net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86
  87 #ifdef CONFIG_IPV6_PRIVACY
  88 #include <linux/random.h>
  89 #endif
  90
  91 #include <linux/uaccess.h>
  92 #include <asm/unaligned.h>
  93
  94 #include <linux/proc_fs.h>
  95 #include <linux/seq_file.h>
  96 #include <linux/export.h>
  97
  98 /* Set to 3 to get tracing... */
  99 #define ACONF_DEBUG 2
 100
 101 #if ACONF_DEBUG >= 3
 102 #define ADBG(x) printk x
 103 #else
 104 #define ADBG(x)
 105 #endif
 106
 107 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 108
 109 static inline u32 cstamp_delta(unsigned long cstamp)
 110 {
 111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112 }
 113
 114 #ifdef CONFIG_SYSCTL
 115 static void addrconf_sysctl_register(struct inet6_dev *idev);
 116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117 #else
 118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 119 {
 120 }
 121
 122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 123 {
 124 }
 125 #endif
 126
 127 #ifdef CONFIG_IPV6_PRIVACY
 128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130 static void ipv6_regen_rndid(unsigned long data);
 131 #endif
 132
 133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134 static int ipv6_count_addresses(struct inet6_dev *idev);
 135
 136 /*
 137  *      Configured unicast address hash table
 138  */
 139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 140 static DEFINE_SPINLOCK(addrconf_hash_lock);
 141
 142 static void addrconf_verify(unsigned long);
 143
 144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 145 static DEFINE_SPINLOCK(addrconf_verify_lock);
 146
 147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 149
 150 static void addrconf_type_change(struct net_device *dev,
 151                                  unsigned long event);
 152 static int addrconf_ifdown(struct net_device *dev, int how);
 153
 154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 155                                                   int plen,
 156                                                   const struct net_device *dev,
 157                                                   u32 flags, u32 noflags);
 158
 159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 160 static void addrconf_dad_timer(unsigned long data);
 161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 162 static void addrconf_dad_run(struct inet6_dev *idev);
 163 static void addrconf_rs_timer(unsigned long data);
 164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 166
 167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 168                                 struct prefix_info *pinfo);
 169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 170                                struct net_device *dev);
 171
 172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 173         .forwarding             = 0,
 174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 175         .mtu6                   = IPV6_MIN_MTU,
 176         .accept_ra              = 1,
 177         .accept_redirects       = 1,
 178         .autoconf               = 1,
 179         .force_mld_version      = 0,
 180         .dad_transmits          = 1,
 181         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 182         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 183         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 184 #ifdef CONFIG_IPV6_PRIVACY
 185         .use_tempaddr           = 0,
 186         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 187         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 188         .regen_max_retry        = REGEN_MAX_RETRY,
 189         .max_desync_factor      = MAX_DESYNC_FACTOR,
 190 #endif
 191         .max_addresses          = IPV6_MAX_ADDRESSES,
 192         .accept_ra_defrtr       = 1,
 193         .accept_ra_pinfo        = 1,
 194 #ifdef CONFIG_IPV6_ROUTER_PREF
 195         .accept_ra_rtr_pref     = 1,
 196         .rtr_probe_interval     = 60 * HZ,
 197 #ifdef CONFIG_IPV6_ROUTE_INFO
 198         .accept_ra_rt_info_max_plen = 0,
 199 #endif
 200 #endif
 201         .proxy_ndp              = 0,
 202         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 203         .disable_ipv6           = 0,
 204         .accept_dad             = 1,
 205 };
 206
 207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 208         .forwarding             = 0,
 209         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 210         .mtu6                   = IPV6_MIN_MTU,
 211         .accept_ra              = 1,
 212         .accept_redirects       = 1,
 213         .autoconf               = 1,
 214         .dad_transmits          = 1,
 215         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 216         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 217         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 218 #ifdef CONFIG_IPV6_PRIVACY
 219         .use_tempaddr           = 0,
 220         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 221         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 222         .regen_max_retry        = REGEN_MAX_RETRY,
 223         .max_desync_factor      = MAX_DESYNC_FACTOR,
 224 #endif
 225         .max_addresses          = IPV6_MAX_ADDRESSES,
 226         .accept_ra_defrtr       = 1,
 227         .accept_ra_pinfo        = 1,
 228 #ifdef CONFIG_IPV6_ROUTER_PREF
 229         .accept_ra_rtr_pref     = 1,
 230         .rtr_probe_interval     = 60 * HZ,
 231 #ifdef CONFIG_IPV6_ROUTE_INFO
 232         .accept_ra_rt_info_max_plen = 0,
 233 #endif
 234 #endif
 235         .proxy_ndp              = 0,
 236         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 237         .disable_ipv6           = 0,
 238         .accept_dad             = 1,
 239 };
 240
 241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 246 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 247 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 248 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 249
 250 /* Check if a valid qdisc is available */
 251 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 252 {
 253         return !qdisc_tx_is_noop(dev);
 254 }
 255
 256 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 257 {
 258         if (del_timer(&ifp->timer))
 259                 __in6_ifa_put(ifp);
 260 }
 261
 262 enum addrconf_timer_t {
 263         AC_NONE,
 264         AC_DAD,
 265         AC_RS,
 266 };
 267
 268 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 269                                enum addrconf_timer_t what,
 270                                unsigned long when)
 271 {
 272         if (!del_timer(&ifp->timer))
 273                 in6_ifa_hold(ifp);
 274
 275         switch (what) {
 276         case AC_DAD:
 277                 ifp->timer.function = addrconf_dad_timer;
 278                 break;
 279         case AC_RS:
 280                 ifp->timer.function = addrconf_rs_timer;
 281                 break;
 282         default:
 283                 break;
 284         }
 285         ifp->timer.expires = jiffies + when;
 286         add_timer(&ifp->timer);
 287 }
 288
 289 static int snmp6_alloc_dev(struct inet6_dev *idev)
 290 {
 291         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 292                           sizeof(struct ipstats_mib),
 293                           __alignof__(struct ipstats_mib)) < 0)
 294                 goto err_ip;
 295         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 296                                         GFP_KERNEL);
 297         if (!idev->stats.icmpv6dev)
 298                 goto err_icmp;
 299         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 300                                            GFP_KERNEL);
 301         if (!idev->stats.icmpv6msgdev)
 302                 goto err_icmpmsg;
 303
 304         return 0;
 305
 306 err_icmpmsg:
 307         kfree(idev->stats.icmpv6dev);
 308 err_icmp:
 309         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 310 err_ip:
 311         return -ENOMEM;
 312 }
 313
 314 static void snmp6_free_dev(struct inet6_dev *idev)
 315 {
 316         kfree(idev->stats.icmpv6msgdev);
 317         kfree(idev->stats.icmpv6dev);
 318         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 319 }
 320
 321 /* Nobody refers to this device, we may destroy it. */
 322
 323 void in6_dev_finish_destroy(struct inet6_dev *idev)
 324 {
 325         struct net_device *dev = idev->dev;
 326
 327         WARN_ON(!list_empty(&idev->addr_list));
 328         WARN_ON(idev->mc_list != NULL);
 329
 330 #ifdef NET_REFCNT_DEBUG
 331         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 332 #endif
 333         dev_put(dev);
 334         if (!idev->dead) {
 335                 pr_warn("Freeing alive inet6 device %p\n", idev);
 336                 return;
 337         }
 338         snmp6_free_dev(idev);
 339         kfree_rcu(idev, rcu);
 340 }
 341 EXPORT_SYMBOL(in6_dev_finish_destroy);
 342
 343 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 344 {
 345         struct inet6_dev *ndev;
 346
 347         ASSERT_RTNL();
 348
 349         if (dev->mtu < IPV6_MIN_MTU)
 350                 return NULL;
 351
 352         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 353
 354         if (ndev == NULL)
 355                 return NULL;
 356
 357         rwlock_init(&ndev->lock);
 358         ndev->dev = dev;
 359         INIT_LIST_HEAD(&ndev->addr_list);
 360
 361         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 362         ndev->cnf.mtu6 = dev->mtu;
 363         ndev->cnf.sysctl = NULL;
 364         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 365         if (ndev->nd_parms == NULL) {
 366                 kfree(ndev);
 367                 return NULL;
 368         }
 369         if (ndev->cnf.forwarding)
 370                 dev_disable_lro(dev);
 371         /* We refer to the device */
 372         dev_hold(dev);
 373
 374         if (snmp6_alloc_dev(ndev) < 0) {
 375                 ADBG((KERN_WARNING
 376                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 377                         __func__, dev->name));
 378                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 379                 dev_put(dev);
 380                 kfree(ndev);
 381                 return NULL;
 382         }
 383
 384         if (snmp6_register_dev(ndev) < 0) {
 385                 ADBG((KERN_WARNING
 386                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 387                         __func__, dev->name));
 388                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 389                 ndev->dead = 1;
 390                 in6_dev_finish_destroy(ndev);
 391                 return NULL;
 392         }
 393
 394         /* One reference from device.  We must do this before
 395          * we invoke __ipv6_regen_rndid().
 396          */
 397         in6_dev_hold(ndev);
 398
 399         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 400                 ndev->cnf.accept_dad = -1;
 401
 402 #if IS_ENABLED(CONFIG_IPV6_SIT)
 403         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 404                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 405                 ndev->cnf.rtr_solicits = 0;
 406         }
 407 #endif
 408
 409 #ifdef CONFIG_IPV6_PRIVACY
 410         INIT_LIST_HEAD(&ndev->tempaddr_list);
 411         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 412         if ((dev->flags&IFF_LOOPBACK) ||
 413             dev->type == ARPHRD_TUNNEL ||
 414             dev->type == ARPHRD_TUNNEL6 ||
 415             dev->type == ARPHRD_SIT ||
 416             dev->type == ARPHRD_NONE) {
 417                 ndev->cnf.use_tempaddr = -1;
 418         } else {
 419                 in6_dev_hold(ndev);
 420                 ipv6_regen_rndid((unsigned long) ndev);
 421         }
 422 #endif
 423         ndev->token = in6addr_any;
 424
 425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 426                 ndev->if_flags |= IF_READY;
 427
 428         ipv6_mc_init_dev(ndev);
 429         ndev->tstamp = jiffies;
 430         addrconf_sysctl_register(ndev);
 431         /* protected by rtnl_lock */
 432         rcu_assign_pointer(dev->ip6_ptr, ndev);
 433
 434         /* Join interface-local all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 436
 437         /* Join all-node multicast group */
 438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 439
 440         /* Join all-router multicast group if forwarding is set */
 441         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 442                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 443
 444         return ndev;
 445 }
 446
 447 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 448 {
 449         struct inet6_dev *idev;
 450
 451         ASSERT_RTNL();
 452
 453         idev = __in6_dev_get(dev);
 454         if (!idev) {
 455                 idev = ipv6_add_dev(dev);
 456                 if (!idev)
 457                         return NULL;
 458         }
 459
 460         if (dev->flags&IFF_UP)
 461                 ipv6_mc_up(idev);
 462         return idev;
 463 }
 464
 465 static int inet6_netconf_msgsize_devconf(int type)
 466 {
 467         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 468                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 469
 470         /* type -1 is used for ALL */
 471         if (type == -1 || type == NETCONFA_FORWARDING)
 472                 size += nla_total_size(4);
 473 #ifdef CONFIG_IPV6_MROUTE
 474         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 475                 size += nla_total_size(4);
 476 #endif
 477
 478         return size;
 479 }
 480
 481 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 482                                       struct ipv6_devconf *devconf, u32 portid,
 483                                       u32 seq, int event, unsigned int flags,
 484                                       int type)
 485 {
 486         struct nlmsghdr  *nlh;
 487         struct netconfmsg *ncm;
 488
 489         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 490                         flags);
 491         if (nlh == NULL)
 492                 return -EMSGSIZE;
 493
 494         ncm = nlmsg_data(nlh);
 495         ncm->ncm_family = AF_INET6;
 496
 497         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 498                 goto nla_put_failure;
 499
 500         /* type -1 is used for ALL */
 501         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 502             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 503                 goto nla_put_failure;
 504 #ifdef CONFIG_IPV6_MROUTE
 505         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 506             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 507                         devconf->mc_forwarding) < 0)
 508                 goto nla_put_failure;
 509 #endif
 510         return nlmsg_end(skb, nlh);
 511
 512 nla_put_failure:
 513         nlmsg_cancel(skb, nlh);
 514         return -EMSGSIZE;
 515 }
 516
 517 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 518                                   struct ipv6_devconf *devconf)
 519 {
 520         struct sk_buff *skb;
 521         int err = -ENOBUFS;
 522
 523         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 524         if (skb == NULL)
 525                 goto errout;
 526
 527         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 528                                          RTM_NEWNETCONF, 0, type);
 529         if (err < 0) {
 530                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 531                 WARN_ON(err == -EMSGSIZE);
 532                 kfree_skb(skb);
 533                 goto errout;
 534         }
 535         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 536         return;
 537 errout:
 538         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 539 }
 540
 541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 542         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 543         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 544 };
 545
 546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 547                                      struct nlmsghdr *nlh)
 548 {
 549         struct net *net = sock_net(in_skb->sk);
 550         struct nlattr *tb[NETCONFA_MAX+1];
 551         struct netconfmsg *ncm;
 552         struct sk_buff *skb;
 553         struct ipv6_devconf *devconf;
 554         struct inet6_dev *in6_dev;
 555         struct net_device *dev;
 556         int ifindex;
 557         int err;
 558
 559         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 560                           devconf_ipv6_policy);
 561         if (err < 0)
 562                 goto errout;
 563
 564         err = EINVAL;
 565         if (!tb[NETCONFA_IFINDEX])
 566                 goto errout;
 567
 568         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 569         switch (ifindex) {
 570         case NETCONFA_IFINDEX_ALL:
 571                 devconf = net->ipv6.devconf_all;
 572                 break;
 573         case NETCONFA_IFINDEX_DEFAULT:
 574                 devconf = net->ipv6.devconf_dflt;
 575                 break;
 576         default:
 577                 dev = __dev_get_by_index(net, ifindex);
 578                 if (dev == NULL)
 579                         goto errout;
 580                 in6_dev = __in6_dev_get(dev);
 581                 if (in6_dev == NULL)
 582                         goto errout;
 583                 devconf = &in6_dev->cnf;
 584                 break;
 585         }
 586
 587         err = -ENOBUFS;
 588         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 589         if (skb == NULL)
 590                 goto errout;
 591
 592         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 593                                          NETLINK_CB(in_skb).portid,
 594                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 595                                          -1);
 596         if (err < 0) {
 597                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 598                 WARN_ON(err == -EMSGSIZE);
 599                 kfree_skb(skb);
 600                 goto errout;
 601         }
 602         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 603 errout:
 604         return err;
 605 }
 606
 607 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 608                                       struct netlink_callback *cb)
 609 {
 610         struct net *net = sock_net(skb->sk);
 611         int h, s_h;
 612         int idx, s_idx;
 613         struct net_device *dev;
 614         struct inet6_dev *idev;
 615         struct hlist_head *head;
 616
 617         s_h = cb->args[0];
 618         s_idx = idx = cb->args[1];
 619
 620         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 621                 idx = 0;
 622                 head = &net->dev_index_head[h];
 623                 rcu_read_lock();
 624                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 625                           net->dev_base_seq;
 626                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 627                         if (idx < s_idx)
 628                                 goto cont;
 629                         idev = __in6_dev_get(dev);
 630                         if (!idev)
 631                                 goto cont;
 632
 633                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 634                                                        &idev->cnf,
 635                                                        NETLINK_CB(cb->skb).portid,
 636                                                        cb->nlh->nlmsg_seq,
 637                                                        RTM_NEWNETCONF,
 638                                                        NLM_F_MULTI,
 639                                                        -1) <= 0) {
 640                                 rcu_read_unlock();
 641                                 goto done;
 642                         }
 643                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 644 cont:
 645                         idx++;
 646                 }
 647                 rcu_read_unlock();
 648         }
 649         if (h == NETDEV_HASHENTRIES) {
 650                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 651                                                net->ipv6.devconf_all,
 652                                                NETLINK_CB(cb->skb).portid,
 653                                                cb->nlh->nlmsg_seq,
 654                                                RTM_NEWNETCONF, NLM_F_MULTI,
 655                                                -1) <= 0)
 656                         goto done;
 657                 else
 658                         h++;
 659         }
 660         if (h == NETDEV_HASHENTRIES + 1) {
 661                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 662                                                net->ipv6.devconf_dflt,
 663                                                NETLINK_CB(cb->skb).portid,
 664                                                cb->nlh->nlmsg_seq,
 665                                                RTM_NEWNETCONF, NLM_F_MULTI,
 666                                                -1) <= 0)
 667                         goto done;
 668                 else
 669                         h++;
 670         }
 671 done:
 672         cb->args[0] = h;
 673         cb->args[1] = idx;
 674
 675         return skb->len;
 676 }
 677
 678 #ifdef CONFIG_SYSCTL
 679 static void dev_forward_change(struct inet6_dev *idev)
 680 {
 681         struct net_device *dev;
 682         struct inet6_ifaddr *ifa;
 683
 684         if (!idev)
 685                 return;
 686         dev = idev->dev;
 687         if (idev->cnf.forwarding)
 688                 dev_disable_lro(dev);
 689         if (dev->flags & IFF_MULTICAST) {
 690                 if (idev->cnf.forwarding) {
 691                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 692                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 693                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 694                 } else {
 695                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 696                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 697                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 698                 }
 699         }
 700
 701         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 702                 if (ifa->flags&IFA_F_TENTATIVE)
 703                         continue;
 704                 if (idev->cnf.forwarding)
 705                         addrconf_join_anycast(ifa);
 706                 else
 707                         addrconf_leave_anycast(ifa);
 708         }
 709         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 710                                      dev->ifindex, &idev->cnf);
 711 }
 712
 713
 714 static void addrconf_forward_change(struct net *net, __s32 newf)
 715 {
 716         struct net_device *dev;
 717         struct inet6_dev *idev;
 718
 719         for_each_netdev(net, dev) {
 720                 idev = __in6_dev_get(dev);
 721                 if (idev) {
 722                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 723                         idev->cnf.forwarding = newf;
 724                         if (changed)
 725                                 dev_forward_change(idev);
 726                 }
 727         }
 728 }
 729
 730 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 731 {
 732         struct net *net;
 733         int old;
 734
 735         if (!rtnl_trylock())
 736                 return restart_syscall();
 737
 738         net = (struct net *)table->extra2;
 739         old = *p;
 740         *p = newf;
 741
 742         if (p == &net->ipv6.devconf_dflt->forwarding) {
 743                 if ((!newf) ^ (!old))
 744                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 745                                                      NETCONFA_IFINDEX_DEFAULT,
 746                                                      net->ipv6.devconf_dflt);
 747                 rtnl_unlock();
 748                 return 0;
 749         }
 750
 751         if (p == &net->ipv6.devconf_all->forwarding) {
 752                 net->ipv6.devconf_dflt->forwarding = newf;
 753                 addrconf_forward_change(net, newf);
 754                 if ((!newf) ^ (!old))
 755                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 756                                                      NETCONFA_IFINDEX_ALL,
 757                                                      net->ipv6.devconf_all);
 758         } else if ((!newf) ^ (!old))
 759                 dev_forward_change((struct inet6_dev *)table->extra1);
 760         rtnl_unlock();
 761
 762         if (newf)
 763                 rt6_purge_dflt_routers(net);
 764         return 1;
 765 }
 766 #endif
 767
 768 /* Nobody refers to this ifaddr, destroy it */
 769 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 770 {
 771         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 772
 773 #ifdef NET_REFCNT_DEBUG
 774         pr_debug("%s\n", __func__);
 775 #endif
 776
 777         in6_dev_put(ifp->idev);
 778
 779         if (del_timer(&ifp->timer))
 780                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 781
 782         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 783                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 784                 return;
 785         }
 786         ip6_rt_put(ifp->rt);
 787
 788         kfree_rcu(ifp, rcu);
 789 }
 790
 791 static void
 792 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 793 {
 794         struct list_head *p;
 795         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 796
 797         /*
 798          * Each device address list is sorted in order of scope -
 799          * global before linklocal.
