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/ipv6.h>
  74 #include <net/protocol.h>
  75 #include <net/ndisc.h>
  76 #include <net/ip6_route.h>
  77 #include <net/addrconf.h>
  78 #include <net/tcp.h>
  79 #include <net/ip.h>
  80 #include <net/netlink.h>
  81 #include <net/pkt_sched.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85
  86 #ifdef CONFIG_IPV6_PRIVACY
  87 #include <linux/random.h>
  88 #endif
  89
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 /* Set to 3 to get tracing... */
  98 #define ACONF_DEBUG 2
  99
 100 #if ACONF_DEBUG >= 3
 101 #define ADBG(x) printk x
 102 #else
 103 #define ADBG(x)
 104 #endif
 105
 106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 107
 108 static inline u32 cstamp_delta(unsigned long cstamp)
 109 {
 110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 111 }
 112
 113 #ifdef CONFIG_SYSCTL
 114 static void addrconf_sysctl_register(struct inet6_dev *idev);
 115 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 116 #else
 117 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 118 {
 119 }
 120
 121 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 122 {
 123 }
 124 #endif
 125
 126 #ifdef CONFIG_IPV6_PRIVACY
 127 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 128 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 129 static void ipv6_regen_rndid(unsigned long data);
 130 #endif
 131
 132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 133 static int ipv6_count_addresses(struct inet6_dev *idev);
 134
 135 /*
 136  *      Configured unicast address hash table
 137  */
 138 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 139 static DEFINE_SPINLOCK(addrconf_hash_lock);
 140
 141 static void addrconf_verify(unsigned long);
 142
 143 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 144 static DEFINE_SPINLOCK(addrconf_verify_lock);
 145
 146 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 147 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 148
 149 static void addrconf_type_change(struct net_device *dev,
 150                                  unsigned long event);
 151 static int addrconf_ifdown(struct net_device *dev, int how);
 152
 153 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 154                                                   int plen,
 155                                                   const struct net_device *dev,
 156                                                   u32 flags, u32 noflags);
 157
 158 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 159 static void addrconf_dad_timer(unsigned long data);
 160 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 161 static void addrconf_dad_run(struct inet6_dev *idev);
 162 static void addrconf_rs_timer(unsigned long data);
 163 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 164 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165
 166 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 167                                 struct prefix_info *pinfo);
 168 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 169                                struct net_device *dev);
 170
 171 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 172
 173 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 174         .forwarding             = 0,
 175         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 176         .mtu6                   = IPV6_MIN_MTU,
 177         .accept_ra              = 1,
 178         .accept_redirects       = 1,
 179         .autoconf               = 1,
 180         .force_mld_version      = 0,
 181         .dad_transmits          = 1,
 182         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 183         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 184         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 185 #ifdef CONFIG_IPV6_PRIVACY
 186         .use_tempaddr           = 0,
 187         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 188         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 189         .regen_max_retry        = REGEN_MAX_RETRY,
 190         .max_desync_factor      = MAX_DESYNC_FACTOR,
 191 #endif
 192         .max_addresses          = IPV6_MAX_ADDRESSES,
 193         .accept_ra_defrtr       = 1,
 194         .accept_ra_pinfo        = 1,
 195 #ifdef CONFIG_IPV6_ROUTER_PREF
 196         .accept_ra_rtr_pref     = 1,
 197         .rtr_probe_interval     = 60 * HZ,
 198 #ifdef CONFIG_IPV6_ROUTE_INFO
 199         .accept_ra_rt_info_max_plen = 0,
 200 #endif
 201 #endif
 202         .proxy_ndp              = 0,
 203         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 204         .disable_ipv6           = 0,
 205         .accept_dad             = 1,
 206 };
 207
 208 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 209         .forwarding             = 0,
 210         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 211         .mtu6                   = IPV6_MIN_MTU,
 212         .accept_ra              = 1,
 213         .accept_redirects       = 1,
 214         .autoconf               = 1,
 215         .dad_transmits          = 1,
 216         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 217         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 218         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 219 #ifdef CONFIG_IPV6_PRIVACY
 220         .use_tempaddr           = 0,
 221         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 222         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 223         .regen_max_retry        = REGEN_MAX_RETRY,
 224         .max_desync_factor      = MAX_DESYNC_FACTOR,
 225 #endif
 226         .max_addresses          = IPV6_MAX_ADDRESSES,
 227         .accept_ra_defrtr       = 1,
 228         .accept_ra_pinfo        = 1,
 229 #ifdef CONFIG_IPV6_ROUTER_PREF
 230         .accept_ra_rtr_pref     = 1,
 231         .rtr_probe_interval     = 60 * HZ,
 232 #ifdef CONFIG_IPV6_ROUTE_INFO
 233         .accept_ra_rt_info_max_plen = 0,
 234 #endif
 235 #endif
 236         .proxy_ndp              = 0,
 237         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 238         .disable_ipv6           = 0,
 239         .accept_dad             = 1,
 240 };
 241
 242 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 243 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 244 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 245 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 246 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 247 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 248 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 249 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 250
 251 /* Check if a valid qdisc is available */
 252 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 253 {
 254         return !qdisc_tx_is_noop(dev);
 255 }
 256
 257 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 258 {
 259         if (del_timer(&ifp->timer))
 260                 __in6_ifa_put(ifp);
 261 }
 262
 263 enum addrconf_timer_t {
 264         AC_NONE,
 265         AC_DAD,
 266         AC_RS,
 267 };
 268
 269 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 270                                enum addrconf_timer_t what,
 271                                unsigned long when)
 272 {
 273         if (!del_timer(&ifp->timer))
 274                 in6_ifa_hold(ifp);
 275
 276         switch (what) {
 277         case AC_DAD:
 278                 ifp->timer.function = addrconf_dad_timer;
 279                 break;
 280         case AC_RS:
 281                 ifp->timer.function = addrconf_rs_timer;
 282                 break;
 283         default:
 284                 break;
 285         }
 286         ifp->timer.expires = jiffies + when;
 287         add_timer(&ifp->timer);
 288 }
 289
 290 static int snmp6_alloc_dev(struct inet6_dev *idev)
 291 {
 292         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 293                           sizeof(struct ipstats_mib),
 294                           __alignof__(struct ipstats_mib)) < 0)
 295                 goto err_ip;
 296         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 297                                         GFP_KERNEL);
 298         if (!idev->stats.icmpv6dev)
 299                 goto err_icmp;
 300         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 301                                            GFP_KERNEL);
 302         if (!idev->stats.icmpv6msgdev)
 303                 goto err_icmpmsg;
 304
 305         return 0;
 306
 307 err_icmpmsg:
 308         kfree(idev->stats.icmpv6dev);
 309 err_icmp:
 310         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 311 err_ip:
 312         return -ENOMEM;
 313 }
 314
 315 static void snmp6_free_dev(struct inet6_dev *idev)
 316 {
 317         kfree(idev->stats.icmpv6msgdev);
 318         kfree(idev->stats.icmpv6dev);
 319         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 320 }
 321
 322 /* Nobody refers to this device, we may destroy it. */
 323
 324 void in6_dev_finish_destroy(struct inet6_dev *idev)
 325 {
 326         struct net_device *dev = idev->dev;
 327
 328         WARN_ON(!list_empty(&idev->addr_list));
 329         WARN_ON(idev->mc_list != NULL);
 330
 331 #ifdef NET_REFCNT_DEBUG
 332         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 333 #endif
 334         dev_put(dev);
 335         if (!idev->dead) {
 336                 pr_warn("Freeing alive inet6 device %p\n", idev);
 337                 return;
 338         }
 339         snmp6_free_dev(idev);
 340         kfree_rcu(idev, rcu);
 341 }
 342 EXPORT_SYMBOL(in6_dev_finish_destroy);
 343
 344 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 345 {
 346         struct inet6_dev *ndev;
 347
 348         ASSERT_RTNL();
 349
 350         if (dev->mtu < IPV6_MIN_MTU)
 351                 return NULL;
 352
 353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 354
 355         if (ndev == NULL)
 356                 return NULL;
 357
 358         rwlock_init(&ndev->lock);
 359         ndev->dev = dev;
 360         INIT_LIST_HEAD(&ndev->addr_list);
 361
 362         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 363         ndev->cnf.mtu6 = dev->mtu;
 364         ndev->cnf.sysctl = NULL;
 365         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 366         if (ndev->nd_parms == NULL) {
 367                 kfree(ndev);
 368                 return NULL;
 369         }
 370         if (ndev->cnf.forwarding)
 371                 dev_disable_lro(dev);
 372         /* We refer to the device */
 373         dev_hold(dev);
 374
 375         if (snmp6_alloc_dev(ndev) < 0) {
 376                 ADBG((KERN_WARNING
 377                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 378                         __func__, dev->name));
 379                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 380                 dev_put(dev);
 381                 kfree(ndev);
 382                 return NULL;
 383         }
 384
 385         if (snmp6_register_dev(ndev) < 0) {
 386                 ADBG((KERN_WARNING
 387                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 388                         __func__, dev->name));
 389                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 390                 ndev->dead = 1;
 391                 in6_dev_finish_destroy(ndev);
 392                 return NULL;
 393         }
 394
 395         /* One reference from device.  We must do this before
 396          * we invoke __ipv6_regen_rndid().
 397          */
 398         in6_dev_hold(ndev);
 399
 400         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 401                 ndev->cnf.accept_dad = -1;
 402
 403 #if IS_ENABLED(CONFIG_IPV6_SIT)
 404         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 405                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 ndev->cnf.use_tempaddr = -1;
 419         } else {
 420                 in6_dev_hold(ndev);
 421                 ipv6_regen_rndid((unsigned long) ndev);
 422         }
 423 #endif
 424
 425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 426                 ndev->if_flags |= IF_READY;
 427
 428         ipv6_mc_init_dev(ndev);
 429         ndev->tstamp = jiffies;
 430         addrconf_sysctl_register(ndev);
 431         /* protected by rtnl_lock */
 432         rcu_assign_pointer(dev->ip6_ptr, ndev);
 433
 434         /* Join interface-local all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 436
 437         /* Join all-node multicast group */
 438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 439
 440         /* Join all-router multicast group if forwarding is set */
 441         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 442                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 443
 444         return ndev;
 445 }
 446
 447 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 448 {
 449         struct inet6_dev *idev;
 450
 451         ASSERT_RTNL();
 452
 453         idev = __in6_dev_get(dev);
 454         if (!idev) {
 455                 idev = ipv6_add_dev(dev);
 456                 if (!idev)
 457                         return NULL;
 458         }
 459
 460         if (dev->flags&IFF_UP)
 461                 ipv6_mc_up(idev);
 462         return idev;
 463 }
 464
 465 static int inet6_netconf_msgsize_devconf(int type)
 466 {
 467         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 468                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 469
 470         /* type -1 is used for ALL */
 471         if (type == -1 || type == NETCONFA_FORWARDING)
 472                 size += nla_total_size(4);
 473 #ifdef CONFIG_IPV6_MROUTE
 474         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 475                 size += nla_total_size(4);
 476 #endif
 477
 478         return size;
 479 }
 480
 481 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 482                                       struct ipv6_devconf *devconf, u32 portid,
 483                                       u32 seq, int event, unsigned int flags,
 484                                       int type)
 485 {
 486         struct nlmsghdr  *nlh;
 487         struct netconfmsg *ncm;
 488
 489         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 490                         flags);
 491         if (nlh == NULL)
 492                 return -EMSGSIZE;
 493
 494         ncm = nlmsg_data(nlh);
 495         ncm->ncm_family = AF_INET6;
 496
 497         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 498                 goto nla_put_failure;
 499
 500         /* type -1 is used for ALL */
 501         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 502             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 503                 goto nla_put_failure;
 504 #ifdef CONFIG_IPV6_MROUTE
 505         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 506             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 507                         devconf->mc_forwarding) < 0)
 508                 goto nla_put_failure;
 509 #endif
 510         return nlmsg_end(skb, nlh);
 511
 512 nla_put_failure:
 513         nlmsg_cancel(skb, nlh);
 514         return -EMSGSIZE;
 515 }
 516
 517 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 518                                   struct ipv6_devconf *devconf)
 519 {
 520         struct sk_buff *skb;
 521         int err = -ENOBUFS;
 522
 523         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 524         if (skb == NULL)
 525                 goto errout;
 526
 527         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 528                                          RTM_NEWNETCONF, 0, type);
 529         if (err < 0) {
 530                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 531                 WARN_ON(err == -EMSGSIZE);
 532                 kfree_skb(skb);
 533                 goto errout;
 534         }
 535         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 536         return;
 537 errout:
 538         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 539 }
 540
 541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 542         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 543         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 544 };
 545
 546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 547                                      struct nlmsghdr *nlh,
 548                                      void *arg)
 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                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 626                         if (idx < s_idx)
 627                                 goto cont;
 628                         idev = __in6_dev_get(dev);
 629                         if (!idev)
 630                                 goto cont;
 631
 632                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 633                                                        &idev->cnf,
 634                                                        NETLINK_CB(cb->skb).portid,
 635                                                        cb->nlh->nlmsg_seq,
 636                                                        RTM_NEWNETCONF,
 637                                                        NLM_F_MULTI,
 638                                                        -1) <= 0) {
 639                                 rcu_read_unlock();
 640                                 goto done;
 641                         }
 642 cont:
 643                         idx++;
 644                 }
 645                 rcu_read_unlock();
 646         }
 647         if (h == NETDEV_HASHENTRIES) {
 648                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 649                                                net->ipv6.devconf_all,
 650                                                NETLINK_CB(cb->skb).portid,
 651                                                cb->nlh->nlmsg_seq,
 652                                                RTM_NEWNETCONF, NLM_F_MULTI,
 653                                                -1) <= 0)
 654                         goto done;
 655                 else
 656                         h++;
 657         }
 658         if (h == NETDEV_HASHENTRIES + 1) {
 659                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 660                                                net->ipv6.devconf_dflt,
 661                                                NETLINK_CB(cb->skb).portid,
 662                                                cb->nlh->nlmsg_seq,
 663                                                RTM_NEWNETCONF, NLM_F_MULTI,
 664                                                -1) <= 0)
 665                         goto done;
 666                 else
 667                         h++;
 668         }
 669 done:
 670         cb->args[0] = h;
 671         cb->args[1] = idx;
 672
 673         return skb->len;
 674 }
 675
 676 #ifdef CONFIG_SYSCTL
 677 static void dev_forward_change(struct inet6_dev *idev)
 678 {
 679         struct net_device *dev;
 680         struct inet6_ifaddr *ifa;
 681
 682         if (!idev)
 683                 return;
 684         dev = idev->dev;
 685         if (idev->cnf.forwarding)
 686                 dev_disable_lro(dev);
 687         if (dev->flags & IFF_MULTICAST) {
 688                 if (idev->cnf.forwarding) {
 689                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 690                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 691                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 692                 } else {
 693                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 694                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 695                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 696                 }
 697         }
 698
 699         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 700                 if (ifa->flags&IFA_F_TENTATIVE)
 701                         continue;
 702                 if (idev->cnf.forwarding)
 703                         addrconf_join_anycast(ifa);
 704                 else
 705                         addrconf_leave_anycast(ifa);
 706         }
 707         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 708                                      dev->ifindex, &idev->cnf);
 709 }
 710
 711
 712 static void addrconf_forward_change(struct net *net, __s32 newf)
 713 {
 714         struct net_device *dev;
 715         struct inet6_dev *idev;
 716
 717         for_each_netdev(net, dev) {
 718                 idev = __in6_dev_get(dev);
 719                 if (idev) {
 720                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 721                         idev->cnf.forwarding = newf;
 722                         if (changed)
 723                                 dev_forward_change(idev);
 724                 }
 725         }
 726 }
 727
 728 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 729 {
 730         struct net *net;
 731         int old;
 732
 733         if (!rtnl_trylock())
 734                 return restart_syscall();
 735
 736         net = (struct net *)table->extra2;
 737         old = *p;
 738         *p = newf;
 739
 740         if (p == &net->ipv6.devconf_dflt->forwarding) {
 741                 if ((!newf) ^ (!old))
 742                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 743                                                      NETCONFA_IFINDEX_DEFAULT,
 744                                                      net->ipv6.devconf_dflt);
 745                 rtnl_unlock();
 746                 return 0;
 747         }
 748
 749         if (p == &net->ipv6.devconf_all->forwarding) {
 750                 net->ipv6.devconf_dflt->forwarding = newf;
 751                 addrconf_forward_change(net, newf);
 752                 if ((!newf) ^ (!old))
 753                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 754                                                      NETCONFA_IFINDEX_ALL,
 755                                                      net->ipv6.devconf_all);
 756         } else if ((!newf) ^ (!old))
 757                 dev_forward_change((struct inet6_dev *)table->extra1);
 758         rtnl_unlock();
 759
 760         if (newf)
 761                 rt6_purge_dflt_routers(net);
 762         return 1;
 763 }
 764 #endif
 765
 766 /* Nobody refers to this ifaddr, destroy it */
 767 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 768 {
 769         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 770
 771 #ifdef NET_REFCNT_DEBUG
 772         pr_debug("%s\n", __func__);
 773 #endif
 774
 775         in6_dev_put(ifp->idev);
 776
 777         if (del_timer(&ifp->timer))
 778                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 779
 780         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 781                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 782                 return;
 783         }
 784         ip6_rt_put(ifp->rt);
 785
 786         kfree_rcu(ifp, rcu);
 787 }
 788
 789 static void
 790 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 791 {
 792         struct list_head *p;
 793         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 794
 795         /*
 796          * Each device address list is sorted in order of scope -
 797          * global before linklocal.
