         return 0;
1764 }
1765
1766 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1767 {
1768         if (dev->addr_len != INFINIBAND_ALEN)
1769                 return -1;
1770         memcpy(eui, dev->dev_addr + 12, 8);
1771         eui[0] |= 2;
1772         return 0;
1773 }
1774
1775 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1776 {
1777         if (addr == 0)
1778                 return -1;
1779         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1780                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1781                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1782                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1783                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1784                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1785         eui[1] = 0;
1786         eui[2] = 0x5E;
1787         eui[3] = 0xFE;
1788         memcpy(eui + 4, &addr, 4);
1789         return 0;
1790 }
1791
1792 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1793 {
1794         if (dev->priv_flags & IFF_ISATAP)
1795                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1796         return -1;
1797 }
1798
1799 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1800 {
1801         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1802 }
1803
1804 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1805 {
1806         switch (dev->type) {
1807         case ARPHRD_ETHER:
1808         case ARPHRD_FDDI:
1809                 return addrconf_ifid_eui48(eui, dev);
1810         case ARPHRD_ARCNET:
1811                 return addrconf_ifid_arcnet(eui, dev);
1812         case ARPHRD_INFINIBAND:
1813                 return addrconf_ifid_infiniband(eui, dev);
1814         case ARPHRD_SIT:
1815                 return addrconf_ifid_sit(eui, dev);
1816         case ARPHRD_IPGRE:
1817                 return addrconf_ifid_gre(eui, dev);
1818         case ARPHRD_IEEE802154:
1819                 return addrconf_ifid_eui64(eui, dev);
1820         case ARPHRD_IEEE1394:
1821                 return addrconf_ifid_ieee1394(eui, dev);
1822         }
1823         return -1;
1824 }
1825
1826 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1827 {
1828         int err = -1;
1829         struct inet6_ifaddr *ifp;
1830
1831         read_lock_bh(&idev->lock);
1832         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1833                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1834                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1835                         err = 0;
1836                         break;
1837                 }
1838         }
1839         read_unlock_bh(&idev->lock);
1840         return err;
1841 }
1842
1843 #ifdef CONFIG_IPV6_PRIVACY
1844 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1845 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1846 {
1847 regen:
1848         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1849         idev->rndid[0] &= ~0x02;
1850
1851         /*
1852          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1853          * check if generated address is not inappropriate
1854          *
1855          *  - Reserved subnet anycast (RFC 2526)
1856          *      11111101 11....11 1xxxxxxx
1857          *  - ISATAP (RFC4214) 6.1
1858          *      00-00-5E-FE-xx-xx-xx-xx
1859          *  - value 0
1860          *  - XXX: already assigned to an address on the device
1861          */
1862         if (idev->rndid[0] == 0xfd &&
1863             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1864             (idev->rndid[7]&0x80))
1865                 goto regen;
1866         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1867                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1868                         goto regen;
1869                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1870                         goto regen;
1871         }
1872 }
1873
1874 static void ipv6_regen_rndid(unsigned long data)
1875 {
1876         struct inet6_dev *idev = (struct inet6_dev *) data;
1877         unsigned long expires;
1878
1879         rcu_read_lock_bh();
1880         write_lock_bh(&idev->lock);
1881
1882         if (idev->dead)
1883                 goto out;
1884
1885         __ipv6_regen_rndid(idev);
1886
1887         expires = jiffies +
1888                 idev->cnf.temp_prefered_lft * HZ -
1889                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1890                 idev->cnf.max_desync_factor * HZ;
1891         if (time_before(expires, jiffies)) {
1892                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1893                         __func__, idev->dev->name);
1894                 goto out;
1895         }
1896
1897         if (!mod_timer(&idev->regen_timer, expires))
1898                 in6_dev_hold(idev);
1899
1900 out:
1901         write_unlock_bh(&idev->lock);
1902         rcu_read_unlock_bh();
1903         in6_dev_put(idev);
1904 }
1905
1906 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1907 {
1908         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1909                 __ipv6_regen_rndid(idev);
1910 }
1911 #endif
1912
1913 /*
1914  *      Add prefix route.
1915  */
1916
1917 static void
1918 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1919                       unsigned long expires, u32 flags)
1920 {
1921         struct fib6_config cfg = {
1922                 .fc_table = RT6_TABLE_PREFIX,
1923                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1924                 .fc_ifindex = dev->ifindex,
1925                 .fc_expires = expires,
1926                 .fc_dst_len = plen,
1927                 .fc_flags = RTF_UP | flags,
1928                 .fc_nlinfo.nl_net = dev_net(dev),
1929                 .fc_protocol = RTPROT_KERNEL,
1930         };
1931
1932         cfg.fc_dst = *pfx;
1933
1934         /* Prevent useless cloning on PtP SIT.
1935            This thing is done here expecting that the whole
1936            class of non-broadcast devices need not cloning.
1937          */
1938 #if IS_ENABLED(CONFIG_IPV6_SIT)
1939         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1940                 cfg.fc_flags |= RTF_NONEXTHOP;
1941 #endif
1942
1943         ip6_route_add(&cfg);
1944 }
1945
1946
1947 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1948                                                   int plen,
1949                                                   const struct net_device *dev,
1950                                                   u32 flags, u32 noflags)
1951 {
1952         struct fib6_node *fn;
1953         struct rt6_info *rt = NULL;
1954         struct fib6_table *table;
1955
1956         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1957         if (table == NULL)
1958                 return NULL;
1959
1960         read_lock_bh(&table->tb6_lock);
1961         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1962         if (!fn)
1963                 goto out;
1964         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1965                 if (rt->dst.dev->ifindex != dev->ifindex)
1966                         continue;
1967                 if ((rt->rt6i_flags & flags) != flags)
1968                         continue;
1969                 if ((rt->rt6i_flags & noflags) != 0)
1970                         continue;
1971                 dst_hold(&rt->dst);
1972                 break;
1973         }
1974 out:
1975         read_unlock_bh(&table->tb6_lock);
1976         return rt;
1977 }
1978
1979
1980 /* Create "default" multicast route to the interface */
1981
1982 static void addrconf_add_mroute(struct net_device *dev)
1983 {
1984         struct fib6_config cfg = {
1985                 .fc_table = RT6_TABLE_LOCAL,
1986                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1987                 .fc_ifindex = dev->ifindex,
1988                 .fc_dst_len = 8,
1989                 .fc_flags = RTF_UP,
1990                 .fc_nlinfo.nl_net = dev_net(dev),
1991         };
1992
1993         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1994
1995         ip6_route_add(&cfg);
1996 }
1997
1998 #if IS_ENABLED(CONFIG_IPV6_SIT)
1999 static void sit_route_add(struct net_device *dev)
2000 {
2001         struct fib6_config cfg = {
2002                 .fc_table = RT6_TABLE_MAIN,
2003                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2004                 .fc_ifindex = dev->ifindex,
2005                 .fc_dst_len = 96,
2006                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2007                 .fc_nlinfo.nl_net = dev_net(dev),
2008         };
2009
2010         /* prefix length - 96 bits "::d.d.d.d" */
2011         ip6_route_add(&cfg);
2012 }
2013 #endif
2014
2015 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2016 {
2017         struct inet6_dev *idev;
2018
2019         ASSERT_RTNL();
2020
2021         idev = ipv6_find_idev(dev);
2022         if (!idev)
2023                 return ERR_PTR(-ENOBUFS);
2024
2025         if (idev->cnf.disable_ipv6)
2026                 return ERR_PTR(-EACCES);
2027
2028         /* Add default multicast route */
2029         if (!(dev->flags & IFF_LOOPBACK))
2030                 addrconf_add_mroute(dev);
2031
2032         return idev;
2033 }
2034
2035 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2036 {
2037         struct prefix_info *pinfo;
2038         __u32 valid_lft;
2039         __u32 prefered_lft;
2040         int addr_type;
2041         struct inet6_dev *in6_dev;
2042         struct net *net = dev_net(dev);
2043
2044         pinfo = (struct prefix_info *) opt;
2045
2046         if (len < sizeof(struct prefix_info)) {
2047                 ADBG(("addrconf: prefix option too short\n"));
2048                 return;
2049         }
2050
2051         /*
2052          *      Validation checks ([ADDRCONF], page 19)
2053          */
2054
2055         addr_type = ipv6_addr_type(&pinfo->prefix);
2056
2057         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2058                 return;
2059
2060         valid_lft = ntohl(pinfo->valid);
2061         prefered_lft = ntohl(pinfo->prefered);
2062
2063         if (prefered_lft > valid_lft) {
2064                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2065                 return;
2066         }
2067
2068         in6_dev = in6_dev_get(dev);
2069
2070         if (in6_dev == NULL) {
2071                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2072                                     dev->name);
2073                 return;
2074         }
2075
2076         /*
2077          *      Two things going on here:
2078          *      1) Add routes for on-link prefixes
2079          *      2) Configure prefixes with the auto flag set
2080          */
2081
2082         if (pinfo->onlink) {
2083                 struct rt6_info *rt;
2084                 unsigned long rt_expires;
2085
2086                 /* Avoid arithmetic overflow. Really, we could
2087                  * save rt_expires in seconds, likely valid_lft,
2088                  * but it would require division in fib gc, that it
2089                  * not good.
2090                  */
2091                 if (HZ > USER_HZ)
2092                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2093                 else
2094                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2095
2096                 if (addrconf_finite_timeout(rt_expires))
2097                         rt_expires *= HZ;
2098
2099                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2100                                                pinfo->prefix_len,
2101                                                dev,
2102                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2103                                                RTF_GATEWAY | RTF_DEFAULT);
2104
2105                 if (rt) {
2106                         /* Autoconf prefix route */
2107                         if (valid_lft == 0) {
2108                                 ip6_del_rt(rt);
2109                                 rt = NULL;
2110                         } else if (addrconf_finite_timeout(rt_expires)) {
2111                                 /* not infinity */
2112                                 rt6_set_expires(rt, jiffies + rt_expires);
2113                         } else {
2114                                 rt6_clean_expires(rt);
2115                         }
2116                 } else if (valid_lft) {
2117                         clock_t expires = 0;
2118                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2119                         if (addrconf_finite_timeout(rt_expires)) {
2120                                 /* not infinity */
2121                                 flags |= RTF_EXPIRES;
2122                                 expires = jiffies_to_clock_t(rt_expires);
2123                         }
2124                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2125                                               dev, expires, flags);
2126                 }
2127                 ip6_rt_put(rt);
2128         }
2129
2130         /* Try to figure out our local address for this prefix */
2131
2132         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2133                 struct inet6_ifaddr *ifp;
2134                 struct in6_addr addr;
2135                 int create = 0, update_lft = 0;
2136
2137                 if (pinfo->prefix_len == 64) {
2138                         memcpy(&addr, &pinfo->prefix, 8);
2139                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2140                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2141                                 in6_dev_put(in6_dev);
2142                                 return;
2143                         }
2144                         goto ok;
2145                 }
2146                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2147                                     pinfo->prefix_len);
2148                 in6_dev_put(in6_dev);
2149                 return;
2150
2151 ok:
2152
2153                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2154
2155                 if (ifp == NULL && valid_lft) {
2156                         int max_addresses = in6_dev->cnf.max_addresses;
2157                         u32 addr_flags = 0;
2158
2159 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2160                         if (in6_dev->cnf.optimistic_dad &&
2161                             !net->ipv6.devconf_all->forwarding && sllao)
2162                                 addr_flags = IFA_F_OPTIMISTIC;
2163 #endif
2164
2165                         /* Do not allow to create too much of autoconfigured
2166                          * addresses; this would be too easy way to crash kernel.