 800          */
 801         list_for_each(p, &idev->addr_list) {
 802                 struct inet6_ifaddr *ifa
 803                         = list_entry(p, struct inet6_ifaddr, if_list);
 804                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 805                         break;
 806         }
 807
 808         list_add_tail(&ifp->if_list, p);
 809 }
 810
 811 static u32 inet6_addr_hash(const struct in6_addr *addr)
 812 {
 813         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 814 }
 815
 816 /* On success it returns ifp with increased reference count */
 817
 818 static struct inet6_ifaddr *
 819 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 820               int scope, u32 flags)
 821 {
 822         struct inet6_ifaddr *ifa = NULL;
 823         struct rt6_info *rt;
 824         unsigned int hash;
 825         int err = 0;
 826         int addr_type = ipv6_addr_type(addr);
 827
 828         if (addr_type == IPV6_ADDR_ANY ||
 829             addr_type & IPV6_ADDR_MULTICAST ||
 830             (!(idev->dev->flags & IFF_LOOPBACK) &&
 831              addr_type & IPV6_ADDR_LOOPBACK))
 832                 return ERR_PTR(-EADDRNOTAVAIL);
 833
 834         rcu_read_lock_bh();
 835         if (idev->dead) {
 836                 err = -ENODEV;                  /*XXX*/
 837                 goto out2;
 838         }
 839
 840         if (idev->cnf.disable_ipv6) {
 841                 err = -EACCES;
 842                 goto out2;
 843         }
 844
 845         spin_lock(&addrconf_hash_lock);
 846
 847         /* Ignore adding duplicate addresses on an interface */
 848         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 849                 ADBG(("ipv6_add_addr: already assigned\n"));
 850                 err = -EEXIST;
 851                 goto out;
 852         }
 853
 854         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 855
 856         if (ifa == NULL) {
 857                 ADBG(("ipv6_add_addr: malloc failed\n"));
 858                 err = -ENOBUFS;
 859                 goto out;
 860         }
 861
 862         rt = addrconf_dst_alloc(idev, addr, false);
 863         if (IS_ERR(rt)) {
 864                 err = PTR_ERR(rt);
 865                 goto out;
 866         }
 867
 868         ifa->addr = *addr;
 869
 870         spin_lock_init(&ifa->lock);
 871         spin_lock_init(&ifa->state_lock);
 872         init_timer(&ifa->timer);
 873         INIT_HLIST_NODE(&ifa->addr_lst);
 874         ifa->timer.data = (unsigned long) ifa;
 875         ifa->scope = scope;
 876         ifa->prefix_len = pfxlen;
 877         ifa->flags = flags | IFA_F_TENTATIVE;
 878         ifa->cstamp = ifa->tstamp = jiffies;
 879         ifa->tokenized = false;
 880
 881         ifa->rt = rt;
 882
 883         ifa->idev = idev;
 884         in6_dev_hold(idev);
 885         /* For caller */
 886         in6_ifa_hold(ifa);
 887
 888         /* Add to big hash table */
 889         hash = inet6_addr_hash(addr);
 890
 891         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 892         spin_unlock(&addrconf_hash_lock);
 893
 894         write_lock(&idev->lock);
 895         /* Add to inet6_dev unicast addr list. */
 896         ipv6_link_dev_addr(idev, ifa);
 897
 898 #ifdef CONFIG_IPV6_PRIVACY
 899         if (ifa->flags&IFA_F_TEMPORARY) {
 900                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 901                 in6_ifa_hold(ifa);
 902         }
 903 #endif
 904
 905         in6_ifa_hold(ifa);
 906         write_unlock(&idev->lock);
 907 out2:
 908         rcu_read_unlock_bh();
 909
 910         if (likely(err == 0))
 911                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 912         else {
 913                 kfree(ifa);
 914                 ifa = ERR_PTR(err);
 915         }
 916
 917         return ifa;
 918 out:
 919         spin_unlock(&addrconf_hash_lock);
 920         goto out2;
 921 }
 922
 923 /* This function wants to get referenced ifp and releases it before return */
 924
 925 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 926 {
 927         struct inet6_ifaddr *ifa, *ifn;
 928         struct inet6_dev *idev = ifp->idev;
 929         int state;
 930         int deleted = 0, onlink = 0;
 931         unsigned long expires = jiffies;
 932
 933         spin_lock_bh(&ifp->state_lock);
 934         state = ifp->state;
 935         ifp->state = INET6_IFADDR_STATE_DEAD;
 936         spin_unlock_bh(&ifp->state_lock);
 937
 938         if (state == INET6_IFADDR_STATE_DEAD)
 939                 goto out;
 940
 941         spin_lock_bh(&addrconf_hash_lock);
 942         hlist_del_init_rcu(&ifp->addr_lst);
 943         spin_unlock_bh(&addrconf_hash_lock);
 944
 945         write_lock_bh(&idev->lock);
 946 #ifdef CONFIG_IPV6_PRIVACY
 947         if (ifp->flags&IFA_F_TEMPORARY) {
 948                 list_del(&ifp->tmp_list);
 949                 if (ifp->ifpub) {
 950                         in6_ifa_put(ifp->ifpub);
 951                         ifp->ifpub = NULL;
 952                 }
 953                 __in6_ifa_put(ifp);
 954         }
 955 #endif
 956
 957         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 958                 if (ifa == ifp) {
 959                         list_del_init(&ifp->if_list);
 960                         __in6_ifa_put(ifp);
 961
 962                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 963                                 break;
 964                         deleted = 1;
 965                         continue;
 966                 } else if (ifp->flags & IFA_F_PERMANENT) {
 967                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 968                                               ifp->prefix_len)) {
 969                                 if (ifa->flags & IFA_F_PERMANENT) {
 970                                         onlink = 1;
 971                                         if (deleted)
 972                                                 break;
 973                                 } else {
 974                                         unsigned long lifetime;
 975
 976                                         if (!onlink)
 977                                                 onlink = -1;
 978
 979                                         spin_lock(&ifa->lock);
 980
 981                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 982                                         /*
 983                                          * Note: Because this address is
 984                                          * not permanent, lifetime <
 985                                          * LONG_MAX / HZ here.
 986                                          */
 987                                         if (time_before(expires,
 988                                                         ifa->tstamp + lifetime * HZ))
 989                                                 expires = ifa->tstamp + lifetime * HZ;
 990                                         spin_unlock(&ifa->lock);
 991                                 }
 992                         }
 993                 }
 994         }
 995         write_unlock_bh(&idev->lock);
 996
 997         addrconf_del_timer(ifp);
 998
 999         ipv6_ifa_notify(RTM_DELADDR, ifp);
1000
1001         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1002
1003         /*
1004          * Purge or update corresponding prefix
1005          *
1006          * 1) we don't purge prefix here if address was not permanent.
1007          *    prefix is managed by its own lifetime.
1008          * 2) if there're no addresses, delete prefix.
1009          * 3) if there're still other permanent address(es),
1010          *    corresponding prefix is still permanent.
1011          * 4) otherwise, update prefix lifetime to the
1012          *    longest valid lifetime among the corresponding
1013          *    addresses on the device.
1014          *    Note: subsequent RA will update lifetime.
1015          *
1016          * --yoshfuji
1017          */
1018         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1019                 struct in6_addr prefix;
1020                 struct rt6_info *rt;
1021
1022                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1023
1024                 rt = addrconf_get_prefix_route(&prefix,
1025                                                ifp->prefix_len,
1026                                                ifp->idev->dev,
1027                                                0, RTF_GATEWAY | RTF_DEFAULT);
1028
1029                 if (rt) {
1030                         if (onlink == 0) {
1031                                 ip6_del_rt(rt);
1032                                 rt = NULL;
1033                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1034                                 rt6_set_expires(rt, expires);
1035                         }
1036                 }
1037                 ip6_rt_put(rt);
1038         }
1039
1040         /* clean up prefsrc entries */
1041         rt6_remove_prefsrc(ifp);
1042 out:
1043         in6_ifa_put(ifp);
1044 }
1045
1046 #ifdef CONFIG_IPV6_PRIVACY
1047 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1048 {
1049         struct inet6_dev *idev = ifp->idev;
1050         struct in6_addr addr, *tmpaddr;
1051         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1052         unsigned long regen_advance;
1053         int tmp_plen;
1054         int ret = 0;
1055         int max_addresses;
1056         u32 addr_flags;
1057         unsigned long now = jiffies;
1058
1059         write_lock(&idev->lock);
1060         if (ift) {
1061                 spin_lock_bh(&ift->lock);
1062                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1063                 spin_unlock_bh(&ift->lock);
1064                 tmpaddr = &addr;
1065         } else {
1066                 tmpaddr = NULL;
1067         }
1068 retry:
1069         in6_dev_hold(idev);
1070         if (idev->cnf.use_tempaddr <= 0) {
1071                 write_unlock(&idev->lock);
1072                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1073                 in6_dev_put(idev);
1074                 ret = -1;
1075                 goto out;
1076         }
1077         spin_lock_bh(&ifp->lock);
1078         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1079                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1080                 spin_unlock_bh(&ifp->lock);
1081                 write_unlock(&idev->lock);
1082                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1083                         __func__);
1084                 in6_dev_put(idev);
1085                 ret = -1;
1086                 goto out;
1087         }
1088         in6_ifa_hold(ifp);
1089         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1090         __ipv6_try_regen_rndid(idev, tmpaddr);
1091         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1092         age = (now - ifp->tstamp) / HZ;
1093         tmp_valid_lft = min_t(__u32,
1094                               ifp->valid_lft,
1095                               idev->cnf.temp_valid_lft + age);
1096         tmp_prefered_lft = min_t(__u32,
1097                                  ifp->prefered_lft,
1098                                  idev->cnf.temp_prefered_lft + age -
1099                                  idev->cnf.max_desync_factor);
1100         tmp_plen = ifp->prefix_len;
1101         max_addresses = idev->cnf.max_addresses;
1102         tmp_tstamp = ifp->tstamp;
1103         spin_unlock_bh(&ifp->lock);
1104
1105         regen_advance = idev->cnf.regen_max_retry *
1106                         idev->cnf.dad_transmits *
1107                         idev->nd_parms->retrans_time / HZ;
1108         write_unlock(&idev->lock);
1109
1110         /* A temporary address is created only if this calculated Preferred
1111          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1112          * an implementation must not create a temporary address with a zero
1113          * Preferred Lifetime.
1114          */
1115         if (tmp_prefered_lft <= regen_advance) {
1116                 in6_ifa_put(ifp);
1117                 in6_dev_put(idev);
1118                 ret = -1;
1119                 goto out;
1120         }
1121
1122         addr_flags = IFA_F_TEMPORARY;
1123         /* set in addrconf_prefix_rcv() */
1124         if (ifp->flags & IFA_F_OPTIMISTIC)
1125                 addr_flags |= IFA_F_OPTIMISTIC;
1126
1127         ift = !max_addresses ||
1128               ipv6_count_addresses(idev) < max_addresses ?
1129                 ipv6_add_addr(idev, &addr, tmp_plen, ipv6_addr_scope(&addr),
1130                               addr_flags) : NULL;
1131         if (IS_ERR_OR_NULL(ift)) {
1132                 in6_ifa_put(ifp);
1133                 in6_dev_put(idev);
1134                 pr_info("%s: retry temporary address regeneration\n", __func__);
1135                 tmpaddr = &addr;
1136                 write_lock(&idev->lock);
1137                 goto retry;
1138         }
1139
1140         spin_lock_bh(&ift->lock);
1141         ift->ifpub = ifp;
1142         ift->valid_lft = tmp_valid_lft;
1143         ift->prefered_lft = tmp_prefered_lft;
1144         ift->cstamp = now;
1145         ift->tstamp = tmp_tstamp;
1146         spin_unlock_bh(&ift->lock);
1147
1148         addrconf_dad_start(ift);
1149         in6_ifa_put(ift);
1150         in6_dev_put(idev);
1151 out:
1152         return ret;
1153 }
1154 #endif
1155
1156 /*
1157  *      Choose an appropriate source address (RFC3484)
1158  */
1159 enum {
1160         IPV6_SADDR_RULE_INIT = 0,
1161         IPV6_SADDR_RULE_LOCAL,
1162         IPV6_SADDR_RULE_SCOPE,
1163         IPV6_SADDR_RULE_PREFERRED,
1164 #ifdef CONFIG_IPV6_MIP6
1165         IPV6_SADDR_RULE_HOA,
1166 #endif
1167         IPV6_SADDR_RULE_OIF,
1168         IPV6_SADDR_RULE_LABEL,
1169 #ifdef CONFIG_IPV6_PRIVACY
1170         IPV6_SADDR_RULE_PRIVACY,
1171 #endif
1172         IPV6_SADDR_RULE_ORCHID,
1173         IPV6_SADDR_RULE_PREFIX,
1174         IPV6_SADDR_RULE_MAX
1175 };
1176
1177 struct ipv6_saddr_score {
1178         int                     rule;
1179         int                     addr_type;
1180         struct inet6_ifaddr     *ifa;
1181         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1182         int                     scopedist;
1183         int                     matchlen;
1184 };
1185
1186 struct ipv6_saddr_dst {
1187         const struct in6_addr *addr;
1188         int ifindex;
1189         int scope;
1190         int label;
1191         unsigned int prefs;
1192 };
1193
1194 static inline int ipv6_saddr_preferred(int type)
1195 {
1196         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1197                 return 1;
1198         return 0;
1199 }
1200
1201 static int ipv6_get_saddr_eval(struct net *net,
1202                                struct ipv6_saddr_score *score,
1203                                struct ipv6_saddr_dst *dst,
1204                                int i)
1205 {
1206         int ret;
1207
1208         if (i <= score->rule) {
1209                 switch (i) {
1210                 case IPV6_SADDR_RULE_SCOPE:
1211                         ret = score->scopedist;
1212                         break;
1213                 case IPV6_SADDR_RULE_PREFIX:
1214                         ret = score->matchlen;
1215                         break;
1216                 default:
1217                         ret = !!test_bit(i, score->scorebits);
1218                 }
1219                 goto out;
1220         }
1221
1222         switch (i) {
1223         case IPV6_SADDR_RULE_INIT:
1224                 /* Rule 0: remember if hiscore is not ready yet */
1225                 ret = !!score->ifa;
1226                 break;
1227         case IPV6_SADDR_RULE_LOCAL:
1228                 /* Rule 1: Prefer same address */
1229                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1230                 break;
1231         case IPV6_SADDR_RULE_SCOPE:
1232                 /* Rule 2: Prefer appropriate scope
1233                  *
1234                  *      ret
1235                  *       ^
1236                  *    -1 |  d 15
1237                  *    ---+--+-+---> scope
1238                  *       |
1239                  *       |             d is scope of the destination.
1240                  *  B-d  |  \
1241                  *       |   \      <- smaller scope is better if
1242                  *  B-15 |    \        if scope is enough for destinaion.
1243                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1244                  * d-C-1 | /
1245                  *       |/         <- greater is better
1246                  *   -C  /             if scope is not enough for destination.
1247                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1248                  *
1249                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1250                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1251                  * Assume B = 0 and we get C > 29.
1252                  */
1253                 ret = __ipv6_addr_src_scope(score->addr_type);
1254                 if (ret >= dst->scope)
1255                         ret = -ret;
1256                 else
1257                         ret -= 128;     /* 30 is enough */
1258                 score->scopedist = ret;
1259                 break;
1260         case IPV6_SADDR_RULE_PREFERRED:
1261                 /* Rule 3: Avoid deprecated and optimistic addresses */
1262                 ret = ipv6_saddr_preferred(score->addr_type) ||
1263                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1264                 break;
1265 #ifdef CONFIG_IPV6_MIP6
1266         case IPV6_SADDR_RULE_HOA:
1267             {
1268                 /* Rule 4: Prefer home address */
1269                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1270                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1271                 break;
1272             }
1273 #endif
1274         case IPV6_SADDR_RULE_OIF:
1275                 /* Rule 5: Prefer outgoing interface */
1276                 ret = (!dst->ifindex ||
1277                        dst->ifindex == score->ifa->idev->dev->ifindex);
1278                 break;
1279         case IPV6_SADDR_RULE_LABEL:
1280                 /* Rule 6: Prefer matching label */
1281                 ret = ipv6_addr_label(net,
1282                                       &score->ifa->addr, score->addr_type,
1283                                       score->ifa->idev->dev->ifindex) == dst->label;
1284                 break;
1285 #ifdef CONFIG_IPV6_PRIVACY
1286         case IPV6_SADDR_RULE_PRIVACY:
1287             {
1288                 /* Rule 7: Prefer public address
1289                  * Note: prefer temporary address if use_tempaddr >= 2
1290                  */
1291                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1292                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1293                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1294                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1295                 break;
1296             }
1297 #endif
1298         case IPV6_SADDR_RULE_ORCHID:
1299                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1300                  *          non-ORCHID vs non-ORCHID
1301                  */
1302                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1303                         ipv6_addr_orchid(dst->addr));
1304                 break;
1305         case IPV6_SADDR_RULE_PREFIX:
1306                 /* Rule 8: Use longest matching prefix */
1307                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1308                 if (ret > score->ifa->prefix_len)
1309                         ret = score->ifa->prefix_len;
1310                 score->matchlen = ret;
1311                 break;
1312         default:
1313                 ret = 0;
1314         }
1315
1316         if (ret)
1317                 __set_bit(i, score->scorebits);
1318         score->rule = i;
1319 out:
1320         return ret;
1321 }
1322
1323 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1324                        const struct in6_addr *daddr, unsigned int prefs,
1325                        struct in6_addr *saddr)
1326 {
1327         struct ipv6_saddr_score scores[2],
1328                                 *score = &scores[0], *hiscore = &scores[1];
1329         struct ipv6_saddr_dst dst;
1330         struct net_device *dev;
1331         int dst_type;
1332
1333         dst_type = __ipv6_addr_type(daddr);
1334         dst.addr = daddr;
1335         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1336         dst.scope = __ipv6_addr_src_scope(dst_type);
1337         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1338         dst.prefs = prefs;
1339
1340         hiscore->rule = -1;
1341         hiscore->ifa = NULL;
1342
1343         rcu_read_lock();
1344
1345         for_each_netdev_rcu(net, dev) {
1346                 struct inet6_dev *idev;
1347
1348                 /* Candidate Source Address (section 4)
1349                  *  - multicast and link-local destination address,
1350                  *    the set of candidate source address MUST only
1351                  *    include addresses assigned to interfaces
1352                  *    belonging to the same link as the outgoing
1353                  *    interface.