 798          */
 799         list_for_each(p, &idev->addr_list) {
 800                 struct inet6_ifaddr *ifa
 801                         = list_entry(p, struct inet6_ifaddr, if_list);
 802                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 803                         break;
 804         }
 805
 806         list_add_tail(&ifp->if_list, p);
 807 }
 808
 809 static u32 inet6_addr_hash(const struct in6_addr *addr)
 810 {
 811         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 812 }
 813
 814 /* On success it returns ifp with increased reference count */
 815
 816 static struct inet6_ifaddr *
 817 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 818               int scope, u32 flags)
 819 {
 820         struct inet6_ifaddr *ifa = NULL;
 821         struct rt6_info *rt;
 822         unsigned int hash;
 823         int err = 0;
 824         int addr_type = ipv6_addr_type(addr);
 825
 826         if (addr_type == IPV6_ADDR_ANY ||
 827             addr_type & IPV6_ADDR_MULTICAST ||
 828             (!(idev->dev->flags & IFF_LOOPBACK) &&
 829              addr_type & IPV6_ADDR_LOOPBACK))
 830                 return ERR_PTR(-EADDRNOTAVAIL);
 831
 832         rcu_read_lock_bh();
 833         if (idev->dead) {
 834                 err = -ENODEV;                  /*XXX*/
 835                 goto out2;
 836         }
 837
 838         if (idev->cnf.disable_ipv6) {
 839                 err = -EACCES;
 840                 goto out2;
 841         }
 842
 843         spin_lock(&addrconf_hash_lock);
 844
 845         /* Ignore adding duplicate addresses on an interface */
 846         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 847                 ADBG(("ipv6_add_addr: already assigned\n"));
 848                 err = -EEXIST;
 849                 goto out;
 850         }
 851
 852         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 853
 854         if (ifa == NULL) {
 855                 ADBG(("ipv6_add_addr: malloc failed\n"));
 856                 err = -ENOBUFS;
 857                 goto out;
 858         }
 859
 860         rt = addrconf_dst_alloc(idev, addr, false);
 861         if (IS_ERR(rt)) {
 862                 err = PTR_ERR(rt);
 863                 goto out;
 864         }
 865
 866         ifa->addr = *addr;
 867
 868         spin_lock_init(&ifa->lock);
 869         spin_lock_init(&ifa->state_lock);
 870         init_timer(&ifa->timer);
 871         INIT_HLIST_NODE(&ifa->addr_lst);
 872         ifa->timer.data = (unsigned long) ifa;
 873         ifa->scope = scope;
 874         ifa->prefix_len = pfxlen;
 875         ifa->flags = flags | IFA_F_TENTATIVE;
 876         ifa->cstamp = ifa->tstamp = jiffies;
 877
 878         ifa->rt = rt;
 879
 880         ifa->idev = idev;
 881         in6_dev_hold(idev);
 882         /* For caller */
 883         in6_ifa_hold(ifa);
 884
 885         /* Add to big hash table */
 886         hash = inet6_addr_hash(addr);
 887
 888         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 889         spin_unlock(&addrconf_hash_lock);
 890
 891         write_lock(&idev->lock);
 892         /* Add to inet6_dev unicast addr list. */
 893         ipv6_link_dev_addr(idev, ifa);
 894
 895 #ifdef CONFIG_IPV6_PRIVACY
 896         if (ifa->flags&IFA_F_TEMPORARY) {
 897                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 898                 in6_ifa_hold(ifa);
 899         }
 900 #endif
 901
 902         in6_ifa_hold(ifa);
 903         write_unlock(&idev->lock);
 904 out2:
 905         rcu_read_unlock_bh();
 906
 907         if (likely(err == 0))
 908                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 909         else {
 910                 kfree(ifa);
 911                 ifa = ERR_PTR(err);
 912         }
 913
 914         return ifa;
 915 out:
 916         spin_unlock(&addrconf_hash_lock);
 917         goto out2;
 918 }
 919
 920 /* This function wants to get referenced ifp and releases it before return */
 921
 922 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 923 {
 924         struct inet6_ifaddr *ifa, *ifn;
 925         struct inet6_dev *idev = ifp->idev;
 926         int state;
 927         int deleted = 0, onlink = 0;
 928         unsigned long expires = jiffies;
 929
 930         spin_lock_bh(&ifp->state_lock);
 931         state = ifp->state;
 932         ifp->state = INET6_IFADDR_STATE_DEAD;
 933         spin_unlock_bh(&ifp->state_lock);
 934
 935         if (state == INET6_IFADDR_STATE_DEAD)
 936                 goto out;
 937
 938         spin_lock_bh(&addrconf_hash_lock);
 939         hlist_del_init_rcu(&ifp->addr_lst);
 940         spin_unlock_bh(&addrconf_hash_lock);
 941
 942         write_lock_bh(&idev->lock);
 943 #ifdef CONFIG_IPV6_PRIVACY
 944         if (ifp->flags&IFA_F_TEMPORARY) {
 945                 list_del(&ifp->tmp_list);
 946                 if (ifp->ifpub) {
 947                         in6_ifa_put(ifp->ifpub);
 948                         ifp->ifpub = NULL;
 949                 }
 950                 __in6_ifa_put(ifp);
 951         }
 952 #endif
 953
 954         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 955                 if (ifa == ifp) {
 956                         list_del_init(&ifp->if_list);
 957                         __in6_ifa_put(ifp);
 958
 959                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 960                                 break;
 961                         deleted = 1;
 962                         continue;
 963                 } else if (ifp->flags & IFA_F_PERMANENT) {
 964                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 965                                               ifp->prefix_len)) {
 966                                 if (ifa->flags & IFA_F_PERMANENT) {
 967                                         onlink = 1;
 968                                         if (deleted)
 969                                                 break;
 970                                 } else {
 971                                         unsigned long lifetime;
 972
 973                                         if (!onlink)
 974                                                 onlink = -1;
 975
 976                                         spin_lock(&ifa->lock);
 977
 978                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 979                                         /*
 980                                          * Note: Because this address is
 981                                          * not permanent, lifetime <
 982                                          * LONG_MAX / HZ here.
 983                                          */
 984                                         if (time_before(expires,
 985                                                         ifa->tstamp + lifetime * HZ))
 986                                                 expires = ifa->tstamp + lifetime * HZ;
 987                                         spin_unlock(&ifa->lock);
 988                                 }
 989                         }
 990                 }
 991         }
 992         write_unlock_bh(&idev->lock);
 993
 994         addrconf_del_timer(ifp);
 995
 996         ipv6_ifa_notify(RTM_DELADDR, ifp);
 997
 998         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 999
1000         /*
1001          * Purge or update corresponding prefix
1002          *
1003          * 1) we don't purge prefix here if address was not permanent.
1004          *    prefix is managed by its own lifetime.
1005          * 2) if there're no addresses, delete prefix.
1006          * 3) if there're still other permanent address(es),
1007          *    corresponding prefix is still permanent.
1008          * 4) otherwise, update prefix lifetime to the
1009          *    longest valid lifetime among the corresponding
1010          *    addresses on the device.
1011          *    Note: subsequent RA will update lifetime.
1012          *
1013          * --yoshfuji
1014          */
1015         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1016                 struct in6_addr prefix;
1017                 struct rt6_info *rt;
1018
1019                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1020
1021                 rt = addrconf_get_prefix_route(&prefix,
1022                                                ifp->prefix_len,
1023                                                ifp->idev->dev,
1024                                                0, RTF_GATEWAY | RTF_DEFAULT);
1025
1026                 if (rt) {
1027                         if (onlink == 0) {
1028                                 ip6_del_rt(rt);
1029                                 rt = NULL;
1030                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1031                                 rt6_set_expires(rt, expires);
1032                         }
1033                 }
1034                 ip6_rt_put(rt);
1035         }
1036
1037         /* clean up prefsrc entries */
1038         rt6_remove_prefsrc(ifp);
1039 out:
1040         in6_ifa_put(ifp);
1041 }
1042
1043 #ifdef CONFIG_IPV6_PRIVACY
1044 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1045 {
1046         struct inet6_dev *idev = ifp->idev;
1047         struct in6_addr addr, *tmpaddr;
1048         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1049         unsigned long regen_advance;
1050         int tmp_plen;
1051         int ret = 0;
1052         int max_addresses;
1053         u32 addr_flags;
1054         unsigned long now = jiffies;
1055
1056         write_lock(&idev->lock);
1057         if (ift) {
1058                 spin_lock_bh(&ift->lock);
1059                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1060                 spin_unlock_bh(&ift->lock);
1061                 tmpaddr = &addr;
1062         } else {
1063                 tmpaddr = NULL;
1064         }
1065 retry:
1066         in6_dev_hold(idev);
1067         if (idev->cnf.use_tempaddr <= 0) {
1068                 write_unlock(&idev->lock);
1069                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1070                 in6_dev_put(idev);
1071                 ret = -1;
1072                 goto out;
1073         }
1074         spin_lock_bh(&ifp->lock);
1075         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1076                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1077                 spin_unlock_bh(&ifp->lock);
1078                 write_unlock(&idev->lock);
1079                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1080                         __func__);
1081                 in6_dev_put(idev);
1082                 ret = -1;
1083                 goto out;
1084         }
1085         in6_ifa_hold(ifp);
1086         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1087         __ipv6_try_regen_rndid(idev, tmpaddr);
1088         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1089         age = (now - ifp->tstamp) / HZ;
1090         tmp_valid_lft = min_t(__u32,
1091                               ifp->valid_lft,
1092                               idev->cnf.temp_valid_lft + age);
1093         tmp_prefered_lft = min_t(__u32,
1094                                  ifp->prefered_lft,
1095                                  idev->cnf.temp_prefered_lft + age -
1096                                  idev->cnf.max_desync_factor);
1097         tmp_plen = ifp->prefix_len;
1098         max_addresses = idev->cnf.max_addresses;
1099         tmp_tstamp = ifp->tstamp;
1100         spin_unlock_bh(&ifp->lock);
1101
1102         regen_advance = idev->cnf.regen_max_retry *
1103                         idev->cnf.dad_transmits *
1104                         idev->nd_parms->retrans_time / HZ;
1105         write_unlock(&idev->lock);
1106
1107         /* A temporary address is created only if this calculated Preferred
1108          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1109          * an implementation must not create a temporary address with a zero
1110          * Preferred Lifetime.
1111          */
1112         if (tmp_prefered_lft <= regen_advance) {
1113                 in6_ifa_put(ifp);
1114                 in6_dev_put(idev);
1115                 ret = -1;
1116                 goto out;
1117         }
1118
1119         addr_flags = IFA_F_TEMPORARY;
1120         /* set in addrconf_prefix_rcv() */
1121         if (ifp->flags & IFA_F_OPTIMISTIC)
1122                 addr_flags |= IFA_F_OPTIMISTIC;
1123
1124         ift = !max_addresses ||
1125               ipv6_count_addresses(idev) < max_addresses ?
1126                 ipv6_add_addr(idev, &addr, tmp_plen,
1127                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1128                               addr_flags) : NULL;
1129         if (IS_ERR_OR_NULL(ift)) {
1130                 in6_ifa_put(ifp);
1131                 in6_dev_put(idev);
1132                 pr_info("%s: retry temporary address regeneration\n", __func__);
1133                 tmpaddr = &addr;
1134                 write_lock(&idev->lock);
1135                 goto retry;
1136         }
1137
1138         spin_lock_bh(&ift->lock);
1139         ift->ifpub = ifp;
1140         ift->valid_lft = tmp_valid_lft;
1141         ift->prefered_lft = tmp_prefered_lft;
1142         ift->cstamp = now;
1143         ift->tstamp = tmp_tstamp;
1144         spin_unlock_bh(&ift->lock);
1145
1146         addrconf_dad_start(ift);
1147         in6_ifa_put(ift);
1148         in6_dev_put(idev);
1149 out:
1150         return ret;
1151 }
1152 #endif
1153
1154 /*
1155  *      Choose an appropriate source address (RFC3484)
1156  */
1157 enum {
1158         IPV6_SADDR_RULE_INIT = 0,
1159         IPV6_SADDR_RULE_LOCAL,
1160         IPV6_SADDR_RULE_SCOPE,
1161         IPV6_SADDR_RULE_PREFERRED,
1162 #ifdef CONFIG_IPV6_MIP6
1163         IPV6_SADDR_RULE_HOA,
1164 #endif
1165         IPV6_SADDR_RULE_OIF,
1166         IPV6_SADDR_RULE_LABEL,
1167 #ifdef CONFIG_IPV6_PRIVACY
1168         IPV6_SADDR_RULE_PRIVACY,
1169 #endif
1170         IPV6_SADDR_RULE_ORCHID,
1171         IPV6_SADDR_RULE_PREFIX,
1172         IPV6_SADDR_RULE_MAX
1173 };
1174
1175 struct ipv6_saddr_score {
1176         int                     rule;
1177         int                     addr_type;
1178         struct inet6_ifaddr     *ifa;
1179         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1180         int                     scopedist;
1181         int                     matchlen;
1182 };
1183
1184 struct ipv6_saddr_dst {
1185         const struct in6_addr *addr;
1186         int ifindex;
1187         int scope;
1188         int label;
1189         unsigned int prefs;
1190 };
1191
1192 static inline int ipv6_saddr_preferred(int type)
1193 {
1194         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1195                 return 1;
1196         return 0;
1197 }
1198
1199 static int ipv6_get_saddr_eval(struct net *net,
1200                                struct ipv6_saddr_score *score,
1201                                struct ipv6_saddr_dst *dst,
1202                                int i)
1203 {
1204         int ret;
1205
1206         if (i <= score->rule) {
1207                 switch (i) {
1208                 case IPV6_SADDR_RULE_SCOPE:
1209                         ret = score->scopedist;
1210                         break;
1211                 case IPV6_SADDR_RULE_PREFIX:
1212                         ret = score->matchlen;
1213                         break;
1214                 default:
1215                         ret = !!test_bit(i, score->scorebits);
1216                 }
1217                 goto out;
1218         }
1219
1220         switch (i) {
1221         case IPV6_SADDR_RULE_INIT:
1222                 /* Rule 0: remember if hiscore is not ready yet */
1223                 ret = !!score->ifa;
1224                 break;
1225         case IPV6_SADDR_RULE_LOCAL:
1226                 /* Rule 1: Prefer same address */
1227                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1228                 break;
1229         case IPV6_SADDR_RULE_SCOPE:
1230                 /* Rule 2: Prefer appropriate scope
1231                  *
1232                  *      ret
1233                  *       ^
1234                  *    -1 |  d 15
1235                  *    ---+--+-+---> scope
1236                  *       |
1237                  *       |             d is scope of the destination.
1238                  *  B-d  |  \
1239                  *       |   \      <- smaller scope is better if
1240                  *  B-15 |    \        if scope is enough for destinaion.
1241                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1242                  * d-C-1 | /
1243                  *       |/         <- greater is better
1244                  *   -C  /             if scope is not enough for destination.
1245                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1246                  *
1247                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1248                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1249                  * Assume B = 0 and we get C > 29.
1250                  */
1251                 ret = __ipv6_addr_src_scope(score->addr_type);
1252                 if (ret >= dst->scope)
1253                         ret = -ret;
1254                 else
1255                         ret -= 128;     /* 30 is enough */
1256                 score->scopedist = ret;
1257                 break;
1258         case IPV6_SADDR_RULE_PREFERRED:
1259                 /* Rule 3: Avoid deprecated and optimistic addresses */
1260                 ret = ipv6_saddr_preferred(score->addr_type) ||
1261                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1262                 break;
1263 #ifdef CONFIG_IPV6_MIP6
1264         case IPV6_SADDR_RULE_HOA:
1265             {
1266                 /* Rule 4: Prefer home address */
1267                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1268                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1269                 break;
1270             }
1271 #endif
1272         case IPV6_SADDR_RULE_OIF:
1273                 /* Rule 5: Prefer outgoing interface */
1274                 ret = (!dst->ifindex ||
1275                        dst->ifindex == score->ifa->idev->dev->ifindex);
1276                 break;
1277         case IPV6_SADDR_RULE_LABEL:
1278                 /* Rule 6: Prefer matching label */
1279                 ret = ipv6_addr_label(net,
1280                                       &score->ifa->addr, score->addr_type,
1281                                       score->ifa->idev->dev->ifindex) == dst->label;
1282                 break;
1283 #ifdef CONFIG_IPV6_PRIVACY
1284         case IPV6_SADDR_RULE_PRIVACY:
1285             {
1286                 /* Rule 7: Prefer public address
1287                  * Note: prefer temporary address if use_tempaddr >= 2
1288                  */
1289                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1290                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1291                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1292                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1293                 break;
1294             }
1295 #endif
1296         case IPV6_SADDR_RULE_ORCHID:
1297                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1298                  *          non-ORCHID vs non-ORCHID
1299                  */
1300                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1301                         ipv6_addr_orchid(dst->addr));
1302                 break;
1303         case IPV6_SADDR_RULE_PREFIX:
1304                 /* Rule 8: Use longest matching prefix */
1305                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1306                 if (ret > score->ifa->prefix_len)
1307                         ret = score->ifa->prefix_len;
1308                 score->matchlen = ret;
1309                 break;
1310         default:
1311                 ret = 0;
1312         }
1313
1314         if (ret)
1315                 __set_bit(i, score->scorebits);
1316         score->rule = i;
1317 out:
1318         return ret;
1319 }
1320
1321 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1322                        const struct in6_addr *daddr, unsigned int prefs,
1323                        struct in6_addr *saddr)
1324 {
1325         struct ipv6_saddr_score scores[2],
1326                                 *score = &scores[0], *hiscore = &scores[1];
1327         struct ipv6_saddr_dst dst;
1328         struct net_device *dev;
1329         int dst_type;
1330
1331         dst_type = __ipv6_addr_type(daddr);
1332         dst.addr = daddr;
1333         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1334         dst.scope = __ipv6_addr_src_scope(dst_type);
1335         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1336         dst.prefs = prefs;
1337
1338         hiscore->rule = -1;
1339         hiscore->ifa = NULL;
1340
1341         rcu_read_lock();
1342
1343         for_each_netdev_rcu(net, dev) {
1344                 struct inet6_dev *idev;
1345
1346                 /* Candidate Source Address (section 4)
1347                  *  - multicast and link-local destination address,
1348                  *    the set of candidate source address MUST only
1349                  *    include addresses assigned to interfaces
1350                  *    belonging to the same link as the outgoing
1351                  *    interface.
1352                  * (- For site-local destination addresses, the
1353                  *    set of candidate source addresses MUST only
1354                  *    include addresses assigned to interfaces
1355                  *    belonging to the same site as the outgoing
1356                  *    interface.)