2167                          */
2168                         if (!max_addresses ||
2169                             ipv6_count_addresses(in6_dev) < max_addresses)
2170                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2171                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2172                                                     addr_flags);
2173
2174                         if (IS_ERR_OR_NULL(ifp)) {
2175                                 in6_dev_put(in6_dev);
2176                                 return;
2177                         }
2178
2179                         update_lft = create = 1;
2180                         ifp->cstamp = jiffies;
2181                         addrconf_dad_start(ifp);
2182                 }
2183
2184                 if (ifp) {
2185                         int flags;
2186                         unsigned long now;
2187 #ifdef CONFIG_IPV6_PRIVACY
2188                         struct inet6_ifaddr *ift;
2189 #endif
2190                         u32 stored_lft;
2191
2192                         /* update lifetime (RFC2462 5.5.3 e) */
2193                         spin_lock(&ifp->lock);
2194                         now = jiffies;
2195                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2196                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2197                         else
2198                                 stored_lft = 0;
2199                         if (!update_lft && stored_lft) {
2200                                 if (valid_lft > MIN_VALID_LIFETIME ||
2201                                     valid_lft > stored_lft)
2202                                         update_lft = 1;
2203                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2204                                         /* valid_lft <= stored_lft is always true */
2205                                         /*
2206                                          * RFC 4862 Section 5.5.3e:
2207                                          * "Note that the preferred lifetime of
2208                                          *  the corresponding address is always
2209                                          *  reset to the Preferred Lifetime in
2210                                          *  the received Prefix Information
2211                                          *  option, regardless of whether the
2212                                          *  valid lifetime is also reset or
2213                                          *  ignored."
2214                                          *
2215                                          *  So if the preferred lifetime in
2216                                          *  this advertisement is different
2217                                          *  than what we have stored, but the
2218                                          *  valid lifetime is invalid, just
2219                                          *  reset prefered_lft.
2220                                          *
2221                                          *  We must set the valid lifetime
2222                                          *  to the stored lifetime since we'll
2223                                          *  be updating the timestamp below,
2224                                          *  else we'll set it back to the
2225                                          *  minimum.
2226                                          */
2227                                         if (prefered_lft != ifp->prefered_lft) {
2228                                                 valid_lft = stored_lft;
2229                                                 update_lft = 1;
2230                                         }
2231                                 } else {
2232                                         valid_lft = MIN_VALID_LIFETIME;
2233                                         if (valid_lft < prefered_lft)
2234                                                 prefered_lft = valid_lft;
2235                                         update_lft = 1;
2236                                 }
2237                         }
2238
2239                         if (update_lft) {
2240                                 ifp->valid_lft = valid_lft;
2241                                 ifp->prefered_lft = prefered_lft;
2242                                 ifp->tstamp = now;
2243                                 flags = ifp->flags;
2244                                 ifp->flags &= ~IFA_F_DEPRECATED;
2245                                 spin_unlock(&ifp->lock);
2246
2247                                 if (!(flags&IFA_F_TENTATIVE))
2248                                         ipv6_ifa_notify(0, ifp);
2249                         } else
2250                                 spin_unlock(&ifp->lock);
2251
2252 #ifdef CONFIG_IPV6_PRIVACY
2253                         read_lock_bh(&in6_dev->lock);
2254                         /* update all temporary addresses in the list */
2255                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2256                                             tmp_list) {
2257                                 int age, max_valid, max_prefered;
2258
2259                                 if (ifp != ift->ifpub)
2260                                         continue;
2261
2262                                 /*
2263                                  * RFC 4941 section 3.3:
2264                                  * If a received option will extend the lifetime
2265                                  * of a public address, the lifetimes of
2266                                  * temporary addresses should be extended,
2267                                  * subject to the overall constraint that no
2268                                  * temporary addresses should ever remain
2269                                  * "valid" or "preferred" for a time longer than
2270                                  * (TEMP_VALID_LIFETIME) or
2271                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2272                                  * respectively.
2273                                  */
2274                                 age = (now - ift->cstamp) / HZ;
2275                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2276                                 if (max_valid < 0)
2277                                         max_valid = 0;
2278
2279                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2280                                                in6_dev->cnf.max_desync_factor -
2281                                                age;
2282                                 if (max_prefered < 0)
2283                                         max_prefered = 0;
2284
2285                                 if (valid_lft > max_valid)
2286                                         valid_lft = max_valid;
2287
2288                                 if (prefered_lft > max_prefered)
2289                                         prefered_lft = max_prefered;
2290
2291                                 spin_lock(&ift->lock);
2292                                 flags = ift->flags;
2293                                 ift->valid_lft = valid_lft;
2294                                 ift->prefered_lft = prefered_lft;
2295                                 ift->tstamp = now;
2296                                 if (prefered_lft > 0)
2297                                         ift->flags &= ~IFA_F_DEPRECATED;
2298
2299                                 spin_unlock(&ift->lock);
2300                                 if (!(flags&IFA_F_TENTATIVE))
2301                                         ipv6_ifa_notify(0, ift);
2302                         }
2303
2304                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2305                                 /*
2306                                  * When a new public address is created as
2307                                  * described in [ADDRCONF], also create a new
2308                                  * temporary address. Also create a temporary
2309                                  * address if it's enabled but no temporary
2310                                  * address currently exists.
2311                                  */
2312                                 read_unlock_bh(&in6_dev->lock);
2313                                 ipv6_create_tempaddr(ifp, NULL);
2314                         } else {
2315                                 read_unlock_bh(&in6_dev->lock);
2316                         }
2317 #endif
2318                         in6_ifa_put(ifp);
2319                         addrconf_verify(0);
2320                 }
2321         }
2322         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2323         in6_dev_put(in6_dev);
2324 }
2325
2326 /*
2327  *      Set destination address.
2328  *      Special case for SIT interfaces where we create a new "virtual"
2329  *      device.
2330  */
2331 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2332 {
2333         struct in6_ifreq ireq;
2334         struct net_device *dev;
2335         int err = -EINVAL;
2336
2337         rtnl_lock();
2338
2339         err = -EFAULT;
2340         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2341                 goto err_exit;
2342
2343         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2344
2345         err = -ENODEV;
2346         if (dev == NULL)
2347                 goto err_exit;
2348
2349 #if IS_ENABLED(CONFIG_IPV6_SIT)
2350         if (dev->type == ARPHRD_SIT) {
2351                 const struct net_device_ops *ops = dev->netdev_ops;
2352                 struct ifreq ifr;
2353                 struct ip_tunnel_parm p;
2354
2355                 err = -EADDRNOTAVAIL;
2356                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2357                         goto err_exit;
2358
2359                 memset(&p, 0, sizeof(p));
2360                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2361                 p.iph.saddr = 0;
2362                 p.iph.version = 4;
2363                 p.iph.ihl = 5;
2364                 p.iph.protocol = IPPROTO_IPV6;
2365                 p.iph.ttl = 64;
2366                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2367
2368                 if (ops->ndo_do_ioctl) {
2369                         mm_segment_t oldfs = get_fs();
2370
2371                         set_fs(KERNEL_DS);
2372                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2373                         set_fs(oldfs);
2374                 } else
2375                         err = -EOPNOTSUPP;
2376
2377                 if (err == 0) {
2378                         err = -ENOBUFS;
2379                         dev = __dev_get_by_name(net, p.name);
2380                         if (!dev)
2381                                 goto err_exit;
2382                         err = dev_open(dev);
2383                 }
2384         }
2385 #endif
2386
2387 err_exit:
2388         rtnl_unlock();
2389         return err;
2390 }
2391
2392 /*
2393  *      Manual configuration of address on an interface
2394  */
2395 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2396                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2397                           __u32 valid_lft)
2398 {
2399         struct inet6_ifaddr *ifp;
2400         struct inet6_dev *idev;
2401         struct net_device *dev;
2402         int scope;
2403         u32 flags;
2404         clock_t expires;
2405         unsigned long timeout;
2406
2407         ASSERT_RTNL();
2408
2409         if (plen > 128)
2410                 return -EINVAL;
2411
2412         /* check the lifetime */
2413         if (!valid_lft || prefered_lft > valid_lft)
2414                 return -EINVAL;
2415
2416         dev = __dev_get_by_index(net, ifindex);
2417         if (!dev)
2418                 return -ENODEV;
2419
2420         idev = addrconf_add_dev(dev);
2421         if (IS_ERR(idev))
2422                 return PTR_ERR(idev);
2423
2424         scope = ipv6_addr_scope(pfx);
2425
2426         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2427         if (addrconf_finite_timeout(timeout)) {
2428                 expires = jiffies_to_clock_t(timeout * HZ);
2429                 valid_lft = timeout;
2430                 flags = RTF_EXPIRES;
2431         } else {
2432                 expires = 0;
2433                 flags = 0;
2434                 ifa_flags |= IFA_F_PERMANENT;
2435         }
2436
2437         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2438         if (addrconf_finite_timeout(timeout)) {
2439                 if (timeout == 0)
2440                         ifa_flags |= IFA_F_DEPRECATED;
2441                 prefered_lft = timeout;
2442         }
2443
2444         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2445
2446         if (!IS_ERR(ifp)) {
2447                 spin_lock_bh(&ifp->lock);
2448                 ifp->valid_lft = valid_lft;
2449                 ifp->prefered_lft = prefered_lft;
2450                 ifp->tstamp = jiffies;
2451                 spin_unlock_bh(&ifp->lock);
2452
2453                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2454                                       expires, flags);
2455                 /*
2456                  * Note that section 3.1 of RFC 4429 indicates
2457                  * that the Optimistic flag should not be set for
2458                  * manually configured addresses
2459                  */
2460                 addrconf_dad_start(ifp);
2461                 in6_ifa_put(ifp);
2462                 addrconf_verify(0);
2463                 return 0;
2464         }
2465
2466         return PTR_ERR(ifp);
2467 }
2468
2469 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2470                           unsigned int plen)
2471 {
2472         struct inet6_ifaddr *ifp;
2473         struct inet6_dev *idev;
2474         struct net_device *dev;
2475
2476         if (plen > 128)
2477                 return -EINVAL;
2478
2479         dev = __dev_get_by_index(net, ifindex);
2480         if (!dev)
2481                 return -ENODEV;
2482
2483         if ((idev = __in6_dev_get(dev)) == NULL)
2484                 return -ENXIO;
2485
2486         read_lock_bh(&idev->lock);
2487         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2488                 if (ifp->prefix_len == plen &&
2489                     ipv6_addr_equal(pfx, &ifp->addr)) {
2490                         in6_ifa_hold(ifp);
2491                         read_unlock_bh(&idev->lock);
2492
2493                         ipv6_del_addr(ifp);
2494
2495                         /* If the last address is deleted administratively,
2496                            disable IPv6 on this interface.
