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_rs_timer(struct inet6_dev *idev)
 257 {
 258         if (del_timer(&idev->rs_timer))
 259                 __in6_dev_put(idev);
 260 }
 261
 262 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 263 {
 264         if (del_timer(&ifp->dad_timer))
 265                 __in6_ifa_put(ifp);
 266 }
 267
 268 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 269                                   unsigned long when)
 270 {
 271         if (!timer_pending(&idev->rs_timer))
 272                 in6_dev_hold(idev);
 273         mod_timer(&idev->rs_timer, jiffies + when);
 274 }
 275
 276 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 277                                    unsigned long when)
 278 {
 279         if (!timer_pending(&ifp->dad_timer))
 280                 in6_ifa_hold(ifp);
 281         mod_timer(&ifp->dad_timer, jiffies + when);
 282 }
 283
 284 static int snmp6_alloc_dev(struct inet6_dev *idev)
 285 {
 286         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 287                           sizeof(struct ipstats_mib),
 288                           __alignof__(struct ipstats_mib)) < 0)
 289                 goto err_ip;
 290         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 291                                         GFP_KERNEL);
 292         if (!idev->stats.icmpv6dev)
 293                 goto err_icmp;
 294         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 295                                            GFP_KERNEL);
 296         if (!idev->stats.icmpv6msgdev)
 297                 goto err_icmpmsg;
 298
 299         return 0;
 300
 301 err_icmpmsg:
 302         kfree(idev->stats.icmpv6dev);
 303 err_icmp:
 304         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 305 err_ip:
 306         return -ENOMEM;
 307 }
 308
 309 static void snmp6_free_dev(struct inet6_dev *idev)
 310 {
 311         kfree(idev->stats.icmpv6msgdev);
 312         kfree(idev->stats.icmpv6dev);
 313         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 314 }
 315
 316 /* Nobody refers to this device, we may destroy it. */
 317
 318 void in6_dev_finish_destroy(struct inet6_dev *idev)
 319 {
 320         struct net_device *dev = idev->dev;
 321
 322         WARN_ON(!list_empty(&idev->addr_list));
 323         WARN_ON(idev->mc_list != NULL);
 324         WARN_ON(timer_pending(&idev->rs_timer));
 325
 326 #ifdef NET_REFCNT_DEBUG
 327         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 328 #endif
 329         dev_put(dev);
 330         if (!idev->dead) {
 331                 pr_warn("Freeing alive inet6 device %p\n", idev);
 332                 return;
 333         }
 334         snmp6_free_dev(idev);
 335         kfree_rcu(idev, rcu);
 336 }
 337 EXPORT_SYMBOL(in6_dev_finish_destroy);
 338
 339 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 340 {
 341         struct inet6_dev *ndev;
 342
 343         ASSERT_RTNL();
 344
 345         if (dev->mtu < IPV6_MIN_MTU)
 346                 return NULL;
 347
 348         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 349
 350         if (ndev == NULL)
 351                 return NULL;
 352
 353         rwlock_init(&ndev->lock);
 354         ndev->dev = dev;
 355         INIT_LIST_HEAD(&ndev->addr_list);
 356         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 357                     (unsigned long)ndev);
 358         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 359         ndev->cnf.mtu6 = dev->mtu;
 360         ndev->cnf.sysctl = NULL;
 361         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 362         if (ndev->nd_parms == NULL) {
 363                 kfree(ndev);
 364                 return NULL;
 365         }
 366         if (ndev->cnf.forwarding)
 367                 dev_disable_lro(dev);
 368         /* We refer to the device */
 369         dev_hold(dev);
 370
 371         if (snmp6_alloc_dev(ndev) < 0) {
 372                 ADBG((KERN_WARNING
 373                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 374                         __func__, dev->name));
 375                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 376                 dev_put(dev);
 377                 kfree(ndev);
 378                 return NULL;
 379         }
 380
 381         if (snmp6_register_dev(ndev) < 0) {
 382                 ADBG((KERN_WARNING
 383                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 384                         __func__, dev->name));
 385                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 386                 ndev->dead = 1;
 387                 in6_dev_finish_destroy(ndev);
 388                 return NULL;
 389         }
 390
 391         /* One reference from device.  We must do this before
 392          * we invoke __ipv6_regen_rndid().