1357                  */
1358                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1359                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1360                     dst.ifindex && dev->ifindex != dst.ifindex)
1361                         continue;
1362
1363                 idev = __in6_dev_get(dev);
1364                 if (!idev)
1365                         continue;
1366
1367                 read_lock_bh(&idev->lock);
1368                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1369                         int i;
1370
1371                         /*
1372                          * - Tentative Address (RFC2462 section 5.4)
1373                          *  - A tentative address is not considered
1374                          *    "assigned to an interface" in the traditional
1375                          *    sense, unless it is also flagged as optimistic.
1376                          * - Candidate Source Address (section 4)
1377                          *  - In any case, anycast addresses, multicast
1378                          *    addresses, and the unspecified address MUST
1379                          *    NOT be included in a candidate set.
1380                          */
1381                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1382                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1383                                 continue;
1384
1385                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1386
1387                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1388                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1389                                 LIMIT_NETDEBUG(KERN_DEBUG
1390                                                "ADDRCONF: unspecified / multicast address "
1391                                                "assigned as unicast address on %s",
1392                                                dev->name);
1393                                 continue;
1394                         }
1395
1396                         score->rule = -1;
1397                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1398
1399                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1400                                 int minihiscore, miniscore;
1401
1402                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1403                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1404
1405                                 if (minihiscore > miniscore) {
1406                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1407                                             score->scopedist > 0) {
1408                                                 /*
1409                                                  * special case:
1410                                                  * each remaining entry
1411                                                  * has too small (not enough)
1412                                                  * scope, because ifa entries
1413                                                  * are sorted by their scope
1414                                                  * values.
1415                                                  */
1416                                                 goto try_nextdev;
1417                                         }
1418                                         break;
1419                                 } else if (minihiscore < miniscore) {
1420                                         if (hiscore->ifa)
1421                                                 in6_ifa_put(hiscore->ifa);
1422
1423                                         in6_ifa_hold(score->ifa);
1424
1425                                         swap(hiscore, score);
1426
1427                                         /* restore our iterator */
1428                                         score->ifa = hiscore->ifa;
1429
1430                                         break;
1431                                 }
1432                         }
1433                 }
1434 try_nextdev:
1435                 read_unlock_bh(&idev->lock);
1436         }
1437         rcu_read_unlock();
1438
1439         if (!hiscore->ifa)
1440                 return -EADDRNOTAVAIL;
1441
1442         *saddr = hiscore->ifa->addr;
1443         in6_ifa_put(hiscore->ifa);
1444         return 0;
1445 }
1446 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1447
1448 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1449                     unsigned char banned_flags)
1450 {
1451         struct inet6_dev *idev;
1452         int err = -EADDRNOTAVAIL;
1453
1454         rcu_read_lock();
1455         idev = __in6_dev_get(dev);
1456         if (idev) {
1457                 struct inet6_ifaddr *ifp;
1458
1459                 read_lock_bh(&idev->lock);
1460                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1461                         if (ifp->scope == IFA_LINK &&
1462                             !(ifp->flags & banned_flags)) {
1463                                 *addr = ifp->addr;
1464                                 err = 0;
1465                                 break;
1466                         }
1467                 }
1468                 read_unlock_bh(&idev->lock);
1469         }
1470         rcu_read_unlock();
1471         return err;
1472 }
1473
1474 static int ipv6_count_addresses(struct inet6_dev *idev)
1475 {
1476         int cnt = 0;
1477         struct inet6_ifaddr *ifp;
1478
1479         read_lock_bh(&idev->lock);
1480         list_for_each_entry(ifp, &idev->addr_list, if_list)
1481                 cnt++;
1482         read_unlock_bh(&idev->lock);
1483         return cnt;
1484 }
1485
1486 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1487                   struct net_device *dev, int strict)
1488 {
1489         struct inet6_ifaddr *ifp;
1490         unsigned int hash = inet6_addr_hash(addr);
1491
1492         rcu_read_lock_bh();
1493         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1494                 if (!net_eq(dev_net(ifp->idev->dev), net))
1495                         continue;
1496                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1497                     !(ifp->flags&IFA_F_TENTATIVE) &&
1498                     (dev == NULL || ifp->idev->dev == dev ||
1499                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1500                         rcu_read_unlock_bh();
1501                         return 1;
1502                 }
1503         }
1504
1505         rcu_read_unlock_bh();
1506         return 0;
1507 }
1508 EXPORT_SYMBOL(ipv6_chk_addr);
1509
1510 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1511                                struct net_device *dev)
1512 {
1513         unsigned int hash = inet6_addr_hash(addr);
1514         struct inet6_ifaddr *ifp;
1515
1516         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1517                 if (!net_eq(dev_net(ifp->idev->dev), net))
1518                         continue;
1519                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1520                         if (dev == NULL || ifp->idev->dev == dev)
1521                                 return true;
1522                 }
1523         }
1524         return false;
1525 }
1526
1527 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1528 {
1529         struct inet6_dev *idev;
1530         struct inet6_ifaddr *ifa;
1531         int     onlink;
1532
1533         onlink = 0;
1534         rcu_read_lock();
1535         idev = __in6_dev_get(dev);
1536         if (idev) {
1537                 read_lock_bh(&idev->lock);
1538                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1539                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1540                                                    ifa->prefix_len);
1541                         if (onlink)
1542                                 break;
1543                 }
1544                 read_unlock_bh(&idev->lock);
1545         }
1546         rcu_read_unlock();
1547         return onlink;
1548 }
1549 EXPORT_SYMBOL(ipv6_chk_prefix);
1550
1551 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1552                                      struct net_device *dev, int strict)
1553 {
1554         struct inet6_ifaddr *ifp, *result = NULL;
1555         unsigned int hash = inet6_addr_hash(addr);
1556
1557         rcu_read_lock_bh();
1558         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1559                 if (!net_eq(dev_net(ifp->idev->dev), net))
1560                         continue;
1561                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1562                         if (dev == NULL || ifp->idev->dev == dev ||
1563                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1564                                 result = ifp;
1565                                 in6_ifa_hold(ifp);
1566                                 break;
1567                         }
1568                 }
1569         }
1570         rcu_read_unlock_bh();
1571
1572         return result;
1573 }
1574
1575 /* Gets referenced address, destroys ifaddr */
1576
1577 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1578 {
1579         if (ifp->flags&IFA_F_PERMANENT) {
1580                 spin_lock_bh(&ifp->lock);
1581                 addrconf_del_timer(ifp);
1582                 ifp->flags |= IFA_F_TENTATIVE;
1583                 if (dad_failed)
1584                         ifp->flags |= IFA_F_DADFAILED;
1585                 spin_unlock_bh(&ifp->lock);
1586                 if (dad_failed)
1587                         ipv6_ifa_notify(0, ifp);
1588                 in6_ifa_put(ifp);
1589 #ifdef CONFIG_IPV6_PRIVACY
1590         } else if (ifp->flags&IFA_F_TEMPORARY) {
1591                 struct inet6_ifaddr *ifpub;
1592                 spin_lock_bh(&ifp->lock);
1593                 ifpub = ifp->ifpub;
1594                 if (ifpub) {
1595                         in6_ifa_hold(ifpub);
1596                         spin_unlock_bh(&ifp->lock);
1597                         ipv6_create_tempaddr(ifpub, ifp);
1598                         in6_ifa_put(ifpub);
1599                 } else {
1600                         spin_unlock_bh(&ifp->lock);
1601                 }
1602                 ipv6_del_addr(ifp);
1603 #endif
1604         } else
1605                 ipv6_del_addr(ifp);
1606 }
1607
1608 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1609 {
1610         int err = -ENOENT;
1611
1612         spin_lock(&ifp->state_lock);
1613         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1614                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1615                 err = 0;
1616         }
1617         spin_unlock(&ifp->state_lock);
1618
1619         return err;
1620 }
1621
1622 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1623 {
1624         struct inet6_dev *idev = ifp->idev;
1625
1626         if (addrconf_dad_end(ifp)) {
1627                 in6_ifa_put(ifp);
1628                 return;
1629         }
1630
1631         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1632                              ifp->idev->dev->name, &ifp->addr);
1633
1634         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1635                 struct in6_addr addr;
1636
1637                 addr.s6_addr32[0] = htonl(0xfe800000);
1638                 addr.s6_addr32[1] = 0;
1639
1640                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1641                     ipv6_addr_equal(&ifp->addr, &addr)) {
1642                         /* DAD failed for link-local based on MAC address */
1643                         idev->cnf.disable_ipv6 = 1;
1644
1645                         pr_info("%s: IPv6 being disabled!\n",
1646                                 ifp->idev->dev->name);
1647                 }
1648         }
1649
1650         addrconf_dad_stop(ifp, 1);
1651 }
1652
1653 /* Join to solicited addr multicast group. */
1654
1655 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1656 {
1657         struct in6_addr maddr;
1658
1659         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1660                 return;
1661
1662         addrconf_addr_solict_mult(addr, &maddr);
1663         ipv6_dev_mc_inc(dev, &maddr);
1664 }
1665
1666 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1667 {
1668         struct in6_addr maddr;
1669
1670         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1671                 return;
1672
1673         addrconf_addr_solict_mult(addr, &maddr);
1674         __ipv6_dev_mc_dec(idev, &maddr);
1675 }
1676
1677 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1678 {
1679         struct in6_addr addr;
1680         if (ifp->prefix_len == 127) /* RFC 6164 */
1681                 return;
1682         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1683         if (ipv6_addr_any(&addr))
1684                 return;
1685         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1686 }
1687
1688 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1689 {
1690         struct in6_addr addr;
1691         if (ifp->prefix_len == 127) /* RFC 6164 */
1692                 return;
1693         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1694         if (ipv6_addr_any(&addr))
1695                 return;
1696         __ipv6_dev_ac_dec(ifp->idev, &addr);
1697 }
1698
1699 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1700 {
1701         if (dev->addr_len != ETH_ALEN)
1702                 return -1;
1703         memcpy(eui, dev->dev_addr, 3);
1704         memcpy(eui + 5, dev->dev_addr + 3, 3);
1705
1706         /*
1707          * The zSeries OSA network cards can be shared among various
1708          * OS instances, but the OSA cards have only one MAC address.
1709          * This leads to duplicate address conflicts in conjunction
1710          * with IPv6 if more than one instance uses the same card.
1711          *
1712          * The driver for these cards can deliver a unique 16-bit
1713          * identifier for each instance sharing the same card.  It is
1714          * placed instead of 0xFFFE in the interface identifier.  The
1715          * "u" bit of the interface identifier is not inverted in this
1716          * case.  Hence the resulting interface identifier has local
1717          * scope according to RFC2373.
1718          */
1719         if (dev->dev_id) {
1720                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1721                 eui[4] = dev->dev_id & 0xFF;
1722         } else {
1723                 eui[3] = 0xFF;
1724                 eui[4] = 0xFE;
1725                 eui[0] ^= 2;
1726         }
1727         return 0;
1728 }
1729
1730 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1731 {
1732         if (dev->addr_len != IEEE802154_ADDR_LEN)
1733                 return -1;
1734         memcpy(eui, dev->dev_addr, 8);
1735         eui[0] ^= 2;
1736         return 0;
1737 }
1738
1739 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1740 {
1741         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1742         if (dev->addr_len != ARCNET_ALEN)
1743                 return -1;
1744         memset(eui, 0, 7);
1745         eui[7] = *(u8 *)dev->dev_addr;
1746         return 0;
1747 }
1748
1749 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1750 {
1751         if (dev->addr_len != INFINIBAND_ALEN)
1752                 return -1;
1753         memcpy(eui, dev->dev_addr + 12, 8);
1754         eui[0] |= 2;
1755         return 0;
1756 }
1757
1758 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1759 {
1760         if (addr == 0)
1761                 return -1;
1762         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1763                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1764                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1765                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1766                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1767                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1768         eui[1] = 0;
1769         eui[2] = 0x5E;
1770         eui[3] = 0xFE;
1771         memcpy(eui + 4, &addr, 4);
1772         return 0;
1773 }
1774
1775 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1776 {
1777         if (dev->priv_flags & IFF_ISATAP)
1778                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1779         return -1;
1780 }
1781
1782 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1783 {
1784         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1785 }
1786
1787 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1788 {
1789         switch (dev->type) {
1790         case ARPHRD_ETHER:
1791         case ARPHRD_FDDI:
1792                 return addrconf_ifid_eui48(eui, dev);
1793         case ARPHRD_ARCNET:
1794                 return addrconf_ifid_arcnet(eui, dev);
1795         case ARPHRD_INFINIBAND:
1796                 return addrconf_ifid_infiniband(eui, dev);
1797         case ARPHRD_SIT:
1798                 return addrconf_ifid_sit(eui, dev);
1799         case ARPHRD_IPGRE:
1800                 return addrconf_ifid_gre(eui, dev);
1801         case ARPHRD_IEEE802154:
1802                 return addrconf_ifid_eui64(eui, dev);
1803         }
1804         return -1;
1805 }
1806
1807 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1808 {
1809         int err = -1;
1810         struct inet6_ifaddr *ifp;
1811
1812         read_lock_bh(&idev->lock);
1813         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1814                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1815                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1816                         err = 0;
1817                         break;
1818                 }
1819         }
1820         read_unlock_bh(&idev->lock);
1821         return err;
1822 }
1823
1824 #ifdef CONFIG_IPV6_PRIVACY
1825 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1826 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1827 {
1828 regen:
1829         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1830         idev->rndid[0] &= ~0x02;
1831
1832         /*
1833          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1834          * check if generated address is not inappropriate
1835          *
1836          *  - Reserved subnet anycast (RFC 2526)
1837          *      11111101 11....11 1xxxxxxx
1838          *  - ISATAP (RFC4214) 6.1
1839          *      00-00-5E-FE-xx-xx-xx-xx
1840          *  - value 0
1841          *  - XXX: already assigned to an address on the device
1842          */
1843         if (idev->rndid[0] == 0xfd &&
1844             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1845             (idev->rndid[7]&0x80))
1846                 goto regen;
1847         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1848                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1849                         goto regen;
1850                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1851                         goto regen;
1852         }
1853 }
1854
1855 static void ipv6_regen_rndid(unsigned long data)
1856 {
1857         struct inet6_dev *idev = (struct inet6_dev *) data;
1858         unsigned long expires;
1859
1860         rcu_read_lock_bh();
1861         write_lock_bh(&idev->lock);
1862
1863         if (idev->dead)
1864                 goto out;
1865
1866         __ipv6_regen_rndid(idev);
1867
1868         expires = jiffies +
1869                 idev->cnf.temp_prefered_lft * HZ -
1870                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1871                 idev->cnf.max_desync_factor * HZ;
1872         if (time_before(expires, jiffies)) {
1873                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1874                         __func__, idev->dev->name);
1875                 goto out;
1876         }
1877
1878         if (!mod_timer(&idev->regen_timer, expires))
1879                 in6_dev_hold(idev);
1880
1881 out:
1882         write_unlock_bh(&idev->lock);
1883         rcu_read_unlock_bh();
1884         in6_dev_put(idev);
1885 }
1886
1887 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1888 {
1889         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1890                 __ipv6_regen_rndid(idev);
1891 }
1892 #endif
1893
1894 /*
1895  *      Add prefix route.
1896  */
1897
1898 static void
1899 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1900                       unsigned long expires, u32 flags)
1901 {
1902         struct fib6_config cfg = {
1903                 .fc_table = RT6_TABLE_PREFIX,
1904                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1905                 .fc_ifindex = dev->ifindex,
1906                 .fc_expires = expires,
1907                 .fc_dst_len = plen,
1908                 .fc_flags = RTF_UP | flags,
1909                 .fc_nlinfo.nl_net = dev_net(dev),
1910                 .fc_protocol = RTPROT_KERNEL,
1911         };
1912
1913         cfg.fc_dst = *pfx;
1914
1915         /* Prevent useless cloning on PtP SIT.
1916            This thing is done here expecting that the whole
1917            class of non-broadcast devices need not cloning.