2497                          */
2498                         if (list_empty(&idev->addr_list))
2499                                 addrconf_ifdown(idev->dev, 1);
2500                         return 0;
2501                 }
2502         }
2503         read_unlock_bh(&idev->lock);
2504         return -EADDRNOTAVAIL;
2505 }
2506
2507
2508 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2509 {
2510         struct in6_ifreq ireq;
2511         int err;
2512
2513         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2514                 return -EPERM;
2515
2516         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2517                 return -EFAULT;
2518
2519         rtnl_lock();
2520         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2521                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2522                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2523         rtnl_unlock();
2524         return err;
2525 }
2526
2527 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2528 {
2529         struct in6_ifreq ireq;
2530         int err;
2531
2532         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2533                 return -EPERM;
2534
2535         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2536                 return -EFAULT;
2537
2538         rtnl_lock();
2539         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2540                              ireq.ifr6_prefixlen);
2541         rtnl_unlock();
2542         return err;
2543 }
2544
2545 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2546                      int plen, int scope)
2547 {
2548         struct inet6_ifaddr *ifp;
2549
2550         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2551         if (!IS_ERR(ifp)) {
2552                 spin_lock_bh(&ifp->lock);
2553                 ifp->flags &= ~IFA_F_TENTATIVE;
2554                 spin_unlock_bh(&ifp->lock);
2555                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2556                 in6_ifa_put(ifp);
2557         }
2558 }
2559
2560 #if IS_ENABLED(CONFIG_IPV6_SIT)
2561 static void sit_add_v4_addrs(struct inet6_dev *idev)
2562 {
2563         struct in6_addr addr;
2564         struct net_device *dev;
2565         struct net *net = dev_net(idev->dev);
2566         int scope;
2567
2568         ASSERT_RTNL();
2569
2570         memset(&addr, 0, sizeof(struct in6_addr));
2571         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2572
2573         if (idev->dev->flags&IFF_POINTOPOINT) {
2574                 addr.s6_addr32[0] = htonl(0xfe800000);
2575                 scope = IFA_LINK;
2576         } else {
2577                 scope = IPV6_ADDR_COMPATv4;
2578         }
2579
2580         if (addr.s6_addr32[3]) {
2581                 add_addr(idev, &addr, 128, scope);
2582                 return;
2583         }
2584
2585         for_each_netdev(net, dev) {
2586                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2587                 if (in_dev && (dev->flags & IFF_UP)) {
2588                         struct in_ifaddr *ifa;
2589
2590                         int flag = scope;
2591
2592                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2593                                 int plen;
2594
2595                                 addr.s6_addr32[3] = ifa->ifa_local;
2596
2597                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2598                                         continue;
2599                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2600                                         if (idev->dev->flags&IFF_POINTOPOINT)
2601                                                 continue;
2602                                         flag |= IFA_HOST;
2603                                 }
2604                                 if (idev->dev->flags&IFF_POINTOPOINT)
2605                                         plen = 64;
2606                                 else
2607                                         plen = 96;
2608
2609                                 add_addr(idev, &addr, plen, flag);
2610                         }
2611                 }
2612         }
2613 }
2614 #endif
2615
2616 static void init_loopback(struct net_device *dev)
2617 {
2618         struct inet6_dev  *idev;
2619
2620         /* ::1 */
2621
2622         ASSERT_RTNL();
2623
2624         if ((idev = ipv6_find_idev(dev)) == NULL) {
2625                 pr_debug("%s: add_dev failed\n", __func__);
2626                 return;
2627         }
2628
2629         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2630 }
2631
2632 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2633 {
2634         struct inet6_ifaddr *ifp;
2635         u32 addr_flags = IFA_F_PERMANENT;
2636
2637 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2638         if (idev->cnf.optimistic_dad &&
2639             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2640                 addr_flags |= IFA_F_OPTIMISTIC;
2641 #endif
2642
2643
2644         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2645         if (!IS_ERR(ifp)) {
2646                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2647                 addrconf_dad_start(ifp);
2648                 in6_ifa_put(ifp);
2649         }
2650 }
2651
2652 static void addrconf_dev_config(struct net_device *dev)
2653 {
2654         struct in6_addr addr;
2655         struct inet6_dev *idev;
2656
2657         ASSERT_RTNL();
2658
2659         if ((dev->type != ARPHRD_ETHER) &&
2660             (dev->type != ARPHRD_FDDI) &&
2661             (dev->type != ARPHRD_ARCNET) &&
2662             (dev->type != ARPHRD_INFINIBAND) &&
2663             (dev->type != ARPHRD_IEEE802154) &&
2664             (dev->type != ARPHRD_IEEE1394)) {
2665                 /* Alas, we support only Ethernet autoconfiguration. */
2666                 return;
2667         }
2668
2669         idev = addrconf_add_dev(dev);
2670         if (IS_ERR(idev))
2671                 return;
2672
2673         memset(&addr, 0, sizeof(struct in6_addr));
2674         addr.s6_addr32[0] = htonl(0xFE800000);
2675
2676         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2677                 addrconf_add_linklocal(idev, &addr);
2678 }
2679
2680 #if IS_ENABLED(CONFIG_IPV6_SIT)
2681 static void addrconf_sit_config(struct net_device *dev)
2682 {
2683         struct inet6_dev *idev;
2684
2685         ASSERT_RTNL();
2686
2687         /*
2688          * Configure the tunnel with one of our IPv4
2689          * addresses... we should configure all of
2690          * our v4 addrs in the tunnel
2691          */
2692
2693         if ((idev = ipv6_find_idev(dev)) == NULL) {
2694                 pr_debug("%s: add_dev failed\n", __func__);
2695                 return;
2696         }
2697
2698         if (dev->priv_flags & IFF_ISATAP) {
2699                 struct in6_addr addr;
2700
2701                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2702                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2703                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2704                         addrconf_add_linklocal(idev, &addr);
2705                 return;
2706         }
2707
2708         sit_add_v4_addrs(idev);
2709
2710         if (dev->flags&IFF_POINTOPOINT)
2711                 addrconf_add_mroute(dev);
2712         else
2713                 sit_route_add(dev);
2714 }
2715 #endif
2716
2717 #if IS_ENABLED(CONFIG_NET_IPGRE)
2718 static void addrconf_gre_config(struct net_device *dev)
2719 {
2720         struct inet6_dev *idev;
2721         struct in6_addr addr;
2722
2723         pr_info("%s(%s)\n", __func__, dev->name);
2724
2725         ASSERT_RTNL();
2726
2727         if ((idev = ipv6_find_idev(dev)) == NULL) {
2728                 pr_debug("%s: add_dev failed\n", __func__);
2729                 return;
2730         }
2731
2732         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2733         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2734
2735         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2736                 addrconf_add_linklocal(idev, &addr);
2737 }
2738 #endif
2739
2740 static inline int
2741 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2742 {
2743         struct in6_addr lladdr;
2744
2745         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2746                 addrconf_add_linklocal(idev, &lladdr);
2747                 return 0;
2748         }
2749         return -1;
2750 }
2751
2752 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2753 {
2754         struct net_device *link_dev;
2755         struct net *net = dev_net(idev->dev);
2756
2757         /* first try to inherit the link-local address from the link device */
2758         if (idev->dev->iflink &&
2759             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2760                 if (!ipv6_inherit_linklocal(idev, link_dev))
2761                         return;
2762         }
2763         /* then try to inherit it from any device */
2764         for_each_netdev(net, link_dev) {
2765                 if (!ipv6_inherit_linklocal(idev, link_dev))
2766                         return;
2767         }
2768         pr_debug("init ip6-ip6: add_linklocal failed\n");
2769 }
2770
2771 /*
2772  * Autoconfigure tunnel with a link-local address so routing protocols,
2773  * DHCPv6, MLD etc. can be run over the virtual link
2774  */
2775
2776 static void addrconf_ip6_tnl_config(struct net_device *dev)
2777 {
2778         struct inet6_dev *idev;
2779
2780         ASSERT_RTNL();
2781
2782         idev = addrconf_add_dev(dev);
2783         if (IS_ERR(idev)) {
2784                 pr_debug("init ip6-ip6: add_dev failed\n");
2785                 return;
2786         }
2787         ip6_tnl_add_linklocal(idev);
2788 }
2789
2790 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2791                            void *data)
2792 {
2793         struct net_device *dev = (struct net_device *) data;
2794         struct inet6_dev *idev = __in6_dev_get(dev);
2795         int run_pending = 0;
2796         int err;
2797
2798         switch (event) {
2799         case NETDEV_REGISTER:
2800                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2801                         idev = ipv6_add_dev(dev);
2802                         if (!idev)
2803                                 return notifier_from_errno(-ENOMEM);
2804                 }
2805                 break;
2806
2807         case NETDEV_UP:
2808         case NETDEV_CHANGE:
2809                 if (dev->flags & IFF_SLAVE)
2810                         break;
2811
2812                 if (event == NETDEV_UP) {
2813                         if (!addrconf_qdisc_ok(dev)) {
2814                                 /* device is not ready yet. */
2815                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2816                                         dev->name);
2817                                 break;
2818                         }
2819
2820                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2821                                 idev = ipv6_add_dev(dev);
2822
2823                         if (idev) {
2824                                 idev->if_flags |= IF_READY;
2825                                 run_pending = 1;
2826                         }
2827                 } else {
2828                         if (!addrconf_qdisc_ok(dev)) {
2829                                 /* device is still not ready. */
2830                                 break;
2831                         }
2832
2833                         if (idev) {
2834                                 if (idev->if_flags & IF_READY)
2835                                         /* device is already configured. */
2836                                         break;
2837                                 idev->if_flags |= IF_READY;
2838                         }
2839
2840                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2841                                 dev->name);
2842
2843                         run_pending = 1;
2844                 }
2845
2846                 switch (dev->type) {
2847 #if IS_ENABLED(CONFIG_IPV6_SIT)
2848                 case ARPHRD_SIT:
2849                         addrconf_sit_config(dev);
2850                         break;
2851 #endif
2852 #if IS_ENABLED(CONFIG_NET_IPGRE)
2853                 case ARPHRD_IPGRE:
2854                         addrconf_gre_config(dev);
2855                         break;
2856 #endif
2857                 case ARPHRD_TUNNEL6:
2858                         addrconf_ip6_tnl_config(dev);
2859                         break;
2860                 case ARPHRD_LOOPBACK:
2861                         init_loopback(dev);
2862                         break;
2863
2864                 default:
2865                         addrconf_dev_config(dev);
2866                         break;
2867                 }
2868
2869                 if (idev) {
2870                         if (run_pending)
2871                                 addrconf_dad_run(idev);
2872
2873                         /*
2874                          * If the MTU changed during the interface down,
2875                          * when the interface up, the changed MTU must be
2876                          * reflected in the idev as well as routers.
2877                          */
2878                         if (idev->cnf.mtu6 != dev->mtu &&
2879                             dev->mtu >= IPV6_MIN_MTU) {
2880                                 rt6_mtu_change(dev, dev->mtu);
2881                                 idev->cnf.mtu6 = dev->mtu;
2882                         }
2883                         idev->tstamp = jiffies;
2884                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2885
2886                         /*
2887                          * If the changed mtu during down is lower than
2888                          * IPV6_MIN_MTU stop IPv6 on this interface.
2889                          */
2890                         if (dev->mtu < IPV6_MIN_MTU)
2891                                 addrconf_ifdown(dev, 1);
2892                 }
2893                 break;
2894
2895         case NETDEV_CHANGEMTU:
2896                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2897                         rt6_mtu_change(dev, dev->mtu);
2898                         idev->cnf.mtu6 = dev->mtu;
2899                         break;
2900                 }
2901
2902                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2903                         idev = ipv6_add_dev(dev);
2904                         if (idev)
2905                                 break;
2906                 }
2907
2908                 /*
2909                  * MTU falled under IPV6_MIN_MTU.
2910                  * Stop IPv6 on this interface.
2911                  */
2912
2913         case NETDEV_DOWN:
2914         case NETDEV_UNREGISTER:
2915                 /*
2916                  *      Remove all addresses from this interface.
2917                  */
2918                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2919                 break;
2920
2921         case NETDEV_CHANGENAME:
2922                 if (idev) {
2923                         snmp6_unregister_dev(idev);
2924                         addrconf_sysctl_unregister(idev);
2925                         addrconf_sysctl_register(idev);
2926                         err = snmp6_register_dev(idev);
2927                         if (err)
2928                                 return notifier_from_errno(err);
2929                 }
2930                 break;
2931
2932         case NETDEV_PRE_TYPE_CHANGE:
2933         case NETDEV_POST_TYPE_CHANGE:
2934                 addrconf_type_change(dev, event);
2935                 break;
2936         }
2937
2938         return NOTIFY_OK;
2939 }
2940
2941 /*
2942  *      addrconf module should be notified of a device going up
2943  */
2944 static struct notifier_block ipv6_dev_notf = {
2945         .notifier_call = addrconf_notify,
2946 };
2947
2948 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2949 {
2950         struct inet6_dev *idev;
2951         ASSERT_RTNL();
2952
2953         idev = __in6_dev_get(dev);
2954
2955         if (event == NETDEV_POST_TYPE_CHANGE)
2956                 ipv6_mc_remap(idev);
2957         else if (event == NETDEV_PRE_TYPE_CHANGE)
2958                 ipv6_mc_unmap(idev);
2959 }
2960
2961 static int addrconf_ifdown(struct net_device *dev, int how)
2962 {
2963         struct net *net = dev_net(dev);
2964         struct inet6_dev *idev;
2965         struct inet6_ifaddr *ifa;
2966         int state, i;
2967
2968         ASSERT_RTNL();
2969
2970         rt6_ifdown(net, dev);
2971         neigh_ifdown(&nd_tbl, dev);
2972
2973         idev = __in6_dev_get(dev);
2974         if (idev == NULL)
2975                 return -ENODEV;
2976
2977         /*
2978          * Step 1: remove reference to ipv6 device from parent device.
2979          *         Do not dev_put!