1354                  * (- For site-local destination addresses, the
1355                  *    set of candidate source addresses MUST only
1356                  *    include addresses assigned to interfaces
1357                  *    belonging to the same site as the outgoing
1358                  *    interface.)
1359                  */
1360                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1361                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1362                     dst.ifindex && dev->ifindex != dst.ifindex)
1363                         continue;
1364
1365                 idev = __in6_dev_get(dev);
1366                 if (!idev)
1367                         continue;
1368
1369                 read_lock_bh(&idev->lock);
1370                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1371                         int i;
1372
1373                         /*
1374                          * - Tentative Address (RFC2462 section 5.4)
1375                          *  - A tentative address is not considered
1376                          *    "assigned to an interface" in the traditional
1377                          *    sense, unless it is also flagged as optimistic.
1378                          * - Candidate Source Address (section 4)
1379                          *  - In any case, anycast addresses, multicast
1380                          *    addresses, and the unspecified address MUST
1381                          *    NOT be included in a candidate set.
1382                          */
1383                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1384                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1385                                 continue;
1386
1387                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1388
1389                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1390                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1391                                 LIMIT_NETDEBUG(KERN_DEBUG
1392                                                "ADDRCONF: unspecified / multicast address "
1393                                                "assigned as unicast address on %s",
1394                                                dev->name);
1395                                 continue;
1396                         }
1397
1398                         score->rule = -1;
1399                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1400
1401                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1402                                 int minihiscore, miniscore;
1403
1404                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1405                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1406
1407                                 if (minihiscore > miniscore) {
1408                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1409                                             score->scopedist > 0) {
1410                                                 /*
1411                                                  * special case:
1412                                                  * each remaining entry
1413                                                  * has too small (not enough)
1414                                                  * scope, because ifa entries
1415                                                  * are sorted by their scope
1416                                                  * values.
1417                                                  */
1418                                                 goto try_nextdev;
1419                                         }
1420                                         break;
1421                                 } else if (minihiscore < miniscore) {
1422                                         if (hiscore->ifa)
1423                                                 in6_ifa_put(hiscore->ifa);
1424
1425                                         in6_ifa_hold(score->ifa);
1426
1427                                         swap(hiscore, score);
1428
1429                                         /* restore our iterator */
1430                                         score->ifa = hiscore->ifa;
1431
1432                                         break;
1433                                 }
1434                         }
1435                 }
1436 try_nextdev:
1437                 read_unlock_bh(&idev->lock);
1438         }
1439         rcu_read_unlock();
1440
1441         if (!hiscore->ifa)
1442                 return -EADDRNOTAVAIL;
1443
1444         *saddr = hiscore->ifa->addr;
1445         in6_ifa_put(hiscore->ifa);
1446         return 0;
1447 }
1448 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1449
1450 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1451                     unsigned char banned_flags)
1452 {
1453         struct inet6_dev *idev;
1454         int err = -EADDRNOTAVAIL;
1455
1456         rcu_read_lock();
1457         idev = __in6_dev_get(dev);
1458         if (idev) {
1459                 struct inet6_ifaddr *ifp;
1460
1461                 read_lock_bh(&idev->lock);
1462                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1463                         if (ifp->scope == IFA_LINK &&
1464                             !(ifp->flags & banned_flags)) {
1465                                 *addr = ifp->addr;
1466                                 err = 0;
1467                                 break;
1468                         }
1469                 }
1470                 read_unlock_bh(&idev->lock);
1471         }
1472         rcu_read_unlock();
1473         return err;
1474 }
1475
1476 static int ipv6_count_addresses(struct inet6_dev *idev)
1477 {
1478         int cnt = 0;
1479         struct inet6_ifaddr *ifp;
1480
1481         read_lock_bh(&idev->lock);
1482         list_for_each_entry(ifp, &idev->addr_list, if_list)
1483                 cnt++;
1484         read_unlock_bh(&idev->lock);
1485         return cnt;
1486 }
1487
1488 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1489                   struct net_device *dev, int strict)
1490 {
1491         struct inet6_ifaddr *ifp;
1492         unsigned int hash = inet6_addr_hash(addr);
1493
1494         rcu_read_lock_bh();
1495         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1496                 if (!net_eq(dev_net(ifp->idev->dev), net))
1497                         continue;
1498                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1499                     !(ifp->flags&IFA_F_TENTATIVE) &&
1500                     (dev == NULL || ifp->idev->dev == dev ||
1501                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1502                         rcu_read_unlock_bh();
1503                         return 1;
1504                 }
1505         }
1506
1507         rcu_read_unlock_bh();
1508         return 0;
1509 }
1510 EXPORT_SYMBOL(ipv6_chk_addr);
1511
1512 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1513                                struct net_device *dev)
1514 {
1515         unsigned int hash = inet6_addr_hash(addr);
1516         struct inet6_ifaddr *ifp;
1517
1518         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1519                 if (!net_eq(dev_net(ifp->idev->dev), net))
1520                         continue;
1521                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1522                         if (dev == NULL || ifp->idev->dev == dev)
1523                                 return true;
1524                 }
1525         }
1526         return false;
1527 }
1528
1529 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1530 {
1531         struct inet6_dev *idev;
1532         struct inet6_ifaddr *ifa;
1533         int     onlink;
1534
1535         onlink = 0;
1536         rcu_read_lock();
1537         idev = __in6_dev_get(dev);
1538         if (idev) {
1539                 read_lock_bh(&idev->lock);
1540                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1541                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1542                                                    ifa->prefix_len);
1543                         if (onlink)
1544                                 break;
1545                 }
1546                 read_unlock_bh(&idev->lock);
1547         }
1548         rcu_read_unlock();
1549         return onlink;
1550 }
1551 EXPORT_SYMBOL(ipv6_chk_prefix);
1552
1553 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1554                                      struct net_device *dev, int strict)
1555 {
1556         struct inet6_ifaddr *ifp, *result = NULL;
1557         unsigned int hash = inet6_addr_hash(addr);
1558
1559         rcu_read_lock_bh();
1560         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1561                 if (!net_eq(dev_net(ifp->idev->dev), net))
1562                         continue;
1563                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1564                         if (dev == NULL || ifp->idev->dev == dev ||
1565                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1566                                 result = ifp;
1567                                 in6_ifa_hold(ifp);
1568                                 break;
1569                         }
1570                 }
1571         }
1572         rcu_read_unlock_bh();
1573
1574         return result;
1575 }
1576
1577 /* Gets referenced address, destroys ifaddr */
1578
1579 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1580 {
1581         if (ifp->flags&IFA_F_PERMANENT) {
1582                 spin_lock_bh(&ifp->lock);
1583                 addrconf_del_timer(ifp);
1584                 ifp->flags |= IFA_F_TENTATIVE;
1585                 if (dad_failed)
1586                         ifp->flags |= IFA_F_DADFAILED;
1587                 spin_unlock_bh(&ifp->lock);
1588                 if (dad_failed)
1589                         ipv6_ifa_notify(0, ifp);
1590                 in6_ifa_put(ifp);
1591 #ifdef CONFIG_IPV6_PRIVACY
1592         } else if (ifp->flags&IFA_F_TEMPORARY) {
1593                 struct inet6_ifaddr *ifpub;
1594                 spin_lock_bh(&ifp->lock);
1595                 ifpub = ifp->ifpub;
1596                 if (ifpub) {
1597                         in6_ifa_hold(ifpub);
1598                         spin_unlock_bh(&ifp->lock);
1599                         ipv6_create_tempaddr(ifpub, ifp);
1600                         in6_ifa_put(ifpub);
1601                 } else {
1602                         spin_unlock_bh(&ifp->lock);
1603                 }
1604                 ipv6_del_addr(ifp);
1605 #endif
1606         } else
1607                 ipv6_del_addr(ifp);
1608 }
1609
1610 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1611 {
1612         int err = -ENOENT;
1613
1614         spin_lock(&ifp->state_lock);
1615         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1616                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1617                 err = 0;
1618         }
1619         spin_unlock(&ifp->state_lock);
1620
1621         return err;
1622 }
1623
1624 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1625 {
1626         struct inet6_dev *idev = ifp->idev;
1627
1628         if (addrconf_dad_end(ifp)) {
1629                 in6_ifa_put(ifp);
1630                 return;
1631         }
1632
1633         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1634                              ifp->idev->dev->name, &ifp->addr);
1635
1636         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1637                 struct in6_addr addr;
1638
1639                 addr.s6_addr32[0] = htonl(0xfe800000);
1640                 addr.s6_addr32[1] = 0;
1641
1642                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1643                     ipv6_addr_equal(&ifp->addr, &addr)) {
1644                         /* DAD failed for link-local based on MAC address */
1645                         idev->cnf.disable_ipv6 = 1;
1646
1647                         pr_info("%s: IPv6 being disabled!\n",
1648                                 ifp->idev->dev->name);
1649                 }
1650         }
1651
1652         addrconf_dad_stop(ifp, 1);
1653 }
1654
1655 /* Join to solicited addr multicast group. */
1656
1657 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1658 {
1659         struct in6_addr maddr;
1660
1661         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1662                 return;
1663
1664         addrconf_addr_solict_mult(addr, &maddr);
1665         ipv6_dev_mc_inc(dev, &maddr);
1666 }
1667
1668 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1669 {
1670         struct in6_addr maddr;
1671
1672         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1673                 return;
1674
1675         addrconf_addr_solict_mult(addr, &maddr);
1676         __ipv6_dev_mc_dec(idev, &maddr);
1677 }
1678
1679 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1680 {
1681         struct in6_addr addr;
1682         if (ifp->prefix_len == 127) /* RFC 6164 */
1683                 return;
1684         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1685         if (ipv6_addr_any(&addr))
1686                 return;
1687         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1688 }
1689
1690 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1691 {
1692         struct in6_addr addr;
1693         if (ifp->prefix_len == 127) /* RFC 6164 */
1694                 return;
1695         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1696         if (ipv6_addr_any(&addr))
1697                 return;
1698         __ipv6_dev_ac_dec(ifp->idev, &addr);
1699 }
1700
1701 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1702 {
1703         if (dev->addr_len != ETH_ALEN)
1704                 return -1;
1705         memcpy(eui, dev->dev_addr, 3);
1706         memcpy(eui + 5, dev->dev_addr + 3, 3);
1707
1708         /*
1709          * The zSeries OSA network cards can be shared among various
1710          * OS instances, but the OSA cards have only one MAC address.
1711          * This leads to duplicate address conflicts in conjunction
1712          * with IPv6 if more than one instance uses the same card.
1713          *
1714          * The driver for these cards can deliver a unique 16-bit
1715          * identifier for each instance sharing the same card.  It is
1716          * placed instead of 0xFFFE in the interface identifier.  The
1717          * "u" bit of the interface identifier is not inverted in this
1718          * case.  Hence the resulting interface identifier has local
1719          * scope according to RFC2373.
1720          */
1721         if (dev->dev_id) {
1722                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1723                 eui[4] = dev->dev_id & 0xFF;
1724         } else {
1725                 eui[3] = 0xFF;
1726                 eui[4] = 0xFE;
1727                 eui[0] ^= 2;
1728         }
1729         return 0;
1730 }
1731
1732 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1733 {
1734         if (dev->addr_len != IEEE802154_ADDR_LEN)
1735                 return -1;
1736         memcpy(eui, dev->dev_addr, 8);
1737         eui[0] ^= 2;
1738         return 0;
1739 }
1740
1741 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1742 {
1743         union fwnet_hwaddr *ha;
1744
1745         if (dev->addr_len != FWNET_ALEN)
1746                 return -1;
1747
1748         ha = (union fwnet_hwaddr *)dev->dev_addr;
1749
1750         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1751         eui[0] ^= 2;
1752         return 0;
1753 }
1754
1755 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1756 {
1757         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1758         if (dev->addr_len != ARCNET_ALEN)
1759                 return -1;
1760         memset(eui, 0, 7);
1761         eui[7] = *(u8 *)dev->dev_addr;
1762         return 0;
1763 }
1764
1765 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1766 {
1767         if (dev->addr_len != INFINIBAND_ALEN)
1768                 return -1;
1769         memcpy(eui, dev->dev_addr + 12, 8);
1770         eui[0] |= 2;
1771         return 0;
1772 }
1773
1774 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1775 {
1776         if (addr == 0)
1777                 return -1;
1778         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1779                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1780                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1781                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1782                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1783                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1784         eui[1] = 0;
1785         eui[2] = 0x5E;
1786         eui[3] = 0xFE;
1787         memcpy(eui + 4, &addr, 4);
1788         return 0;
1789 }
1790
1791 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1792 {
1793         if (dev->priv_flags & IFF_ISATAP)
1794                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1795         return -1;
1796 }
1797
1798 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1799 {
1800         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1801 }
1802
1803 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1804 {
1805         switch (dev->type) {
1806         case ARPHRD_ETHER:
1807         case ARPHRD_FDDI:
1808                 return addrconf_ifid_eui48(eui, dev);
1809         case ARPHRD_ARCNET:
1810                 return addrconf_ifid_arcnet(eui, dev);
1811         case ARPHRD_INFINIBAND:
1812                 return addrconf_ifid_infiniband(eui, dev);
1813         case ARPHRD_SIT:
1814                 return addrconf_ifid_sit(eui, dev);
1815         case ARPHRD_IPGRE:
1816                 return addrconf_ifid_gre(eui, dev);
1817         case ARPHRD_IEEE802154:
1818                 return addrconf_ifid_eui64(eui, dev);
1819         case ARPHRD_IEEE1394:
1820                 return addrconf_ifid_ieee1394(eui, dev);
1821         }
1822         return -1;
1823 }
1824
1825 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1826 {
1827         int err = -1;
1828         struct inet6_ifaddr *ifp;
1829
1830         read_lock_bh(&idev->lock);
1831         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1832                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1833                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1834                         err = 0;
1835                         break;
1836                 }
1837         }
1838         read_unlock_bh(&idev->lock);
1839         return err;
1840 }
1841
1842 #ifdef CONFIG_IPV6_PRIVACY
1843 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1844 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1845 {
1846 regen:
1847         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1848         idev->rndid[0] &= ~0x02;
1849
1850         /*
1851          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1852          * check if generated address is not inappropriate
1853          *
1854          *  - Reserved subnet anycast (RFC 2526)
1855          *      11111101 11....11 1xxxxxxx
1856          *  - ISATAP (RFC4214) 6.1
1857          *      00-00-5E-FE-xx-xx-xx-xx
1858          *  - value 0
1859          *  - XXX: already assigned to an address on the device
1860          */
1861         if (idev->rndid[0] == 0xfd &&
1862             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1863             (idev->rndid[7]&0x80))
1864                 goto regen;
1865         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1866                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1867                         goto regen;
1868                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1869                         goto regen;
1870         }
1871 }
1872
1873 static void ipv6_regen_rndid(unsigned long data)
1874 {
1875         struct inet6_dev *idev = (struct inet6_dev *) data;
1876         unsigned long expires;
1877
1878         rcu_read_lock_bh();
1879         write_lock_bh(&idev->lock);
1880
1881         if (idev->dead)
1882                 goto out;
1883
1884         __ipv6_regen_rndid(idev);
1885
1886         expires = jiffies +
1887                 idev->cnf.temp_prefered_lft * HZ -
1888                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1889                 idev->cnf.max_desync_factor * HZ;
1890         if (time_before(expires, jiffies)) {
1891                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1892                         __func__, idev->dev->name);
1893                 goto out;
1894         }
1895
1896         if (!mod_timer(&idev->regen_timer, expires))
1897                 in6_dev_hold(idev);
1898
1899 out:
1900         write_unlock_bh(&idev->lock);
1901         rcu_read_unlock_bh();
1902         in6_dev_put(idev);
1903 }
1904
1905 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1906 {
1907         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1908                 __ipv6_regen_rndid(idev);
1909 }
1910 #endif
1911
1912 /*
1913  *      Add prefix route.
1914  */
1915
1916 static void
1917 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1918                       unsigned long expires, u32 flags)
1919 {
1920         struct fib6_config cfg = {
1921                 .fc_table = RT6_TABLE_PREFIX,
1922                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1923                 .fc_ifindex = dev->ifindex,
1924                 .fc_expires = expires,
1925                 .fc_dst_len = plen,
1926                 .fc_flags = RTF_UP | flags,
1927                 .fc_nlinfo.nl_net = dev_net(dev),
1928                 .fc_protocol = RTPROT_KERNEL,
1929         };
1930
1931         cfg.fc_dst = *pfx;
1932
1933         /* Prevent useless cloning on PtP SIT.
1934            This thing is done here expecting that the whole
1935            class of non-broadcast devices need not cloning.