1918          */
1919 #if IS_ENABLED(CONFIG_IPV6_SIT)
1920         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1921                 cfg.fc_flags |= RTF_NONEXTHOP;
1922 #endif
1923
1924         ip6_route_add(&cfg);
1925 }
1926
1927
1928 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1929                                                   int plen,
1930                                                   const struct net_device *dev,
1931                                                   u32 flags, u32 noflags)
1932 {
1933         struct fib6_node *fn;
1934         struct rt6_info *rt = NULL;
1935         struct fib6_table *table;
1936
1937         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1938         if (table == NULL)
1939                 return NULL;
1940
1941         read_lock_bh(&table->tb6_lock);
1942         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1943         if (!fn)
1944                 goto out;
1945         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1946                 if (rt->dst.dev->ifindex != dev->ifindex)
1947                         continue;
1948                 if ((rt->rt6i_flags & flags) != flags)
1949                         continue;
1950                 if ((rt->rt6i_flags & noflags) != 0)
1951                         continue;
1952                 dst_hold(&rt->dst);
1953                 break;
1954         }
1955 out:
1956         read_unlock_bh(&table->tb6_lock);
1957         return rt;
1958 }
1959
1960
1961 /* Create "default" multicast route to the interface */
1962
1963 static void addrconf_add_mroute(struct net_device *dev)
1964 {
1965         struct fib6_config cfg = {
1966                 .fc_table = RT6_TABLE_LOCAL,
1967                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1968                 .fc_ifindex = dev->ifindex,
1969                 .fc_dst_len = 8,
1970                 .fc_flags = RTF_UP,
1971                 .fc_nlinfo.nl_net = dev_net(dev),
1972         };
1973
1974         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1975
1976         ip6_route_add(&cfg);
1977 }
1978
1979 #if IS_ENABLED(CONFIG_IPV6_SIT)
1980 static void sit_route_add(struct net_device *dev)
1981 {
1982         struct fib6_config cfg = {
1983                 .fc_table = RT6_TABLE_MAIN,
1984                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1985                 .fc_ifindex = dev->ifindex,
1986                 .fc_dst_len = 96,
1987                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1988                 .fc_nlinfo.nl_net = dev_net(dev),
1989         };
1990
1991         /* prefix length - 96 bits "::d.d.d.d" */
1992         ip6_route_add(&cfg);
1993 }
1994 #endif
1995
1996 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1997 {
1998         struct inet6_dev *idev;
1999
2000         ASSERT_RTNL();
2001
2002         idev = ipv6_find_idev(dev);
2003         if (!idev)
2004                 return ERR_PTR(-ENOBUFS);
2005
2006         if (idev->cnf.disable_ipv6)
2007                 return ERR_PTR(-EACCES);
2008
2009         /* Add default multicast route */
2010         if (!(dev->flags & IFF_LOOPBACK))
2011                 addrconf_add_mroute(dev);
2012
2013         return idev;
2014 }
2015
2016 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2017 {
2018         struct prefix_info *pinfo;
2019         __u32 valid_lft;
2020         __u32 prefered_lft;
2021         int addr_type;
2022         struct inet6_dev *in6_dev;
2023         struct net *net = dev_net(dev);
2024
2025         pinfo = (struct prefix_info *) opt;
2026
2027         if (len < sizeof(struct prefix_info)) {
2028                 ADBG(("addrconf: prefix option too short\n"));
2029                 return;
2030         }
2031
2032         /*
2033          *      Validation checks ([ADDRCONF], page 19)
2034          */
2035
2036         addr_type = ipv6_addr_type(&pinfo->prefix);
2037
2038         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2039                 return;
2040
2041         valid_lft = ntohl(pinfo->valid);
2042         prefered_lft = ntohl(pinfo->prefered);
2043
2044         if (prefered_lft > valid_lft) {
2045                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2046                 return;
2047         }
2048
2049         in6_dev = in6_dev_get(dev);
2050
2051         if (in6_dev == NULL) {
2052                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2053                                     dev->name);
2054                 return;
2055         }
2056
2057         /*
2058          *      Two things going on here:
2059          *      1) Add routes for on-link prefixes
2060          *      2) Configure prefixes with the auto flag set
2061          */
2062
2063         if (pinfo->onlink) {
2064                 struct rt6_info *rt;
2065                 unsigned long rt_expires;
2066
2067                 /* Avoid arithmetic overflow. Really, we could
2068                  * save rt_expires in seconds, likely valid_lft,
2069                  * but it would require division in fib gc, that it
2070                  * not good.
2071                  */
2072                 if (HZ > USER_HZ)
2073                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2074                 else
2075                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2076
2077                 if (addrconf_finite_timeout(rt_expires))
2078                         rt_expires *= HZ;
2079
2080                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2081                                                pinfo->prefix_len,
2082                                                dev,
2083                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2084                                                RTF_GATEWAY | RTF_DEFAULT);
2085
2086                 if (rt) {
2087                         /* Autoconf prefix route */
2088                         if (valid_lft == 0) {
2089                                 ip6_del_rt(rt);
2090                                 rt = NULL;
2091                         } else if (addrconf_finite_timeout(rt_expires)) {
2092                                 /* not infinity */
2093                                 rt6_set_expires(rt, jiffies + rt_expires);
2094                         } else {
2095                                 rt6_clean_expires(rt);
2096                         }
2097                 } else if (valid_lft) {
2098                         clock_t expires = 0;
2099                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2100                         if (addrconf_finite_timeout(rt_expires)) {
2101                                 /* not infinity */
2102                                 flags |= RTF_EXPIRES;
2103                                 expires = jiffies_to_clock_t(rt_expires);
2104                         }
2105                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2106                                               dev, expires, flags);
2107                 }
2108                 ip6_rt_put(rt);
2109         }
2110
2111         /* Try to figure out our local address for this prefix */
2112
2113         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2114                 struct inet6_ifaddr *ifp;
2115                 struct in6_addr addr;
2116                 int create = 0, update_lft = 0;
2117
2118                 if (pinfo->prefix_len == 64) {
2119                         memcpy(&addr, &pinfo->prefix, 8);
2120                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2121                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2122                                 in6_dev_put(in6_dev);
2123                                 return;
2124                         }
2125                         goto ok;
2126                 }
2127                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2128                                     pinfo->prefix_len);
2129                 in6_dev_put(in6_dev);
2130                 return;
2131
2132 ok:
2133
2134                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2135
2136                 if (ifp == NULL && valid_lft) {
2137                         int max_addresses = in6_dev->cnf.max_addresses;
2138                         u32 addr_flags = 0;
2139
2140 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2141                         if (in6_dev->cnf.optimistic_dad &&
2142                             !net->ipv6.devconf_all->forwarding && sllao)
2143                                 addr_flags = IFA_F_OPTIMISTIC;
2144 #endif
2145
2146                         /* Do not allow to create too much of autoconfigured
2147                          * addresses; this would be too easy way to crash kernel.
2148                          */
2149                         if (!max_addresses ||
2150                             ipv6_count_addresses(in6_dev) < max_addresses)
2151                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2152                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2153                                                     addr_flags);
2154
2155                         if (IS_ERR_OR_NULL(ifp)) {
2156                                 in6_dev_put(in6_dev);
2157                                 return;
2158                         }
2159
2160                         update_lft = create = 1;
2161                         ifp->cstamp = jiffies;
2162                         addrconf_dad_start(ifp);
2163                 }
2164
2165                 if (ifp) {
2166                         int flags;
2167                         unsigned long now;
2168 #ifdef CONFIG_IPV6_PRIVACY
2169                         struct inet6_ifaddr *ift;
2170 #endif
2171                         u32 stored_lft;
2172
2173                         /* update lifetime (RFC2462 5.5.3 e) */
2174                         spin_lock(&ifp->lock);
2175                         now = jiffies;
2176                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2177                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2178                         else
2179                                 stored_lft = 0;
2180                         if (!update_lft && stored_lft) {
2181                                 if (valid_lft > MIN_VALID_LIFETIME ||
2182                                     valid_lft > stored_lft)
2183                                         update_lft = 1;
2184                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2185                                         /* valid_lft <= stored_lft is always true */
2186                                         /*
2187                                          * RFC 4862 Section 5.5.3e:
2188                                          * "Note that the preferred lifetime of
2189                                          *  the corresponding address is always
2190                                          *  reset to the Preferred Lifetime in
2191                                          *  the received Prefix Information
2192                                          *  option, regardless of whether the
2193                                          *  valid lifetime is also reset or
2194                                          *  ignored."
2195                                          *
2196                                          *  So if the preferred lifetime in
2197                                          *  this advertisement is different
2198                                          *  than what we have stored, but the
2199                                          *  valid lifetime is invalid, just
2200                                          *  reset prefered_lft.
2201                                          *
2202                                          *  We must set the valid lifetime
2203                                          *  to the stored lifetime since we'll
2204                                          *  be updating the timestamp below,
2205                                          *  else we'll set it back to the
2206                                          *  minimum.
2207                                          */
2208                                         if (prefered_lft != ifp->prefered_lft) {
2209                                                 valid_lft = stored_lft;
2210                                                 update_lft = 1;
2211                                         }
2212                                 } else {
2213                                         valid_lft = MIN_VALID_LIFETIME;
2214                                         if (valid_lft < prefered_lft)
2215                                                 prefered_lft = valid_lft;
2216                                         update_lft = 1;
2217                                 }
2218                         }
2219
2220                         if (update_lft) {
2221                                 ifp->valid_lft = valid_lft;
2222                                 ifp->prefered_lft = prefered_lft;
2223                                 ifp->tstamp = now;
2224                                 flags = ifp->flags;
2225                                 ifp->flags &= ~IFA_F_DEPRECATED;
2226                                 spin_unlock(&ifp->lock);
2227
2228                                 if (!(flags&IFA_F_TENTATIVE))
2229                                         ipv6_ifa_notify(0, ifp);
2230                         } else
2231                                 spin_unlock(&ifp->lock);
2232
2233 #ifdef CONFIG_IPV6_PRIVACY
2234                         read_lock_bh(&in6_dev->lock);
2235                         /* update all temporary addresses in the list */
2236                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2237                                             tmp_list) {
2238                                 int age, max_valid, max_prefered;
2239
2240                                 if (ifp != ift->ifpub)
2241                                         continue;
2242
2243                                 /*
2244                                  * RFC 4941 section 3.3:
2245                                  * If a received option will extend the lifetime
2246                                  * of a public address, the lifetimes of
2247                                  * temporary addresses should be extended,
2248                                  * subject to the overall constraint that no
2249                                  * temporary addresses should ever remain
2250                                  * "valid" or "preferred" for a time longer than
2251                                  * (TEMP_VALID_LIFETIME) or
2252                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2253                                  * respectively.
2254                                  */
2255                                 age = (now - ift->cstamp) / HZ;
2256                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2257                                 if (max_valid < 0)
2258                                         max_valid = 0;
2259
2260                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2261                                                in6_dev->cnf.max_desync_factor -
2262                                                age;
2263                                 if (max_prefered < 0)
2264                                         max_prefered = 0;
2265
2266                                 if (valid_lft > max_valid)
2267                                         valid_lft = max_valid;
2268
2269                                 if (prefered_lft > max_prefered)
2270                                         prefered_lft = max_prefered;
2271
2272                                 spin_lock(&ift->lock);
2273                                 flags = ift->flags;
2274                                 ift->valid_lft = valid_lft;
2275                                 ift->prefered_lft = prefered_lft;
2276                                 ift->tstamp = now;
2277                                 if (prefered_lft > 0)
2278                                         ift->flags &= ~IFA_F_DEPRECATED;
2279
2280                                 spin_unlock(&ift->lock);
2281                                 if (!(flags&IFA_F_TENTATIVE))
2282                                         ipv6_ifa_notify(0, ift);
2283                         }
2284
2285                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2286                                 /*
2287                                  * When a new public address is created as
2288                                  * described in [ADDRCONF], also create a new
2289                                  * temporary address. Also create a temporary
2290                                  * address if it's enabled but no temporary
2291                                  * address currently exists.
2292                                  */
2293                                 read_unlock_bh(&in6_dev->lock);
2294                                 ipv6_create_tempaddr(ifp, NULL);
2295                         } else {
2296                                 read_unlock_bh(&in6_dev->lock);
2297                         }
2298 #endif
2299                         in6_ifa_put(ifp);
2300                         addrconf_verify(0);
2301                 }
2302         }
2303         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2304         in6_dev_put(in6_dev);
2305 }
2306
2307 /*
2308  *      Set destination address.
2309  *      Special case for SIT interfaces where we create a new "virtual"
2310  *      device.
2311  */
2312 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2313 {
2314         struct in6_ifreq ireq;
2315         struct net_device *dev;
2316         int err = -EINVAL;
2317
2318         rtnl_lock();
2319
2320         err = -EFAULT;
2321         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2322                 goto err_exit;
2323
2324         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2325
2326         err = -ENODEV;
2327         if (dev == NULL)
2328                 goto err_exit;
2329
2330 #if IS_ENABLED(CONFIG_IPV6_SIT)
2331         if (dev->type == ARPHRD_SIT) {
2332                 const struct net_device_ops *ops = dev->netdev_ops;
2333                 struct ifreq ifr;
2334                 struct ip_tunnel_parm p;
2335
2336                 err = -EADDRNOTAVAIL;
2337                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2338                         goto err_exit;
2339
2340                 memset(&p, 0, sizeof(p));
2341                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2342                 p.iph.saddr = 0;
2343                 p.iph.version = 4;
2344                 p.iph.ihl = 5;
2345                 p.iph.protocol = IPPROTO_IPV6;
2346                 p.iph.ttl = 64;
2347                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2348
2349                 if (ops->ndo_do_ioctl) {
2350                         mm_segment_t oldfs = get_fs();
2351
2352                         set_fs(KERNEL_DS);
2353                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2354                         set_fs(oldfs);
2355                 } else
2356                         err = -EOPNOTSUPP;
2357
2358                 if (err == 0) {
2359                         err = -ENOBUFS;
2360                         dev = __dev_get_by_name(net, p.name);
2361                         if (!dev)
2362                                 goto err_exit;
2363                         err = dev_open(dev);
2364                 }
2365         }
2366 #endif
2367
2368 err_exit:
2369         rtnl_unlock();
2370         return err;
2371 }
2372
2373 /*
2374  *      Manual configuration of address on an interface
2375  */
2376 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2377                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2378                           __u32 valid_lft)
2379 {
2380         struct inet6_ifaddr *ifp;
2381         struct inet6_dev *idev;
2382         struct net_device *dev;
2383         int scope;
2384         u32 flags;
2385         clock_t expires;
2386         unsigned long timeout;
2387
2388         ASSERT_RTNL();
2389
2390         if (plen > 128)
2391                 return -EINVAL;
2392
2393         /* check the lifetime */
2394         if (!valid_lft || prefered_lft > valid_lft)
2395                 return -EINVAL;
2396
2397         dev = __dev_get_by_index(net, ifindex);
2398         if (!dev)
2399                 return -ENODEV;
2400
2401         idev = addrconf_add_dev(dev);
2402         if (IS_ERR(idev))
2403                 return PTR_ERR(idev);
2404
2405         scope = ipv6_addr_scope(pfx);
2406
2407         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2408         if (addrconf_finite_timeout(timeout)) {
2409                 expires = jiffies_to_clock_t(timeout * HZ);
2410                 valid_lft = timeout;
2411                 flags = RTF_EXPIRES;
2412         } else {
2413                 expires = 0;
2414                 flags = 0;
2415                 ifa_flags |= IFA_F_PERMANENT;
2416         }
2417
2418         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2419         if (addrconf_finite_timeout(timeout)) {
2420                 if (timeout == 0)
2421                         ifa_flags |= IFA_F_DEPRECATED;
2422                 prefered_lft = timeout;
2423         }
2424
2425         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2426
2427         if (!IS_ERR(ifp)) {
2428                 spin_lock_bh(&ifp->lock);
2429                 ifp->valid_lft = valid_lft;
2430                 ifp->prefered_lft = prefered_lft;
2431                 ifp->tstamp = jiffies;
2432                 spin_unlock_bh(&ifp->lock);
2433
2434                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2435                                       expires, flags);
2436                 /*
2437                  * Note that section 3.1 of RFC 4429 indicates
2438                  * that the Optimistic flag should not be set for
2439                  * manually configured addresses
2440                  */
2441                 addrconf_dad_start(ifp);
2442                 in6_ifa_put(ifp);
2443                 addrconf_verify(0);
2444                 return 0;
2445         }
2446
2447         return PTR_ERR(ifp);
2448 }
2449
2450 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2451                           unsigned int plen)
2452 {
2453         struct inet6_ifaddr *ifp;
2454         struct inet6_dev *idev;
2455         struct net_device *dev;
2456
2457         if (plen > 128)
2458                 return -EINVAL;
2459
2460         dev = __dev_get_by_index(net, ifindex);
2461         if (!dev)
2462                 return -ENODEV;
2463
2464         if ((idev = __in6_dev_get(dev)) == NULL)
2465                 return -ENXIO;
2466
2467         read_lock_bh(&idev->lock);
2468         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2469                 if (ifp->prefix_len == plen &&
2470                     ipv6_addr_equal(pfx, &ifp->addr)) {
2471                         in6_ifa_hold(ifp);
2472                         read_unlock_bh(&idev->lock);
2473
2474                         ipv6_del_addr(ifp);
2475
2476                         /* If the last address is deleted administratively,
2477                            disable IPv6 on this interface.