2980          */
2981         if (how) {
2982                 idev->dead = 1;
2983
2984                 /* protected by rtnl_lock */
2985                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2986
2987                 /* Step 1.5: remove snmp6 entry */
2988                 snmp6_unregister_dev(idev);
2989
2990         }
2991
2992         /* Step 2: clear hash table */
2993         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2994                 struct hlist_head *h = &inet6_addr_lst[i];
2995
2996                 spin_lock_bh(&addrconf_hash_lock);
2997         restart:
2998                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2999                         if (ifa->idev == idev) {
3000                                 hlist_del_init_rcu(&ifa->addr_lst);
3001                                 addrconf_del_timer(ifa);
3002                                 goto restart;
3003                         }
3004                 }
3005                 spin_unlock_bh(&addrconf_hash_lock);
3006         }
3007
3008         write_lock_bh(&idev->lock);
3009
3010         /* Step 2: clear flags for stateless addrconf */
3011         if (!how)
3012                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3013
3014 #ifdef CONFIG_IPV6_PRIVACY
3015         if (how && del_timer(&idev->regen_timer))
3016                 in6_dev_put(idev);
3017
3018         /* Step 3: clear tempaddr list */
3019         while (!list_empty(&idev->tempaddr_list)) {
3020                 ifa = list_first_entry(&idev->tempaddr_list,
3021                                        struct inet6_ifaddr, tmp_list);
3022                 list_del(&ifa->tmp_list);
3023                 write_unlock_bh(&idev->lock);
3024                 spin_lock_bh(&ifa->lock);
3025
3026                 if (ifa->ifpub) {
3027                         in6_ifa_put(ifa->ifpub);
3028                         ifa->ifpub = NULL;
3029                 }
3030                 spin_unlock_bh(&ifa->lock);
3031                 in6_ifa_put(ifa);
3032                 write_lock_bh(&idev->lock);
3033         }
3034 #endif
3035
3036         while (!list_empty(&idev->addr_list)) {
3037                 ifa = list_first_entry(&idev->addr_list,
3038                                        struct inet6_ifaddr, if_list);
3039                 addrconf_del_timer(ifa);
3040
3041                 list_del(&ifa->if_list);
3042
3043                 write_unlock_bh(&idev->lock);
3044
3045                 spin_lock_bh(&ifa->state_lock);
3046                 state = ifa->state;
3047                 ifa->state = INET6_IFADDR_STATE_DEAD;
3048                 spin_unlock_bh(&ifa->state_lock);
3049
3050                 if (state != INET6_IFADDR_STATE_DEAD) {
3051                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3052                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
3053                 }
3054                 in6_ifa_put(ifa);
3055
3056                 write_lock_bh(&idev->lock);
3057         }
3058
3059         write_unlock_bh(&idev->lock);
3060
3061         /* Step 5: Discard multicast list */
3062         if (how)
3063                 ipv6_mc_destroy_dev(idev);
3064         else
3065                 ipv6_mc_down(idev);
3066
3067         idev->tstamp = jiffies;
3068
3069         /* Last: Shot the device (if unregistered) */
3070         if (how) {
3071                 addrconf_sysctl_unregister(idev);
3072                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3073                 neigh_ifdown(&nd_tbl, dev);
3074                 in6_dev_put(idev);
3075         }
3076         return 0;
3077 }
3078
3079 static void addrconf_rs_timer(unsigned long data)
3080 {
3081         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3082         struct inet6_dev *idev = ifp->idev;
3083
3084         read_lock(&idev->lock);
3085         if (idev->dead || !(idev->if_flags & IF_READY))
3086                 goto out;
3087
3088         if (!ipv6_accept_ra(idev))
3089                 goto out;
3090
3091         /* Announcement received after solicitation was sent */
3092         if (idev->if_flags & IF_RA_RCVD)
3093                 goto out;
3094
3095         spin_lock(&ifp->lock);
3096         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3097                 /* The wait after the last probe can be shorter */
3098                 addrconf_mod_timer(ifp, AC_RS,
3099                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3100                                    idev->cnf.rtr_solicit_delay :
3101                                    idev->cnf.rtr_solicit_interval);
3102                 spin_unlock(&ifp->lock);
3103
3104                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3105         } else {
3106                 spin_unlock(&ifp->lock);
3107                 /*
3108                  * Note: we do not support deprecated "all on-link"
3109                  * assumption any longer.
3110                  */
3111                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3112         }
3113
3114 out:
3115         read_unlock(&idev->lock);
3116         in6_ifa_put(ifp);
3117 }
3118
3119 /*
3120  *      Duplicate Address Detection
3121  */
3122 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3123 {
3124         unsigned long rand_num;
3125         struct inet6_dev *idev = ifp->idev;
3126
3127         if (ifp->flags & IFA_F_OPTIMISTIC)
3128                 rand_num = 0;
3129         else
3130                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3131
3132         ifp->probes = idev->cnf.dad_transmits;
3133         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3134 }
3135
3136 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3137 {
3138         struct inet6_dev *idev = ifp->idev;
3139         struct net_device *dev = idev->dev;
3140
3141         addrconf_join_solict(dev, &ifp->addr);
3142
3143         net_srandom(ifp->addr.s6_addr32[3]);
3144
3145         read_lock_bh(&idev->lock);
3146         spin_lock(&ifp->lock);
3147         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3148                 goto out;
3149
3150         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3151             idev->cnf.accept_dad < 1 ||
3152             !(ifp->flags&IFA_F_TENTATIVE) ||
3153             ifp->flags & IFA_F_NODAD) {
3154                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3155                 spin_unlock(&ifp->lock);
3156                 read_unlock_bh(&idev->lock);
3157
3158                 addrconf_dad_completed(ifp);
3159                 return;
3160         }
3161
3162         if (!(idev->if_flags & IF_READY)) {
3163                 spin_unlock(&ifp->lock);
3164                 read_unlock_bh(&idev->lock);
3165                 /*
3166                  * If the device is not ready:
3167                  * - keep it tentative if it is a permanent address.
3168                  * - otherwise, kill it.
3169                  */
3170                 in6_ifa_hold(ifp);
3171                 addrconf_dad_stop(ifp, 0);
3172                 return;
3173         }
3174
3175         /*
3176          * Optimistic nodes can start receiving
3177          * Frames right away
3178          */
3179         if (ifp->flags & IFA_F_OPTIMISTIC)
3180                 ip6_ins_rt(ifp->rt);
3181
3182         addrconf_dad_kick(ifp);
3183 out:
3184         spin_unlock(&ifp->lock);
3185         read_unlock_bh(&idev->lock);
3186 }
3187
3188 static void addrconf_dad_timer(unsigned long data)
3189 {
3190         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3191         struct inet6_dev *idev = ifp->idev;
3192         struct in6_addr mcaddr;
3193
3194         if (!ifp->probes && addrconf_dad_end(ifp))
3195                 goto out;
3196
3197         read_lock(&idev->lock);
3198         if (idev->dead || !(idev->if_flags & IF_READY)) {
3199                 read_unlock(&idev->lock);
3200                 goto out;
3201         }
3202
3203         spin_lock(&ifp->lock);
3204         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3205                 spin_unlock(&ifp->lock);
3206                 read_unlock(&idev->lock);
3207                 goto out;
3208         }
3209
3210         if (ifp->probes == 0) {
3211                 /*
3212                  * DAD was successful
3213                  */
3214
3215                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3216                 spin_unlock(&ifp->lock);
3217                 read_unlock(&idev->lock);
3218
3219                 addrconf_dad_completed(ifp);
3220
3221                 goto out;
3222         }
3223
3224         ifp->probes--;
3225         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3226         spin_unlock(&ifp->lock);
3227         read_unlock(&idev->lock);
3228
3229         /* send a neighbour solicitation for our addr */
3230         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3231         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3232 out:
3233         in6_ifa_put(ifp);
3234 }
3235
3236 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3237 {
3238         struct net_device *dev = ifp->idev->dev;
3239
3240         /*
3241          *      Configure the address for reception. Now it is valid.
3242          */
3243
3244         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3245
3246         /* If added prefix is link local and we are prepared to process
3247            router advertisements, start sending router solicitations.
3248          */
3249
3250         if (ipv6_accept_ra(ifp->idev) &&
3251             ifp->idev->cnf.rtr_solicits > 0 &&
3252             (dev->flags&IFF_LOOPBACK) == 0 &&
3253             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3254                 /*
3255                  *      If a host as already performed a random delay
3256                  *      [...] as part of DAD [...] there is no need
3257                  *      to delay again before sending the first RS
3258                  */
3259                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3260
3261                 spin_lock_bh(&ifp->lock);
3262                 ifp->probes = 1;
3263                 ifp->idev->if_flags |= IF_RS_SENT;
3264                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3265                 spin_unlock_bh(&ifp->lock);
3266         }
3267 }
3268
3269 static void addrconf_dad_run(struct inet6_dev *idev)
3270 {
3271         struct inet6_ifaddr *ifp;
3272
3273         read_lock_bh(&idev->lock);
3274         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3275                 spin_lock(&ifp->lock);
3276                 if (ifp->flags & IFA_F_TENTATIVE &&
3277                     ifp->state == INET6_IFADDR_STATE_DAD)
3278                         addrconf_dad_kick(ifp);
3279                 spin_unlock(&ifp->lock);
3280         }
3281         read_unlock_bh(&idev->lock);
3282 }
3283
3284 #ifdef CONFIG_PROC_FS
3285 struct if6_iter_state {
3286         struct seq_net_private p;
3287         int bucket;
3288         int offset;
3289 };
3290
3291 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3292 {
3293         struct inet6_ifaddr *ifa = NULL;
3294         struct if6_iter_state *state = seq->private;
3295         struct net *net = seq_file_net(seq);
3296         int p = 0;
3297
3298         /* initial bucket if pos is 0 */
3299         if (pos == 0) {
3300                 state->bucket = 0;
3301                 state->offset = 0;
3302         }
3303
3304         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3305                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3306                                          addr_lst) {
3307                         if (!net_eq(dev_net(ifa->idev->dev), net))
3308                                 continue;
3309                         /* sync with offset */
3310                         if (p < state->offset) {
3311                                 p++;
3312                                 continue;
3313                         }
3314                         state->offset++;
3315                         return ifa;
3316                 }
3317
3318                 /* prepare for next bucket */
3319                 state->offset = 0;
3320                 p = 0;
3321         }
3322         return NULL;
3323 }
3324
3325 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3326                                          struct inet6_ifaddr *ifa)
3327 {
3328         struct if6_iter_state *state = seq->private;
3329         struct net *net = seq_file_net(seq);
3330
3331         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3332                 if (!net_eq(dev_net(ifa->idev->dev), net))
3333                         continue;
3334                 state->offset++;
3335                 return ifa;
3336         }
3337
3338         while (++state->bucket < IN6_ADDR_HSIZE) {
3339                 state->offset = 0;
3340                 hlist_for_each_entry_rcu_bh(ifa,
3341                                      &inet6_addr_lst[state->bucket], addr_lst) {
3342                         if (!net_eq(dev_net(ifa->idev->dev), net))
3343                                 continue;
3344                         state->offset++;
3345                         return ifa;
3346                 }
3347         }
3348
3349         return NULL;
3350 }
3351
3352 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3353         __acquires(rcu_bh)
3354 {
3355         rcu_read_lock_bh();
3356         return if6_get_first(seq, *pos);