1936          */
1937 #if IS_ENABLED(CONFIG_IPV6_SIT)
1938         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1939                 cfg.fc_flags |= RTF_NONEXTHOP;
1940 #endif
1941
1942         ip6_route_add(&cfg);
1943 }
1944
1945
1946 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1947                                                   int plen,
1948                                                   const struct net_device *dev,
1949                                                   u32 flags, u32 noflags)
1950 {
1951         struct fib6_node *fn;
1952         struct rt6_info *rt = NULL;
1953         struct fib6_table *table;
1954
1955         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1956         if (table == NULL)
1957                 return NULL;
1958
1959         read_lock_bh(&table->tb6_lock);
1960         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1961         if (!fn)
1962                 goto out;
1963         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1964                 if (rt->dst.dev->ifindex != dev->ifindex)
1965                         continue;
1966                 if ((rt->rt6i_flags & flags) != flags)
1967                         continue;
1968                 if ((rt->rt6i_flags & noflags) != 0)
1969                         continue;
1970                 dst_hold(&rt->dst);
1971                 break;
1972         }
1973 out:
1974         read_unlock_bh(&table->tb6_lock);
1975         return rt;
1976 }
1977
1978
1979 /* Create "default" multicast route to the interface */
1980
1981 static void addrconf_add_mroute(struct net_device *dev)
1982 {
1983         struct fib6_config cfg = {
1984                 .fc_table = RT6_TABLE_LOCAL,
1985                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1986                 .fc_ifindex = dev->ifindex,
1987                 .fc_dst_len = 8,
1988                 .fc_flags = RTF_UP,
1989                 .fc_nlinfo.nl_net = dev_net(dev),
1990         };
1991
1992         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1993
1994         ip6_route_add(&cfg);
1995 }
1996
1997 #if IS_ENABLED(CONFIG_IPV6_SIT)
1998 static void sit_route_add(struct net_device *dev)
1999 {
2000         struct fib6_config cfg = {
2001                 .fc_table = RT6_TABLE_MAIN,
2002                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2003                 .fc_ifindex = dev->ifindex,
2004                 .fc_dst_len = 96,
2005                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2006                 .fc_nlinfo.nl_net = dev_net(dev),
2007         };
2008
2009         /* prefix length - 96 bits "::d.d.d.d" */
2010         ip6_route_add(&cfg);
2011 }
2012 #endif
2013
2014 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2015 {
2016         struct inet6_dev *idev;
2017
2018         ASSERT_RTNL();
2019
2020         idev = ipv6_find_idev(dev);
2021         if (!idev)
2022                 return ERR_PTR(-ENOBUFS);
2023
2024         if (idev->cnf.disable_ipv6)
2025                 return ERR_PTR(-EACCES);
2026
2027         /* Add default multicast route */
2028         if (!(dev->flags & IFF_LOOPBACK))
2029                 addrconf_add_mroute(dev);
2030
2031         return idev;
2032 }
2033
2034 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2035 {
2036         struct prefix_info *pinfo;
2037         __u32 valid_lft;
2038         __u32 prefered_lft;
2039         int addr_type;
2040         struct inet6_dev *in6_dev;
2041         struct net *net = dev_net(dev);
2042
2043         pinfo = (struct prefix_info *) opt;
2044
2045         if (len < sizeof(struct prefix_info)) {
2046                 ADBG(("addrconf: prefix option too short\n"));
2047                 return;
2048         }
2049
2050         /*
2051          *      Validation checks ([ADDRCONF], page 19)
2052          */
2053
2054         addr_type = ipv6_addr_type(&pinfo->prefix);
2055
2056         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2057                 return;
2058
2059         valid_lft = ntohl(pinfo->valid);
2060         prefered_lft = ntohl(pinfo->prefered);
2061
2062         if (prefered_lft > valid_lft) {
2063                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2064                 return;
2065         }
2066
2067         in6_dev = in6_dev_get(dev);
2068
2069         if (in6_dev == NULL) {
2070                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2071                                     dev->name);
2072                 return;
2073         }
2074
2075         /*
2076          *      Two things going on here:
2077          *      1) Add routes for on-link prefixes
2078          *      2) Configure prefixes with the auto flag set
2079          */
2080
2081         if (pinfo->onlink) {
2082                 struct rt6_info *rt;
2083                 unsigned long rt_expires;
2084
2085                 /* Avoid arithmetic overflow. Really, we could
2086                  * save rt_expires in seconds, likely valid_lft,
2087                  * but it would require division in fib gc, that it
2088                  * not good.
2089                  */
2090                 if (HZ > USER_HZ)
2091                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2092                 else
2093                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2094
2095                 if (addrconf_finite_timeout(rt_expires))
2096                         rt_expires *= HZ;
2097
2098                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2099                                                pinfo->prefix_len,
2100                                                dev,
2101                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2102                                                RTF_GATEWAY | RTF_DEFAULT);
2103
2104                 if (rt) {
2105                         /* Autoconf prefix route */
2106                         if (valid_lft == 0) {
2107                                 ip6_del_rt(rt);
2108                                 rt = NULL;
2109                         } else if (addrconf_finite_timeout(rt_expires)) {
2110                                 /* not infinity */
2111                                 rt6_set_expires(rt, jiffies + rt_expires);
2112                         } else {
2113                                 rt6_clean_expires(rt);
2114                         }
2115                 } else if (valid_lft) {
2116                         clock_t expires = 0;
2117                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2118                         if (addrconf_finite_timeout(rt_expires)) {
2119                                 /* not infinity */
2120                                 flags |= RTF_EXPIRES;
2121                                 expires = jiffies_to_clock_t(rt_expires);
2122                         }
2123                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2124                                               dev, expires, flags);
2125                 }
2126                 ip6_rt_put(rt);
2127         }
2128
2129         /* Try to figure out our local address for this prefix */
2130
2131         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2132                 struct inet6_ifaddr *ifp;
2133                 struct in6_addr addr;
2134                 int create = 0, update_lft = 0;
2135                 bool tokenized = false;
2136
2137                 if (pinfo->prefix_len == 64) {
2138                         memcpy(&addr, &pinfo->prefix, 8);
2139
2140                         if (!ipv6_addr_any(&in6_dev->token)) {
2141                                 read_lock_bh(&in6_dev->lock);
2142                                 memcpy(addr.s6_addr + 8,
2143                                        in6_dev->token.s6_addr + 8, 8);
2144                                 read_unlock_bh(&in6_dev->lock);
2145                                 tokenized = true;
2146                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2147                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2148                                 in6_dev_put(in6_dev);
2149                                 return;
2150                         }
2151                         goto ok;
2152                 }
2153                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2154                                     pinfo->prefix_len);
2155                 in6_dev_put(in6_dev);
2156                 return;
2157
2158 ok:
2159
2160                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2161
2162                 if (ifp == NULL && valid_lft) {
2163                         int max_addresses = in6_dev->cnf.max_addresses;
2164                         u32 addr_flags = 0;
2165
2166 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2167                         if (in6_dev->cnf.optimistic_dad &&
2168                             !net->ipv6.devconf_all->forwarding && sllao)
2169                                 addr_flags = IFA_F_OPTIMISTIC;
2170 #endif
2171
2172                         /* Do not allow to create too much of autoconfigured
2173                          * addresses; this would be too easy way to crash kernel.
2174                          */
2175                         if (!max_addresses ||
2176                             ipv6_count_addresses(in6_dev) < max_addresses)
2177                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2178                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2179                                                     addr_flags);
2180
2181                         if (IS_ERR_OR_NULL(ifp)) {
2182                                 in6_dev_put(in6_dev);
2183                                 return;
2184                         }
2185
2186                         update_lft = create = 1;
2187                         ifp->cstamp = jiffies;
2188                         ifp->tokenized = tokenized;
2189                         addrconf_dad_start(ifp);
2190                 }
2191
2192                 if (ifp) {
2193                         int flags;
2194                         unsigned long now;
2195 #ifdef CONFIG_IPV6_PRIVACY
2196                         struct inet6_ifaddr *ift;
2197 #endif
2198                         u32 stored_lft;
2199
2200                         /* update lifetime (RFC2462 5.5.3 e) */
2201                         spin_lock(&ifp->lock);
2202                         now = jiffies;
2203                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2204                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2205                         else
2206                                 stored_lft = 0;
2207                         if (!update_lft && stored_lft) {
2208                                 if (valid_lft > MIN_VALID_LIFETIME ||
2209                                     valid_lft > stored_lft)
2210                                         update_lft = 1;
2211                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2212                                         /* valid_lft <= stored_lft is always true */
2213                                         /*
2214                                          * RFC 4862 Section 5.5.3e:
2215                                          * "Note that the preferred lifetime of
2216                                          *  the corresponding address is always
2217                                          *  reset to the Preferred Lifetime in
2218                                          *  the received Prefix Information
2219                                          *  option, regardless of whether the
2220                                          *  valid lifetime is also reset or
2221                                          *  ignored."
2222                                          *
2223                                          *  So if the preferred lifetime in
2224                                          *  this advertisement is different
2225                                          *  than what we have stored, but the
2226                                          *  valid lifetime is invalid, just
2227                                          *  reset prefered_lft.
2228                                          *
2229                                          *  We must set the valid lifetime
2230                                          *  to the stored lifetime since we'll
2231                                          *  be updating the timestamp below,
2232                                          *  else we'll set it back to the
2233                                          *  minimum.
2234                                          */
2235                                         if (prefered_lft != ifp->prefered_lft) {
2236                                                 valid_lft = stored_lft;
2237                                                 update_lft = 1;
2238                                         }
2239                                 } else {
2240                                         valid_lft = MIN_VALID_LIFETIME;
2241                                         if (valid_lft < prefered_lft)
2242                                                 prefered_lft = valid_lft;
2243                                         update_lft = 1;
2244                                 }
2245                         }
2246
2247                         if (update_lft) {
2248                                 ifp->valid_lft = valid_lft;
2249                                 ifp->prefered_lft = prefered_lft;
2250                                 ifp->tstamp = now;
2251                                 flags = ifp->flags;
2252                                 ifp->flags &= ~IFA_F_DEPRECATED;
2253                                 spin_unlock(&ifp->lock);
2254
2255                                 if (!(flags&IFA_F_TENTATIVE))
2256                                         ipv6_ifa_notify(0, ifp);
2257                         } else
2258                                 spin_unlock(&ifp->lock);
2259
2260 #ifdef CONFIG_IPV6_PRIVACY
2261                         read_lock_bh(&in6_dev->lock);
2262                         /* update all temporary addresses in the list */
2263                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2264                                             tmp_list) {
2265                                 int age, max_valid, max_prefered;
2266
2267                                 if (ifp != ift->ifpub)
2268                                         continue;
2269
2270                                 /*
2271                                  * RFC 4941 section 3.3:
2272                                  * If a received option will extend the lifetime
2273                                  * of a public address, the lifetimes of
2274                                  * temporary addresses should be extended,
2275                                  * subject to the overall constraint that no
2276                                  * temporary addresses should ever remain
2277                                  * "valid" or "preferred" for a time longer than
2278                                  * (TEMP_VALID_LIFETIME) or
2279                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2280                                  * respectively.
2281                                  */
2282                                 age = (now - ift->cstamp) / HZ;
2283                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2284                                 if (max_valid < 0)
2285                                         max_valid = 0;
2286
2287                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2288                                                in6_dev->cnf.max_desync_factor -
2289                                                age;
2290                                 if (max_prefered < 0)
2291                                         max_prefered = 0;
2292
2293                                 if (valid_lft > max_valid)
2294                                         valid_lft = max_valid;
2295
2296                                 if (prefered_lft > max_prefered)
2297                                         prefered_lft = max_prefered;
2298
2299                                 spin_lock(&ift->lock);
2300                                 flags = ift->flags;
2301                                 ift->valid_lft = valid_lft;
2302                                 ift->prefered_lft = prefered_lft;
2303                                 ift->tstamp = now;
2304                                 if (prefered_lft > 0)
2305                                         ift->flags &= ~IFA_F_DEPRECATED;
2306
2307                                 spin_unlock(&ift->lock);
2308                                 if (!(flags&IFA_F_TENTATIVE))
2309                                         ipv6_ifa_notify(0, ift);
2310                         }
2311
2312                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2313                                 /*
2314                                  * When a new public address is created as
2315                                  * described in [ADDRCONF], also create a new
2316                                  * temporary address. Also create a temporary
2317                                  * address if it's enabled but no temporary
2318                                  * address currently exists.
2319                                  */
2320                                 read_unlock_bh(&in6_dev->lock);
2321                                 ipv6_create_tempaddr(ifp, NULL);
2322                         } else {
2323                                 read_unlock_bh(&in6_dev->lock);
2324                         }
2325 #endif
2326                         in6_ifa_put(ifp);
2327                         addrconf_verify(0);
2328                 }
2329         }
2330         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2331         in6_dev_put(in6_dev);
2332 }
2333
2334 /*
2335  *      Set destination address.
2336  *      Special case for SIT interfaces where we create a new "virtual"
2337  *      device.
2338  */
2339 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2340 {
2341         struct in6_ifreq ireq;
2342         struct net_device *dev;
2343         int err = -EINVAL;
2344
2345         rtnl_lock();
2346
2347         err = -EFAULT;
2348         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2349                 goto err_exit;
2350
2351         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2352
2353         err = -ENODEV;
2354         if (dev == NULL)
2355                 goto err_exit;
2356
2357 #if IS_ENABLED(CONFIG_IPV6_SIT)
2358         if (dev->type == ARPHRD_SIT) {
2359                 const struct net_device_ops *ops = dev->netdev_ops;
2360                 struct ifreq ifr;
2361                 struct ip_tunnel_parm p;
2362
2363                 err = -EADDRNOTAVAIL;
2364                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2365                         goto err_exit;
2366
2367                 memset(&p, 0, sizeof(p));
2368                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2369                 p.iph.saddr = 0;
2370                 p.iph.version = 4;
2371                 p.iph.ihl = 5;
2372                 p.iph.protocol = IPPROTO_IPV6;
2373                 p.iph.ttl = 64;
2374                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2375
2376                 if (ops->ndo_do_ioctl) {
2377                         mm_segment_t oldfs = get_fs();
2378
2379                         set_fs(KERNEL_DS);
2380                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2381                         set_fs(oldfs);
2382                 } else
2383                         err = -EOPNOTSUPP;
2384
2385                 if (err == 0) {
2386                         err = -ENOBUFS;
2387                         dev = __dev_get_by_name(net, p.name);
2388                         if (!dev)
2389                                 goto err_exit;
2390                         err = dev_open(dev);
2391                 }
2392         }
2393 #endif
2394
2395 err_exit:
2396         rtnl_unlock();
2397         return err;
2398 }
2399
2400 /*
2401  *      Manual configuration of address on an interface
2402  */
2403 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2404                           const struct in6_addr *peer_pfx,
2405                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2406                           __u32 valid_lft)
2407 {
2408         struct inet6_ifaddr *ifp;
2409         struct inet6_dev *idev;
2410         struct net_device *dev;
2411         int scope;
2412         u32 flags;
2413         clock_t expires;
2414         unsigned long timeout;
2415
2416         ASSERT_RTNL();
2417
2418         if (plen > 128)
2419                 return -EINVAL;
2420
2421         /* check the lifetime */
2422         if (!valid_lft || prefered_lft > valid_lft)
2423                 return -EINVAL;
2424
2425         dev = __dev_get_by_index(net, ifindex);
2426         if (!dev)
2427                 return -ENODEV;
2428
2429         idev = addrconf_add_dev(dev);
2430         if (IS_ERR(idev))
2431                 return PTR_ERR(idev);
2432
2433         scope = ipv6_addr_scope(pfx);
2434
2435         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2436         if (addrconf_finite_timeout(timeout)) {
2437                 expires = jiffies_to_clock_t(timeout * HZ);
2438                 valid_lft = timeout;
2439                 flags = RTF_EXPIRES;
2440         } else {
2441                 expires = 0;
2442                 flags = 0;
2443                 ifa_flags |= IFA_F_PERMANENT;
2444         }
2445
2446         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2447         if (addrconf_finite_timeout(timeout)) {
2448                 if (timeout == 0)
2449                         ifa_flags |= IFA_F_DEPRECATED;
2450                 prefered_lft = timeout;
2451         }
2452
2453         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2454
2455         if (!IS_ERR(ifp)) {
2456                 spin_lock_bh(&ifp->lock);
2457                 ifp->valid_lft = valid_lft;
2458                 ifp->prefered_lft = prefered_lft;
2459                 ifp->tstamp = jiffies;
2460                 if (peer_pfx)
2461                         ifp->peer_addr = *peer_pfx;
2462                 spin_unlock_bh(&ifp->lock);
2463
2464                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2465                                       expires, flags);
2466                 /*
2467                  * Note that section 3.1 of RFC 4429 indicates
2468                  * that the Optimistic flag should not be set for
2469                  * manually configured addresses
2470                  */
2471                 addrconf_dad_start(ifp);
2472                 in6_ifa_put(ifp);
2473                 addrconf_verify(0);
2474                 return 0;
2475         }
2476
2477         return PTR_ERR(ifp);
2478 }
2479
2480 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2481                           unsigned int plen)
2482 {
2483         struct inet6_ifaddr *ifp;
2484         struct inet6_dev *idev;
2485         struct net_device *dev;
2486
2487         if (plen > 128)
2488                 return -EINVAL;
2489
2490         dev = __dev_get_by_index(net, ifindex);
2491         if (!dev)
2492                 return -ENODEV;
2493
2494         if ((idev = __in6_dev_get(dev)) == NULL)
2495                 return -ENXIO;
2496
2497         read_lock_bh(&idev->lock);
2498         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2499                 if (ifp->prefix_len == plen &&
2500                     ipv6_addr_equal(pfx, &ifp->addr)) {
2501                         in6_ifa_hold(ifp);
2502                         read_unlock_bh(&idev->lock);
2503
2504                         ipv6_del_addr(ifp);
2505
2506                         /* If the last address is deleted administratively,
2507                            disable IPv6 on this interface.