2478                          */
2479                         if (list_empty(&idev->addr_list))
2480                                 addrconf_ifdown(idev->dev, 1);
2481                         return 0;
2482                 }
2483         }
2484         read_unlock_bh(&idev->lock);
2485         return -EADDRNOTAVAIL;
2486 }
2487
2488
2489 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2490 {
2491         struct in6_ifreq ireq;
2492         int err;
2493
2494         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2495                 return -EPERM;
2496
2497         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2498                 return -EFAULT;
2499
2500         rtnl_lock();
2501         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2502                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2503                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2504         rtnl_unlock();
2505         return err;
2506 }
2507
2508 int addrconf_del_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_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2521                              ireq.ifr6_prefixlen);
2522         rtnl_unlock();
2523         return err;
2524 }
2525
2526 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2527                      int plen, int scope)
2528 {
2529         struct inet6_ifaddr *ifp;
2530
2531         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2532         if (!IS_ERR(ifp)) {
2533                 spin_lock_bh(&ifp->lock);
2534                 ifp->flags &= ~IFA_F_TENTATIVE;
2535                 spin_unlock_bh(&ifp->lock);
2536                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2537                 in6_ifa_put(ifp);
2538         }
2539 }
2540
2541 #if IS_ENABLED(CONFIG_IPV6_SIT)
2542 static void sit_add_v4_addrs(struct inet6_dev *idev)
2543 {
2544         struct in6_addr addr;
2545         struct net_device *dev;
2546         struct net *net = dev_net(idev->dev);
2547         int scope;
2548
2549         ASSERT_RTNL();
2550
2551         memset(&addr, 0, sizeof(struct in6_addr));
2552         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2553
2554         if (idev->dev->flags&IFF_POINTOPOINT) {
2555                 addr.s6_addr32[0] = htonl(0xfe800000);
2556                 scope = IFA_LINK;
2557         } else {
2558                 scope = IPV6_ADDR_COMPATv4;
2559         }
2560
2561         if (addr.s6_addr32[3]) {
2562                 add_addr(idev, &addr, 128, scope);
2563                 return;
2564         }
2565
2566         for_each_netdev(net, dev) {
2567                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2568                 if (in_dev && (dev->flags & IFF_UP)) {
2569                         struct in_ifaddr *ifa;
2570
2571                         int flag = scope;
2572
2573                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2574                                 int plen;
2575
2576                                 addr.s6_addr32[3] = ifa->ifa_local;
2577
2578                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2579                                         continue;
2580                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2581                                         if (idev->dev->flags&IFF_POINTOPOINT)
2582                                                 continue;
2583                                         flag |= IFA_HOST;
2584                                 }
2585                                 if (idev->dev->flags&IFF_POINTOPOINT)
2586                                         plen = 64;
2587                                 else
2588                                         plen = 96;
2589
2590                                 add_addr(idev, &addr, plen, flag);
2591                         }
2592                 }
2593         }
2594 }
2595 #endif
2596
2597 static void init_loopback(struct net_device *dev)
2598 {
2599         struct inet6_dev  *idev;
2600
2601         /* ::1 */
2602
2603         ASSERT_RTNL();
2604
2605         if ((idev = ipv6_find_idev(dev)) == NULL) {
2606                 pr_debug("%s: add_dev failed\n", __func__);
2607                 return;
2608         }
2609
2610         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2611 }
2612
2613 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2614 {
2615         struct inet6_ifaddr *ifp;
2616         u32 addr_flags = IFA_F_PERMANENT;
2617
2618 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2619         if (idev->cnf.optimistic_dad &&
2620             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2621                 addr_flags |= IFA_F_OPTIMISTIC;
2622 #endif
2623
2624
2625         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2626         if (!IS_ERR(ifp)) {
2627                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2628                 addrconf_dad_start(ifp);
2629                 in6_ifa_put(ifp);
2630         }
2631 }
2632
2633 static void addrconf_dev_config(struct net_device *dev)
2634 {
2635         struct in6_addr addr;
2636         struct inet6_dev *idev;
2637
2638         ASSERT_RTNL();
2639
2640         if ((dev->type != ARPHRD_ETHER) &&
2641             (dev->type != ARPHRD_FDDI) &&
2642             (dev->type != ARPHRD_ARCNET) &&
2643             (dev->type != ARPHRD_INFINIBAND) &&
2644             (dev->type != ARPHRD_IEEE802154)) {
2645                 /* Alas, we support only Ethernet autoconfiguration. */
2646                 return;
2647         }
2648
2649         idev = addrconf_add_dev(dev);
2650         if (IS_ERR(idev))
2651                 return;
2652
2653         memset(&addr, 0, sizeof(struct in6_addr));
2654         addr.s6_addr32[0] = htonl(0xFE800000);
2655
2656         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2657                 addrconf_add_linklocal(idev, &addr);
2658 }
2659
2660 #if IS_ENABLED(CONFIG_IPV6_SIT)
2661 static void addrconf_sit_config(struct net_device *dev)
2662 {
2663         struct inet6_dev *idev;
2664
2665         ASSERT_RTNL();
2666
2667         /*
2668          * Configure the tunnel with one of our IPv4
2669          * addresses... we should configure all of
2670          * our v4 addrs in the tunnel
2671          */
2672
2673         if ((idev = ipv6_find_idev(dev)) == NULL) {
2674                 pr_debug("%s: add_dev failed\n", __func__);
2675                 return;
2676         }
2677
2678         if (dev->priv_flags & IFF_ISATAP) {
2679                 struct in6_addr addr;
2680
2681                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2682                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2683                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2684                         addrconf_add_linklocal(idev, &addr);
2685                 return;
2686         }
2687
2688         sit_add_v4_addrs(idev);
2689
2690         if (dev->flags&IFF_POINTOPOINT)
2691                 addrconf_add_mroute(dev);
2692         else
2693                 sit_route_add(dev);
2694 }
2695 #endif
2696
2697 #if IS_ENABLED(CONFIG_NET_IPGRE)
2698 static void addrconf_gre_config(struct net_device *dev)
2699 {
2700         struct inet6_dev *idev;
2701         struct in6_addr addr;
2702
2703         pr_info("%s(%s)\n", __func__, dev->name);
2704
2705         ASSERT_RTNL();
2706
2707         if ((idev = ipv6_find_idev(dev)) == NULL) {
2708                 pr_debug("%s: add_dev failed\n", __func__);
2709                 return;
2710         }
2711
2712         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2713         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2714
2715         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2716                 addrconf_add_linklocal(idev, &addr);
2717 }
2718 #endif
2719
2720 static inline int
2721 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2722 {
2723         struct in6_addr lladdr;
2724
2725         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2726                 addrconf_add_linklocal(idev, &lladdr);
2727                 return 0;
2728         }
2729         return -1;
2730 }
2731
2732 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2733 {
2734         struct net_device *link_dev;
2735         struct net *net = dev_net(idev->dev);
2736
2737         /* first try to inherit the link-local address from the link device */
2738         if (idev->dev->iflink &&
2739             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2740                 if (!ipv6_inherit_linklocal(idev, link_dev))
2741                         return;
2742         }
2743         /* then try to inherit it from any device */
2744         for_each_netdev(net, link_dev) {
2745                 if (!ipv6_inherit_linklocal(idev, link_dev))
2746                         return;
2747         }
2748         pr_debug("init ip6-ip6: add_linklocal failed\n");
2749 }
2750
2751 /*
2752  * Autoconfigure tunnel with a link-local address so routing protocols,
2753  * DHCPv6, MLD etc. can be run over the virtual link
2754  */
2755
2756 static void addrconf_ip6_tnl_config(struct net_device *dev)
2757 {
2758         struct inet6_dev *idev;
2759
2760         ASSERT_RTNL();
2761
2762         idev = addrconf_add_dev(dev);
2763         if (IS_ERR(idev)) {
2764                 pr_debug("init ip6-ip6: add_dev failed\n");
2765                 return;
2766         }
2767         ip6_tnl_add_linklocal(idev);
2768 }
2769
2770 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2771                            void *data)
2772 {
2773         struct net_device *dev = (struct net_device *) data;
2774         struct inet6_dev *idev = __in6_dev_get(dev);
2775         int run_pending = 0;
2776         int err;
2777
2778         switch (event) {
2779         case NETDEV_REGISTER:
2780                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2781                         idev = ipv6_add_dev(dev);
2782                         if (!idev)
2783                                 return notifier_from_errno(-ENOMEM);
2784                 }
2785                 break;
2786
2787         case NETDEV_UP:
2788         case NETDEV_CHANGE:
2789                 if (dev->flags & IFF_SLAVE)
2790                         break;
2791
2792                 if (event == NETDEV_UP) {
2793                         if (!addrconf_qdisc_ok(dev)) {
2794                                 /* device is not ready yet. */
2795                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2796                                         dev->name);
2797                                 break;
2798                         }
2799
2800                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2801                                 idev = ipv6_add_dev(dev);
2802
2803                         if (idev) {
2804                                 idev->if_flags |= IF_READY;
2805                                 run_pending = 1;
2806                         }
2807                 } else {
2808                         if (!addrconf_qdisc_ok(dev)) {
2809                                 /* device is still not ready. */
2810                                 break;
2811                         }
2812
2813                         if (idev) {
2814                                 if (idev->if_flags & IF_READY)
2815                                         /* device is already configured. */
2816                                         break;
2817                                 idev->if_flags |= IF_READY;
2818                         }
2819
2820                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2821                                 dev->name);
2822
2823                         run_pending = 1;
2824                 }
2825
2826                 switch (dev->type) {
2827 #if IS_ENABLED(CONFIG_IPV6_SIT)
2828                 case ARPHRD_SIT:
2829                         addrconf_sit_config(dev);
2830                         break;
2831 #endif
2832 #if IS_ENABLED(CONFIG_NET_IPGRE)
2833                 case ARPHRD_IPGRE:
2834                         addrconf_gre_config(dev);
2835                         break;
2836 #endif
2837                 case ARPHRD_TUNNEL6:
2838                         addrconf_ip6_tnl_config(dev);
2839                         break;
2840                 case ARPHRD_LOOPBACK:
2841                         init_loopback(dev);
2842                         break;
2843
2844                 default:
2845                         addrconf_dev_config(dev);
2846                         break;
2847                 }
2848
2849                 if (idev) {
2850                         if (run_pending)
2851                                 addrconf_dad_run(idev);
2852
2853                         /*
2854                          * If the MTU changed during the interface down,
2855                          * when the interface up, the changed MTU must be
2856                          * reflected in the idev as well as routers.
2857                          */
2858                         if (idev->cnf.mtu6 != dev->mtu &&
2859                             dev->mtu >= IPV6_MIN_MTU) {
2860                                 rt6_mtu_change(dev, dev->mtu);
2861                                 idev->cnf.mtu6 = dev->mtu;
2862                         }
2863                         idev->tstamp = jiffies;
2864                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2865
2866                         /*
2867                          * If the changed mtu during down is lower than
2868                          * IPV6_MIN_MTU stop IPv6 on this interface.
2869                          */
2870                         if (dev->mtu < IPV6_MIN_MTU)
2871                                 addrconf_ifdown(dev, 1);
2872                 }
2873                 break;
2874
2875         case NETDEV_CHANGEMTU:
2876                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2877                         rt6_mtu_change(dev, dev->mtu);
2878                         idev->cnf.mtu6 = dev->mtu;
2879                         break;
2880                 }
2881
2882                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2883                         idev = ipv6_add_dev(dev);
2884                         if (idev)
2885                                 break;
2886                 }
2887
2888                 /*
2889                  * MTU falled under IPV6_MIN_MTU.
2890                  * Stop IPv6 on this interface.
2891                  */
2892
2893         case NETDEV_DOWN:
2894         case NETDEV_UNREGISTER:
2895                 /*
2896                  *      Remove all addresses from this interface.
2897                  */
2898                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2899                 break;
2900
2901         case NETDEV_CHANGENAME:
2902                 if (idev) {
2903                         snmp6_unregister_dev(idev);
2904                         addrconf_sysctl_unregister(idev);
2905                         addrconf_sysctl_register(idev);
2906                         err = snmp6_register_dev(idev);
2907                         if (err)
2908                                 return notifier_from_errno(err);
2909                 }
2910                 break;
2911
2912         case NETDEV_PRE_TYPE_CHANGE:
2913         case NETDEV_POST_TYPE_CHANGE:
2914                 addrconf_type_change(dev, event);
2915                 break;
2916         }
2917
2918         return NOTIFY_OK;
2919 }
2920
2921 /*
2922  *      addrconf module should be notified of a device going up
2923  */
2924 static struct notifier_block ipv6_dev_notf = {
2925         .notifier_call = addrconf_notify,
2926 };
2927
2928 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2929 {
2930         struct inet6_dev *idev;
2931         ASSERT_RTNL();
2932
2933         idev = __in6_dev_get(dev);
2934
2935         if (event == NETDEV_POST_TYPE_CHANGE)
2936                 ipv6_mc_remap(idev);
2937         else if (event == NETDEV_PRE_TYPE_CHANGE)
2938                 ipv6_mc_unmap(idev);
2939 }
2940
2941 static int addrconf_ifdown(struct net_device *dev, int how)
2942 {
2943         struct net *net = dev_net(dev);
2944         struct inet6_dev *idev;
2945         struct inet6_ifaddr *ifa;
2946         int state, i;
2947
2948         ASSERT_RTNL();
2949
2950         rt6_ifdown(net, dev);
2951         neigh_ifdown(&nd_tbl, dev);
2952
2953         idev = __in6_dev_get(dev);
2954         if (idev == NULL)
2955                 return -ENODEV;
2956
2957         /*
2958          * Step 1: remove reference to ipv6 device from parent device.
2959          *         Do not dev_put!
2960          */
2961         if (how) {
2962                 idev->dead = 1;
2963
2964                 /* protected by rtnl_lock */
2965                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2966
2967                 /* Step 1.5: remove snmp6 entry */
2968                 snmp6_unregister_dev(idev);
2969
2970         }
2971
2972         /* Step 2: clear hash table */
2973         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2974                 struct hlist_head *h = &inet6_addr_lst[i];
2975
2976                 spin_lock_bh(&addrconf_hash_lock);
2977         restart:
2978                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2979                         if (ifa->idev == idev) {
2980                                 hlist_del_init_rcu(&ifa->addr_lst);
2981                                 addrconf_del_timer(ifa);
2982                                 goto restart;
2983                         }
2984                 }
2985                 spin_unlock_bh(&addrconf_hash_lock);
2986         }
2987
2988         write_lock_bh(&idev->lock);
2989
2990         /* Step 2: clear flags for stateless addrconf */
2991         if (!how)
2992                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2993
2994 #ifdef CONFIG_IPV6_PRIVACY
2995         if (how && del_timer(&idev->regen_timer))
2996                 in6_dev_put(idev);
2997
2998         /* Step 3: clear tempaddr list */
2999         while (!list_empty(&idev->tempaddr_list)) {
3000                 ifa = list_first_entry(&idev->tempaddr_list,
3001                                        struct inet6_ifaddr, tmp_list);
3002                 list_del(&ifa->tmp_list);
3003                 write_unlock_bh(&idev->lock);
3004                 spin_lock_bh(&ifa->lock);
3005
3006                 if (ifa->ifpub) {
3007                         in6_ifa_put(ifa->ifpub);
3008                         ifa->ifpub = NULL;
3009                 }
3010                 spin_unlock_bh(&ifa->lock);
3011                 in6_ifa_put(ifa);
3012                 write_lock_bh(&idev->lock);
3013         }
3014 #endif
3015
3016         while (!list_empty(&idev->addr_list)) {
3017                 ifa = list_first_entry(&idev->addr_list,
3018                                        struct inet6_ifaddr, if_list);
3019                 addrconf_del_timer(ifa);
3020
3021                 list_del(&ifa->if_list);
3022
3023                 write_unlock_bh(&idev->lock);
3024
3025                 spin_lock_bh(&ifa->state_lock);
3026                 state = ifa->state;
3027                 ifa->state = INET6_IFADDR_STATE_DEAD;
3028                 spin_unlock_bh(&ifa->state_lock);
3029
3030                 if (state != INET6_IFADDR_STATE_DEAD) {
3031                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3032                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
3033                 }
3034                 in6_ifa_put(ifa);
3035
3036                 write_lock_bh(&idev->lock);
3037         }
3038
3039         write_unlock_bh(&idev->lock);
3040
3041         /* Step 5: Discard multicast list */
3042         if (how)
3043                 ipv6_mc_destroy_dev(idev);
3044         else
3045                 ipv6_mc_down(idev);
3046
3047         idev->tstamp = jiffies;
3048
3049         /* Last: Shot the device (if unregistered) */
3050         if (how) {
3051                 addrconf_sysctl_unregister(idev);
3052                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3053                 neigh_ifdown(&nd_tbl, dev);
3054                 in6_dev_put(idev);
3055         }
3056         return 0;
3057 }
3058
3059 static void addrconf_rs_timer(unsigned long data)
3060 {
3061         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3062         struct inet6_dev *idev = ifp->idev;
3063
3064         read_lock(&idev->lock);
3065         if (idev->dead || !(idev->if_flags & IF_READY))
3066                 goto out;
3067
3068         if (!ipv6_accept_ra(idev))
3069                 goto out;
3070
3071         /* Announcement received after solicitation was sent */
3072         if (idev->if_flags & IF_RA_RCVD)
3073                 goto out;
3074
3075         spin_lock(&ifp->lock);
3076         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3077                 /* The wait after the last probe can be shorter */
3078                 addrconf_mod_timer(ifp, AC_RS,
3079                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3080                                    idev->cnf.rtr_solicit_delay :
3081                                    idev->cnf.rtr_solicit_interval);
3082                 spin_unlock(&ifp->lock);
3083
3084                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3085         } else {
3086                 spin_unlock(&ifp->lock);
3087                 /*
3088                  * Note: we do not support deprecated "all on-link"
3089                  * assumption any longer.
3090                  */
3091                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3092         }
3093
3094 out:
3095         read_unlock(&idev->lock);
3096         in6_ifa_put(ifp);
3097 }
3098
3099 /*
3100  *      Duplicate Address Detection
3101  */
3102 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3103 {
3104         unsigned long rand_num;
3105         struct inet6_dev *idev = ifp->idev;
3106
3107         if (ifp->flags & IFA_F_OPTIMISTIC)
3108                 rand_num = 0;
3109         else
3110                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3111
3112         ifp->probes = idev->cnf.dad_transmits;
3113         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3114 }
3115
3116 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3117 {
3118         struct inet6_dev *idev = ifp->idev;
3119         struct net_device *dev = idev->dev;
3120
3121         addrconf_join_solict(dev, &ifp->addr);
3122
3123         net_srandom(ifp->addr.s6_addr32[3]);
3124
3125         read_lock_bh(&idev->lock);
3126         spin_lock(&ifp->lock);
3127         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3128                 goto out;
3129
3130         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3131             idev->cnf.accept_dad < 1 ||
3132             !(ifp->flags&IFA_F_TENTATIVE) ||
3133             ifp->flags & IFA_F_NODAD) {
3134                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3135                 spin_unlock(&ifp->lock);
3136                 read_unlock_bh(&idev->lock);
3137
3138                 addrconf_dad_completed(ifp);
3139                 return;
3140         }
3141
3142         if (!(idev->if_flags & IF_READY)) {
3143                 spin_unlock(&ifp->lock);
3144                 read_unlock_bh(&idev->lock);
3145                 /*
3146                  * If the device is not ready:
3147                  * - keep it tentative if it is a permanent address.
3148                  * - otherwise, kill it.
3149                  */
3150                 in6_ifa_hold(ifp);
3151                 addrconf_dad_stop(ifp, 0);
3152                 return;
3153         }
3154
3155         /*
3156          * Optimistic nodes can start receiving
3157          * Frames right away
3158          */
3159         if (ifp->flags & IFA_F_OPTIMISTIC)
3160                 ip6_ins_rt(ifp->rt);
3161
3162         addrconf_dad_kick(ifp);
3163 out:
3164         spin_unlock(&ifp->lock);
3165         read_unlock_bh(&idev->lock);
3166 }
3167
3168 static void addrconf_dad_timer(unsigned long data)
3169 {
3170         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3171         struct inet6_dev *idev = ifp->idev;
3172         struct in6_addr mcaddr;
3173
3174         if (!ifp->probes && addrconf_dad_end(ifp))
3175                 goto out;
3176
3177         read_lock(&idev->lock);
3178         if (idev->dead || !(idev->if_flags & IF_READY)) {
3179                 read_unlock(&idev->lock);
3180                 goto out;
3181         }
3182
3183         spin_lock(&ifp->lock);
3184         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3185                 spin_unlock(&ifp->lock);
3186                 read_unlock(&idev->lock);
3187                 goto out;
3188         }
3189
3190         if (ifp->probes == 0) {
3191                 /*
3192                  * DAD was successful
3193                  */
3194
3195                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3196                 spin_unlock(&ifp->lock);
3197                 read_unlock(&idev->lock);
3198
3199                 addrconf_dad_completed(ifp);
3200
3201                 goto out;
3202         }
3203
3204         ifp->probes--;
3205         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3206         spin_unlock(&ifp->lock);
3207         read_unlock(&idev->lock);
3208
3209         /* send a neighbour solicitation for our addr */
3210         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3211         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3212 out:
3213         in6_ifa_put(ifp);
3214 }
3215
3216 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3217 {
3218         struct net_device *dev = ifp->idev->dev;
3219
3220         /*
3221          *      Configure the address for reception. Now it is valid.