3357 }
3358
3359 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3360 {
3361         struct inet6_ifaddr *ifa;
3362
3363         ifa = if6_get_next(seq, v);
3364         ++*pos;
3365         return ifa;
3366 }
3367
3368 static void if6_seq_stop(struct seq_file *seq, void *v)
3369         __releases(rcu_bh)
3370 {
3371         rcu_read_unlock_bh();
3372 }
3373
3374 static int if6_seq_show(struct seq_file *seq, void *v)
3375 {
3376         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3377         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3378                    &ifp->addr,
3379                    ifp->idev->dev->ifindex,
3380                    ifp->prefix_len,
3381                    ifp->scope,
3382                    ifp->flags,
3383                    ifp->idev->dev->name);
3384         return 0;
3385 }
3386
3387 static const struct seq_operations if6_seq_ops = {
3388         .start  = if6_seq_start,
3389         .next   = if6_seq_next,
3390         .show   = if6_seq_show,
3391         .stop   = if6_seq_stop,
3392 };
3393
3394 static int if6_seq_open(struct inode *inode, struct file *file)
3395 {
3396         return seq_open_net(inode, file, &if6_seq_ops,
3397                             sizeof(struct if6_iter_state));
3398 }
3399
3400 static const struct file_operations if6_fops = {
3401         .owner          = THIS_MODULE,
3402         .open           = if6_seq_open,
3403         .read           = seq_read,
3404         .llseek         = seq_lseek,
3405         .release        = seq_release_net,
3406 };
3407
3408 static int __net_init if6_proc_net_init(struct net *net)
3409 {
3410         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3411                 return -ENOMEM;
3412         return 0;
3413 }
3414
3415 static void __net_exit if6_proc_net_exit(struct net *net)
3416 {
3417         remove_proc_entry("if_inet6", net->proc_net);
3418 }
3419
3420 static struct pernet_operations if6_proc_net_ops = {
3421        .init = if6_proc_net_init,
3422        .exit = if6_proc_net_exit,
3423 };
3424
3425 int __init if6_proc_init(void)
3426 {
3427         return register_pernet_subsys(&if6_proc_net_ops);
3428 }
3429
3430 void if6_proc_exit(void)
3431 {
3432         unregister_pernet_subsys(&if6_proc_net_ops);
3433 }
3434 #endif  /* CONFIG_PROC_FS */
3435
3436 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3437 /* Check if address is a home address configured on any interface. */
3438 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3439 {
3440         int ret = 0;
3441         struct inet6_ifaddr *ifp = NULL;
3442         unsigned int hash = inet6_addr_hash(addr);
3443
3444         rcu_read_lock_bh();
3445         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3446                 if (!net_eq(dev_net(ifp->idev->dev), net))
3447                         continue;
3448                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3449                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3450                         ret = 1;
3451                         break;
3452                 }
3453         }
3454         rcu_read_unlock_bh();
3455         return ret;
3456 }
3457 #endif
3458
3459 /*
3460  *      Periodic address status verification
3461  */
3462
3463 static void addrconf_verify(unsigned long foo)
3464 {
3465         unsigned long now, next, next_sec, next_sched;
3466         struct inet6_ifaddr *ifp;
3467         int i;
3468
3469         rcu_read_lock_bh();
3470         spin_lock(&addrconf_verify_lock);
3471         now = jiffies;
3472         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3473
3474         del_timer(&addr_chk_timer);
3475
3476         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3477 restart:
3478                 hlist_for_each_entry_rcu_bh(ifp,
3479                                          &inet6_addr_lst[i], addr_lst) {
3480                         unsigned long age;
3481
3482                         if (ifp->flags & IFA_F_PERMANENT)
3483                                 continue;
3484
3485                         spin_lock(&ifp->lock);
3486                         /* We try to batch several events at once. */
3487                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3488
3489                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3490                             age >= ifp->valid_lft) {
3491                                 spin_unlock(&ifp->lock);
3492                                 in6_ifa_hold(ifp);
3493                                 ipv6_del_addr(ifp);
3494                                 goto restart;
3495                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3496                                 spin_unlock(&ifp->lock);
3497                                 continue;
3498                         } else if (age >= ifp->prefered_lft) {
3499                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3500                                 int deprecate = 0;
3501
3502                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3503                                         deprecate = 1;
3504                                         ifp->flags |= IFA_F_DEPRECATED;
3505                                 }
3506
3507                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3508                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3509
3510                                 spin_unlock(&ifp->lock);
3511
3512                                 if (deprecate) {
3513                                         in6_ifa_hold(ifp);
3514
3515                                         ipv6_ifa_notify(0, ifp);
3516                                         in6_ifa_put(ifp);
3517                                         goto restart;
3518                                 }
3519 #ifdef CONFIG_IPV6_PRIVACY
3520                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3521                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3522                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3523                                         ifp->idev->cnf.dad_transmits *
3524                                         ifp->idev->nd_parms->retrans_time / HZ;
3525
3526                                 if (age >= ifp->prefered_lft - regen_advance) {
3527                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3528                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3529                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3530                                         if (!ifp->regen_count && ifpub) {
3531                                                 ifp->regen_count++;
3532                                                 in6_ifa_hold(ifp);
3533                                                 in6_ifa_hold(ifpub);
3534                                                 spin_unlock(&ifp->lock);
3535
3536                                                 spin_lock(&ifpub->lock);
3537                                                 ifpub->regen_count = 0;
3538                                                 spin_unlock(&ifpub->lock);
3539                                                 ipv6_create_tempaddr(ifpub, ifp);
3540                                                 in6_ifa_put(ifpub);
3541                                                 in6_ifa_put(ifp);
3542                                                 goto restart;
3543                                         }
3544                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3545                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3546                                 spin_unlock(&ifp->lock);
3547 #endif
3548                         } else {
3549                                 /* ifp->prefered_lft <= ifp->valid_lft */
3550                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3551                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3552                                 spin_unlock(&ifp->lock);
3553                         }
3554                 }
3555         }
3556
3557         next_sec = round_jiffies_up(next);
3558         next_sched = next;
3559
3560         /* If rounded timeout is accurate enough, accept it. */
3561         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3562                 next_sched = next_sec;
3563
3564         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3565         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3566                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3567
3568         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3569               now, next, next_sec, next_sched));
3570
3571         addr_chk_timer.expires = next_sched;
3572         add_timer(&addr_chk_timer);
3573         spin_unlock(&addrconf_verify_lock);
3574         rcu_read_unlock_bh();
3575 }
3576
3577 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3578 {
3579         struct in6_addr *pfx = NULL;
3580
3581         if (addr)
3582                 pfx = nla_data(addr);
3583
3584         if (local) {
3585                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3586                         pfx = NULL;
3587                 else
3588                         pfx = nla_data(local);
3589         }
3590
3591         return pfx;
3592 }
3593
3594 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3595         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3596         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3597         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3598 };
3599
3600 static int
3601 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3602 {
3603         struct net *net = sock_net(skb->sk);
3604         struct ifaddrmsg *ifm;
3605         struct nlattr *tb[IFA_MAX+1];
3606         struct in6_addr *pfx;
3607         int err;
3608
3609         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3610         if (err < 0)
3611                 return err;
3612
3613         ifm = nlmsg_data(nlh);
3614         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3615         if (pfx == NULL)
3616                 return -EINVAL;
3617
3618         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3619 }
3620
3621 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3622                              u32 prefered_lft, u32 valid_lft)
3623 {
3624         u32 flags;
3625         clock_t expires;
3626         unsigned long timeout;
3627
3628         if (!valid_lft || (prefered_lft > valid_lft))
3629                 return -EINVAL;
3630
3631         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3632         if (addrconf_finite_timeout(timeout)) {
3633                 expires = jiffies_to_clock_t(timeout * HZ);
3634                 valid_lft = timeout;
3635                 flags = RTF_EXPIRES;
3636         } else {
3637                 expires = 0;
3638                 flags = 0;
3639                 ifa_flags |= IFA_F_PERMANENT;
3640         }
3641
3642         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3643         if (addrconf_finite_timeout(timeout)) {
3644                 if (timeout == 0)
3645                         ifa_flags |= IFA_F_DEPRECATED;
3646                 prefered_lft = timeout;
3647         }
3648
3649         spin_lock_bh(&ifp->lock);
3650         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3651         ifp->tstamp = jiffies;
3652         ifp->valid_lft = valid_lft;
3653         ifp->prefered_lft = prefered_lft;
3654
3655         spin_unlock_bh(&ifp->lock);
3656         if (!(ifp->flags&IFA_F_TENTATIVE))
3657                 ipv6_ifa_notify(0, ifp);
3658
3659         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3660                               expires, flags);
3661         addrconf_verify(0);
3662
3663         return 0;
3664 }
3665
3666 static int
3667 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3668 {
3669         struct net *net = sock_net(skb->sk);
3670         struct ifaddrmsg *ifm;
3671         struct nlattr *tb[IFA_MAX+1];
3672         struct in6_addr *pfx;
3673         struct inet6_ifaddr *ifa;
3674         struct net_device *dev;
3675         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3676         u8 ifa_flags;
3677         int err;
3678
3679         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3680         if (err < 0)
3681                 return err;
3682
3683         ifm = nlmsg_data(nlh);
3684         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3685         if (pfx == NULL)
3686                 return -EINVAL;
3687
3688         if (tb[IFA_CACHEINFO]) {
3689                 struct ifa_cacheinfo *ci;
3690
3691                 ci = nla_data(tb[IFA_CACHEINFO]);
3692                 valid_lft = ci->ifa_valid;
3693                 preferred_lft = ci->ifa_prefered;
3694         } else {
3695                 preferred_lft = INFINITY_LIFE_TIME;
3696                 valid_lft = INFINITY_LIFE_TIME;
3697         }
3698
3699         dev =  __dev_get_by_index(net, ifm->ifa_index);
3700         if (dev == NULL)
3701                 return -ENODEV;
3702
3703         /* We ignore other flags so far. */
3704         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3705
3706         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3707         if (ifa == NULL) {
3708                 /*
3709                  * It would be best to check for !NLM_F_CREATE here but
3710                  * userspace alreay relies on not having to provide this.