2508                          */
2509                         if (list_empty(&idev->addr_list))
2510                                 addrconf_ifdown(idev->dev, 1);
2511                         return 0;
2512                 }
2513         }
2514         read_unlock_bh(&idev->lock);
2515         return -EADDRNOTAVAIL;
2516 }
2517
2518
2519 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2520 {
2521         struct in6_ifreq ireq;
2522         int err;
2523
2524         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2525                 return -EPERM;
2526
2527         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2528                 return -EFAULT;
2529
2530         rtnl_lock();
2531         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2532                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2533                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2534         rtnl_unlock();
2535         return err;
2536 }
2537
2538 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2539 {
2540         struct in6_ifreq ireq;
2541         int err;
2542
2543         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2544                 return -EPERM;
2545
2546         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2547                 return -EFAULT;
2548
2549         rtnl_lock();
2550         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2551                              ireq.ifr6_prefixlen);
2552         rtnl_unlock();
2553         return err;
2554 }
2555
2556 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2557                      int plen, int scope)
2558 {
2559         struct inet6_ifaddr *ifp;
2560
2561         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2562         if (!IS_ERR(ifp)) {
2563                 spin_lock_bh(&ifp->lock);
2564                 ifp->flags &= ~IFA_F_TENTATIVE;
2565                 spin_unlock_bh(&ifp->lock);
2566                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2567                 in6_ifa_put(ifp);
2568         }
2569 }
2570
2571 #if IS_ENABLED(CONFIG_IPV6_SIT)
2572 static void sit_add_v4_addrs(struct inet6_dev *idev)
2573 {
2574         struct in6_addr addr;
2575         struct net_device *dev;
2576         struct net *net = dev_net(idev->dev);
2577         int scope;
2578
2579         ASSERT_RTNL();
2580
2581         memset(&addr, 0, sizeof(struct in6_addr));
2582         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2583
2584         if (idev->dev->flags&IFF_POINTOPOINT) {
2585                 addr.s6_addr32[0] = htonl(0xfe800000);
2586                 scope = IFA_LINK;
2587         } else {
2588                 scope = IPV6_ADDR_COMPATv4;
2589         }
2590
2591         if (addr.s6_addr32[3]) {
2592                 add_addr(idev, &addr, 128, scope);
2593                 return;
2594         }
2595
2596         for_each_netdev(net, dev) {
2597                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2598                 if (in_dev && (dev->flags & IFF_UP)) {
2599                         struct in_ifaddr *ifa;
2600
2601                         int flag = scope;
2602
2603                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2604                                 int plen;
2605
2606                                 addr.s6_addr32[3] = ifa->ifa_local;
2607
2608                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2609                                         continue;
2610                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2611                                         if (idev->dev->flags&IFF_POINTOPOINT)
2612                                                 continue;
2613                                         flag |= IFA_HOST;
2614                                 }
2615                                 if (idev->dev->flags&IFF_POINTOPOINT)
2616                                         plen = 64;
2617                                 else
2618                                         plen = 96;
2619
2620                                 add_addr(idev, &addr, plen, flag);
2621                         }
2622                 }
2623         }
2624 }
2625 #endif
2626
2627 static void init_loopback(struct net_device *dev)
2628 {
2629         struct inet6_dev  *idev;
2630         struct net_device *sp_dev;
2631         struct inet6_ifaddr *sp_ifa;
2632         struct rt6_info *sp_rt;
2633
2634         /* ::1 */
2635
2636         ASSERT_RTNL();
2637
2638         if ((idev = ipv6_find_idev(dev)) == NULL) {
2639                 pr_debug("%s: add_dev failed\n", __func__);
2640                 return;
2641         }
2642
2643         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2644
2645         /* Add routes to other interface's IPv6 addresses */
2646         for_each_netdev(dev_net(dev), sp_dev) {
2647                 if (!strcmp(sp_dev->name, dev->name))
2648                         continue;
2649
2650                 idev = __in6_dev_get(sp_dev);
2651                 if (!idev)
2652                         continue;
2653
2654                 read_lock_bh(&idev->lock);
2655                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2656
2657                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2658                                 continue;
2659
2660                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2661
2662                         /* Failure cases are ignored */
2663                         if (!IS_ERR(sp_rt))
2664                                 ip6_ins_rt(sp_rt);
2665                 }
2666                 read_unlock_bh(&idev->lock);
2667         }
2668 }
2669
2670 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2671 {
2672         struct inet6_ifaddr *ifp;
2673         u32 addr_flags = IFA_F_PERMANENT;
2674
2675 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2676         if (idev->cnf.optimistic_dad &&
2677             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2678                 addr_flags |= IFA_F_OPTIMISTIC;
2679 #endif
2680
2681
2682         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2683         if (!IS_ERR(ifp)) {
2684                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2685                 addrconf_dad_start(ifp);
2686                 in6_ifa_put(ifp);
2687         }
2688 }
2689
2690 static void addrconf_dev_config(struct net_device *dev)
2691 {
2692         struct in6_addr addr;
2693         struct inet6_dev *idev;
2694
2695         ASSERT_RTNL();
2696
2697         if ((dev->type != ARPHRD_ETHER) &&
2698             (dev->type != ARPHRD_FDDI) &&
2699             (dev->type != ARPHRD_ARCNET) &&
2700             (dev->type != ARPHRD_INFINIBAND) &&
2701             (dev->type != ARPHRD_IEEE802154) &&
2702             (dev->type != ARPHRD_IEEE1394)) {
2703                 /* Alas, we support only Ethernet autoconfiguration. */
2704                 return;
2705         }
2706
2707         idev = addrconf_add_dev(dev);
2708         if (IS_ERR(idev))
2709                 return;
2710
2711         memset(&addr, 0, sizeof(struct in6_addr));
2712         addr.s6_addr32[0] = htonl(0xFE800000);
2713
2714         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2715                 addrconf_add_linklocal(idev, &addr);
2716 }
2717
2718 #if IS_ENABLED(CONFIG_IPV6_SIT)
2719 static void addrconf_sit_config(struct net_device *dev)
2720 {
2721         struct inet6_dev *idev;
2722
2723         ASSERT_RTNL();
2724
2725         /*
2726          * Configure the tunnel with one of our IPv4
2727          * addresses... we should configure all of
2728          * our v4 addrs in the tunnel
2729          */
2730
2731         if ((idev = ipv6_find_idev(dev)) == NULL) {
2732                 pr_debug("%s: add_dev failed\n", __func__);
2733                 return;
2734         }
2735
2736         if (dev->priv_flags & IFF_ISATAP) {
2737                 struct in6_addr addr;
2738
2739                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2740                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2741                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2742                         addrconf_add_linklocal(idev, &addr);
2743                 return;
2744         }
2745
2746         sit_add_v4_addrs(idev);
2747
2748         if (dev->flags&IFF_POINTOPOINT)
2749                 addrconf_add_mroute(dev);
2750         else
2751                 sit_route_add(dev);
2752 }
2753 #endif
2754
2755 #if IS_ENABLED(CONFIG_NET_IPGRE)
2756 static void addrconf_gre_config(struct net_device *dev)
2757 {
2758         struct inet6_dev *idev;
2759         struct in6_addr addr;
2760
2761         pr_info("%s(%s)\n", __func__, dev->name);
2762
2763         ASSERT_RTNL();
2764
2765         if ((idev = ipv6_find_idev(dev)) == NULL) {
2766                 pr_debug("%s: add_dev failed\n", __func__);
2767                 return;
2768         }
2769
2770         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2771         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2772
2773         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2774                 addrconf_add_linklocal(idev, &addr);
2775 }
2776 #endif
2777
2778 static inline int
2779 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2780 {
2781         struct in6_addr lladdr;
2782
2783         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2784                 addrconf_add_linklocal(idev, &lladdr);
2785                 return 0;
2786         }
2787         return -1;
2788 }
2789
2790 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2791 {
2792         struct net_device *link_dev;
2793         struct net *net = dev_net(idev->dev);
2794
2795         /* first try to inherit the link-local address from the link device */
2796         if (idev->dev->iflink &&
2797             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2798                 if (!ipv6_inherit_linklocal(idev, link_dev))
2799                         return;
2800         }
2801         /* then try to inherit it from any device */
2802         for_each_netdev(net, link_dev) {
2803                 if (!ipv6_inherit_linklocal(idev, link_dev))
2804                         return;
2805         }
2806         pr_debug("init ip6-ip6: add_linklocal failed\n");
2807 }
2808
2809 /*
2810  * Autoconfigure tunnel with a link-local address so routing protocols,
2811  * DHCPv6, MLD etc. can be run over the virtual link
2812  */
2813
2814 static void addrconf_ip6_tnl_config(struct net_device *dev)
2815 {
2816         struct inet6_dev *idev;
2817
2818         ASSERT_RTNL();
2819
2820         idev = addrconf_add_dev(dev);
2821         if (IS_ERR(idev)) {
2822                 pr_debug("init ip6-ip6: add_dev failed\n");
2823                 return;
2824         }
2825         ip6_tnl_add_linklocal(idev);
2826 }
2827
2828 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2829                            void *ptr)
2830 {
2831         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2832         struct inet6_dev *idev = __in6_dev_get(dev);
2833         int run_pending = 0;
2834         int err;
2835
2836         switch (event) {
2837         case NETDEV_REGISTER:
2838                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2839                         idev = ipv6_add_dev(dev);
2840                         if (!idev)
2841                                 return notifier_from_errno(-ENOMEM);
2842                 }
2843                 break;
2844
2845         case NETDEV_UP:
2846         case NETDEV_CHANGE:
2847                 if (dev->flags & IFF_SLAVE)
2848                         break;
2849
2850                 if (event == NETDEV_UP) {
2851                         if (!addrconf_qdisc_ok(dev)) {
2852                                 /* device is not ready yet. */
2853                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2854                                         dev->name);
2855                                 break;
2856                         }
2857
2858                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2859                                 idev = ipv6_add_dev(dev);
2860
2861                         if (idev) {
2862                                 idev->if_flags |= IF_READY;
2863                                 run_pending = 1;
2864                         }
2865                 } else {
2866                         if (!addrconf_qdisc_ok(dev)) {
2867                                 /* device is still not ready. */
2868                                 break;
2869                         }
2870
2871                         if (idev) {
2872                                 if (idev->if_flags & IF_READY)
2873                                         /* device is already configured. */
2874                                         break;
2875                                 idev->if_flags |= IF_READY;
2876                         }
2877
2878                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2879                                 dev->name);
2880
2881                         run_pending = 1;
2882                 }
2883
2884                 switch (dev->type) {
2885 #if IS_ENABLED(CONFIG_IPV6_SIT)
2886                 case ARPHRD_SIT:
2887                         addrconf_sit_config(dev);
2888                         break;
2889 #endif
2890 #if IS_ENABLED(CONFIG_NET_IPGRE)
2891                 case ARPHRD_IPGRE:
2892                         addrconf_gre_config(dev);
2893                         break;
2894 #endif
2895                 case ARPHRD_TUNNEL6:
2896                         addrconf_ip6_tnl_config(dev);
2897                         break;
2898                 case ARPHRD_LOOPBACK:
2899                         init_loopback(dev);
2900                         break;
2901
2902                 default:
2903                         addrconf_dev_config(dev);
2904                         break;
2905                 }
2906
2907                 if (idev) {
2908                         if (run_pending)
2909                                 addrconf_dad_run(idev);
2910
2911                         /*
2912                          * If the MTU changed during the interface down,
2913                          * when the interface up, the changed MTU must be
2914                          * reflected in the idev as well as routers.
2915                          */
2916                         if (idev->cnf.mtu6 != dev->mtu &&
2917                             dev->mtu >= IPV6_MIN_MTU) {
2918                                 rt6_mtu_change(dev, dev->mtu);
2919                                 idev->cnf.mtu6 = dev->mtu;
2920                         }
2921                         idev->tstamp = jiffies;
2922                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2923
2924                         /*
2925                          * If the changed mtu during down is lower than
2926                          * IPV6_MIN_MTU stop IPv6 on this interface.
2927                          */
2928                         if (dev->mtu < IPV6_MIN_MTU)
2929                                 addrconf_ifdown(dev, 1);
2930                 }
2931                 break;
2932
2933         case NETDEV_CHANGEMTU:
2934                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2935                         rt6_mtu_change(dev, dev->mtu);
2936                         idev->cnf.mtu6 = dev->mtu;
2937                         break;
2938                 }
2939
2940                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2941                         idev = ipv6_add_dev(dev);
2942                         if (idev)
2943                                 break;
2944                 }
2945
2946                 /*
2947                  * MTU falled under IPV6_MIN_MTU.
2948                  * Stop IPv6 on this interface.
2949                  */
2950
2951         case NETDEV_DOWN:
2952         case NETDEV_UNREGISTER:
2953                 /*
2954                  *      Remove all addresses from this interface.
2955                  */
2956                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2957                 break;
2958
2959         case NETDEV_CHANGENAME:
2960                 if (idev) {
2961                         snmp6_unregister_dev(idev);
2962                         addrconf_sysctl_unregister(idev);
2963                         addrconf_sysctl_register(idev);
2964                         err = snmp6_register_dev(idev);
2965                         if (err)
2966                                 return notifier_from_errno(err);
2967                 }
2968                 break;
2969
2970         case NETDEV_PRE_TYPE_CHANGE:
2971         case NETDEV_POST_TYPE_CHANGE:
2972                 addrconf_type_change(dev, event);
2973                 break;
2974         }
2975
2976         return NOTIFY_OK;
2977 }
2978
2979 /*
2980  *      addrconf module should be notified of a device going up
2981  */
2982 static struct notifier_block ipv6_dev_notf = {
2983         .notifier_call = addrconf_notify,
2984 };
2985
2986 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2987 {
2988         struct inet6_dev *idev;
2989         ASSERT_RTNL();
2990
2991         idev = __in6_dev_get(dev);
2992
2993         if (event == NETDEV_POST_TYPE_CHANGE)
2994                 ipv6_mc_remap(idev);
2995         else if (event == NETDEV_PRE_TYPE_CHANGE)
2996                 ipv6_mc_unmap(idev);
2997 }
2998
2999 static int addrconf_ifdown(struct net_device *dev, int how)
3000 {
3001         struct net *net = dev_net(dev);
3002         struct inet6_dev *idev;
3003         struct inet6_ifaddr *ifa;
3004         int state, i;
3005
3006         ASSERT_RTNL();
3007
3008         rt6_ifdown(net, dev);
3009         neigh_ifdown(&nd_tbl, dev);
3010
3011         idev = __in6_dev_get(dev);
3012         if (idev == NULL)
3013                 return -ENODEV;
3014
3015         /*
3016          * Step 1: remove reference to ipv6 device from parent device.
3017          *         Do not dev_put!
3018          */
3019         if (how) {
3020                 idev->dead = 1;
3021
3022                 /* protected by rtnl_lock */
3023                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3024
3025                 /* Step 1.5: remove snmp6 entry */
3026                 snmp6_unregister_dev(idev);
3027
3028         }
3029
3030         /* Step 2: clear hash table */
3031         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3032                 struct hlist_head *h = &inet6_addr_lst[i];
3033
3034                 spin_lock_bh(&addrconf_hash_lock);
3035         restart:
3036                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3037                         if (ifa->idev == idev) {
3038                                 hlist_del_init_rcu(&ifa->addr_lst);
3039                                 addrconf_del_timer(ifa);
3040                                 goto restart;
3041                         }
3042                 }
3043                 spin_unlock_bh(&addrconf_hash_lock);
3044         }
3045
3046         write_lock_bh(&idev->lock);
3047
3048         /* Step 2: clear flags for stateless addrconf */
3049         if (!how)
3050                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3051
3052 #ifdef CONFIG_IPV6_PRIVACY
3053         if (how && del_timer(&idev->regen_timer))
3054                 in6_dev_put(idev);
3055
3056         /* Step 3: clear tempaddr list */
3057         while (!list_empty(&idev->tempaddr_list)) {
3058                 ifa = list_first_entry(&idev->tempaddr_list,
3059                                        struct inet6_ifaddr, tmp_list);
3060                 list_del(&ifa->tmp_list);
3061                 write_unlock_bh(&idev->lock);
3062                 spin_lock_bh(&ifa->lock);
3063
3064                 if (ifa->ifpub) {
3065                         in6_ifa_put(ifa->ifpub);
3066                         ifa->ifpub = NULL;
3067                 }
3068                 spin_unlock_bh(&ifa->lock);
3069                 in6_ifa_put(ifa);
3070                 write_lock_bh(&idev->lock);
3071         }
3072 #endif
3073
3074         while (!list_empty(&idev->addr_list)) {
3075                 ifa = list_first_entry(&idev->addr_list,
3076                                        struct inet6_ifaddr, if_list);
3077                 addrconf_del_timer(ifa);
3078
3079                 list_del(&ifa->if_list);
3080
3081                 write_unlock_bh(&idev->lock);
3082
3083                 spin_lock_bh(&ifa->state_lock);
3084                 state = ifa->state;
3085                 ifa->state = INET6_IFADDR_STATE_DEAD;
3086                 spin_unlock_bh(&ifa->state_lock);
3087
3088                 if (state != INET6_IFADDR_STATE_DEAD) {
3089                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3090                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3091                 }
3092                 in6_ifa_put(ifa);
3093
3094                 write_lock_bh(&idev->lock);
3095         }
3096
3097         write_unlock_bh(&idev->lock);
3098
3099         /* Step 5: Discard multicast list */
3100         if (how)
3101                 ipv6_mc_destroy_dev(idev);
3102         else
3103                 ipv6_mc_down(idev);
3104
3105         idev->tstamp = jiffies;
3106
3107         /* Last: Shot the device (if unregistered) */
3108         if (how) {
3109                 addrconf_sysctl_unregister(idev);
3110                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3111                 neigh_ifdown(&nd_tbl, dev);
3112                 in6_dev_put(idev);
3113         }
3114         return 0;
3115 }
3116
3117 static void addrconf_rs_timer(unsigned long data)
3118 {
3119         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3120         struct inet6_dev *idev = ifp->idev;
3121
3122         read_lock(&idev->lock);
3123         if (idev->dead || !(idev->if_flags & IF_READY))
3124                 goto out;
3125
3126         if (!ipv6_accept_ra(idev))
3127                 goto out;
3128
3129         /* Announcement received after solicitation was sent */
3130         if (idev->if_flags & IF_RA_RCVD)
3131                 goto out;
3132
3133         spin_lock(&ifp->lock);
3134         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3135                 /* The wait after the last probe can be shorter */
3136                 addrconf_mod_timer(ifp, AC_RS,
3137                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3138                                    idev->cnf.rtr_solicit_delay :
3139                                    idev->cnf.rtr_solicit_interval);
3140                 spin_unlock(&ifp->lock);
3141
3142                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3143         } else {
3144                 spin_unlock(&ifp->lock);
3145                 /*
3146                  * Note: we do not support deprecated "all on-link"
3147                  * assumption any longer.
3148                  */
3149                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3150         }
3151
3152 out:
3153         read_unlock(&idev->lock);
3154         in6_ifa_put(ifp);
3155 }
3156
3157 /*
3158  *      Duplicate Address Detection
3159  */
3160 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3161 {
3162         unsigned long rand_num;
3163         struct inet6_dev *idev = ifp->idev;
3164
3165         if (ifp->flags & IFA_F_OPTIMISTIC)
3166                 rand_num = 0;
3167         else
3168                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3169
3170         ifp->probes = idev->cnf.dad_transmits;
3171         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3172 }
3173
3174 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3175 {
3176         struct inet6_dev *idev = ifp->idev;
3177         struct net_device *dev = idev->dev;
3178
3179         addrconf_join_solict(dev, &ifp->addr);
3180
3181         net_srandom(ifp->addr.s6_addr32[3]);
3182
3183         read_lock_bh(&idev->lock);
3184         spin_lock(&ifp->lock);
3185         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3186                 goto out;
3187
3188         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3189             idev->cnf.accept_dad < 1 ||
3190             !(ifp->flags&IFA_F_TENTATIVE) ||
3191             ifp->flags & IFA_F_NODAD) {
3192                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3193                 spin_unlock(&ifp->lock);
3194                 read_unlock_bh(&idev->lock);
3195
3196                 addrconf_dad_completed(ifp);
3197                 return;
3198         }
3199
3200         if (!(idev->if_flags & IF_READY)) {
3201                 spin_unlock(&ifp->lock);
3202                 read_unlock_bh(&idev->lock);
3203                 /*
3204                  * If the device is not ready:
3205                  * - keep it tentative if it is a permanent address.
3206                  * - otherwise, kill it.
3207                  */
3208                 in6_ifa_hold(ifp);
3209                 addrconf_dad_stop(ifp, 0);
3210                 return;
3211         }
3212
3213         /*
3214          * Optimistic nodes can start receiving
3215          * Frames right away
3216          */
3217         if (ifp->flags & IFA_F_OPTIMISTIC)
3218                 ip6_ins_rt(ifp->rt);
3219
3220         addrconf_dad_kick(ifp);
3221 out:
3222         spin_unlock(&ifp->lock);
3223         read_unlock_bh(&idev->lock);
3224 }
3225
3226 static void addrconf_dad_timer(unsigned long data)
3227 {
3228         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3229         struct inet6_dev *idev = ifp->idev;
3230         struct in6_addr mcaddr;
3231
3232         if (!ifp->probes && addrconf_dad_end(ifp))
3233                 goto out;
3234
3235         read_lock(&idev->lock);
3236         if (idev->dead || !(idev->if_flags & IF_READY)) {
3237                 read_unlock(&idev->lock);
3238                 goto out;
3239         }
3240
3241         spin_lock(&ifp->lock);
3242         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3243                 spin_unlock(&ifp->lock);
3244                 read_unlock(&idev->lock);
3245                 goto out;
3246         }
3247
3248         if (ifp->probes == 0) {
3249                 /*
3250                  * DAD was successful
3251                  */
3252
3253                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3254                 spin_unlock(&ifp->lock);
3255                 read_unlock(&idev->lock);
3256
3257                 addrconf_dad_completed(ifp);
3258
3259                 goto out;
3260         }
3261
3262         ifp->probes--;
3263         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3264         spin_unlock(&ifp->lock);
3265         read_unlock(&idev->lock);
3266
3267         /* send a neighbour solicitation for our addr */
3268         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3269         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3270 out:
3271         in6_ifa_put(ifp);
3272 }
3273
3274 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3275 {
3276         struct net_device *dev = ifp->idev->dev;
3277
3278         /*
3279          *      Configure the address for reception. Now it is valid.