3222          */
3223
3224         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3225
3226         /* If added prefix is link local and we are prepared to process
3227            router advertisements, start sending router solicitations.
3228          */
3229
3230         if (ipv6_accept_ra(ifp->idev) &&
3231             ifp->idev->cnf.rtr_solicits > 0 &&
3232             (dev->flags&IFF_LOOPBACK) == 0 &&
3233             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3234                 /*
3235                  *      If a host as already performed a random delay
3236                  *      [...] as part of DAD [...] there is no need
3237                  *      to delay again before sending the first RS
3238                  */
3239                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3240
3241                 spin_lock_bh(&ifp->lock);
3242                 ifp->probes = 1;
3243                 ifp->idev->if_flags |= IF_RS_SENT;
3244                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3245                 spin_unlock_bh(&ifp->lock);
3246         }
3247 }
3248
3249 static void addrconf_dad_run(struct inet6_dev *idev)
3250 {
3251         struct inet6_ifaddr *ifp;
3252
3253         read_lock_bh(&idev->lock);
3254         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3255                 spin_lock(&ifp->lock);
3256                 if (ifp->flags & IFA_F_TENTATIVE &&
3257                     ifp->state == INET6_IFADDR_STATE_DAD)
3258                         addrconf_dad_kick(ifp);
3259                 spin_unlock(&ifp->lock);
3260         }
3261         read_unlock_bh(&idev->lock);
3262 }
3263
3264 #ifdef CONFIG_PROC_FS
3265 struct if6_iter_state {
3266         struct seq_net_private p;
3267         int bucket;
3268         int offset;
3269 };
3270
3271 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3272 {
3273         struct inet6_ifaddr *ifa = NULL;
3274         struct if6_iter_state *state = seq->private;
3275         struct net *net = seq_file_net(seq);
3276         int p = 0;
3277
3278         /* initial bucket if pos is 0 */
3279         if (pos == 0) {
3280                 state->bucket = 0;
3281                 state->offset = 0;
3282         }
3283
3284         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3285                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3286                                          addr_lst) {
3287                         if (!net_eq(dev_net(ifa->idev->dev), net))
3288                                 continue;
3289                         /* sync with offset */
3290                         if (p < state->offset) {
3291                                 p++;
3292                                 continue;
3293                         }
3294                         state->offset++;
3295                         return ifa;
3296                 }
3297
3298                 /* prepare for next bucket */
3299                 state->offset = 0;
3300                 p = 0;
3301         }
3302         return NULL;
3303 }
3304
3305 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3306                                          struct inet6_ifaddr *ifa)
3307 {
3308         struct if6_iter_state *state = seq->private;
3309         struct net *net = seq_file_net(seq);
3310
3311         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3312                 if (!net_eq(dev_net(ifa->idev->dev), net))
3313                         continue;
3314                 state->offset++;
3315                 return ifa;
3316         }
3317
3318         while (++state->bucket < IN6_ADDR_HSIZE) {
3319                 state->offset = 0;
3320                 hlist_for_each_entry_rcu_bh(ifa,
3321                                      &inet6_addr_lst[state->bucket], addr_lst) {
3322                         if (!net_eq(dev_net(ifa->idev->dev), net))
3323                                 continue;
3324                         state->offset++;
3325                         return ifa;
3326                 }
3327         }
3328
3329         return NULL;
3330 }
3331
3332 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3333         __acquires(rcu_bh)
3334 {
3335         rcu_read_lock_bh();
3336         return if6_get_first(seq, *pos);
3337 }
3338
3339 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3340 {
3341         struct inet6_ifaddr *ifa;
3342
3343         ifa = if6_get_next(seq, v);
3344         ++*pos;
3345         return ifa;
3346 }
3347
3348 static void if6_seq_stop(struct seq_file *seq, void *v)
3349         __releases(rcu_bh)
3350 {
3351         rcu_read_unlock_bh();
3352 }
3353
3354 static int if6_seq_show(struct seq_file *seq, void *v)
3355 {
3356         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3357         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3358                    &ifp->addr,
3359                    ifp->idev->dev->ifindex,
3360                    ifp->prefix_len,
3361                    ifp->scope,
3362                    ifp->flags,
3363                    ifp->idev->dev->name);
3364         return 0;
3365 }
3366
3367 static const struct seq_operations if6_seq_ops = {
3368         .start  = if6_seq_start,
3369         .next   = if6_seq_next,
3370         .show   = if6_seq_show,
3371         .stop   = if6_seq_stop,
3372 };
3373
3374 static int if6_seq_open(struct inode *inode, struct file *file)
3375 {
3376         return seq_open_net(inode, file, &if6_seq_ops,
3377                             sizeof(struct if6_iter_state));
3378 }
3379
3380 static const struct file_operations if6_fops = {
3381         .owner          = THIS_MODULE,
3382         .open           = if6_seq_open,
3383         .read           = seq_read,
3384         .llseek         = seq_lseek,
3385         .release        = seq_release_net,
3386 };
3387
3388 static int __net_init if6_proc_net_init(struct net *net)
3389 {
3390         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3391                 return -ENOMEM;
3392         return 0;
3393 }
3394
3395 static void __net_exit if6_proc_net_exit(struct net *net)
3396 {
3397         remove_proc_entry("if_inet6", net->proc_net);
3398 }
3399
3400 static struct pernet_operations if6_proc_net_ops = {
3401        .init = if6_proc_net_init,
3402        .exit = if6_proc_net_exit,
3403 };
3404
3405 int __init if6_proc_init(void)
3406 {
3407         return register_pernet_subsys(&if6_proc_net_ops);
3408 }
3409
3410 void if6_proc_exit(void)
3411 {
3412         unregister_pernet_subsys(&if6_proc_net_ops);
3413 }
3414 #endif  /* CONFIG_PROC_FS */
3415
3416 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3417 /* Check if address is a home address configured on any interface. */
3418 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3419 {
3420         int ret = 0;
3421         struct inet6_ifaddr *ifp = NULL;
3422         unsigned int hash = inet6_addr_hash(addr);
3423
3424         rcu_read_lock_bh();
3425         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3426                 if (!net_eq(dev_net(ifp->idev->dev), net))
3427                         continue;
3428                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3429                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3430                         ret = 1;
3431                         break;
3432                 }
3433         }
3434         rcu_read_unlock_bh();
3435         return ret;
3436 }
3437 #endif
3438
3439 /*
3440  *      Periodic address status verification
3441  */
3442
3443 static void addrconf_verify(unsigned long foo)
3444 {
3445         unsigned long now, next, next_sec, next_sched;
3446         struct inet6_ifaddr *ifp;
3447         int i;
3448
3449         rcu_read_lock_bh();
3450         spin_lock(&addrconf_verify_lock);
3451         now = jiffies;
3452         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3453
3454         del_timer(&addr_chk_timer);
3455
3456         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3457 restart:
3458                 hlist_for_each_entry_rcu_bh(ifp,
3459                                          &inet6_addr_lst[i], addr_lst) {
3460                         unsigned long age;
3461
3462                         if (ifp->flags & IFA_F_PERMANENT)
3463                                 continue;
3464
3465                         spin_lock(&ifp->lock);
3466                         /* We try to batch several events at once. */
3467                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3468
3469                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3470                             age >= ifp->valid_lft) {
3471                                 spin_unlock(&ifp->lock);
3472                                 in6_ifa_hold(ifp);
3473                                 ipv6_del_addr(ifp);
3474                                 goto restart;
3475                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3476                                 spin_unlock(&ifp->lock);
3477                                 continue;
3478                         } else if (age >= ifp->prefered_lft) {
3479                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3480                                 int deprecate = 0;
3481
3482                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3483                                         deprecate = 1;
3484                                         ifp->flags |= IFA_F_DEPRECATED;
3485                                 }
3486
3487                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3488                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3489
3490                                 spin_unlock(&ifp->lock);
3491
3492                                 if (deprecate) {
3493                                         in6_ifa_hold(ifp);
3494
3495                                         ipv6_ifa_notify(0, ifp);
3496                                         in6_ifa_put(ifp);
3497                                         goto restart;
3498                                 }
3499 #ifdef CONFIG_IPV6_PRIVACY
3500                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3501                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3502                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3503                                         ifp->idev->cnf.dad_transmits *
3504                                         ifp->idev->nd_parms->retrans_time / HZ;
3505
3506                                 if (age >= ifp->prefered_lft - regen_advance) {
3507                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3508                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3509                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3510                                         if (!ifp->regen_count && ifpub) {
3511                                                 ifp->regen_count++;
3512                                                 in6_ifa_hold(ifp);
3513                                                 in6_ifa_hold(ifpub);
3514                                                 spin_unlock(&ifp->lock);
3515
3516                                                 spin_lock(&ifpub->lock);
3517                                                 ifpub->regen_count = 0;
3518                                                 spin_unlock(&ifpub->lock);
3519                                                 ipv6_create_tempaddr(ifpub, ifp);
3520                                                 in6_ifa_put(ifpub);
3521                                                 in6_ifa_put(ifp);
3522                                                 goto restart;
3523                                         }
3524                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3525                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3526                                 spin_unlock(&ifp->lock);
3527 #endif
3528                         } else {
3529                                 /* ifp->prefered_lft <= ifp->valid_lft */
3530                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3531                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3532                                 spin_unlock(&ifp->lock);
3533                         }
3534                 }
3535         }
3536
3537         next_sec = round_jiffies_up(next);
3538         next_sched = next;
3539
3540         /* If rounded timeout is accurate enough, accept it. */
3541         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3542                 next_sched = next_sec;
3543
3544         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3545         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3546                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3547
3548         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3549               now, next, next_sec, next_sched));
3550
3551         addr_chk_timer.expires = next_sched;
3552         add_timer(&addr_chk_timer);
3553         spin_unlock(&addrconf_verify_lock);
3554         rcu_read_unlock_bh();
3555 }
3556
3557 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3558 {
3559         struct in6_addr *pfx = NULL;
3560
3561         if (addr)
3562                 pfx = nla_data(addr);
3563
3564         if (local) {
3565                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3566                         pfx = NULL;
3567                 else
3568                         pfx = nla_data(local);
3569         }
3570
3571         return pfx;
3572 }
3573
3574 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3575         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3576         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3577         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3578 };
3579
3580 static int
3581 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3582 {
3583         struct net *net = sock_net(skb->sk);
3584         struct ifaddrmsg *ifm;
3585         struct nlattr *tb[IFA_MAX+1];
3586         struct in6_addr *pfx;
3587         int err;
3588
3589         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3590         if (err < 0)
3591                 return err;
3592
3593         ifm = nlmsg_data(nlh);
3594         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3595         if (pfx == NULL)
3596                 return -EINVAL;
3597
3598         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3599 }
3600
3601 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3602                              u32 prefered_lft, u32 valid_lft)
3603 {
3604         u32 flags;
3605         clock_t expires;
3606         unsigned long timeout;
3607
3608         if (!valid_lft || (prefered_lft > valid_lft))
3609                 return -EINVAL;
3610
3611         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3612         if (addrconf_finite_timeout(timeout)) {
3613                 expires = jiffies_to_clock_t(timeout * HZ);
3614                 valid_lft = timeout;
3615                 flags = RTF_EXPIRES;
3616         } else {
3617                 expires = 0;
3618                 flags = 0;
3619                 ifa_flags |= IFA_F_PERMANENT;
3620         }
3621
3622         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3623         if (addrconf_finite_timeout(timeout)) {
3624                 if (timeout == 0)
3625                         ifa_flags |= IFA_F_DEPRECATED;
3626                 prefered_lft = timeout;
3627         }
3628
3629         spin_lock_bh(&ifp->lock);
3630         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3631         ifp->tstamp = jiffies;
3632         ifp->valid_lft = valid_lft;
3633         ifp->prefered_lft = prefered_lft;
3634
3635         spin_unlock_bh(&ifp->lock);
3636         if (!(ifp->flags&IFA_F_TENTATIVE))
3637                 ipv6_ifa_notify(0, ifp);
3638
3639         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3640                               expires, flags);
3641         addrconf_verify(0);
3642
3643         return 0;
3644 }
3645
3646 static int
3647 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3648 {
3649         struct net *net = sock_net(skb->sk);
3650         struct ifaddrmsg *ifm;
3651         struct nlattr *tb[IFA_MAX+1];
3652         struct in6_addr *pfx;
3653         struct inet6_ifaddr *ifa;
3654         struct net_device *dev;
3655         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3656         u8 ifa_flags;
3657         int err;
3658
3659         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3660         if (err < 0)
3661                 return err;
3662
3663         ifm = nlmsg_data(nlh);
3664         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3665         if (pfx == NULL)
3666                 return -EINVAL;
3667
3668         if (tb[IFA_CACHEINFO]) {
3669                 struct ifa_cacheinfo *ci;
3670
3671                 ci = nla_data(tb[IFA_CACHEINFO]);
3672                 valid_lft = ci->ifa_valid;
3673                 preferred_lft = ci->ifa_prefered;
3674         } else {
3675                 preferred_lft = INFINITY_LIFE_TIME;
3676                 valid_lft = INFINITY_LIFE_TIME;
3677         }
3678
3679         dev =  __dev_get_by_index(net, ifm->ifa_index);
3680         if (dev == NULL)
3681                 return -ENODEV;
3682
3683         /* We ignore other flags so far. */
3684         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3685
3686         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3687         if (ifa == NULL) {
3688                 /*
3689                  * It would be best to check for !NLM_F_CREATE here but
3690                  * userspace alreay relies on not having to provide this.