3711                  */
3712                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3713                                       ifm->ifa_prefixlen, ifa_flags,
3714                                       preferred_lft, valid_lft);
3715         }
3716
3717         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3718             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3719                 err = -EEXIST;
3720         else
3721                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3722
3723         in6_ifa_put(ifa);
3724
3725         return err;
3726 }
3727
3728 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3729                           u8 scope, int ifindex)
3730 {
3731         struct ifaddrmsg *ifm;
3732
3733         ifm = nlmsg_data(nlh);
3734         ifm->ifa_family = AF_INET6;
3735         ifm->ifa_prefixlen = prefixlen;
3736         ifm->ifa_flags = flags;
3737         ifm->ifa_scope = scope;
3738         ifm->ifa_index = ifindex;
3739 }
3740
3741 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3742                          unsigned long tstamp, u32 preferred, u32 valid)
3743 {
3744         struct ifa_cacheinfo ci;
3745
3746         ci.cstamp = cstamp_delta(cstamp);
3747         ci.tstamp = cstamp_delta(tstamp);
3748         ci.ifa_prefered = preferred;
3749         ci.ifa_valid = valid;
3750
3751         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3752 }
3753
3754 static inline int rt_scope(int ifa_scope)
3755 {
3756         if (ifa_scope & IFA_HOST)
3757                 return RT_SCOPE_HOST;
3758         else if (ifa_scope & IFA_LINK)
3759                 return RT_SCOPE_LINK;
3760         else if (ifa_scope & IFA_SITE)
3761                 return RT_SCOPE_SITE;
3762         else
3763                 return RT_SCOPE_UNIVERSE;
3764 }
3765
3766 static inline int inet6_ifaddr_msgsize(void)
3767 {
3768         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3769                + nla_total_size(16) /* IFA_ADDRESS */
3770                + nla_total_size(sizeof(struct ifa_cacheinfo));
3771 }
3772
3773 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3774                              u32 portid, u32 seq, int event, unsigned int flags)
3775 {
3776         struct nlmsghdr  *nlh;
3777         u32 preferred, valid;
3778
3779         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3780         if (nlh == NULL)
3781                 return -EMSGSIZE;
3782
3783         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3784                       ifa->idev->dev->ifindex);
3785
3786         if (!(ifa->flags&IFA_F_PERMANENT)) {
3787                 preferred = ifa->prefered_lft;
3788                 valid = ifa->valid_lft;
3789                 if (preferred != INFINITY_LIFE_TIME) {
3790                         long tval = (jiffies - ifa->tstamp)/HZ;
3791                         if (preferred > tval)
3792                                 preferred -= tval;
3793                         else
3794                                 preferred = 0;
3795                         if (valid != INFINITY_LIFE_TIME) {
3796                                 if (valid > tval)
3797                                         valid -= tval;
3798                                 else
3799                                         valid = 0;
3800                         }
3801                 }
3802         } else {
3803                 preferred = INFINITY_LIFE_TIME;
3804                 valid = INFINITY_LIFE_TIME;
3805         }
3806
3807         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3808             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3809                 nlmsg_cancel(skb, nlh);
3810                 return -EMSGSIZE;
3811         }
3812
3813         return nlmsg_end(skb, nlh);
3814 }
3815
3816 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3817                                 u32 portid, u32 seq, int event, u16 flags)
3818 {
3819         struct nlmsghdr  *nlh;
3820         u8 scope = RT_SCOPE_UNIVERSE;
3821         int ifindex = ifmca->idev->dev->ifindex;
3822
3823         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3824                 scope = RT_SCOPE_SITE;
3825
3826         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3827         if (nlh == NULL)
3828                 return -EMSGSIZE;
3829
3830         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3831         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3832             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3833                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3834                 nlmsg_cancel(skb, nlh);
3835                 return -EMSGSIZE;
3836         }
3837
3838         return nlmsg_end(skb, nlh);
3839 }
3840
3841 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3842                                 u32 portid, u32 seq, int event, unsigned int flags)
3843 {
3844         struct nlmsghdr  *nlh;
3845         u8 scope = RT_SCOPE_UNIVERSE;
3846         int ifindex = ifaca->aca_idev->dev->ifindex;
3847
3848         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3849                 scope = RT_SCOPE_SITE;
3850
3851         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3852         if (nlh == NULL)
3853                 return -EMSGSIZE;
3854
3855         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3856         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3857             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3858                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3859                 nlmsg_cancel(skb, nlh);
3860                 return -EMSGSIZE;
3861         }
3862
3863         return nlmsg_end(skb, nlh);
3864 }
3865
3866 enum addr_type_t {
3867         UNICAST_ADDR,
3868         MULTICAST_ADDR,
3869         ANYCAST_ADDR,
3870 };
3871
3872 /* called with rcu_read_lock() */
3873 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3874                           struct netlink_callback *cb, enum addr_type_t type,
3875                           int s_ip_idx, int *p_ip_idx)
3876 {
3877         struct ifmcaddr6 *ifmca;
3878         struct ifacaddr6 *ifaca;
3879         int err = 1;
3880         int ip_idx = *p_ip_idx;
3881
3882         read_lock_bh(&idev->lock);
3883         switch (type) {
3884         case UNICAST_ADDR: {
3885                 struct inet6_ifaddr *ifa;
3886
3887                 /* unicast address incl. temp addr */
3888                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3889                         if (++ip_idx < s_ip_idx)
3890                                 continue;
3891                         err = inet6_fill_ifaddr(skb, ifa,
3892                                                 NETLINK_CB(cb->skb).portid,
3893                                                 cb->nlh->nlmsg_seq,
3894                                                 RTM_NEWADDR,
3895                                                 NLM_F_MULTI);
3896                         if (err <= 0)
3897                                 break;
3898                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3899                 }
3900                 break;
3901         }
3902         case MULTICAST_ADDR:
3903                 /* multicast address */
3904                 for (ifmca = idev->mc_list; ifmca;
3905                      ifmca = ifmca->next, ip_idx++) {
3906                         if (ip_idx < s_ip_idx)
3907                                 continue;
3908                         err = inet6_fill_ifmcaddr(skb, ifmca,
3909                                                   NETLINK_CB(cb->skb).portid,
3910                                                   cb->nlh->nlmsg_seq,
3911                                                   RTM_GETMULTICAST,
3912                                                   NLM_F_MULTI);
3913                         if (err <= 0)
3914                                 break;
3915                 }
3916                 break;
3917         case ANYCAST_ADDR:
3918                 /* anycast address */
3919                 for (ifaca = idev->ac_list; ifaca;
3920                      ifaca = ifaca->aca_next, ip_idx++) {
3921                         if (ip_idx < s_ip_idx)
3922                                 continue;
3923                         err = inet6_fill_ifacaddr(skb, ifaca,
3924                                                   NETLINK_CB(cb->skb).portid,
3925                                                   cb->nlh->nlmsg_seq,
3926                                                   RTM_GETANYCAST,
3927                                                   NLM_F_MULTI);
3928                         if (err <= 0)
3929                                 break;
3930                 }
3931                 break;
3932         default:
3933                 break;
3934         }
3935         read_unlock_bh(&idev->lock);
3936         *p_ip_idx = ip_idx;
3937         return err;
3938 }
3939
3940 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3941                            enum addr_type_t type)
3942 {
3943         struct net *net = sock_net(skb->sk);
3944         int h, s_h;
3945         int idx, ip_idx;
3946         int s_idx, s_ip_idx;
3947         struct net_device *dev;
3948         struct inet6_dev *idev;
3949         struct hlist_head *head;
3950
3951         s_h = cb->args[0];
3952         s_idx = idx = cb->args[1];
3953         s_ip_idx = ip_idx = cb->args[2];
3954
3955         rcu_read_lock();
3956         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3957         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3958                 idx = 0;
3959                 head = &net->dev_index_head[h];
3960                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3961                         if (idx < s_idx)
3962                                 goto cont;
3963                         if (h > s_h || idx > s_idx)
3964                                 s_ip_idx = 0;
3965                         ip_idx = 0;
3966                         idev = __in6_dev_get(dev);
3967                         if (!idev)
3968                                 goto cont;
3969
3970                         if (in6_dump_addrs(idev, skb, cb, type,
3971                                            s_ip_idx, &ip_idx) <= 0)
3972                                 goto done;
3973 cont:
3974                         idx++;
3975                 }
3976         }
3977 done:
3978         rcu_read_unlock();
3979         cb->args[0] = h;
3980         cb->args[1] = idx;
3981         cb->args[2] = ip_idx;
3982
3983         return skb->len;
3984 }
3985
3986 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3987 {
3988         enum addr_type_t type = UNICAST_ADDR;
3989
3990         return inet6_dump_addr(skb, cb, type);
3991 }
3992
3993 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3994 {
3995         enum addr_type_t type = MULTICAST_ADDR;
3996
3997         return inet6_dump_addr(skb, cb, type);
3998 }
3999
4000
4001 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4002 {
4003         enum addr_type_t type = ANYCAST_ADDR;
4004
4005         return inet6_dump_addr(skb, cb, type);
4006 }
4007
4008 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4009 {
4010         struct net *net = sock_net(in_skb->sk);
4011         struct ifaddrmsg *ifm;
4012         struct nlattr *tb[IFA_MAX+1];
4013         struct in6_addr *addr = NULL;
4014         struct net_device *dev = NULL;
4015         struct inet6_ifaddr *ifa;
4016         struct sk_buff *skb;
4017         int err;
4018
4019         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4020         if (err < 0)
4021                 goto errout;
4022
4023         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4024         if (addr == NULL) {
4025                 err = -EINVAL;
4026                 goto errout;
4027         }
4028
4029         ifm = nlmsg_data(nlh);
4030         if (ifm->ifa_index)
4031                 dev = __dev_get_by_index(net, ifm->ifa_index);
4032
4033         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4034         if (!ifa) {
4035                 err = -EADDRNOTAVAIL;
4036                 goto errout;
4037         }
4038
4039         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4040         if (!skb) {
4041                 err = -ENOBUFS;
4042                 goto errout_ifa;
4043         }
4044
4045         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4046                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4047         if (err < 0) {
4048                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4049                 WARN_ON(err == -EMSGSIZE);
4050                 kfree_skb(skb);
4051                 goto errout_ifa;
4052         }
4053         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4054 errout_ifa:
4055         in6_ifa_put(ifa);
4056 errout:
4057         return err;
4058 }
4059
4060 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4061 {
4062         struct sk_buff *skb;
4063         struct net *net = dev_net(ifa->idev->dev);
4064         int err = -ENOBUFS;
4065
4066         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4067         if (skb == NULL)
4068                 goto errout;
4069
4070         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4071         if (err < 0) {
4072                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4073                 WARN_ON(err == -EMSGSIZE);
4074                 kfree_skb(skb);
4075                 goto errout;
4076         }
4077         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4078         return;
4079 errout:
4080         if (err < 0)
4081                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4082 }
4083
4084 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4085                                 __s32 *array, int bytes)
4086 {
4087         BUG_ON(bytes < (DEVCONF_MAX * 4));
4088
4089         memset(array, 0, bytes);
4090         array[DEVCONF_FORWARDING] = cnf->forwarding;
4091         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4092         array[DEVCONF_MTU6] = cnf->mtu6;
4093         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4094         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4095         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4096         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4097         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4098         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4099                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4100         array[DEVCONF_RTR_SOLICIT_DELAY] =
4101                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4102         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4103 #ifdef CONFIG_IPV6_PRIVACY
4104         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4105         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4106         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4107         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4108         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4109 #endif
4110         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4111         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4112         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4113 #ifdef CONFIG_IPV6_ROUTER_PREF
4114         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4115         array[DEVCONF_RTR_PROBE_INTERVAL] =
4116                 jiffies_to_msecs(cnf->rtr_probe_interval);
4117 #ifdef CONFIG_IPV6_ROUTE_INFO
4118         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4119 #endif
4120 #endif
4121         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4122         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4123 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4124         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4125 #endif
4126 #ifdef CONFIG_IPV6_MROUTE
4127         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4128 #endif
4129         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4130         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4131         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4132         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4133 }
4134
4135 static inline size_t inet6_ifla6_size(void)
4136 {
4137         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4138              + nla_total_size(sizeof(struct ifla_cacheinfo))
4139              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4140              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4141              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4142 }
4143
4144 static inline size_t inet6_if_nlmsg_size(void)
4145 {
4146         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4147                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4148                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4149                + nla_total_size(4) /* IFLA_MTU */
4150                + nla_total_size(4) /* IFLA_LINK */
4151                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4152 }
4153
4154 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4155                                       int items, int bytes)
4156 {
4157         int i;
4158         int pad = bytes - sizeof(u64) * items;
4159         BUG_ON(pad < 0);
4160