3280          */
3281
3282         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3283
3284         /* If added prefix is link local and we are prepared to process
3285            router advertisements, start sending router solicitations.
3286          */
3287
3288         if (ipv6_accept_ra(ifp->idev) &&
3289             ifp->idev->cnf.rtr_solicits > 0 &&
3290             (dev->flags&IFF_LOOPBACK) == 0 &&
3291             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3292                 /*
3293                  *      If a host as already performed a random delay
3294                  *      [...] as part of DAD [...] there is no need
3295                  *      to delay again before sending the first RS
3296                  */
3297                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3298
3299                 spin_lock_bh(&ifp->lock);
3300                 ifp->probes = 1;
3301                 ifp->idev->if_flags |= IF_RS_SENT;
3302                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3303                 spin_unlock_bh(&ifp->lock);
3304         }
3305 }
3306
3307 static void addrconf_dad_run(struct inet6_dev *idev)
3308 {
3309         struct inet6_ifaddr *ifp;
3310
3311         read_lock_bh(&idev->lock);
3312         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3313                 spin_lock(&ifp->lock);
3314                 if (ifp->flags & IFA_F_TENTATIVE &&
3315                     ifp->state == INET6_IFADDR_STATE_DAD)
3316                         addrconf_dad_kick(ifp);
3317                 spin_unlock(&ifp->lock);
3318         }
3319         read_unlock_bh(&idev->lock);
3320 }
3321
3322 #ifdef CONFIG_PROC_FS
3323 struct if6_iter_state {
3324         struct seq_net_private p;
3325         int bucket;
3326         int offset;
3327 };
3328
3329 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3330 {
3331         struct inet6_ifaddr *ifa = NULL;
3332         struct if6_iter_state *state = seq->private;
3333         struct net *net = seq_file_net(seq);
3334         int p = 0;
3335
3336         /* initial bucket if pos is 0 */
3337         if (pos == 0) {
3338                 state->bucket = 0;
3339                 state->offset = 0;
3340         }
3341
3342         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3343                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3344                                          addr_lst) {
3345                         if (!net_eq(dev_net(ifa->idev->dev), net))
3346                                 continue;
3347                         /* sync with offset */
3348                         if (p < state->offset) {
3349                                 p++;
3350                                 continue;
3351                         }
3352                         state->offset++;
3353                         return ifa;
3354                 }
3355
3356                 /* prepare for next bucket */
3357                 state->offset = 0;
3358                 p = 0;
3359         }
3360         return NULL;
3361 }
3362
3363 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3364                                          struct inet6_ifaddr *ifa)
3365 {
3366         struct if6_iter_state *state = seq->private;
3367         struct net *net = seq_file_net(seq);
3368
3369         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3370                 if (!net_eq(dev_net(ifa->idev->dev), net))
3371                         continue;
3372                 state->offset++;
3373                 return ifa;
3374         }
3375
3376         while (++state->bucket < IN6_ADDR_HSIZE) {
3377                 state->offset = 0;
3378                 hlist_for_each_entry_rcu_bh(ifa,
3379                                      &inet6_addr_lst[state->bucket], addr_lst) {
3380                         if (!net_eq(dev_net(ifa->idev->dev), net))
3381                                 continue;
3382                         state->offset++;
3383                         return ifa;
3384                 }
3385         }
3386
3387         return NULL;
3388 }
3389
3390 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3391         __acquires(rcu_bh)
3392 {
3393         rcu_read_lock_bh();
3394         return if6_get_first(seq, *pos);
3395 }
3396
3397 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3398 {
3399         struct inet6_ifaddr *ifa;
3400
3401         ifa = if6_get_next(seq, v);
3402         ++*pos;
3403         return ifa;
3404 }
3405
3406 static void if6_seq_stop(struct seq_file *seq, void *v)
3407         __releases(rcu_bh)
3408 {
3409         rcu_read_unlock_bh();
3410 }
3411
3412 static int if6_seq_show(struct seq_file *seq, void *v)
3413 {
3414         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3415         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3416                    &ifp->addr,
3417                    ifp->idev->dev->ifindex,
3418                    ifp->prefix_len,
3419                    ifp->scope,
3420                    ifp->flags,
3421                    ifp->idev->dev->name);
3422         return 0;
3423 }
3424
3425 static const struct seq_operations if6_seq_ops = {
3426         .start  = if6_seq_start,
3427         .next   = if6_seq_next,
3428         .show   = if6_seq_show,
3429         .stop   = if6_seq_stop,
3430 };
3431
3432 static int if6_seq_open(struct inode *inode, struct file *file)
3433 {
3434         return seq_open_net(inode, file, &if6_seq_ops,
3435                             sizeof(struct if6_iter_state));
3436 }
3437
3438 static const struct file_operations if6_fops = {
3439         .owner          = THIS_MODULE,
3440         .open           = if6_seq_open,
3441         .read           = seq_read,
3442         .llseek         = seq_lseek,
3443         .release        = seq_release_net,
3444 };
3445
3446 static int __net_init if6_proc_net_init(struct net *net)
3447 {
3448         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3449                 return -ENOMEM;
3450         return 0;
3451 }
3452
3453 static void __net_exit if6_proc_net_exit(struct net *net)
3454 {
3455         remove_proc_entry("if_inet6", net->proc_net);
3456 }
3457
3458 static struct pernet_operations if6_proc_net_ops = {
3459        .init = if6_proc_net_init,
3460        .exit = if6_proc_net_exit,
3461 };
3462
3463 int __init if6_proc_init(void)
3464 {
3465         return register_pernet_subsys(&if6_proc_net_ops);
3466 }
3467
3468 void if6_proc_exit(void)
3469 {
3470         unregister_pernet_subsys(&if6_proc_net_ops);
3471 }
3472 #endif  /* CONFIG_PROC_FS */
3473
3474 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3475 /* Check if address is a home address configured on any interface. */
3476 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3477 {
3478         int ret = 0;
3479         struct inet6_ifaddr *ifp = NULL;
3480         unsigned int hash = inet6_addr_hash(addr);
3481
3482         rcu_read_lock_bh();
3483         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3484                 if (!net_eq(dev_net(ifp->idev->dev), net))
3485                         continue;
3486                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3487                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3488                         ret = 1;
3489                         break;
3490                 }
3491         }
3492         rcu_read_unlock_bh();
3493         return ret;
3494 }
3495 #endif
3496
3497 /*
3498  *      Periodic address status verification
3499  */
3500
3501 static void addrconf_verify(unsigned long foo)
3502 {
3503         unsigned long now, next, next_sec, next_sched;
3504         struct inet6_ifaddr *ifp;
3505         int i;
3506
3507         rcu_read_lock_bh();
3508         spin_lock(&addrconf_verify_lock);
3509         now = jiffies;
3510         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3511
3512         del_timer(&addr_chk_timer);
3513
3514         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3515 restart:
3516                 hlist_for_each_entry_rcu_bh(ifp,
3517                                          &inet6_addr_lst[i], addr_lst) {
3518                         unsigned long age;
3519
3520                         if (ifp->flags & IFA_F_PERMANENT)
3521                                 continue;
3522
3523                         spin_lock(&ifp->lock);
3524                         /* We try to batch several events at once. */
3525                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3526
3527                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3528                             age >= ifp->valid_lft) {
3529                                 spin_unlock(&ifp->lock);
3530                                 in6_ifa_hold(ifp);
3531                                 ipv6_del_addr(ifp);
3532                                 goto restart;
3533                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3534                                 spin_unlock(&ifp->lock);
3535                                 continue;
3536                         } else if (age >= ifp->prefered_lft) {
3537                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3538                                 int deprecate = 0;
3539
3540                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3541                                         deprecate = 1;
3542                                         ifp->flags |= IFA_F_DEPRECATED;
3543                                 }
3544
3545                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3546                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3547
3548                                 spin_unlock(&ifp->lock);
3549
3550                                 if (deprecate) {
3551                                         in6_ifa_hold(ifp);
3552
3553                                         ipv6_ifa_notify(0, ifp);
3554                                         in6_ifa_put(ifp);
3555                                         goto restart;
3556                                 }
3557 #ifdef CONFIG_IPV6_PRIVACY
3558                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3559                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3560                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3561                                         ifp->idev->cnf.dad_transmits *
3562                                         ifp->idev->nd_parms->retrans_time / HZ;
3563
3564                                 if (age >= ifp->prefered_lft - regen_advance) {
3565                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3566                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3567                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3568                                         if (!ifp->regen_count && ifpub) {
3569                                                 ifp->regen_count++;
3570                                                 in6_ifa_hold(ifp);
3571                                                 in6_ifa_hold(ifpub);
3572                                                 spin_unlock(&ifp->lock);
3573
3574                                                 spin_lock(&ifpub->lock);
3575                                                 ifpub->regen_count = 0;
3576                                                 spin_unlock(&ifpub->lock);
3577                                                 ipv6_create_tempaddr(ifpub, ifp);
3578                                                 in6_ifa_put(ifpub);
3579                                                 in6_ifa_put(ifp);
3580                                                 goto restart;
3581                                         }
3582                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3583                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3584                                 spin_unlock(&ifp->lock);
3585 #endif
3586                         } else {
3587                                 /* ifp->prefered_lft <= ifp->valid_lft */
3588                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3589                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3590                                 spin_unlock(&ifp->lock);
3591                         }
3592                 }
3593         }
3594
3595         next_sec = round_jiffies_up(next);
3596         next_sched = next;
3597
3598         /* If rounded timeout is accurate enough, accept it. */
3599         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3600                 next_sched = next_sec;
3601
3602         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3603         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3604                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3605
3606         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3607               now, next, next_sec, next_sched));
3608
3609         addr_chk_timer.expires = next_sched;
3610         add_timer(&addr_chk_timer);
3611         spin_unlock(&addrconf_verify_lock);
3612         rcu_read_unlock_bh();
3613 }
3614
3615 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3616                                      struct in6_addr **peer_pfx)
3617 {
3618         struct in6_addr *pfx = NULL;
3619
3620         *peer_pfx = NULL;
3621
3622         if (addr)
3623                 pfx = nla_data(addr);
3624
3625         if (local) {
3626                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3627                         *peer_pfx = pfx;
3628                 pfx = nla_data(local);
3629         }
3630
3631         return pfx;
3632 }
3633
3634 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3635         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3636         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3637         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3638 };
3639
3640 static int
3641 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3642 {
3643         struct net *net = sock_net(skb->sk);
3644         struct ifaddrmsg *ifm;
3645         struct nlattr *tb[IFA_MAX+1];
3646         struct in6_addr *pfx, *peer_pfx;
3647         int err;
3648
3649         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3650         if (err < 0)
3651                 return err;
3652
3653         ifm = nlmsg_data(nlh);
3654         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3655         if (pfx == NULL)
3656                 return -EINVAL;
3657
3658         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3659 }
3660
3661 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3662                              u32 prefered_lft, u32 valid_lft)
3663 {
3664         u32 flags;
3665         clock_t expires;
3666         unsigned long timeout;
3667
3668         if (!valid_lft || (prefered_lft > valid_lft))
3669                 return -EINVAL;
3670
3671         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3672         if (addrconf_finite_timeout(timeout)) {
3673                 expires = jiffies_to_clock_t(timeout * HZ);
3674                 valid_lft = timeout;
3675                 flags = RTF_EXPIRES;
3676         } else {
3677                 expires = 0;
3678                 flags = 0;
3679                 ifa_flags |= IFA_F_PERMANENT;
3680         }
3681
3682         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3683         if (addrconf_finite_timeout(timeout)) {
3684                 if (timeout == 0)
3685                         ifa_flags |= IFA_F_DEPRECATED;
3686                 prefered_lft = timeout;
3687         }
3688
3689         spin_lock_bh(&ifp->lock);
3690         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3691         ifp->tstamp = jiffies;
3692         ifp->valid_lft = valid_lft;
3693         ifp->prefered_lft = prefered_lft;
3694
3695         spin_unlock_bh(&ifp->lock);
3696         if (!(ifp->flags&IFA_F_TENTATIVE))
3697                 ipv6_ifa_notify(0, ifp);
3698
3699         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3700                               expires, flags);
3701         addrconf_verify(0);
3702
3703         return 0;
3704 }
3705
3706 static int
3707 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3708 {
3709         struct net *net = sock_net(skb->sk);
3710         struct ifaddrmsg *ifm;
3711         struct nlattr *tb[IFA_MAX+1];
3712         struct in6_addr *pfx, *peer_pfx;
3713         struct inet6_ifaddr *ifa;
3714         struct net_device *dev;
3715         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3716         u8 ifa_flags;
3717         int err;
3718
3719         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3720         if (err < 0)
3721                 return err;
3722
3723         ifm = nlmsg_data(nlh);
3724         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3725         if (pfx == NULL)
3726                 return -EINVAL;
3727
3728         if (tb[IFA_CACHEINFO]) {
3729                 struct ifa_cacheinfo *ci;
3730
3731                 ci = nla_data(tb[IFA_CACHEINFO]);
3732                 valid_lft = ci->ifa_valid;
3733                 preferred_lft = ci->ifa_prefered;
3734         } else {
3735                 preferred_lft = INFINITY_LIFE_TIME;
3736                 valid_lft = INFINITY_LIFE_TIME;
3737         }
3738
3739         dev =  __dev_get_by_index(net, ifm->ifa_index);
3740         if (dev == NULL)
3741                 return -ENODEV;
3742
3743         /* We ignore other flags so far. */
3744         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3745
3746         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3747         if (ifa == NULL) {
3748                 /*
3749                  * It would be best to check for !NLM_F_CREATE here but
3750                  * userspace alreay relies on not having to provide this.