3691                  */
3692                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3693                                       ifm->ifa_prefixlen, ifa_flags,
3694                                       preferred_lft, valid_lft);
3695         }
3696
3697         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3698             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3699                 err = -EEXIST;
3700         else
3701                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3702
3703         in6_ifa_put(ifa);
3704
3705         return err;
3706 }
3707
3708 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3709                           u8 scope, int ifindex)
3710 {
3711         struct ifaddrmsg *ifm;
3712
3713         ifm = nlmsg_data(nlh);
3714         ifm->ifa_family = AF_INET6;
3715         ifm->ifa_prefixlen = prefixlen;
3716         ifm->ifa_flags = flags;
3717         ifm->ifa_scope = scope;
3718         ifm->ifa_index = ifindex;
3719 }
3720
3721 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3722                          unsigned long tstamp, u32 preferred, u32 valid)
3723 {
3724         struct ifa_cacheinfo ci;
3725
3726         ci.cstamp = cstamp_delta(cstamp);
3727         ci.tstamp = cstamp_delta(tstamp);
3728         ci.ifa_prefered = preferred;
3729         ci.ifa_valid = valid;
3730
3731         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3732 }
3733
3734 static inline int rt_scope(int ifa_scope)
3735 {
3736         if (ifa_scope & IFA_HOST)
3737                 return RT_SCOPE_HOST;
3738         else if (ifa_scope & IFA_LINK)
3739                 return RT_SCOPE_LINK;
3740         else if (ifa_scope & IFA_SITE)
3741                 return RT_SCOPE_SITE;
3742         else
3743                 return RT_SCOPE_UNIVERSE;
3744 }
3745
3746 static inline int inet6_ifaddr_msgsize(void)
3747 {
3748         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3749                + nla_total_size(16) /* IFA_ADDRESS */
3750                + nla_total_size(sizeof(struct ifa_cacheinfo));
3751 }
3752
3753 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3754                              u32 portid, u32 seq, int event, unsigned int flags)
3755 {
3756         struct nlmsghdr  *nlh;
3757         u32 preferred, valid;
3758
3759         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3760         if (nlh == NULL)
3761                 return -EMSGSIZE;
3762
3763         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3764                       ifa->idev->dev->ifindex);
3765
3766         if (!(ifa->flags&IFA_F_PERMANENT)) {
3767                 preferred = ifa->prefered_lft;
3768                 valid = ifa->valid_lft;
3769                 if (preferred != INFINITY_LIFE_TIME) {
3770                         long tval = (jiffies - ifa->tstamp)/HZ;
3771                         if (preferred > tval)
3772                                 preferred -= tval;
3773                         else
3774                                 preferred = 0;
3775                         if (valid != INFINITY_LIFE_TIME) {
3776                                 if (valid > tval)
3777                                         valid -= tval;
3778                                 else
3779                                         valid = 0;
3780                         }
3781                 }
3782         } else {
3783                 preferred = INFINITY_LIFE_TIME;
3784                 valid = INFINITY_LIFE_TIME;
3785         }
3786
3787         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3788             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3789                 nlmsg_cancel(skb, nlh);
3790                 return -EMSGSIZE;
3791         }
3792
3793         return nlmsg_end(skb, nlh);
3794 }
3795
3796 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3797                                 u32 portid, u32 seq, int event, u16 flags)
3798 {
3799         struct nlmsghdr  *nlh;
3800         u8 scope = RT_SCOPE_UNIVERSE;
3801         int ifindex = ifmca->idev->dev->ifindex;
3802
3803         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3804                 scope = RT_SCOPE_SITE;
3805
3806         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3807         if (nlh == NULL)
3808                 return -EMSGSIZE;
3809
3810         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3811         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3812             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3813                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3814                 nlmsg_cancel(skb, nlh);
3815                 return -EMSGSIZE;
3816         }
3817
3818         return nlmsg_end(skb, nlh);
3819 }
3820
3821 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3822                                 u32 portid, u32 seq, int event, unsigned int flags)
3823 {
3824         struct nlmsghdr  *nlh;
3825         u8 scope = RT_SCOPE_UNIVERSE;
3826         int ifindex = ifaca->aca_idev->dev->ifindex;
3827
3828         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3829                 scope = RT_SCOPE_SITE;
3830
3831         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3832         if (nlh == NULL)
3833                 return -EMSGSIZE;
3834
3835         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3836         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3837             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3838                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3839                 nlmsg_cancel(skb, nlh);
3840                 return -EMSGSIZE;
3841         }
3842
3843         return nlmsg_end(skb, nlh);
3844 }
3845
3846 enum addr_type_t {
3847         UNICAST_ADDR,
3848         MULTICAST_ADDR,
3849         ANYCAST_ADDR,
3850 };
3851
3852 /* called with rcu_read_lock() */
3853 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3854                           struct netlink_callback *cb, enum addr_type_t type,
3855                           int s_ip_idx, int *p_ip_idx)
3856 {
3857         struct ifmcaddr6 *ifmca;
3858         struct ifacaddr6 *ifaca;
3859         int err = 1;
3860         int ip_idx = *p_ip_idx;
3861
3862         read_lock_bh(&idev->lock);
3863         switch (type) {
3864         case UNICAST_ADDR: {
3865                 struct inet6_ifaddr *ifa;
3866
3867                 /* unicast address incl. temp addr */
3868                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3869                         if (++ip_idx < s_ip_idx)
3870                                 continue;
3871                         err = inet6_fill_ifaddr(skb, ifa,
3872                                                 NETLINK_CB(cb->skb).portid,
3873                                                 cb->nlh->nlmsg_seq,
3874                                                 RTM_NEWADDR,
3875                                                 NLM_F_MULTI);
3876                         if (err <= 0)
3877                                 break;
3878                 }
3879                 break;
3880         }
3881         case MULTICAST_ADDR:
3882                 /* multicast address */
3883                 for (ifmca = idev->mc_list; ifmca;
3884                      ifmca = ifmca->next, ip_idx++) {
3885                         if (ip_idx < s_ip_idx)
3886                                 continue;
3887                         err = inet6_fill_ifmcaddr(skb, ifmca,
3888                                                   NETLINK_CB(cb->skb).portid,
3889                                                   cb->nlh->nlmsg_seq,
3890                                                   RTM_GETMULTICAST,
3891                                                   NLM_F_MULTI);
3892                         if (err <= 0)
3893                                 break;
3894                 }
3895                 break;
3896         case ANYCAST_ADDR:
3897                 /* anycast address */
3898                 for (ifaca = idev->ac_list; ifaca;
3899                      ifaca = ifaca->aca_next, ip_idx++) {
3900                         if (ip_idx < s_ip_idx)
3901                                 continue;
3902                         err = inet6_fill_ifacaddr(skb, ifaca,
3903                                                   NETLINK_CB(cb->skb).portid,
3904                                                   cb->nlh->nlmsg_seq,
3905                                                   RTM_GETANYCAST,
3906                                                   NLM_F_MULTI);
3907                         if (err <= 0)
3908                                 break;
3909                 }
3910                 break;
3911         default:
3912                 break;
3913         }
3914         read_unlock_bh(&idev->lock);
3915         *p_ip_idx = ip_idx;
3916         return err;
3917 }
3918
3919 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3920                            enum addr_type_t type)
3921 {
3922         struct net *net = sock_net(skb->sk);
3923         int h, s_h;
3924         int idx, ip_idx;
3925         int s_idx, s_ip_idx;
3926         struct net_device *dev;
3927         struct inet6_dev *idev;
3928         struct hlist_head *head;
3929
3930         s_h = cb->args[0];
3931         s_idx = idx = cb->args[1];
3932         s_ip_idx = ip_idx = cb->args[2];
3933
3934         rcu_read_lock();
3935         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3936                 idx = 0;
3937                 head = &net->dev_index_head[h];
3938                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3939                         if (idx < s_idx)
3940                                 goto cont;
3941                         if (h > s_h || idx > s_idx)
3942                                 s_ip_idx = 0;
3943                         ip_idx = 0;
3944                         idev = __in6_dev_get(dev);
3945                         if (!idev)
3946                                 goto cont;
3947
3948                         if (in6_dump_addrs(idev, skb, cb, type,
3949                                            s_ip_idx, &ip_idx) <= 0)
3950                                 goto done;
3951 cont:
3952                         idx++;
3953                 }
3954         }
3955 done:
3956         rcu_read_unlock();
3957         cb->args[0] = h;
3958         cb->args[1] = idx;
3959         cb->args[2] = ip_idx;
3960
3961         return skb->len;
3962 }
3963
3964 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3965 {
3966         enum addr_type_t type = UNICAST_ADDR;
3967
3968         return inet6_dump_addr(skb, cb, type);
3969 }
3970
3971 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3972 {
3973         enum addr_type_t type = MULTICAST_ADDR;
3974
3975         return inet6_dump_addr(skb, cb, type);
3976 }
3977
3978
3979 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3980 {
3981         enum addr_type_t type = ANYCAST_ADDR;
3982
3983         return inet6_dump_addr(skb, cb, type);
3984 }
3985
3986 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3987                              void *arg)
3988 {
3989         struct net *net = sock_net(in_skb->sk);
3990         struct ifaddrmsg *ifm;
3991         struct nlattr *tb[IFA_MAX+1];
3992         struct in6_addr *addr = NULL;
3993         struct net_device *dev = NULL;
3994         struct inet6_ifaddr *ifa;
3995         struct sk_buff *skb;
3996         int err;
3997
3998         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3999         if (err < 0)
4000                 goto errout;
4001
4002         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4003         if (addr == NULL) {
4004                 err = -EINVAL;
4005                 goto errout;
4006         }
4007
4008         ifm = nlmsg_data(nlh);
4009         if (ifm->ifa_index)
4010                 dev = __dev_get_by_index(net, ifm->ifa_index);
4011
4012         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4013         if (!ifa) {
4014                 err = -EADDRNOTAVAIL;
4015                 goto errout;
4016         }
4017
4018         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4019         if (!skb) {
4020                 err = -ENOBUFS;
4021                 goto errout_ifa;
4022         }
4023
4024         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4025                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4026         if (err < 0) {
4027                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4028                 WARN_ON(err == -EMSGSIZE);
4029                 kfree_skb(skb);
4030                 goto errout_ifa;
4031         }
4032         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4033 errout_ifa:
4034         in6_ifa_put(ifa);
4035 errout:
4036         return err;
4037 }
4038
4039 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4040 {
4041         struct sk_buff *skb;
4042         struct net *net = dev_net(ifa->idev->dev);
4043         int err = -ENOBUFS;
4044
4045         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4046         if (skb == NULL)
4047                 goto errout;
4048
4049         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4050         if (err < 0) {
4051                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4052                 WARN_ON(err == -EMSGSIZE);
4053                 kfree_skb(skb);
4054                 goto errout;
4055         }
4056         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4057         return;
4058 errout:
4059         if (err < 0)
4060                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4061 }
4062
4063 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4064                                 __s32 *array, int bytes)
4065 {
4066         BUG_ON(bytes < (DEVCONF_MAX * 4));
4067
4068         memset(array, 0, bytes);
4069         array[DEVCONF_FORWARDING] = cnf->forwarding;
4070         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4071         array[DEVCONF_MTU6] = cnf->mtu6;
4072         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4073         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4074         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4075         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4076         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4077         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4078                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4079         array[DEVCONF_RTR_SOLICIT_DELAY] =
4080                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4081         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4082 #ifdef CONFIG_IPV6_PRIVACY
4083         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4084         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4085         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4086         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4087         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4088 #endif
4089         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4090         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4091         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4092 #ifdef CONFIG_IPV6_ROUTER_PREF
4093         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4094         array[DEVCONF_RTR_PROBE_INTERVAL] =
4095                 jiffies_to_msecs(cnf->rtr_probe_interval);
4096 #ifdef CONFIG_IPV6_ROUTE_INFO
4097         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4098 #endif
4099 #endif
4100         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4101         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4102 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4103         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4104 #endif
4105 #ifdef CONFIG_IPV6_MROUTE
4106         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4107 #endif
4108         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4109         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4110         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4111         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4112 }
4113
4114 static inline size_t inet6_ifla6_size(void)
4115 {
4116         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4117              + nla_total_size(sizeof(struct ifla_cacheinfo))
4118              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4119              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4120              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4121 }
4122
4123 static inline size_t inet6_if_nlmsg_size(void)
4124 {
4125         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4126                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4127                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4128                + nla_total_size(4) /* IFLA_MTU */
4129                + nla_total_size(4) /* IFLA_LINK */
4130                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4131 }
4132
4133 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4134                                       int items, int bytes)
4135 {
4136         int i;
4137         int pad = bytes - sizeof(u64) * items;
4138         BUG_ON(pad < 0);
4139
4140         /* Use put_unaligned() because stats may not be aligned for u64. */
4141         put_unaligned(items, &stats[0]);
4142         for (i = 1; i < items; i++)
4143                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4144
4145         memset(&stats[items], 0, pad);
4146 }
4147
4148 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4149                                       int items, int bytes, size_t syncpoff)
4150 {
4151         int i;
4152         int pad = bytes - sizeof(u64) * items;
4153         BUG_ON(pad < 0);
4154
4155         /* Use put_unaligned() because stats may not be aligned for u64. */
4156         put_unaligned(items, &stats[0]);
4157         for (i = 1; i < items; i++)
4158                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4159
4160         memset(&stats[items], 0, pad);
4161 }
4162
4163 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4164                              int bytes)
4165 {
4166         switch (attrtype) {
4167         case IFLA_INET6_STATS:
4168                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4169                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4170                 break;
4171         case IFLA_INET6_ICMP6STATS:
4172                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4173                 break;
4174         }
4175 }
4176
4177 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4178 {
4179         struct nlattr *nla;
4180         struct ifla_cacheinfo ci;
4181
4182         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4183                 goto nla_put_failure;
4184         ci.max_reasm_len = IPV6_MAXPLEN;
4185         ci.tstamp = cstamp_delta(idev->tstamp);
4186         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4187         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4188         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4189                 goto nla_put_failure;
4190         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4191         if (nla == NULL)
4192                 goto nla_put_failure;
4193         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4194
4195         /* XXX - MC not implemented */
4196
4197         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4198         if (nla == NULL)
4199                 goto nla_put_failure;
4200         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4201
4202         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4203         if (nla == NULL)
4204                 goto nla_put_failure;
4205         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4206
4207         return 0;
4208
4209 nla_put_failure:
4210         return -EMSGSIZE;
4211 }
4212
4213 static size_t inet6_get_link_af_size(const struct net_device *dev)
4214 {
4215         if (!__in6_dev_get(dev))
4216                 return 0;
4217
4218         return inet6_ifla6_size();
4219 }
4220
4221 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4222 {
4223         struct inet6_dev *idev = __in6_dev_get(dev);
4224
4225         if (!idev)
4226                 return -ENODATA;
4227
4228         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4229                 return -EMSGSIZE;
4230
4231         return 0;
4232 }
4233
4234 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4235                              u32 portid, u32 seq, int event, unsigned int flags)
4236 {
4237         struct net_device *dev = idev->dev;
4238         struct ifinfomsg *hdr;
4239         struct nlmsghdr *nlh;
4240         void *protoinfo;
4241
4242         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4243         if (nlh == NULL)
4244                 return -EMSGSIZE;
4245
4246         hdr = nlmsg_data(nlh);
4247         hdr->ifi_family = AF_INET6;
4248         hdr->__ifi_pad = 0;
4249         hdr->ifi_type = dev->type;
4250         hdr->ifi_index = dev->ifindex;
4251         hdr->ifi_flags = dev_get_flags(dev);
4252         hdr->ifi_change = 0;
4253
4254         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4255             (dev->addr_len &&
4256              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4257             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4258             (dev->ifindex != dev->iflink &&
4259              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4260                 goto nla_put_failure;
4261         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4262         if (protoinfo == NULL)
4263                 goto nla_put_failure;
4264
4265         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4266                 goto nla_put_failure;
4267
4268         nla_nest_end(skb, protoinfo);
4269         return nlmsg_end(skb, nlh);
4270
4271 nla_put_failure:
4272         nlmsg_cancel(skb, nlh);
4273         return -EMSGSIZE;
4274 }
4275
4276 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4277 {
4278         struct net *net = sock_net(skb->sk);
4279         int h, s_h;
4280         int idx = 0, s_idx;
4281         struct net_device *dev;
4282         struct inet6_dev *idev;
4283         struct hlist_head *head;
4284
4285         s_h = cb->args[0];
4286         s_idx = cb->args[1];
4287
4288         rcu_read_lock();
4289         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4290                 idx = 0;
4291                 head = &net->dev_index_head[h];
4292                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4293                         if (idx < s_idx)
4294                                 goto cont;
4295                         idev = __in6_dev_get(dev);
4296                         if (!idev)
4297                                 goto cont;
4298                         if (inet6_fill_ifinfo(skb, idev,
4299                                               NETLINK_CB(cb->skb).portid,
4300                                               cb->nlh->nlmsg_seq,
4301                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4302                                 goto out;
4303 cont:
4304                         idx++;
4305                 }
4306         }
4307 out:
4308         rcu_read_unlock();
4309         cb->args[1] = idx;
4310         cb->args[0] = h;
4311
4312         return skb->len;
4313 }
4314
4315 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4316 {
4317         struct sk_buff *skb;
4318         struct net *net = dev_net(idev->dev);
4319         int err = -ENOBUFS;
4320
4321         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4322         if (skb == NULL)
4323                 goto errout;
4324
4325         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4326         if (err < 0) {
4327                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4328                 WARN_ON(err == -EMSGSIZE);
4329                 kfree_skb(skb);
4330                 goto errout;
4331         }
4332         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4333         return;
4334 errout:
4335         if (err < 0)
4336                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4337 }
4338
4339 static inline size_t inet6_prefix_nlmsg_size(void)
4340 {
4341         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4342                + nla_total_size(sizeof(struct in6_addr))
4343                + nla_total_size(sizeof(struct prefix_cacheinfo));
4344 }
4345
4346 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4347                              struct prefix_info *pinfo, u32 portid, u32 seq,
4348                              int event, unsigned int flags)
4349 {
4350         struct prefixmsg *pmsg;
4351         struct nlmsghdr *nlh;
4352         struct prefix_cacheinfo ci;
4353
4354         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4355         if (nlh == NULL)
4356                 return -EMSGSIZE;
4357
4358         pmsg = nlmsg_data(nlh);
4359         pmsg->prefix_family = AF_INET6;
4360         pmsg->prefix_pad1 = 0;
4361         pmsg->prefix_pad2 = 0;
4362         pmsg->prefix_ifindex = idev->dev->ifindex;
4363         pmsg->prefix_len = pinfo->prefix_len;
4364         pmsg->prefix_type = pinfo->type;
4365         pmsg->prefix_pad3 = 0;
4366         pmsg->prefix_flags = 0;
4367         if (pinfo->onlink)
4368                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4369         if (pinfo->autoconf)
4370                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4371
4372         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4373                 goto nla_put_failure;
4374         ci.preferred_time = ntohl(pinfo->prefered);
4375         ci.valid_time = ntohl(pinfo->valid);
4376         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4377                 goto nla_put_failure;
4378         return nlmsg_end(skb, nlh);
4379
4380 nla_put_failure:
4381         nlmsg_cancel(skb, nlh);
4382         return -EMSGSIZE;
4383 }
4384
4385 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4386                          struct prefix_info *pinfo)
4387 {
4388         struct sk_buff *skb;
4389         struct net *net = dev_net(idev->dev);
4390         int err = -ENOBUFS;
4391
4392         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4393         if (skb == NULL)
4394                 goto errout;
4395
4396         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4397         if (err < 0) {
4398                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4399                 WARN_ON(err == -EMSGSIZE);
4400                 kfree_skb(skb);
4401                 goto errout;
4402         }
4403         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4404         return;
4405 errout:
4406         if (err < 0)
4407                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4408 }
4409
4410 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4411 {
4412         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4413
4414         switch (event) {
4415         case RTM_NEWADDR:
4416                 /*
4417                  * If the address was optimistic
4418                  * we inserted the route at the start of
4419                  * our DAD process, so we don't need
4420                  * to do it again
4421                  */
4422                 if (!(ifp->rt->rt6i_node))
4423                         ip6_ins_rt(ifp->rt);
4424                 if (ifp->idev->cnf.forwarding)
4425                         addrconf_join_anycast(ifp);
4426                 break;
4427         case RTM_DELADDR:
4428                 if (ifp->idev->cnf.forwarding)
4429                         addrconf_leave_anycast(ifp);
4430                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4431                 dst_hold(&ifp->rt->dst);
4432
4433                 if (ip6_del_rt(ifp->rt))
4434                         dst_free(&ifp->rt->dst);
4435                 break;
4436         }
4437 }
4438
4439 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4440 {
4441         rcu_read_lock_bh();
4442         if (likely(ifp->idev->dead == 0))
4443                 __ipv6_ifa_notify(event, ifp);
4444         rcu_read_unlock_bh();
4445 }
4446
4447 #ifdef CONFIG_SYSCTL
4448
4449 static
4450 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4451                            void __user *buffer, size_t *lenp, loff_t *ppos)
4452 {
4453         int *valp = ctl->data;
4454         int val = *valp;
4455         loff_t pos = *ppos;
4456         ctl_table lctl;
4457         int ret;
4458
4459         /*
4460          * ctl->data points to idev->cnf.forwarding, we should
4461          * not modify it until we get the rtnl lock.