4161         /* Use put_unaligned() because stats may not be aligned for u64. */
4162         put_unaligned(items, &stats[0]);
4163         for (i = 1; i < items; i++)
4164                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4165
4166         memset(&stats[items], 0, pad);
4167 }
4168
4169 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4170                                       int items, int bytes, size_t syncpoff)
4171 {
4172         int i;
4173         int pad = bytes - sizeof(u64) * items;
4174         BUG_ON(pad < 0);
4175
4176         /* Use put_unaligned() because stats may not be aligned for u64. */
4177         put_unaligned(items, &stats[0]);
4178         for (i = 1; i < items; i++)
4179                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4180
4181         memset(&stats[items], 0, pad);
4182 }
4183
4184 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4185                              int bytes)
4186 {
4187         switch (attrtype) {
4188         case IFLA_INET6_STATS:
4189                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4190                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4191                 break;
4192         case IFLA_INET6_ICMP6STATS:
4193                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4194                 break;
4195         }
4196 }
4197
4198 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4199 {
4200         struct nlattr *nla;
4201         struct ifla_cacheinfo ci;
4202
4203         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4204                 goto nla_put_failure;
4205         ci.max_reasm_len = IPV6_MAXPLEN;
4206         ci.tstamp = cstamp_delta(idev->tstamp);
4207         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4208         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4209         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4210                 goto nla_put_failure;
4211         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4212         if (nla == NULL)
4213                 goto nla_put_failure;
4214         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4215
4216         /* XXX - MC not implemented */
4217
4218         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4219         if (nla == NULL)
4220                 goto nla_put_failure;
4221         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4222
4223         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4224         if (nla == NULL)
4225                 goto nla_put_failure;
4226         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4227
4228         return 0;
4229
4230 nla_put_failure:
4231         return -EMSGSIZE;
4232 }
4233
4234 static size_t inet6_get_link_af_size(const struct net_device *dev)
4235 {
4236         if (!__in6_dev_get(dev))
4237                 return 0;
4238
4239         return inet6_ifla6_size();
4240 }
4241
4242 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4243 {
4244         struct inet6_dev *idev = __in6_dev_get(dev);
4245
4246         if (!idev)
4247                 return -ENODATA;
4248
4249         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4250                 return -EMSGSIZE;
4251
4252         return 0;
4253 }
4254
4255 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4256                              u32 portid, u32 seq, int event, unsigned int flags)
4257 {
4258         struct net_device *dev = idev->dev;
4259         struct ifinfomsg *hdr;
4260         struct nlmsghdr *nlh;
4261         void *protoinfo;
4262
4263         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4264         if (nlh == NULL)
4265                 return -EMSGSIZE;
4266
4267         hdr = nlmsg_data(nlh);
4268         hdr->ifi_family = AF_INET6;
4269         hdr->__ifi_pad = 0;
4270         hdr->ifi_type = dev->type;
4271         hdr->ifi_index = dev->ifindex;
4272         hdr->ifi_flags = dev_get_flags(dev);
4273         hdr->ifi_change = 0;
4274
4275         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4276             (dev->addr_len &&
4277              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4278             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4279             (dev->ifindex != dev->iflink &&
4280              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4281                 goto nla_put_failure;
4282         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4283         if (protoinfo == NULL)
4284                 goto nla_put_failure;
4285
4286         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4287                 goto nla_put_failure;
4288
4289         nla_nest_end(skb, protoinfo);
4290         return nlmsg_end(skb, nlh);
4291
4292 nla_put_failure:
4293         nlmsg_cancel(skb, nlh);
4294         return -EMSGSIZE;
4295 }
4296
4297 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4298 {
4299         struct net *net = sock_net(skb->sk);
4300         int h, s_h;
4301         int idx = 0, s_idx;
4302         struct net_device *dev;
4303         struct inet6_dev *idev;
4304         struct hlist_head *head;
4305
4306         s_h = cb->args[0];
4307         s_idx = cb->args[1];
4308
4309         rcu_read_lock();
4310         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4311                 idx = 0;
4312                 head = &net->dev_index_head[h];
4313                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4314                         if (idx < s_idx)
4315                                 goto cont;
4316                         idev = __in6_dev_get(dev);
4317                         if (!idev)
4318                                 goto cont;
4319                         if (inet6_fill_ifinfo(skb, idev,
4320                                               NETLINK_CB(cb->skb).portid,
4321                                               cb->nlh->nlmsg_seq,
4322                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4323                                 goto out;
4324 cont:
4325                         idx++;
4326                 }
4327         }
4328 out:
4329         rcu_read_unlock();
4330         cb->args[1] = idx;
4331         cb->args[0] = h;
4332
4333         return skb->len;
4334 }
4335
4336 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4337 {
4338         struct sk_buff *skb;
4339         struct net *net = dev_net(idev->dev);
4340         int err = -ENOBUFS;
4341
4342         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4343         if (skb == NULL)
4344                 goto errout;
4345
4346         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4347         if (err < 0) {
4348                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4349                 WARN_ON(err == -EMSGSIZE);
4350                 kfree_skb(skb);
4351                 goto errout;
4352         }
4353         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4354         return;
4355 errout:
4356         if (err < 0)
4357                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4358 }
4359
4360 static inline size_t inet6_prefix_nlmsg_size(void)
4361 {
4362         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4363                + nla_total_size(sizeof(struct in6_addr))
4364                + nla_total_size(sizeof(struct prefix_cacheinfo));
4365 }
4366
4367 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4368                              struct prefix_info *pinfo, u32 portid, u32 seq,
4369                              int event, unsigned int flags)
4370 {
4371         struct prefixmsg *pmsg;
4372         struct nlmsghdr *nlh;
4373         struct prefix_cacheinfo ci;
4374
4375         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4376         if (nlh == NULL)
4377                 return -EMSGSIZE;
4378
4379         pmsg = nlmsg_data(nlh);
4380         pmsg->prefix_family = AF_INET6;
4381         pmsg->prefix_pad1 = 0;
4382         pmsg->prefix_pad2 = 0;
4383         pmsg->prefix_ifindex = idev->dev->ifindex;
4384         pmsg->prefix_len = pinfo->prefix_len;
4385         pmsg->prefix_type = pinfo->type;
4386         pmsg->prefix_pad3 = 0;
4387         pmsg->prefix_flags = 0;
4388         if (pinfo->onlink)
4389                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4390         if (pinfo->autoconf)
4391                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4392
4393         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4394                 goto nla_put_failure;
4395         ci.preferred_time = ntohl(pinfo->prefered);
4396         ci.valid_time = ntohl(pinfo->valid);
4397         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4398                 goto nla_put_failure;
4399         return nlmsg_end(skb, nlh);
4400
4401 nla_put_failure:
4402         nlmsg_cancel(skb, nlh);
4403         return -EMSGSIZE;
4404 }
4405
4406 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4407                          struct prefix_info *pinfo)
4408 {
4409         struct sk_buff *skb;
4410         struct net *net = dev_net(idev->dev);
4411         int err = -ENOBUFS;
4412
4413         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4414         if (skb == NULL)
4415                 goto errout;
4416
4417         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4418         if (err < 0) {
4419                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4420                 WARN_ON(err == -EMSGSIZE);
4421                 kfree_skb(skb);
4422                 goto errout;
4423         }
4424         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4425         return;
4426 errout:
4427         if (err < 0)
4428                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4429 }
4430
4431 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4432 {
4433         struct net *net = dev_net(ifp->idev->dev);
4434
4435         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4436
4437         switch (event) {
4438         case RTM_NEWADDR:
4439                 /*
4440                  * If the address was optimistic
4441                  * we inserted the route at the start of
4442                  * our DAD process, so we don't need
4443                  * to do it again
4444                  */
4445                 if (!(ifp->rt->rt6i_node))
4446                         ip6_ins_rt(ifp->rt);
4447                 if (ifp->idev->cnf.forwarding)
4448                         addrconf_join_anycast(ifp);
4449                 break;
4450         case RTM_DELADDR:
4451                 if (ifp->idev->cnf.forwarding)
4452                         addrconf_leave_anycast(ifp);
4453                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4454                 dst_hold(&ifp->rt->dst);
4455
4456                 if (ip6_del_rt(ifp->rt))
4457                         dst_free(&ifp->rt->dst);
4458                 break;
4459         }
4460         atomic_inc(&net->ipv6.dev_addr_genid);
4461 }
4462
4463 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4464 {
4465         rcu_read_lock_bh();
4466         if (likely(ifp->idev->dead == 0))
4467                 __ipv6_ifa_notify(event, ifp);
4468         rcu_read_unlock_bh();
4469 }
4470
4471 #ifdef CONFIG_SYSCTL
4472
4473 static
4474 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4475                            void __user *buffer, size_t *lenp, loff_t *ppos)
4476 {
4477         int *valp = ctl->data;
4478         int val = *valp;
4479         loff_t pos = *ppos;
4480         ctl_table lctl;
4481         int ret;
4482
4483         /*
4484          * ctl->data points to idev->cnf.forwarding, we should
4485          * not modify it until we get the rtnl lock.
4486          */
4487         lctl = *ctl;
4488         lctl.data = &val;
4489
4490         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4491
4492         if (write)
4493                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4494         if (ret)
4495                 *ppos = pos;
4496         return ret;
4497 }
4498
4499 static void dev_disable_change(struct inet6_dev *idev)
4500 {
4501         if (!idev || !idev->dev)
4502                 return;
4503
4504         if (idev->cnf.disable_ipv6)
4505                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4506         else
4507                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4508 }
4509
4510 static void addrconf_disable_change(struct net *net, __s32 newf)
4511 {
4512         struct net_device *dev;
4513         struct inet6_dev *idev;
4514
4515         rcu_read_lock();
4516         for_each_netdev_rcu(net, dev) {
4517                 idev = __in6_dev_get(dev);
4518                 if (idev) {
4519                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4520                         idev->cnf.disable_ipv6 = newf;
4521                         if (changed)
4522                                 dev_disable_change(idev);
4523                 }
4524         }
4525         rcu_read_unlock();
4526 }
4527
4528 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4529 {
4530         struct net *net;
4531         int old;
4532
4533         if (!rtnl_trylock())
4534                 return restart_syscall();
4535
4536         net = (struct net *)table->extra2;
4537         old = *p;
4538         *p = newf;
4539
4540         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4541                 rtnl_unlock();
4542                 return 0;
4543         }
4544
4545         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4546                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4547                 addrconf_disable_change(net, newf);
4548         } else if ((!newf) ^ (!old))
4549                 dev_disable_change((struct inet6_dev *)table->extra1);
4550
4551         rtnl_unlock();
4552         return 0;
4553 }
4554
4555 static
4556 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4557                             void __user *buffer, size_t *lenp, loff_t *ppos)
4558 {
4559         int *valp = ctl->data;
4560         int val = *valp;
4561         loff_t pos = *ppos;
4562         ctl_table lctl;
4563         int ret;
4564
4565         /*
4566          * ctl->data points to idev->cnf.disable_ipv6, we should
4567          * not modify it until we get the rtnl lock.
4568          */
4569         lctl = *ctl;
4570         lctl.data = &val;
4571
4572         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4573
4574         if (write)
4575                 ret = addrconf_disable_ipv6(ctl, valp, val);
4576         if (ret)
4577                 *ppos = pos;
4578         return ret;
4579 }
4580
4581 static struct addrconf_sysctl_table
4582 {
4583         struct ctl_table_header *sysctl_header;
4584         ctl_table addrconf_vars[DEVCONF_MAX+1];
4585 } addrconf_sysctl __read_mostly = {
4586         .sysctl_header = NULL,
4587         .addrconf_vars = {
4588                 {
4589                         .procname       = "forwarding",
4590                         .data           = &ipv6_devconf.forwarding,
4591                         .maxlen         = sizeof(int),
4592                         .mode           = 0644,
4593                         .proc_handler   = addrconf_sysctl_forward,
4594                 },
4595                 {
4596                         .procname       = "hop_limit",
4597                         .data           = &ipv6_devconf.hop_limit,
4598                         .maxlen         = sizeof(int),
4599                         .mode           = 0644,
4600                         .proc_handler   = proc_dointvec,
4601                 },
4602                 {
4603                         .procname       = "mtu",
4604                         .data           = &ipv6_devconf.mtu6,
4605                         .maxlen         = sizeof(int),
4606                         .mode           = 0644,
4607                         .proc_handler   = proc_dointvec,
4608                 },
4609                 {
4610                         .procname       = "accept_ra",
4611                         .data           = &ipv6_devconf.accept_ra,
4612                         .maxlen         = sizeof(int),
4613                         .mode           = 0644,
4614                         .proc_handler   = proc_dointvec,
4615                 },
4616                 {
4617                         .procname       = "accept_redirects",
4618                         .data           = &ipv6_devconf.accept_redirects,
4619                         .maxlen         = sizeof(int),
4620                         .mode           = 0644,
4621                         .proc_handler   = proc_dointvec,
4622                 },
4623                 {
4624                         .procname       = "autoconf",
4625                         .data           = &ipv6_devconf.autoconf,
4626                         .maxlen         = sizeof(int),
4627                         .mode           = 0644,
4628                         .proc_handler   = proc_dointvec,
4629                 },
4630                 {
4631                         .procname       = "dad_transmits",
4632                         .data           = &ipv6_devconf.dad_transmits,
4633                         .maxlen         = sizeof(int),
4634                         .mode           = 0644,
4635                         .proc_handler   = proc_dointvec,
4636                 },
4637                 {
4638                         .procname       = "router_solicitations",