3751                  */
3752                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3753                                       ifm->ifa_prefixlen, ifa_flags,
3754                                       preferred_lft, valid_lft);
3755         }
3756
3757         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3758             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3759                 err = -EEXIST;
3760         else
3761                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3762
3763         in6_ifa_put(ifa);
3764
3765         return err;
3766 }
3767
3768 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3769                           u8 scope, int ifindex)
3770 {
3771         struct ifaddrmsg *ifm;
3772
3773         ifm = nlmsg_data(nlh);
3774         ifm->ifa_family = AF_INET6;
3775         ifm->ifa_prefixlen = prefixlen;
3776         ifm->ifa_flags = flags;
3777         ifm->ifa_scope = scope;
3778         ifm->ifa_index = ifindex;
3779 }
3780
3781 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3782                          unsigned long tstamp, u32 preferred, u32 valid)
3783 {
3784         struct ifa_cacheinfo ci;
3785
3786         ci.cstamp = cstamp_delta(cstamp);
3787         ci.tstamp = cstamp_delta(tstamp);
3788         ci.ifa_prefered = preferred;
3789         ci.ifa_valid = valid;
3790
3791         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3792 }
3793
3794 static inline int rt_scope(int ifa_scope)
3795 {
3796         if (ifa_scope & IFA_HOST)
3797                 return RT_SCOPE_HOST;
3798         else if (ifa_scope & IFA_LINK)
3799                 return RT_SCOPE_LINK;
3800         else if (ifa_scope & IFA_SITE)
3801                 return RT_SCOPE_SITE;
3802         else
3803                 return RT_SCOPE_UNIVERSE;
3804 }
3805
3806 static inline int inet6_ifaddr_msgsize(void)
3807 {
3808         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3809                + nla_total_size(16) /* IFA_LOCAL */
3810                + nla_total_size(16) /* IFA_ADDRESS */
3811                + nla_total_size(sizeof(struct ifa_cacheinfo));
3812 }
3813
3814 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3815                              u32 portid, u32 seq, int event, unsigned int flags)
3816 {
3817         struct nlmsghdr  *nlh;
3818         u32 preferred, valid;
3819
3820         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3821         if (nlh == NULL)
3822                 return -EMSGSIZE;
3823
3824         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3825                       ifa->idev->dev->ifindex);
3826
3827         if (!(ifa->flags&IFA_F_PERMANENT)) {
3828                 preferred = ifa->prefered_lft;
3829                 valid = ifa->valid_lft;
3830                 if (preferred != INFINITY_LIFE_TIME) {
3831                         long tval = (jiffies - ifa->tstamp)/HZ;
3832                         if (preferred > tval)
3833                                 preferred -= tval;
3834                         else
3835                                 preferred = 0;
3836                         if (valid != INFINITY_LIFE_TIME) {
3837                                 if (valid > tval)
3838                                         valid -= tval;
3839                                 else
3840                                         valid = 0;
3841                         }
3842                 }
3843         } else {
3844                 preferred = INFINITY_LIFE_TIME;
3845                 valid = INFINITY_LIFE_TIME;
3846         }
3847
3848         if (!ipv6_addr_any(&ifa->peer_addr)) {
3849                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3850                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3851                         goto error;
3852         } else
3853                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3854                         goto error;
3855
3856         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3857                 goto error;
3858
3859         return nlmsg_end(skb, nlh);
3860
3861 error:
3862         nlmsg_cancel(skb, nlh);
3863         return -EMSGSIZE;
3864 }
3865
3866 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3867                                 u32 portid, u32 seq, int event, u16 flags)
3868 {
3869         struct nlmsghdr  *nlh;
3870         u8 scope = RT_SCOPE_UNIVERSE;
3871         int ifindex = ifmca->idev->dev->ifindex;
3872
3873         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3874                 scope = RT_SCOPE_SITE;
3875
3876         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3877         if (nlh == NULL)
3878                 return -EMSGSIZE;
3879
3880         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3881         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3882             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3883                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3884                 nlmsg_cancel(skb, nlh);
3885                 return -EMSGSIZE;
3886         }
3887
3888         return nlmsg_end(skb, nlh);
3889 }
3890
3891 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3892                                 u32 portid, u32 seq, int event, unsigned int flags)
3893 {
3894         struct nlmsghdr  *nlh;
3895         u8 scope = RT_SCOPE_UNIVERSE;
3896         int ifindex = ifaca->aca_idev->dev->ifindex;
3897
3898         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3899                 scope = RT_SCOPE_SITE;
3900
3901         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3902         if (nlh == NULL)
3903                 return -EMSGSIZE;
3904
3905         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3906         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3907             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3908                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3909                 nlmsg_cancel(skb, nlh);
3910                 return -EMSGSIZE;
3911         }
3912
3913         return nlmsg_end(skb, nlh);
3914 }
3915
3916 enum addr_type_t {
3917         UNICAST_ADDR,
3918         MULTICAST_ADDR,
3919         ANYCAST_ADDR,
3920 };
3921
3922 /* called with rcu_read_lock() */
3923 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3924                           struct netlink_callback *cb, enum addr_type_t type,
3925                           int s_ip_idx, int *p_ip_idx)
3926 {
3927         struct ifmcaddr6 *ifmca;
3928         struct ifacaddr6 *ifaca;
3929         int err = 1;
3930         int ip_idx = *p_ip_idx;
3931
3932         read_lock_bh(&idev->lock);
3933         switch (type) {
3934         case UNICAST_ADDR: {
3935                 struct inet6_ifaddr *ifa;
3936
3937                 /* unicast address incl. temp addr */
3938                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3939                         if (++ip_idx < s_ip_idx)
3940                                 continue;
3941                         err = inet6_fill_ifaddr(skb, ifa,
3942                                                 NETLINK_CB(cb->skb).portid,
3943                                                 cb->nlh->nlmsg_seq,
3944                                                 RTM_NEWADDR,
3945                                                 NLM_F_MULTI);
3946                         if (err <= 0)
3947                                 break;
3948                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3949                 }
3950                 break;
3951         }
3952         case MULTICAST_ADDR:
3953                 /* multicast address */
3954                 for (ifmca = idev->mc_list; ifmca;
3955                      ifmca = ifmca->next, ip_idx++) {
3956                         if (ip_idx < s_ip_idx)
3957                                 continue;
3958                         err = inet6_fill_ifmcaddr(skb, ifmca,
3959                                                   NETLINK_CB(cb->skb).portid,
3960                                                   cb->nlh->nlmsg_seq,
3961                                                   RTM_GETMULTICAST,
3962                                                   NLM_F_MULTI);
3963                         if (err <= 0)
3964                                 break;
3965                 }
3966                 break;
3967         case ANYCAST_ADDR:
3968                 /* anycast address */
3969                 for (ifaca = idev->ac_list; ifaca;
3970                      ifaca = ifaca->aca_next, ip_idx++) {
3971                         if (ip_idx < s_ip_idx)
3972                                 continue;
3973                         err = inet6_fill_ifacaddr(skb, ifaca,
3974                                                   NETLINK_CB(cb->skb).portid,
3975                                                   cb->nlh->nlmsg_seq,
3976                                                   RTM_GETANYCAST,
3977                                                   NLM_F_MULTI);
3978                         if (err <= 0)
3979                                 break;
3980                 }
3981                 break;
3982         default:
3983                 break;
3984         }
3985         read_unlock_bh(&idev->lock);
3986         *p_ip_idx = ip_idx;
3987         return err;
3988 }
3989
3990 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3991                            enum addr_type_t type)
3992 {
3993         struct net *net = sock_net(skb->sk);
3994         int h, s_h;
3995         int idx, ip_idx;
3996         int s_idx, s_ip_idx;
3997         struct net_device *dev;
3998         struct inet6_dev *idev;
3999         struct hlist_head *head;
4000
4001         s_h = cb->args[0];
4002         s_idx = idx = cb->args[1];
4003         s_ip_idx = ip_idx = cb->args[2];
4004
4005         rcu_read_lock();
4006         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4007         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4008                 idx = 0;
4009                 head = &net->dev_index_head[h];
4010                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4011                         if (idx < s_idx)
4012                                 goto cont;
4013                         if (h > s_h || idx > s_idx)
4014                                 s_ip_idx = 0;
4015                         ip_idx = 0;
4016                         idev = __in6_dev_get(dev);
4017                         if (!idev)
4018                                 goto cont;
4019
4020                         if (in6_dump_addrs(idev, skb, cb, type,
4021                                            s_ip_idx, &ip_idx) <= 0)
4022                                 goto done;
4023 cont:
4024                         idx++;
4025                 }
4026         }
4027 done:
4028         rcu_read_unlock();
4029         cb->args[0] = h;
4030         cb->args[1] = idx;
4031         cb->args[2] = ip_idx;
4032
4033         return skb->len;
4034 }
4035
4036 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4037 {
4038         enum addr_type_t type = UNICAST_ADDR;
4039
4040         return inet6_dump_addr(skb, cb, type);
4041 }
4042
4043 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4044 {
4045         enum addr_type_t type = MULTICAST_ADDR;
4046
4047         return inet6_dump_addr(skb, cb, type);
4048 }
4049
4050
4051 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4052 {
4053         enum addr_type_t type = ANYCAST_ADDR;
4054
4055         return inet6_dump_addr(skb, cb, type);
4056 }
4057
4058 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4059 {
4060         struct net *net = sock_net(in_skb->sk);
4061         struct ifaddrmsg *ifm;
4062         struct nlattr *tb[IFA_MAX+1];
4063         struct in6_addr *addr = NULL, *peer;
4064         struct net_device *dev = NULL;
4065         struct inet6_ifaddr *ifa;
4066         struct sk_buff *skb;
4067         int err;
4068
4069         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4070         if (err < 0)
4071                 goto errout;
4072
4073         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4074         if (addr == NULL) {
4075                 err = -EINVAL;
4076                 goto errout;
4077         }
4078
4079         ifm = nlmsg_data(nlh);
4080         if (ifm->ifa_index)
4081                 dev = __dev_get_by_index(net, ifm->ifa_index);
4082
4083         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4084         if (!ifa) {
4085                 err = -EADDRNOTAVAIL;
4086                 goto errout;
4087         }
4088
4089         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4090         if (!skb) {
4091                 err = -ENOBUFS;
4092                 goto errout_ifa;
4093         }
4094
4095         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4096                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4097         if (err < 0) {
4098                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4099                 WARN_ON(err == -EMSGSIZE);
4100                 kfree_skb(skb);
4101                 goto errout_ifa;
4102         }
4103         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4104 errout_ifa:
4105         in6_ifa_put(ifa);
4106 errout:
4107         return err;
4108 }
4109
4110 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4111 {
4112         struct sk_buff *skb;
4113         struct net *net = dev_net(ifa->idev->dev);
4114         int err = -ENOBUFS;
4115
4116         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4117         if (skb == NULL)
4118                 goto errout;
4119
4120         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4121         if (err < 0) {
4122                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4123                 WARN_ON(err == -EMSGSIZE);
4124                 kfree_skb(skb);
4125                 goto errout;
4126         }
4127         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4128         return;
4129 errout:
4130         if (err < 0)
4131                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4132 }
4133
4134 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4135                                 __s32 *array, int bytes)
4136 {
4137         BUG_ON(bytes < (DEVCONF_MAX * 4));
4138
4139         memset(array, 0, bytes);
4140         array[DEVCONF_FORWARDING] = cnf->forwarding;
4141         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4142         array[DEVCONF_MTU6] = cnf->mtu6;
4143         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4144         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4145         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4146         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4147         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4148         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4149                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4150         array[DEVCONF_RTR_SOLICIT_DELAY] =
4151                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4152         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4153 #ifdef CONFIG_IPV6_PRIVACY
4154         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4155         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4156         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4157         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4158         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4159 #endif
4160         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4161         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4162         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4163 #ifdef CONFIG_IPV6_ROUTER_PREF
4164         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4165         array[DEVCONF_RTR_PROBE_INTERVAL] =
4166                 jiffies_to_msecs(cnf->rtr_probe_interval);
4167 #ifdef CONFIG_IPV6_ROUTE_INFO
4168         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4169 #endif
4170 #endif
4171         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4172         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4173 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4174         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4175 #endif
4176 #ifdef CONFIG_IPV6_MROUTE
4177         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4178 #endif
4179         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4180         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4181         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4182         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4183 }
4184
4185 static inline size_t inet6_ifla6_size(void)
4186 {
4187         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4188              + nla_total_size(sizeof(struct ifla_cacheinfo))
4189              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4190              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4191              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4192              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4193 }
4194
4195 static inline size_t inet6_if_nlmsg_size(void)
4196 {
4197         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4198                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4199                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4200                + nla_total_size(4) /* IFLA_MTU */
4201                + nla_total_size(4) /* IFLA_LINK */
4202                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4203 }
4204
4205 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4206                                       int items, int bytes)
4207 {
4208         int i;
4209         int pad = bytes - sizeof(u64) * items;
4210         BUG_ON(pad < 0);
4211
4212         /* Use put_unaligned() because stats may not be aligned for u64. */
4213         put_unaligned(items, &stats[0]);
4214         for (i = 1; i < items; i++)
4215                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4216
4217         memset(&stats[items], 0, pad);
4218 }
4219
4220 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4221                                       int items, int bytes, size_t syncpoff)
4222 {
4223         int i;
4224         int pad = bytes - sizeof(u64) * items;
4225         BUG_ON(pad < 0);
4226
4227         /* Use put_unaligned() because stats may not be aligned for u64. */
4228         put_unaligned(items, &stats[0]);
4229         for (i = 1; i < items; i++)
4230                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4231
4232         memset(&stats[items], 0, pad);
4233 }
4234
4235 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4236                              int bytes)
4237 {
4238         switch (attrtype) {
4239         case IFLA_INET6_STATS:
4240                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4241                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4242                 break;
4243         case IFLA_INET6_ICMP6STATS:
4244                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4245                 break;
4246         }
4247 }
4248
4249 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4250 {
4251         struct nlattr *nla;
4252         struct ifla_cacheinfo ci;
4253
4254         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4255                 goto nla_put_failure;
4256         ci.max_reasm_len = IPV6_MAXPLEN;
4257         ci.tstamp = cstamp_delta(idev->tstamp);
4258         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4259         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4260         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4261                 goto nla_put_failure;
4262         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4263         if (nla == NULL)
4264                 goto nla_put_failure;
4265         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4266
4267         /* XXX - MC not implemented */
4268
4269         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4270         if (nla == NULL)
4271                 goto nla_put_failure;
4272         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4273
4274         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4275         if (nla == NULL)
4276                 goto nla_put_failure;
4277         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4278
4279         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4280         if (nla == NULL)
4281                 goto nla_put_failure;
4282         read_lock_bh(&idev->lock);
4283         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4284         read_unlock_bh(&idev->lock);
4285
4286         return 0;
4287
4288 nla_put_failure:
4289         return -EMSGSIZE;
4290 }
4291
4292 static size_t inet6_get_link_af_size(const struct net_device *dev)
4293 {
4294         if (!__in6_dev_get(dev))
4295                 return 0;
4296
4297         return inet6_ifla6_size();
4298 }
4299
4300 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4301 {
4302         struct inet6_dev *idev = __in6_dev_get(dev);
4303
4304         if (!idev)
4305                 return -ENODATA;
4306
4307         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4308                 return -EMSGSIZE;
4309
4310         return 0;
4311 }
4312
4313 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4314 {
4315         struct inet6_ifaddr *ifp;
4316         struct net_device *dev = idev->dev;
4317         bool update_rs = false;
4318
4319         if (token == NULL)
4320                 return -EINVAL;
4321         if (ipv6_addr_any(token))
4322                 return -EINVAL;
4323         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4324                 return -EINVAL;
4325         if (!ipv6_accept_ra(idev))
4326                 return -EINVAL;
4327         if (idev->cnf.rtr_solicits <= 0)
4328                 return -EINVAL;
4329
4330         write_lock_bh(&idev->lock);
4331
4332         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4333         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4334
4335         write_unlock_bh(&idev->lock);
4336
4337         if (!idev->dead && (idev->if_flags & IF_READY)) {
4338                 struct in6_addr ll_addr;
4339
4340                 ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4341                                 IFA_F_OPTIMISTIC);
4342
4343                 /* If we're not ready, then normal ifup will take care
4344                  * of this. Otherwise, we need to request our rs here.
4345                  */
4346                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4347                 update_rs = true;
4348         }
4349
4350         write_lock_bh(&idev->lock);
4351
4352         if (update_rs)
4353                 idev->if_flags |= IF_RS_SENT;
4354
4355         /* Well, that's kinda nasty ... */
4356         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4357                 spin_lock(&ifp->lock);
4358                 if (ifp->tokenized) {
4359                         ifp->valid_lft = 0;
4360                         ifp->prefered_lft = 0;
4361                 }
4362                 spin_unlock(&ifp->lock);
4363         }
4364
4365         write_unlock_bh(&idev->lock);
4366         return 0;
4367 }
4368
4369 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4370 {
4371         int err = -EINVAL;
4372         struct inet6_dev *idev = __in6_dev_get(dev);
4373         struct nlattr *tb[IFLA_INET6_MAX + 1];
4374
4375         if (!idev)
4376                 return -EAFNOSUPPORT;
4377
4378         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4379                 BUG();
4380
4381         if (tb[IFLA_INET6_TOKEN])
4382                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4383
4384         return err;
4385 }
4386
4387 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4388                              u32 portid, u32 seq, int event, unsigned int flags)
4389 {
4390         struct net_device *dev = idev->dev;
4391         struct ifinfomsg *hdr;
4392         struct nlmsghdr *nlh;
4393         void *protoinfo;
4394
4395         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4396         if (nlh == NULL)
4397                 return -EMSGSIZE;
4398
4399         hdr = nlmsg_data(nlh);
4400         hdr->ifi_family = AF_INET6;
4401         hdr->__ifi_pad = 0;
4402         hdr->ifi_type = dev->type;
4403         hdr->ifi_index = dev->ifindex;
4404         hdr->ifi_flags = dev_get_flags(dev);
4405         hdr->ifi_change = 0;
4406
4407         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4408             (dev->addr_len &&
4409              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4410             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4411             (dev->ifindex != dev->iflink &&
4412              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4413                 goto nla_put_failure;
4414         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4415         if (protoinfo == NULL)
4416                 goto nla_put_failure;
4417
4418         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4419                 goto nla_put_failure;
4420
4421         nla_nest_end(skb, protoinfo);
4422         return nlmsg_end(skb, nlh);
4423
4424 nla_put_failure:
4425         nlmsg_cancel(skb, nlh);
4426         return -EMSGSIZE;
4427 }
4428
4429 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4430 {
4431         struct net *net = sock_net(skb->sk);
4432         int h, s_h;
4433         int idx = 0, s_idx;
4434         struct net_device *dev;
4435         struct inet6_dev *idev;
4436         struct hlist_head *head;
4437
4438         s_h = cb->args[0];
4439         s_idx = cb->args[1];
4440
4441         rcu_read_lock();
4442         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4443                 idx = 0;
4444                 head = &net->dev_index_head[h];
4445                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4446                         if (idx < s_idx)
4447                                 goto cont;
4448                         idev = __in6_dev_get(dev);
4449                         if (!idev)
4450                                 goto cont;
4451                         if (inet6_fill_ifinfo(skb, idev,
4452                                               NETLINK_CB(cb->skb).portid,
4453                                               cb->nlh->nlmsg_seq,
4454                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4455                                 goto out;
4456 cont:
4457                         idx++;
4458                 }
4459         }
4460 out:
4461         rcu_read_unlock();
4462         cb->args[1] = idx;
4463         cb->args[0] = h;
4464
4465         return skb->len;
4466 }
4467
4468 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4469 {
4470         struct sk_buff *skb;
4471         struct net *net = dev_net(idev->dev);
4472         int err = -ENOBUFS;
4473
4474         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4475         if (skb == NULL)
4476                 goto errout;
4477
4478         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4479         if (err < 0) {
4480                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4481                 WARN_ON(err == -EMSGSIZE);
4482                 kfree_skb(skb);
4483                 goto errout;
4484         }
4485         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4486         return;
4487 errout:
4488         if (err < 0)
4489                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4490 }
4491
4492 static inline size_t inet6_prefix_nlmsg_size(void)
4493 {
4494         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4495                + nla_total_size(sizeof(struct in6_addr))
4496                + nla_total_size(sizeof(struct prefix_cacheinfo));
4497 }
4498
4499 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4500                              struct prefix_info *pinfo, u32 portid, u32 seq,
4501                              int event, unsigned int flags)
4502 {
4503         struct prefixmsg *pmsg;
4504         struct nlmsghdr *nlh;
4505         struct prefix_cacheinfo ci;
4506
4507         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4508         if (nlh == NULL)
4509                 return -EMSGSIZE;
4510
4511         pmsg = nlmsg_data(nlh);
4512         pmsg->prefix_family = AF_INET6;
4513         pmsg->prefix_pad1 = 0;
4514         pmsg->prefix_pad2 = 0;
4515         pmsg->prefix_ifindex = idev->dev->ifindex;
4516         pmsg->prefix_len = pinfo->prefix_len;
4517         pmsg->prefix_type = pinfo->type;
4518         pmsg->prefix_pad3 = 0;
4519         pmsg->prefix_flags = 0;
4520         if (pinfo->onlink)
4521                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4522         if (pinfo->autoconf)
4523                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4524
4525         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4526                 goto nla_put_failure;
4527         ci.preferred_time = ntohl(pinfo->prefered);
4528         ci.valid_time = ntohl(pinfo->valid);
4529         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4530                 goto nla_put_failure;
4531         return nlmsg_end(skb, nlh);
4532
4533 nla_put_failure:
4534         nlmsg_cancel(skb, nlh);
4535         return -EMSGSIZE;
4536 }
4537
4538 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4539                          struct prefix_info *pinfo)
4540 {
4541         struct sk_buff *skb;
4542         struct net *net = dev_net(idev->dev);
4543         int err = -ENOBUFS;
4544
4545         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4546         if (skb == NULL)
4547                 goto errout;
4548
4549         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4550         if (err < 0) {
4551                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4552                 WARN_ON(err == -EMSGSIZE);
4553                 kfree_skb(skb);
4554                 goto errout;
4555         }
4556         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4557         return;
4558 errout:
4559         if (err < 0)
4560                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4561 }
4562
4563 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4564 {
4565         struct net *net = dev_net(ifp->idev->dev);
4566
4567         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4568
4569         switch (event) {
4570         case RTM_NEWADDR:
4571                 /*
4572                  * If the address was optimistic
4573                  * we inserted the route at the start of
4574                  * our DAD process, so we don't need
4575                  * to do it again
4576                  */
4577                 if (!(ifp->rt->rt6i_node))
4578                         ip6_ins_rt(ifp->rt);
4579                 if (ifp->idev->cnf.forwarding)
4580                         addrconf_join_anycast(ifp);
4581                 if (!ipv6_addr_any(&ifp->peer_addr))
4582                         addrconf_prefix_route(&ifp->peer_addr, 128,
4583                                               ifp->idev->dev, 0, 0);
4584                 break;
4585         case RTM_DELADDR:
4586                 if (ifp->idev->cnf.forwarding)
4587                         addrconf_leave_anycast(ifp);
4588                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4589                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4590                         struct rt6_info *rt;
4591                         struct net_device *dev = ifp->idev->dev;
4592
4593                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4594                                         dev->ifindex, 1);
4595                         if (rt) {
4596                                 dst_hold(&rt->dst);
4597                                 if (ip6_del_rt(rt))
4598                                         dst_free(&rt->dst);
4599                         }
4600                 }
4601                 dst_hold(&ifp->rt->dst);
4602
4603                 if (ip6_del_rt(ifp->rt))
4604                         dst_free(&ifp->rt->dst);
4605                 break;
4606         }
4607         atomic_inc(&net->ipv6.dev_addr_genid);
4608 }
4609
4610 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4611 {
4612         rcu_read_lock_bh();
4613         if (likely(ifp->idev->dead == 0))
4614                 __ipv6_ifa_notify(event, ifp);
4615         rcu_read_unlock_bh();
4616 }
4617
4618 #ifdef CONFIG_SYSCTL
4619
4620 static
4621 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4622                            void __user *buffer, size_t *lenp, loff_t *ppos)
4623 {
4624         int *valp = ctl->data;
4625         int val = *valp;
4626         loff_t pos = *ppos;
4627         ctl_table lctl;
4628         int ret;
4629
4630         /*
4631          * ctl->data points to idev->cnf.forwarding, we should
4632          * not modify it until we get the rtnl lock.