 393          */
 394         in6_dev_hold(ndev);
 395
 396         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 397                 ndev->cnf.accept_dad = -1;
 398
 399 #if IS_ENABLED(CONFIG_IPV6_SIT)
 400         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 401                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 402                 ndev->cnf.rtr_solicits = 0;
 403         }
 404 #endif
 405
 406 #ifdef CONFIG_IPV6_PRIVACY
 407         INIT_LIST_HEAD(&ndev->tempaddr_list);
 408         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 409         if ((dev->flags&IFF_LOOPBACK) ||
 410             dev->type == ARPHRD_TUNNEL ||
 411             dev->type == ARPHRD_TUNNEL6 ||
 412             dev->type == ARPHRD_SIT ||
 413             dev->type == ARPHRD_NONE) {
 414                 ndev->cnf.use_tempaddr = -1;
 415         } else {
 416                 in6_dev_hold(ndev);
 417                 ipv6_regen_rndid((unsigned long) ndev);
 418         }
 419 #endif
 420         ndev->token = in6addr_any;
 421
 422         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 423                 ndev->if_flags |= IF_READY;
 424
 425         ipv6_mc_init_dev(ndev);
 426         ndev->tstamp = jiffies;
 427         addrconf_sysctl_register(ndev);
 428         /* protected by rtnl_lock */
 429         rcu_assign_pointer(dev->ip6_ptr, ndev);
 430
 431         /* Join interface-local all-node multicast group */
 432         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         /* Join all-router multicast group if forwarding is set */
 438         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 439                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 static int inet6_netconf_msgsize_devconf(int type)
 463 {
 464         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 465                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 466
 467         /* type -1 is used for ALL */
 468         if (type == -1 || type == NETCONFA_FORWARDING)
 469                 size += nla_total_size(4);
 470 #ifdef CONFIG_IPV6_MROUTE
 471         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 472                 size += nla_total_size(4);
 473 #endif
 474
 475         return size;
 476 }
 477
 478 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 479                                       struct ipv6_devconf *devconf, u32 portid,
 480                                       u32 seq, int event, unsigned int flags,
 481                                       int type)
 482 {
 483         struct nlmsghdr  *nlh;
 484         struct netconfmsg *ncm;
 485
 486         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 487                         flags);
 488         if (nlh == NULL)
 489                 return -EMSGSIZE;
 490
 491         ncm = nlmsg_data(nlh);
 492         ncm->ncm_family = AF_INET6;
 493
 494         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 495                 goto nla_put_failure;
 496
 497         /* type -1 is used for ALL */
 498         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 499             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 500                 goto nla_put_failure;
 501 #ifdef CONFIG_IPV6_MROUTE
 502         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 503             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 504                         devconf->mc_forwarding) < 0)
 505                 goto nla_put_failure;
 506 #endif
 507         return nlmsg_end(skb, nlh);
 508
 509 nla_put_failure:
 510         nlmsg_cancel(skb, nlh);
 511         return -EMSGSIZE;
 512 }
 513
 514 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 515                                   struct ipv6_devconf *devconf)
 516 {
 517         struct sk_buff *skb;
 518         int err = -ENOBUFS;
 519
 520         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 521         if (skb == NULL)
 522                 goto errout;
 523
 524         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 525                                          RTM_NEWNETCONF, 0, type);
 526         if (err < 0) {
 527                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 528                 WARN_ON(err == -EMSGSIZE);
 529                 kfree_skb(skb);
 530                 goto errout;
 531         }
 532         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 533         return;
 534 errout:
 535         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 536 }
 537
 538 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 539         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 540         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 541 };
 542
 543 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 544                                      struct nlmsghdr *nlh)
 545 {
 546         struct net *net = sock_net(in_skb->sk);