4639                         .data           = &ipv6_devconf.rtr_solicits,
4640                         .maxlen         = sizeof(int),
4641                         .mode           = 0644,
4642                         .proc_handler   = proc_dointvec,
4643                 },
4644                 {
4645                         .procname       = "router_solicitation_interval",
4646                         .data           = &ipv6_devconf.rtr_solicit_interval,
4647                         .maxlen         = sizeof(int),
4648                         .mode           = 0644,
4649                         .proc_handler   = proc_dointvec_jiffies,
4650                 },
4651                 {
4652                         .procname       = "router_solicitation_delay",
4653                         .data           = &ipv6_devconf.rtr_solicit_delay,
4654                         .maxlen         = sizeof(int),
4655                         .mode           = 0644,
4656                         .proc_handler   = proc_dointvec_jiffies,
4657                 },
4658                 {
4659                         .procname       = "force_mld_version",
4660                         .data           = &ipv6_devconf.force_mld_version,
4661                         .maxlen         = sizeof(int),
4662                         .mode           = 0644,
4663                         .proc_handler   = proc_dointvec,
4664                 },
4665 #ifdef CONFIG_IPV6_PRIVACY
4666                 {
4667                         .procname       = "use_tempaddr",
4668                         .data           = &ipv6_devconf.use_tempaddr,
4669                         .maxlen         = sizeof(int),
4670                         .mode           = 0644,
4671                         .proc_handler   = proc_dointvec,
4672                 },
4673                 {
4674                         .procname       = "temp_valid_lft",
4675                         .data           = &ipv6_devconf.temp_valid_lft,
4676                         .maxlen         = sizeof(int),
4677                         .mode           = 0644,
4678                         .proc_handler   = proc_dointvec,
4679                 },
4680                 {
4681                         .procname       = "temp_prefered_lft",
4682                         .data           = &ipv6_devconf.temp_prefered_lft,
4683                         .maxlen         = sizeof(int),
4684                         .mode           = 0644,
4685                         .proc_handler   = proc_dointvec,
4686                 },
4687                 {
4688                         .procname       = "regen_max_retry",
4689                         .data           = &ipv6_devconf.regen_max_retry,
4690                         .maxlen         = sizeof(int),
4691                         .mode           = 0644,
4692                         .proc_handler   = proc_dointvec,
4693                 },
4694                 {
4695                         .procname       = "max_desync_factor",
4696                         .data           = &ipv6_devconf.max_desync_factor,
4697                         .maxlen         = sizeof(int),
4698                         .mode           = 0644,
4699                         .proc_handler   = proc_dointvec,
4700                 },
4701 #endif
4702                 {
4703                         .procname       = "max_addresses",
4704                         .data           = &ipv6_devconf.max_addresses,
4705                         .maxlen         = sizeof(int),
4706                         .mode           = 0644,
4707                         .proc_handler   = proc_dointvec,
4708                 },
4709                 {
4710                         .procname       = "accept_ra_defrtr",
4711                         .data           = &ipv6_devconf.accept_ra_defrtr,
4712                         .maxlen         = sizeof(int),
4713                         .mode           = 0644,
4714                         .proc_handler   = proc_dointvec,
4715                 },
4716                 {
4717                         .procname       = "accept_ra_pinfo",
4718                         .data           = &ipv6_devconf.accept_ra_pinfo,
4719                         .maxlen         = sizeof(int),
4720                         .mode           = 0644,
4721                         .proc_handler   = proc_dointvec,
4722                 },
4723 #ifdef CONFIG_IPV6_ROUTER_PREF
4724                 {
4725                         .procname       = "accept_ra_rtr_pref",
4726                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4727                         .maxlen         = sizeof(int),
4728                         .mode           = 0644,
4729                         .proc_handler   = proc_dointvec,
4730                 },
4731                 {
4732                         .procname       = "router_probe_interval",
4733                         .data           = &ipv6_devconf.rtr_probe_interval,
4734                         .maxlen         = sizeof(int),
4735                         .mode           = 0644,
4736                         .proc_handler   = proc_dointvec_jiffies,
4737                 },
4738 #ifdef CONFIG_IPV6_ROUTE_INFO
4739                 {
4740                         .procname       = "accept_ra_rt_info_max_plen",
4741                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4742                         .maxlen         = sizeof(int),
4743                         .mode           = 0644,
4744                         .proc_handler   = proc_dointvec,
4745                 },
4746 #endif
4747 #endif
4748                 {
4749                         .procname       = "proxy_ndp",
4750                         .data           = &ipv6_devconf.proxy_ndp,
4751                         .maxlen         = sizeof(int),
4752                         .mode           = 0644,
4753                         .proc_handler   = proc_dointvec,
4754                 },
4755                 {
4756                         .procname       = "accept_source_route",
4757                         .data           = &ipv6_devconf.accept_source_route,
4758                         .maxlen         = sizeof(int),
4759                         .mode           = 0644,
4760                         .proc_handler   = proc_dointvec,
4761                 },
4762 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4763                 {
4764                         .procname       = "optimistic_dad",
4765                         .data           = &ipv6_devconf.optimistic_dad,
4766                         .maxlen         = sizeof(int),
4767                         .mode           = 0644,
4768                         .proc_handler   = proc_dointvec,
4769
4770                 },
4771 #endif
4772 #ifdef CONFIG_IPV6_MROUTE
4773                 {
4774                         .procname       = "mc_forwarding",
4775                         .data           = &ipv6_devconf.mc_forwarding,
4776                         .maxlen         = sizeof(int),
4777                         .mode           = 0444,
4778                         .proc_handler   = proc_dointvec,
4779                 },
4780 #endif
4781                 {
4782                         .procname       = "disable_ipv6",
4783                         .data           = &ipv6_devconf.disable_ipv6,
4784                         .maxlen         = sizeof(int),
4785                         .mode           = 0644,
4786                         .proc_handler   = addrconf_sysctl_disable,
4787                 },
4788                 {
4789                         .procname       = "accept_dad",
4790                         .data           = &ipv6_devconf.accept_dad,
4791                         .maxlen         = sizeof(int),
4792                         .mode           = 0644,
4793                         .proc_handler   = proc_dointvec,
4794                 },
4795                 {
4796                         .procname       = "force_tllao",
4797                         .data           = &ipv6_devconf.force_tllao,
4798                         .maxlen         = sizeof(int),
4799                         .mode           = 0644,
4800                         .proc_handler   = proc_dointvec
4801                 },
4802                 {
4803                         .procname       = "ndisc_notify",
4804                         .data           = &ipv6_devconf.ndisc_notify,
4805                         .maxlen         = sizeof(int),
4806                         .mode           = 0644,
4807                         .proc_handler   = proc_dointvec
4808                 },
4809                 {
4810                         /* sentinel */
4811                 }
4812         },
4813 };
4814
4815 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4816                 struct inet6_dev *idev, struct ipv6_devconf *p)
4817 {
4818         int i;
4819         struct addrconf_sysctl_table *t;
4820         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4821
4822         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4823         if (t == NULL)
4824                 goto out;
4825
4826         for (i = 0; t->addrconf_vars[i].data; i++) {
4827                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4828                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4829                 t->addrconf_vars[i].extra2 = net;
4830         }
4831
4832         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4833
4834         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4835         if (t->sysctl_header == NULL)
4836                 goto free;
4837
4838         p->sysctl = t;
4839         return 0;
4840
4841 free:
4842         kfree(t);
4843 out:
4844         return -ENOBUFS;
4845 }
4846
4847 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4848 {
4849         struct addrconf_sysctl_table *t;
4850
4851         if (p->sysctl == NULL)
4852                 return;
4853
4854         t = p->sysctl;
4855         p->sysctl = NULL;
4856         unregister_net_sysctl_table(t->sysctl_header);
4857         kfree(t);
4858 }
4859
4860 static void addrconf_sysctl_register(struct inet6_dev *idev)
4861 {
4862         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4863                               &ndisc_ifinfo_sysctl_change);
4864         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4865                                         idev, &idev->cnf);
4866 }
4867
4868 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4869 {
4870         __addrconf_sysctl_unregister(&idev->cnf);
4871         neigh_sysctl_unregister(idev->nd_parms);
4872 }
4873
4874
4875 #endif
4876
4877 static int __net_init addrconf_init_net(struct net *net)
4878 {
4879         int err = -ENOMEM;
4880         struct ipv6_devconf *all, *dflt;
4881
4882         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
4883         if (all == NULL)
4884                 goto err_alloc_all;
4885
4886         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4887         if (dflt == NULL)
4888                 goto err_alloc_dflt;
4889
4890         /* these will be inherited by all namespaces */
4891         dflt->autoconf = ipv6_defaults.autoconf;
4892         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4893
4894         net->ipv6.devconf_all = all;
4895         net->ipv6.devconf_dflt = dflt;
4896
4897 #ifdef CONFIG_SYSCTL
4898         err = __addrconf_sysctl_register(net, "all", NULL, all);
4899         if (err < 0)
4900                 goto err_reg_all;
4901
4902         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4903         if (err < 0)
4904                 goto err_reg_dflt;
4905 #endif
4906         return 0;
4907
4908 #ifdef CONFIG_SYSCTL
4909 err_reg_dflt:
4910         __addrconf_sysctl_unregister(all);
4911 err_reg_all:
4912         kfree(dflt);
4913 #endif
4914 err_alloc_dflt:
4915         kfree(all);
4916 err_alloc_all:
4917         return err;
4918 }
4919
4920 static void __net_exit addrconf_exit_net(struct net *net)
4921 {
4922 #ifdef CONFIG_SYSCTL
4923         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4924         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4925 #endif
4926         if (!net_eq(net, &init_net)) {
4927                 kfree(net->ipv6.devconf_dflt);
4928                 kfree(net->ipv6.devconf_all);
4929         }
4930 }
4931
4932 static struct pernet_operations addrconf_ops = {
4933         .init = addrconf_init_net,
4934         .exit = addrconf_exit_net,
4935 };
4936
4937 /*
4938  *      Device notifier
4939  */
4940
4941 int register_inet6addr_notifier(struct notifier_block *nb)
4942 {
4943         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4944 }
4945 EXPORT_SYMBOL(register_inet6addr_notifier);
4946
4947 int unregister_inet6addr_notifier(struct notifier_block *nb)
4948 {
4949         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4950 }
4951 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4952
4953 static struct rtnl_af_ops inet6_ops = {
4954         .family           = AF_INET6,
4955         .fill_link_af     = inet6_fill_link_af,
4956         .get_link_af_size = inet6_get_link_af_size,
4957 };
4958
4959 /*
4960  *      Init / cleanup code
4961  */
4962
4963 int __init addrconf_init(void)
4964 {
4965         int i, err;
4966
4967         err = ipv6_addr_label_init();
4968         if (err < 0) {
4969                 pr_crit("%s: cannot initialize default policy table: %d\n",
4970                         __func__, err);
4971                 goto out;
4972         }
4973
4974         err = register_pernet_subsys(&addrconf_ops);
4975         if (err < 0)
4976                 goto out_addrlabel;
4977
4978         /* The addrconf netdev notifier requires that loopback_dev
4979          * has it's ipv6 private information allocated and setup
4980          * before it can bring up and give link-local addresses
4981          * to other devices which are up.
4982          *
4983          * Unfortunately, loopback_dev is not necessarily the first
4984          * entry in the global dev_base list of net devices.  In fact,
4985          * it is likely to be the very last entry on that list.
4986          * So this causes the notifier registry below to try and
4987          * give link-local addresses to all devices besides loopback_dev
4988          * first, then loopback_dev, which cases all the non-loopback_dev
4989          * devices to fail to get a link-local address.
4990          *
4991          * So, as a temporary fix, allocate the ipv6 structure for
4992          * loopback_dev first by hand.
4993          * Longer term, all of the dependencies ipv6 has upon the loopback
4994          * device and it being up should be removed.
4995          */
4996         rtnl_lock();
4997         if (!ipv6_add_dev(init_net.loopback_dev))
4998                 err = -ENOMEM;
4999         rtnl_unlock();
5000         if (err)
5001                 goto errlo;
5002
5003         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5004                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5005
5006         register_netdevice_notifier(&ipv6_dev_notf);
5007
5008         addrconf_verify(0);
5009
5010         err = rtnl_af_register(&inet6_ops);
5011         if (err < 0)
5012                 goto errout_af;
5013
5014         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5015                               NULL);
5016         if (err < 0)
5017                 goto errout;
5018
5019         /* Only the first call to __rtnl_register can fail */
5020         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5021         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5022         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5023                         inet6_dump_ifaddr, NULL);
5024         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5025                         inet6_dump_ifmcaddr, NULL);
5026         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5027                         inet6_dump_ifacaddr, NULL);
5028         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5029                         inet6_netconf_dump_devconf, NULL);
5030
5031         ipv6_addr_label_rtnl_register();
5032
5033         return 0;
5034 errout:
5035         rtnl_af_unregister(&inet6_ops);
5036 errout_af:
5037         unregister_netdevice_notifier(&ipv6_dev_notf);
5038 errlo:
5039         unregister_pernet_subsys(&addrconf_ops);
5040 out_addrlabel:
5041         ipv6_addr_label_cleanup();
5042 out:
5043         return err;
5044 }
5045
5046 void addrconf_cleanup(void)
5047 {
5048         struct net_device *dev;
5049         int i;
5050
5051         unregister_netdevice_notifier(&ipv6_dev_notf);
5052         unregister_pernet_subsys(&addrconf_ops);
5053         ipv6_addr_label_cleanup();
5054
5055         rtnl_lock();
5056
5057         __rtnl_af_unregister(&inet6_ops);
5058
5059         /* clean dev list */
5060         for_each_netdev(&init_net, dev) {
5061                 if (__in6_dev_get(dev) == NULL)
5062                         continue;
5063                 addrconf_ifdown(dev, 1);
5064         }
5065         addrconf_ifdown(init_net.loopback_dev, 2);
5066
5067         /*
5068          *      Check hash table.
5069          */
5070         spin_lock_bh(&addrconf_hash_lock);
5071         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5072                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5073         spin_unlock_bh(&addrconf_hash_lock);
5074
5075         del_timer(&addr_chk_timer);
5076         rtnl_unlock();
5077 }