4633          */
4634         lctl = *ctl;
4635         lctl.data = &val;
4636
4637         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4638
4639         if (write)
4640                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4641         if (ret)
4642                 *ppos = pos;
4643         return ret;
4644 }
4645
4646 static void dev_disable_change(struct inet6_dev *idev)
4647 {
4648         struct netdev_notifier_info info;
4649
4650         if (!idev || !idev->dev)
4651                 return;
4652
4653         netdev_notifier_info_init(&info, idev->dev);
4654         if (idev->cnf.disable_ipv6)
4655                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4656         else
4657                 addrconf_notify(NULL, NETDEV_UP, &info);
4658 }
4659
4660 static void addrconf_disable_change(struct net *net, __s32 newf)
4661 {
4662         struct net_device *dev;
4663         struct inet6_dev *idev;
4664
4665         rcu_read_lock();
4666         for_each_netdev_rcu(net, dev) {
4667                 idev = __in6_dev_get(dev);
4668                 if (idev) {
4669                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4670                         idev->cnf.disable_ipv6 = newf;
4671                         if (changed)
4672                                 dev_disable_change(idev);
4673                 }
4674         }
4675         rcu_read_unlock();
4676 }
4677
4678 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4679 {
4680         struct net *net;
4681         int old;
4682
4683         if (!rtnl_trylock())
4684                 return restart_syscall();
4685
4686         net = (struct net *)table->extra2;
4687         old = *p;
4688         *p = newf;
4689
4690         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4691                 rtnl_unlock();
4692                 return 0;
4693         }
4694
4695         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4696                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4697                 addrconf_disable_change(net, newf);
4698         } else if ((!newf) ^ (!old))
4699                 dev_disable_change((struct inet6_dev *)table->extra1);
4700
4701         rtnl_unlock();
4702         return 0;
4703 }
4704
4705 static
4706 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4707                             void __user *buffer, size_t *lenp, loff_t *ppos)
4708 {
4709         int *valp = ctl->data;
4710         int val = *valp;
4711         loff_t pos = *ppos;
4712         ctl_table lctl;
4713         int ret;
4714
4715         /*
4716          * ctl->data points to idev->cnf.disable_ipv6, we should
4717          * not modify it until we get the rtnl lock.
4718          */
4719         lctl = *ctl;
4720         lctl.data = &val;
4721
4722         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4723
4724         if (write)
4725                 ret = addrconf_disable_ipv6(ctl, valp, val);
4726         if (ret)
4727                 *ppos = pos;
4728         return ret;
4729 }
4730
4731 static struct addrconf_sysctl_table
4732 {
4733         struct ctl_table_header *sysctl_header;
4734         ctl_table addrconf_vars[DEVCONF_MAX+1];
4735 } addrconf_sysctl __read_mostly = {
4736         .sysctl_header = NULL,
4737         .addrconf_vars = {
4738                 {
4739                         .procname       = "forwarding",
4740                         .data           = &ipv6_devconf.forwarding,
4741                         .maxlen         = sizeof(int),
4742                         .mode           = 0644,
4743                         .proc_handler   = addrconf_sysctl_forward,
4744                 },
4745                 {
4746                         .procname       = "hop_limit",
4747                         .data           = &ipv6_devconf.hop_limit,
4748                         .maxlen         = sizeof(int),
4749                         .mode           = 0644,
4750                         .proc_handler   = proc_dointvec,
4751                 },
4752                 {
4753                         .procname       = "mtu",
4754                         .data           = &ipv6_devconf.mtu6,
4755                         .maxlen         = sizeof(int),
4756                         .mode           = 0644,
4757                         .proc_handler   = proc_dointvec,
4758                 },
4759                 {
4760                         .procname       = "accept_ra",
4761                         .data           = &ipv6_devconf.accept_ra,
4762                         .maxlen         = sizeof(int),
4763                         .mode           = 0644,
4764                         .proc_handler   = proc_dointvec,
4765                 },
4766                 {
4767                         .procname       = "accept_redirects",
4768                         .data           = &ipv6_devconf.accept_redirects,
4769                         .maxlen         = sizeof(int),
4770                         .mode           = 0644,
4771                         .proc_handler   = proc_dointvec,
4772                 },
4773                 {
4774                         .procname       = "autoconf",
4775                         .data           = &ipv6_devconf.autoconf,
4776                         .maxlen         = sizeof(int),
4777                         .mode           = 0644,
4778                         .proc_handler   = proc_dointvec,
4779                 },
4780                 {
4781                         .procname       = "dad_transmits",
4782                         .data           = &ipv6_devconf.dad_transmits,
4783                         .maxlen         = sizeof(int),
4784                         .mode           = 0644,
4785                         .proc_handler   = proc_dointvec,
4786                 },
4787                 {
4788                         .procname       = "router_solicitations",
4789                         .data           = &ipv6_devconf.rtr_solicits,
4790                         .maxlen         = sizeof(int),
4791                         .mode           = 0644,
4792                         .proc_handler   = proc_dointvec,
4793                 },
4794                 {
4795                         .procname       = "router_solicitation_interval",
4796                         .data           = &ipv6_devconf.rtr_solicit_interval,
4797                         .maxlen         = sizeof(int),
4798                         .mode           = 0644,
4799                         .proc_handler   = proc_dointvec_jiffies,
4800                 },
4801                 {
4802                         .procname       = "router_solicitation_delay",
4803                         .data           = &ipv6_devconf.rtr_solicit_delay,
4804                         .maxlen         = sizeof(int),
4805                         .mode           = 0644,
4806                         .proc_handler   = proc_dointvec_jiffies,
4807                 },
4808                 {
4809                         .procname       = "force_mld_version",
4810                         .data           = &ipv6_devconf.force_mld_version,
4811                         .maxlen         = sizeof(int),
4812                         .mode           = 0644,
4813                         .proc_handler   = proc_dointvec,
4814                 },
4815 #ifdef CONFIG_IPV6_PRIVACY
4816                 {
4817                         .procname       = "use_tempaddr",
4818                         .data           = &ipv6_devconf.use_tempaddr,
4819                         .maxlen         = sizeof(int),
4820                         .mode           = 0644,
4821                         .proc_handler   = proc_dointvec,
4822                 },
4823                 {
4824                         .procname       = "temp_valid_lft",
4825                         .data           = &ipv6_devconf.temp_valid_lft,
4826                         .maxlen         = sizeof(int),
4827                         .mode           = 0644,
4828                         .proc_handler   = proc_dointvec,
4829                 },
4830                 {
4831                         .procname       = "temp_prefered_lft",
4832                         .data           = &ipv6_devconf.temp_prefered_lft,
4833                         .maxlen         = sizeof(int),
4834                         .mode           = 0644,
4835                         .proc_handler   = proc_dointvec,
4836                 },
4837                 {
4838                         .procname       = "regen_max_retry",
4839                         .data           = &ipv6_devconf.regen_max_retry,
4840                         .maxlen         = sizeof(int),
4841                         .mode           = 0644,
4842                         .proc_handler   = proc_dointvec,
4843                 },
4844                 {
4845                         .procname       = "max_desync_factor",
4846                         .data           = &ipv6_devconf.max_desync_factor,
4847                         .maxlen         = sizeof(int),
4848                         .mode           = 0644,
4849                         .proc_handler   = proc_dointvec,
4850                 },
4851 #endif
4852                 {
4853                         .procname       = "max_addresses",
4854                         .data           = &ipv6_devconf.max_addresses,
4855                         .maxlen         = sizeof(int),
4856                         .mode           = 0644,
4857                         .proc_handler   = proc_dointvec,
4858                 },
4859                 {
4860                         .procname       = "accept_ra_defrtr",
4861                         .data           = &ipv6_devconf.accept_ra_defrtr,
4862                         .maxlen         = sizeof(int),
4863                         .mode           = 0644,
4864                         .proc_handler   = proc_dointvec,
4865                 },
4866                 {
4867                         .procname       = "accept_ra_pinfo",
4868                         .data           = &ipv6_devconf.accept_ra_pinfo,
4869                         .maxlen         = sizeof(int),
4870                         .mode           = 0644,
4871                         .proc_handler   = proc_dointvec,
4872                 },
4873 #ifdef CONFIG_IPV6_ROUTER_PREF
4874                 {
4875                         .procname       = "accept_ra_rtr_pref",
4876                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4877                         .maxlen         = sizeof(int),
4878                         .mode           = 0644,
4879                         .proc_handler   = proc_dointvec,
4880                 },
4881                 {
4882                         .procname       = "router_probe_interval",
4883                         .data           = &ipv6_devconf.rtr_probe_interval,
4884                         .maxlen         = sizeof(int),
4885                         .mode           = 0644,
4886                         .proc_handler   = proc_dointvec_jiffies,
4887                 },
4888 #ifdef CONFIG_IPV6_ROUTE_INFO
4889                 {
4890                         .procname       = "accept_ra_rt_info_max_plen",
4891                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4892                         .maxlen         = sizeof(int),
4893                         .mode           = 0644,
4894                         .proc_handler   = proc_dointvec,
4895                 },
4896 #endif
4897 #endif
4898                 {
4899                         .procname       = "proxy_ndp",
4900                         .data           = &ipv6_devconf.proxy_ndp,
4901                         .maxlen         = sizeof(int),
4902                         .mode           = 0644,
4903                         .proc_handler   = proc_dointvec,
4904                 },
4905                 {
4906                         .procname       = "accept_source_route",
4907                         .data           = &ipv6_devconf.accept_source_route,
4908                         .maxlen         = sizeof(int),
4909                         .mode           = 0644,
4910                         .proc_handler   = proc_dointvec,
4911                 },
4912 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4913                 {
4914                         .procname       = "optimistic_dad",
4915                         .data           = &ipv6_devconf.optimistic_dad,
4916                         .maxlen         = sizeof(int),
4917                         .mode           = 0644,
4918                         .proc_handler   = proc_dointvec,
4919
4920                 },
4921 #endif
4922 #ifdef CONFIG_IPV6_MROUTE
4923                 {
4924                         .procname       = "mc_forwarding",
4925                         .data           = &ipv6_devconf.mc_forwarding,
4926                         .maxlen         = sizeof(int),
4927                         .mode           = 0444,
4928                         .proc_handler   = proc_dointvec,
4929                 },
4930 #endif
4931                 {
4932                         .procname       = "disable_ipv6",
4933                         .data           = &ipv6_devconf.disable_ipv6,
4934                         .maxlen         = sizeof(int),
4935                         .mode           = 0644,
4936                         .proc_handler   = addrconf_sysctl_disable,
4937                 },
4938                 {
4939                         .procname       = "accept_dad",
4940                         .data           = &ipv6_devconf.accept_dad,
4941                         .maxlen         = sizeof(int),
4942                         .mode           = 0644,
4943                         .proc_handler   = proc_dointvec,
4944                 },
4945                 {
4946                         .procname       = "force_tllao",
4947                         .data           = &ipv6_devconf.force_tllao,
4948                         .maxlen         = sizeof(int),
4949                         .mode           = 0644,
4950                         .proc_handler   = proc_dointvec
4951                 },
4952                 {
4953                         .procname       = "ndisc_notify",
4954                         .data           = &ipv6_devconf.ndisc_notify,
4955                         .maxlen         = sizeof(int),
4956                         .mode           = 0644,
4957                         .proc_handler   = proc_dointvec
4958                 },
4959                 {
4960                         /* sentinel */
4961                 }
4962         },
4963 };
4964
4965 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4966                 struct inet6_dev *idev, struct ipv6_devconf *p)
4967 {
4968         int i;
4969         struct addrconf_sysctl_table *t;
4970         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4971
4972         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4973         if (t == NULL)
4974                 goto out;
4975
4976         for (i = 0; t->addrconf_vars[i].data; i++) {
4977                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4978                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4979                 t->addrconf_vars[i].extra2 = net;
4980         }
4981
4982         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4983
4984         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4985         if (t->sysctl_header == NULL)
4986                 goto free;
4987
4988         p->sysctl = t;
4989         return 0;
4990
4991 free:
4992         kfree(t);
4993 out:
4994         return -ENOBUFS;
4995 }
4996
4997 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4998 {
4999         struct addrconf_sysctl_table *t;
5000
5001         if (p->sysctl == NULL)
5002                 return;
5003
5004         t = p->sysctl;
5005         p->sysctl = NULL;
5006         unregister_net_sysctl_table(t->sysctl_header);
5007         kfree(t);
5008 }
5009
5010 static void addrconf_sysctl_register(struct inet6_dev *idev)
5011 {
5012         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5013                               &ndisc_ifinfo_sysctl_change);
5014         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5015                                         idev, &idev->cnf);
5016 }
5017
5018 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5019 {
5020         __addrconf_sysctl_unregister(&idev->cnf);
5021         neigh_sysctl_unregister(idev->nd_parms);
5022 }
5023
5024
5025 #endif
5026
5027 static int __net_init addrconf_init_net(struct net *net)
5028 {
5029         int err = -ENOMEM;
5030         struct ipv6_devconf *all, *dflt;
5031
5032         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5033         if (all == NULL)
5034                 goto err_alloc_all;
5035
5036         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5037         if (dflt == NULL)
5038                 goto err_alloc_dflt;
5039
5040         /* these will be inherited by all namespaces */
5041         dflt->autoconf = ipv6_defaults.autoconf;
5042         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5043
5044         net->ipv6.devconf_all = all;
5045         net->ipv6.devconf_dflt = dflt;
5046
5047 #ifdef CONFIG_SYSCTL
5048         err = __addrconf_sysctl_register(net, "all", NULL, all);
5049         if (err < 0)
5050                 goto err_reg_all;
5051
5052         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5053         if (err < 0)
5054                 goto err_reg_dflt;
5055 #endif
5056         return 0;
5057
5058 #ifdef CONFIG_SYSCTL
5059 err_reg_dflt:
5060         __addrconf_sysctl_unregister(all);
5061 err_reg_all:
5062         kfree(dflt);
5063 #endif
5064 err_alloc_dflt:
5065         kfree(all);
5066 err_alloc_all:
5067         return err;
5068 }
5069
5070 static void __net_exit addrconf_exit_net(struct net *net)
5071 {
5072 #ifdef CONFIG_SYSCTL
5073         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5074         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5075 #endif
5076         if (!net_eq(net, &init_net)) {
5077                 kfree(net->ipv6.devconf_dflt);
5078                 kfree(net->ipv6.devconf_all);
5079         }
5080 }
5081
5082 static struct pernet_operations addrconf_ops = {
5083         .init = addrconf_init_net,
5084         .exit = addrconf_exit_net,
5085 };
5086
5087 static struct rtnl_af_ops inet6_ops = {
5088         .family           = AF_INET6,
5089         .fill_link_af     = inet6_fill_link_af,
5090         .get_link_af_size = inet6_get_link_af_size,
5091         .set_link_af      = inet6_set_link_af,
5092 };
5093
5094 /*
5095  *      Init / cleanup code
5096  */
5097
5098 int __init addrconf_init(void)
5099 {
5100         int i, err;
5101
5102         err = ipv6_addr_label_init();
5103         if (err < 0) {
5104                 pr_crit("%s: cannot initialize default policy table: %d\n",
5105                         __func__, err);
5106                 goto out;
5107         }
5108
5109         err = register_pernet_subsys(&addrconf_ops);
5110         if (err < 0)
5111                 goto out_addrlabel;
5112
5113         /* The addrconf netdev notifier requires that loopback_dev
5114          * has it's ipv6 private information allocated and setup
5115          * before it can bring up and give link-local addresses
5116          * to other devices which are up.
5117          *
5118          * Unfortunately, loopback_dev is not necessarily the first
5119          * entry in the global dev_base list of net devices.  In fact,
5120          * it is likely to be the very last entry on that list.
5121          * So this causes the notifier registry below to try and
5122          * give link-local addresses to all devices besides loopback_dev
5123          * first, then loopback_dev, which cases all the non-loopback_dev
5124          * devices to fail to get a link-local address.
5125          *
5126          * So, as a temporary fix, allocate the ipv6 structure for
5127          * loopback_dev first by hand.
5128          * Longer term, all of the dependencies ipv6 has upon the loopback
5129          * device and it being up should be removed.
5130          */
5131         rtnl_lock();
5132         if (!ipv6_add_dev(init_net.loopback_dev))
5133                 err = -ENOMEM;
5134         rtnl_unlock();
5135         if (err)
5136                 goto errlo;
5137
5138         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5139                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5140
5141         register_netdevice_notifier(&ipv6_dev_notf);
5142
5143         addrconf_verify(0);
5144
5145         err = rtnl_af_register(&inet6_ops);
5146         if (err < 0)
5147                 goto errout_af;
5148
5149         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5150                               NULL);
5151         if (err < 0)
5152                 goto errout;
5153
5154         /* Only the first call to __rtnl_register can fail */
5155         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5156         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5157         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5158                         inet6_dump_ifaddr, NULL);
5159         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5160                         inet6_dump_ifmcaddr, NULL);
5161         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5162                         inet6_dump_ifacaddr, NULL);
5163         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5164                         inet6_netconf_dump_devconf, NULL);
5165
5166         ipv6_addr_label_rtnl_register();
5167
5168         return 0;
5169 errout:
5170         rtnl_af_unregister(&inet6_ops);
5171 errout_af:
5172         unregister_netdevice_notifier(&ipv6_dev_notf);
5173 errlo:
5174         unregister_pernet_subsys(&addrconf_ops);
5175 out_addrlabel:
5176         ipv6_addr_label_cleanup();
5177 out:
5178         return err;
5179 }
5180
5181 void addrconf_cleanup(void)
5182 {
5183         struct net_device *dev;
5184         int i;
5185
5186         unregister_netdevice_notifier(&ipv6_dev_notf);
5187         unregister_pernet_subsys(&addrconf_ops);
5188         ipv6_addr_label_cleanup();
5189
5190         rtnl_lock();
5191
5192         __rtnl_af_unregister(&inet6_ops);
5193
5194         /* clean dev list */
5195         for_each_netdev(&init_net, dev) {
5196                 if (__in6_dev_get(dev) == NULL)
5197                         continue;
5198                 addrconf_ifdown(dev, 1);
5199         }
5200         addrconf_ifdown(init_net.loopback_dev, 2);
5201
5202         /*
5203          *      Check hash table.
5204          */
5205         spin_lock_bh(&addrconf_hash_lock);
5206         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5207                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5208         spin_unlock_bh(&addrconf_hash_lock);
5209
5210         del_timer(&addr_chk_timer);
5211         rtnl_unlock();
5212 }