 547         struct nlattr *tb[NETCONFA_MAX+1];
 548         struct netconfmsg *ncm;
 549         struct sk_buff *skb;
 550         struct ipv6_devconf *devconf;
 551         struct inet6_dev *in6_dev;
 552         struct net_device *dev;
 553         int ifindex;
 554         int err;
 555
 556         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 557                           devconf_ipv6_policy);
 558         if (err < 0)
 559                 goto errout;
 560
 561         err = EINVAL;
 562         if (!tb[NETCONFA_IFINDEX])
 563                 goto errout;
 564
 565         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 566         switch (ifindex) {
 567         case NETCONFA_IFINDEX_ALL:
 568                 devconf = net->ipv6.devconf_all;
 569                 break;
 570         case NETCONFA_IFINDEX_DEFAULT:
 571                 devconf = net->ipv6.devconf_dflt;
 572                 break;
 573         default:
 574                 dev = __dev_get_by_index(net, ifindex);
 575                 if (dev == NULL)
 576                         goto errout;
 577                 in6_dev = __in6_dev_get(dev);
 578                 if (in6_dev == NULL)
 579                         goto errout;
 580                 devconf = &in6_dev->cnf;
 581                 break;
 582         }
 583
 584         err = -ENOBUFS;
 585         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 586         if (skb == NULL)
 587                 goto errout;
 588
 589         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 590                                          NETLINK_CB(in_skb).portid,
 591                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 592                                          -1);
 593         if (err < 0) {
 594                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 595                 WARN_ON(err == -EMSGSIZE);
 596                 kfree_skb(skb);
 597                 goto errout;
 598         }
 599         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 600 errout:
 601         return err;
 602 }
 603
 604 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 605                                       struct netlink_callback *cb)
 606 {
 607         struct net *net = sock_net(skb->sk);
 608         int h, s_h;
 609         int idx, s_idx;
 610         struct net_device *dev;
 611         struct inet6_dev *idev;
 612         struct hlist_head *head;
 613
 614         s_h = cb->args[0];
 615         s_idx = idx = cb->args[1];
 616
 617         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 618                 idx = 0;
 619                 head = &net->dev_index_head[h];
 620                 rcu_read_lock();
 621                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 622                           net->dev_base_seq;
 623                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 624                         if (idx < s_idx)
 625                                 goto cont;
 626                         idev = __in6_dev_get(dev);
 627                         if (!idev)
 628                                 goto cont;
 629
 630                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 631                                                        &idev->cnf,
 632                                                        NETLINK_CB(cb->skb).portid,
 633                                                        cb->nlh->nlmsg_seq,
 634                                                        RTM_NEWNETCONF,
 635                                                        NLM_F_MULTI,
 636                                                        -1) <= 0) {
 637                                 rcu_read_unlock();
 638                                 goto done;
 639                         }
 640                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 641 cont:
 642                         idx++;
 643                 }
 644                 rcu_read_unlock();
 645         }
 646         if (h == NETDEV_HASHENTRIES) {
 647                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 648                                                net->ipv6.devconf_all,
 649                                                NETLINK_CB(cb->skb).portid,
 650                                                cb->nlh->nlmsg_seq,
 651                                                RTM_NEWNETCONF, NLM_F_MULTI,
 652                                                -1) <= 0)
 653                         goto done;
 654                 else
 655                         h++;
 656         }
 657         if (h == NETDEV_HASHENTRIES + 1) {
 658                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 659                                                net->ipv6.devconf_dflt,
 660                                                NETLINK_CB(cb->skb).portid,