4462          */
4463         lctl = *ctl;
4464         lctl.data = &val;
4465
4466         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4467
4468         if (write)
4469                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4470         if (ret)
4471                 *ppos = pos;
4472         return ret;
4473 }
4474
4475 static void dev_disable_change(struct inet6_dev *idev)
4476 {
4477         if (!idev || !idev->dev)
4478                 return;
4479
4480         if (idev->cnf.disable_ipv6)
4481                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4482         else
4483                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4484 }
4485
4486 static void addrconf_disable_change(struct net *net, __s32 newf)
4487 {
4488         struct net_device *dev;
4489         struct inet6_dev *idev;
4490
4491         rcu_read_lock();
4492         for_each_netdev_rcu(net, dev) {
4493                 idev = __in6_dev_get(dev);
4494                 if (idev) {
4495                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4496                         idev->cnf.disable_ipv6 = newf;
4497                         if (changed)
4498                                 dev_disable_change(idev);
4499                 }
4500         }
4501         rcu_read_unlock();
4502 }
4503
4504 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4505 {
4506         struct net *net;
4507         int old;
4508
4509         if (!rtnl_trylock())
4510                 return restart_syscall();
4511
4512         net = (struct net *)table->extra2;
4513         old = *p;
4514         *p = newf;
4515
4516         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4517                 rtnl_unlock();
4518                 return 0;
4519         }
4520
4521         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4522                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4523                 addrconf_disable_change(net, newf);
4524         } else if ((!newf) ^ (!old))
4525                 dev_disable_change((struct inet6_dev *)table->extra1);
4526
4527         rtnl_unlock();
4528         return 0;
4529 }
4530
4531 static
4532 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4533                             void __user *buffer, size_t *lenp, loff_t *ppos)
4534 {
4535         int *valp = ctl->data;
4536         int val = *valp;
4537         loff_t pos = *ppos;
4538         ctl_table lctl;
4539         int ret;
4540
4541         /*
4542          * ctl->data points to idev->cnf.disable_ipv6, we should
4543          * not modify it until we get the rtnl lock.
4544          */
4545         lctl = *ctl;
4546         lctl.data = &val;
4547
4548         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4549
4550         if (write)
4551                 ret = addrconf_disable_ipv6(ctl, valp, val);
4552         if (ret)
4553                 *ppos = pos;
4554         return ret;
4555 }
4556
4557 static struct addrconf_sysctl_table
4558 {
4559         struct ctl_table_header *sysctl_header;
4560         ctl_table addrconf_vars[DEVCONF_MAX+1];
4561 } addrconf_sysctl __read_mostly = {
4562         .sysctl_header = NULL,
4563         .addrconf_vars = {
4564                 {
4565                         .procname       = "forwarding",
4566                         .data           = &ipv6_devconf.forwarding,
4567                         .maxlen         = sizeof(int),
4568                         .mode           = 0644,
4569                         .proc_handler   = addrconf_sysctl_forward,
4570                 },
4571                 {
4572                         .procname       = "hop_limit",
4573                         .data           = &ipv6_devconf.hop_limit,
4574                         .maxlen         = sizeof(int),
4575                         .mode           = 0644,
4576                         .proc_handler   = proc_dointvec,
4577                 },
4578                 {
4579                         .procname       = "mtu",
4580                         .data           = &ipv6_devconf.mtu6,
4581                         .maxlen         = sizeof(int),
4582                         .mode           = 0644,
4583                         .proc_handler   = proc_dointvec,
4584                 },
4585                 {
4586                         .procname       = "accept_ra",
4587                         .data           = &ipv6_devconf.accept_ra,
4588                         .maxlen         = sizeof(int),
4589                         .mode           = 0644,
4590                         .proc_handler   = proc_dointvec,
4591                 },
4592                 {
4593                         .procname       = "accept_redirects",
4594                         .data           = &ipv6_devconf.accept_redirects,
4595                         .maxlen         = sizeof(int),
4596                         .mode           = 0644,
4597                         .proc_handler   = proc_dointvec,
4598                 },
4599                 {
4600                         .procname       = "autoconf",
4601                         .data           = &ipv6_devconf.autoconf,
4602                         .maxlen         = sizeof(int),
4603                         .mode           = 0644,
4604                         .proc_handler   = proc_dointvec,
4605                 },
4606                 {
4607                         .procname       = "dad_transmits",
4608                         .data           = &ipv6_devconf.dad_transmits,
4609                         .maxlen         = sizeof(int),
4610                         .mode           = 0644,
4611                         .proc_handler   = proc_dointvec,
4612                 },
4613                 {
4614                         .procname       = "router_solicitations",
4615                         .data           = &ipv6_devconf.rtr_solicits,
4616                         .maxlen         = sizeof(int),
4617                         .mode           = 0644,
4618                         .proc_handler   = proc_dointvec,
4619                 },
4620                 {
4621                         .procname       = "router_solicitation_interval",
4622                         .data           = &ipv6_devconf.rtr_solicit_interval,
4623                         .maxlen         = sizeof(int),
4624                         .mode           = 0644,
4625                         .proc_handler   = proc_dointvec_jiffies,
4626                 },
4627                 {
4628                         .procname       = "router_solicitation_delay",
4629                         .data           = &ipv6_devconf.rtr_solicit_delay,
4630                         .maxlen         = sizeof(int),
4631                         .mode           = 0644,
4632                         .proc_handler   = proc_dointvec_jiffies,
4633                 },
4634                 {
4635                         .procname       = "force_mld_version",
4636                         .data           = &ipv6_devconf.force_mld_version,
4637                         .maxlen         = sizeof(int),
4638                         .mode           = 0644,
4639                         .proc_handler   = proc_dointvec,
4640                 },
4641 #ifdef CONFIG_IPV6_PRIVACY
4642                 {
4643                         .procname       = "use_tempaddr",
4644                         .data           = &ipv6_devconf.use_tempaddr,
4645                         .maxlen         = sizeof(int),
4646                         .mode           = 0644,
4647                         .proc_handler   = proc_dointvec,
4648                 },
4649                 {
4650                         .procname       = "temp_valid_lft",
4651                         .data           = &ipv6_devconf.temp_valid_lft,
4652                         .maxlen         = sizeof(int),
4653                         .mode           = 0644,
4654                         .proc_handler   = proc_dointvec,
4655                 },
4656                 {
4657                         .procname       = "temp_prefered_lft",
4658                         .data           = &ipv6_devconf.temp_prefered_lft,
4659                         .maxlen         = sizeof(int),
4660                         .mode           = 0644,
4661                         .proc_handler   = proc_dointvec,
4662                 },
4663                 {
4664                         .procname       = "regen_max_retry",
4665                         .data           = &ipv6_devconf.regen_max_retry,
4666                         .maxlen         = sizeof(int),
4667                         .mode           = 0644,
4668                         .proc_handler   = proc_dointvec,
4669                 },
4670                 {
4671                         .procname       = "max_desync_factor",
4672                         .data           = &ipv6_devconf.max_desync_factor,
4673                         .maxlen         = sizeof(int),
4674                         .mode           = 0644,
4675                         .proc_handler   = proc_dointvec,
4676                 },
4677 #endif
4678                 {
4679                         .procname       = "max_addresses",
4680                         .data           = &ipv6_devconf.max_addresses,
4681                         .maxlen         = sizeof(int),
4682                         .mode           = 0644,
4683                         .proc_handler   = proc_dointvec,
4684                 },
4685                 {
4686                         .procname       = "accept_ra_defrtr",
4687                         .data           = &ipv6_devconf.accept_ra_defrtr,
4688                         .maxlen         = sizeof(int),
4689                         .mode           = 0644,
4690                         .proc_handler   = proc_dointvec,
4691                 },
4692                 {
4693                         .procname       = "accept_ra_pinfo",
4694                         .data           = &ipv6_devconf.accept_ra_pinfo,
4695                         .maxlen         = sizeof(int),
4696                         .mode           = 0644,
4697                         .proc_handler   = proc_dointvec,
4698                 },
4699 #ifdef CONFIG_IPV6_ROUTER_PREF
4700                 {
4701                         .procname       = "accept_ra_rtr_pref",
4702                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4703                         .maxlen         = sizeof(int),
4704                         .mode           = 0644,
4705                         .proc_handler   = proc_dointvec,
4706                 },
4707                 {
4708                         .procname       = "router_probe_interval",
4709                         .data           = &ipv6_devconf.rtr_probe_interval,
4710                         .maxlen         = sizeof(int),
4711                         .mode           = 0644,
4712                         .proc_handler   = proc_dointvec_jiffies,
4713                 },
4714 #ifdef CONFIG_IPV6_ROUTE_INFO
4715                 {
4716                         .procname       = "accept_ra_rt_info_max_plen",
4717                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4718                         .maxlen         = sizeof(int),
4719                         .mode           = 0644,
4720                         .proc_handler   = proc_dointvec,
4721                 },
4722 #endif
4723 #endif
4724                 {
4725                         .procname       = "proxy_ndp",
4726                         .data           = &ipv6_devconf.proxy_ndp,
4727                         .maxlen         = sizeof(int),
4728                         .mode           = 0644,
4729                         .proc_handler   = proc_dointvec,
4730                 },
4731                 {
4732                         .procname       = "accept_source_route",
4733                         .data           = &ipv6_devconf.accept_source_route,
4734                         .maxlen         = sizeof(int),
4735                         .mode           = 0644,
4736                         .proc_handler   = proc_dointvec,
4737                 },
4738 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4739                 {
4740                         .procname       = "optimistic_dad",
4741                         .data           = &ipv6_devconf.optimistic_dad,
4742                         .maxlen         = sizeof(int),
4743                         .mode           = 0644,
4744                         .proc_handler   = proc_dointvec,
4745
4746                 },
4747 #endif
4748 #ifdef CONFIG_IPV6_MROUTE
4749                 {
4750                         .procname       = "mc_forwarding",
4751                         .data           = &ipv6_devconf.mc_forwarding,
4752                         .maxlen         = sizeof(int),
4753                         .mode           = 0444,
4754                         .proc_handler   = proc_dointvec,
4755                 },
4756 #endif
4757                 {
4758                         .procname       = "disable_ipv6",
4759                         .data           = &ipv6_devconf.disable_ipv6,
4760                         .maxlen         = sizeof(int),
4761                         .mode           = 0644,
4762                         .proc_handler   = addrconf_sysctl_disable,
4763                 },
4764                 {
4765                         .procname       = "accept_dad",
4766                         .data           = &ipv6_devconf.accept_dad,
4767                         .maxlen         = sizeof(int),
4768                         .mode           = 0644,
4769                         .proc_handler   = proc_dointvec,
4770                 },
4771                 {
4772                         .procname       = "force_tllao",
4773                         .data           = &ipv6_devconf.force_tllao,
4774                         .maxlen         = sizeof(int),
4775                         .mode           = 0644,
4776                         .proc_handler   = proc_dointvec
4777                 },
4778                 {
4779                         .procname       = "ndisc_notify",
4780                         .data           = &ipv6_devconf.ndisc_notify,
4781                         .maxlen         = sizeof(int),
4782                         .mode           = 0644,
4783                         .proc_handler   = proc_dointvec
4784                 },
4785                 {
4786                         /* sentinel */
4787                 }
4788         },
4789 };
4790
4791 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4792                 struct inet6_dev *idev, struct ipv6_devconf *p)
4793 {
4794         int i;
4795         struct addrconf_sysctl_table *t;
4796         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4797
4798         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4799         if (t == NULL)
4800                 goto out;
4801
4802         for (i = 0; t->addrconf_vars[i].data; i++) {
4803                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4804                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4805                 t->addrconf_vars[i].extra2 = net;
4806         }
4807
4808         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4809
4810         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4811         if (t->sysctl_header == NULL)
4812                 goto free;
4813
4814         p->sysctl = t;
4815         return 0;
4816
4817 free:
4818         kfree(t);
4819 out:
4820         return -ENOBUFS;
4821 }
4822
4823 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4824 {
4825         struct addrconf_sysctl_table *t;
4826
4827         if (p->sysctl == NULL)
4828                 return;
4829
4830         t = p->sysctl;
4831         p->sysctl = NULL;
4832         unregister_net_sysctl_table(t->sysctl_header);
4833         kfree(t);
4834 }
4835
4836 static void addrconf_sysctl_register(struct inet6_dev *idev)
4837 {
4838         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4839                               &ndisc_ifinfo_sysctl_change);
4840         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4841                                         idev, &idev->cnf);
4842 }
4843
4844 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4845 {
4846         __addrconf_sysctl_unregister(&idev->cnf);
4847         neigh_sysctl_unregister(idev->nd_parms);
4848 }
4849
4850
4851 #endif
4852
4853 static int __net_init addrconf_init_net(struct net *net)
4854 {
4855         int err;
4856         struct ipv6_devconf *all, *dflt;
4857
4858         err = -ENOMEM;
4859         all = &ipv6_devconf;
4860         dflt = &ipv6_devconf_dflt;
4861
4862         if (!net_eq(net, &init_net)) {
4863                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4864                 if (all == NULL)
4865                         goto err_alloc_all;
4866
4867                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4868                 if (dflt == NULL)
4869                         goto err_alloc_dflt;
4870         } else {
4871                 /* these will be inherited by all namespaces */
4872                 dflt->autoconf = ipv6_defaults.autoconf;
4873                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4874         }
4875
4876         net->ipv6.devconf_all = all;
4877         net->ipv6.devconf_dflt = dflt;
4878
4879 #ifdef CONFIG_SYSCTL
4880         err = __addrconf_sysctl_register(net, "all", NULL, all);
4881         if (err < 0)
4882                 goto err_reg_all;
4883
4884         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4885         if (err < 0)
4886                 goto err_reg_dflt;
4887 #endif
4888         return 0;
4889
4890 #ifdef CONFIG_SYSCTL
4891 err_reg_dflt:
4892         __addrconf_sysctl_unregister(all);
4893 err_reg_all:
4894         kfree(dflt);
4895 #endif
4896 err_alloc_dflt:
4897         kfree(all);
4898 err_alloc_all:
4899         return err;
4900 }
4901
4902 static void __net_exit addrconf_exit_net(struct net *net)
4903 {
4904 #ifdef CONFIG_SYSCTL
4905         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4906         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4907 #endif
4908         if (!net_eq(net, &init_net)) {
4909                 kfree(net->ipv6.devconf_dflt);
4910                 kfree(net->ipv6.devconf_all);
4911         }
4912 }
4913
4914 static struct pernet_operations addrconf_ops = {
4915         .init = addrconf_init_net,
4916         .exit = addrconf_exit_net,
4917 };
4918
4919 /*
4920  *      Device notifier
4921  */
4922
4923 int register_inet6addr_notifier(struct notifier_block *nb)
4924 {
4925         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4926 }
4927 EXPORT_SYMBOL(register_inet6addr_notifier);
4928
4929 int unregister_inet6addr_notifier(struct notifier_block *nb)
4930 {
4931         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4932 }
4933 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4934
4935 static struct rtnl_af_ops inet6_ops = {
4936         .family           = AF_INET6,
4937         .fill_link_af     = inet6_fill_link_af,
4938         .get_link_af_size = inet6_get_link_af_size,
4939 };
4940
4941 /*
4942  *      Init / cleanup code
4943  */
4944
4945 int __init addrconf_init(void)
4946 {
4947         int i, err;
4948
4949         err = ipv6_addr_label_init();
4950         if (err < 0) {
4951                 pr_crit("%s: cannot initialize default policy table: %d\n",
4952                         __func__, err);
4953                 goto out;
4954         }
4955
4956         err = register_pernet_subsys(&addrconf_ops);
4957         if (err < 0)
4958                 goto out_addrlabel;
4959
4960         /* The addrconf netdev notifier requires that loopback_dev
4961          * has it's ipv6 private information allocated and setup
4962          * before it can bring up and give link-local addresses
4963          * to other devices which are up.
4964          *
4965          * Unfortunately, loopback_dev is not necessarily the first
4966          * entry in the global dev_base list of net devices.  In fact,
4967          * it is likely to be the very last entry on that list.
4968          * So this causes the notifier registry below to try and
4969          * give link-local addresses to all devices besides loopback_dev
4970          * first, then loopback_dev, which cases all the non-loopback_dev
4971          * devices to fail to get a link-local address.
4972          *
4973          * So, as a temporary fix, allocate the ipv6 structure for
4974          * loopback_dev first by hand.
4975          * Longer term, all of the dependencies ipv6 has upon the loopback
4976          * device and it being up should be removed.
4977          */
4978         rtnl_lock();
4979         if (!ipv6_add_dev(init_net.loopback_dev))
4980                 err = -ENOMEM;
4981         rtnl_unlock();
4982         if (err)
4983                 goto errlo;
4984
4985         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4986                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4987
4988         register_netdevice_notifier(&ipv6_dev_notf);
4989
4990         addrconf_verify(0);
4991
4992         err = rtnl_af_register(&inet6_ops);
4993         if (err < 0)
4994                 goto errout_af;
4995
4996         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4997                               NULL);
4998         if (err < 0)
4999                 goto errout;
5000
5001         /* Only the first call to __rtnl_register can fail */
5002         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5003         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5004         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5005                         inet6_dump_ifaddr, NULL);
5006         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5007                         inet6_dump_ifmcaddr, NULL);
5008         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5009                         inet6_dump_ifacaddr, NULL);
5010         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5011                         inet6_netconf_dump_devconf, NULL);
5012
5013         ipv6_addr_label_rtnl_register();
5014
5015         return 0;
5016 errout:
5017         rtnl_af_unregister(&inet6_ops);
5018 errout_af:
5019         unregister_netdevice_notifier(&ipv6_dev_notf);
5020 errlo:
5021         unregister_pernet_subsys(&addrconf_ops);
5022 out_addrlabel:
5023         ipv6_addr_label_cleanup();
5024 out:
5025         return err;
5026 }
5027
5028 void addrconf_cleanup(void)
5029 {
5030         struct net_device *dev;
5031         int i;
5032
5033         unregister_netdevice_notifier(&ipv6_dev_notf);
5034         unregister_pernet_subsys(&addrconf_ops);
5035         ipv6_addr_label_cleanup();
5036
5037         rtnl_lock();
5038
5039         __rtnl_af_unregister(&inet6_ops);
5040
5041         /* clean dev list */
5042         for_each_netdev(&init_net, dev) {
5043                 if (__in6_dev_get(dev) == NULL)
5044                         continue;
5045                 addrconf_ifdown(dev, 1);
5046         }
5047         addrconf_ifdown(init_net.loopback_dev, 2);
5048
5049         /*
5050          *      Check hash table.
5051          */
5052         spin_lock_bh(&addrconf_hash_lock);
5053         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5054                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5055         spin_unlock_bh(&addrconf_hash_lock);
5056
5057         del_timer(&addr_chk_timer);
5058         rtnl_unlock();
5059 }