 661                                                cb->nlh->nlmsg_seq,
 662                                                RTM_NEWNETCONF, NLM_F_MULTI,
 663                                                -1) <= 0)
 664                         goto done;
 665                 else
 666                         h++;
 667         }
 668 done:
 669         cb->args[0] = h;
 670         cb->args[1] = idx;
 671
 672         return skb->len;
 673 }
 674
 675 #ifdef CONFIG_SYSCTL
 676 static void dev_forward_change(struct inet6_dev *idev)
 677 {
 678         struct net_device *dev;
 679         struct inet6_ifaddr *ifa;
 680
 681         if (!idev)
 682                 return;
 683         dev = idev->dev;
 684         if (idev->cnf.forwarding)
 685                 dev_disable_lro(dev);
 686         if (dev->flags & IFF_MULTICAST) {
 687                 if (idev->cnf.forwarding) {
 688                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 689                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 690                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 691                 } else {
 692                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 693                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 694                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 695                 }
 696         }
 697
 698         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 699                 if (ifa->flags&IFA_F_TENTATIVE)
 700                         continue;
 701                 if (idev->cnf.forwarding)
 702                         addrconf_join_anycast(ifa);
 703                 else
 704                         addrconf_leave_anycast(ifa);
 705         }
 706         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 707                                      dev->ifindex, &idev->cnf);
 708 }
 709
 710
 711 static void addrconf_forward_change(struct net *net, __s32 newf)
 712 {
 713         struct net_device *dev;
 714         struct inet6_dev *idev;
 715
 716         for_each_netdev(net, dev) {
 717                 idev = __in6_dev_get(dev);
 718                 if (idev) {
 719                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 720                         idev->cnf.forwarding = newf;
 721                         if (changed)
 722                                 dev_forward_change(idev);
 723                 }
 724         }
 725 }
 726
 727 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 728 {
 729         struct net *net;
 730         int old;
 731
 732         if (!rtnl_trylock())
 733                 return restart_syscall();
 734
 735         net = (struct net *)table->extra2;
 736         old = *p;
 737         *p = newf;
 738
 739         if (p == &net->ipv6.devconf_dflt->forwarding) {
 740                 if ((!newf) ^ (!old))
 741                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 742                                                      NETCONFA_IFINDEX_DEFAULT,
 743                                                      net->ipv6.devconf_dflt);
 744                 rtnl_unlock();
 745                 return 0;
 746         }
 747
 748         if (p == &net->ipv6.devconf_all->forwarding) {
 749                 net->ipv6.devconf_dflt->forwarding = newf;
 750                 addrconf_forward_change(net, newf);
 751                 if ((!newf) ^ (!old))
 752                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 753                                                      NETCONFA_IFINDEX_ALL,
 754                                                      net->ipv6.devconf_all);
 755         } else if ((!newf) ^ (!old))
 756                 dev_forward_change((struct inet6_dev *)table->extra1);
 757         rtnl_unlock();
 758
 759         if (newf)
 760                 rt6_purge_dflt_routers(net);
 761         return 1;
 762 }
 763 #endif
 764
 765 /* Nobody refers to this ifaddr, destroy it */
 766 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 767 {
 768         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 769
 770 #ifdef NET_REFCNT_DEBUG
 771         pr_debug("%s\n", __func__);
 772 #endif
 773
 774         in6_dev_put(ifp->idev);
 775
 776         if (del_timer(&ifp->dad_timer))
 777                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 778
 779         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 780                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 781                 return;
 782         }
 783         ip6_rt_put(ifp->rt);
 784
 785         kfree_rcu(ifp, rcu);
 786 }
 787
 788 static void
 789 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 790 {
 791         struct list_head *p;
 792         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 793
 794         /*
 795          * Each device address list is sorted in order of scope -
 796          * global before linklocal.
 797          */
 798         list_for_each(p, &idev->addr_list) {
 799                 struct inet6_ifaddr *ifa
 800                         = list_entry(p, struct inet6_ifaddr, if_list);
 801                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 802                         break;
 803         }
 804
 805         list_add_tail(&ifp->if_list, p);
 806 }
 807
 808 static u32 inet6_addr_hash(const struct in6_addr *addr)
 809 {
 810         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 811 }
 812
 813 /* On success it returns ifp with increased reference count */
 814
 815 static struct inet6_ifaddr *
 816 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 817               const struct in6_addr *peer_addr, int pfxlen,
 818               int scope, u32 flags)
 819 {
 820         struct inet6_ifaddr *ifa = NULL;
 821         struct rt6_info *rt;
 822         unsigned int hash;
 823         int err = 0;
 824         int addr_type = ipv6_addr_type(addr);
 825
 826         if (addr_type == IPV6_ADDR_ANY ||
 827             addr_type & IPV6_ADDR_MULTICAST ||
 828             (!(idev->dev->flags & IFF_LOOPBACK) &&
 829              addr_type & IPV6_ADDR_LOOPBACK))
 830                 return ERR_PTR(-EADDRNOTAVAIL);
 831
 832         rcu_read_lock_bh();
 833         if (idev->dead) {
 834                 err = -ENODEV;                  /*XXX*/
 835                 goto out2;
 836         }
 837
 838         if (idev->cnf.disable_ipv6) {
 839                 err = -EACCES;
 840                 goto out2;
 841         }
 842
 843         spin_lock(&addrconf_hash_lock);
 844
 845         /* Ignore adding duplicate addresses on an interface */
 846         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 847                 ADBG(("ipv6_add_addr: already assigned\n"));
 848                 err = -EEXIST;
 849                 goto out;
 850         }
 851
 852         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 853
 854         if (ifa == NULL) {
 855                 ADBG(("ipv6_add_addr: malloc failed\n"));
 856                 err = -ENOBUFS;
 857                 goto out;
 858         }
 859
 860         rt = addrconf_dst_alloc(idev, addr, false);
 861         if (IS_ERR(rt)) {
 862                 err = PTR_ERR(rt);
 863                 goto out;
 864         }
 865
 866         ifa->addr = *addr;
 867         if (peer_addr)
 868                 ifa->peer_addr = *peer_addr;
 869
 870         spin_lock_init(&ifa->lock);
 871         spin_lock_init(&ifa->state_lock);
 872         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 873                     (unsigned long)ifa);
 874         INIT_HLIST_NODE(&ifa->addr_lst);
 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_dad_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, NULL, tmp_plen,
1130                               ipv6_addr_scope(&addr), 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 inet6_dev *idev, struct in6_addr *addr,
1451                       unsigned char banned_flags)
1452 {
1453         struct inet6_ifaddr *ifp;
1454         int err = -EADDRNOTAVAIL;
1455
1456         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1457                 if (ifp->scope == IFA_LINK &&
1458                     !(ifp->flags & banned_flags)) {
1459                         *addr = ifp->addr;
1460                         err = 0;
1461                         break;
1462                 }
1463         }
1464         return err;
1465 }
1466
1467 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1468                     unsigned char banned_flags)
1469 {
1470         struct inet6_dev *idev;
1471         int err = -EADDRNOTAVAIL;
1472
1473         rcu_read_lock();
1474         idev = __in6_dev_get(dev);
1475         if (idev) {
1476                 read_lock_bh(&idev->lock);
1477                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1478                 read_unlock_bh(&idev->lock);
1479         }
1480         rcu_read_unlock();
1481         return err;
1482 }
1483
1484 static int ipv6_count_addresses(struct inet6_dev *idev)
1485 {
1486         int cnt = 0;
1487         struct inet6_ifaddr *ifp;
1488
1489         read_lock_bh(&idev->lock);
1490         list_for_each_entry(ifp, &idev->addr_list, if_list)
1491                 cnt++;
1492         read_unlock_bh(&idev->lock);
1493         return cnt;
1494 }
1495
1496 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1497                   const struct net_device *dev, int strict)
1498 {
1499         struct inet6_ifaddr *ifp;
1500         unsigned int hash = inet6_addr_hash(addr);
1501
1502         rcu_read_lock_bh();
1503         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1504                 if (!net_eq(dev_net(ifp->idev->dev), net))
1505                         continue;
1506                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1507                     !(ifp->flags&IFA_F_TENTATIVE) &&
1508                     (dev == NULL || ifp->idev->dev == dev ||
1509                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1510                         rcu_read_unlock_bh();
1511                         return 1;
1512                 }
1513         }
1514
1515         rcu_read_unlock_bh();
1516         return 0;
1517 }
1518 EXPORT_SYMBOL(ipv6_chk_addr);
1519
1520 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1521                                struct net_device *dev)
1522 {
1523         unsigned int hash = inet6_addr_hash(addr);
1524         struct inet6_ifaddr *ifp;
1525
1526         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1527                 if (!net_eq(dev_net(ifp->idev->dev), net))
1528                         continue;
1529                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1530                         if (dev == NULL || ifp->idev->dev == dev)
1531                                 return true;
1532                 }
1533         }
1534         return false;
1535 }
1536
1537 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1538 {
1539         struct inet6_dev *idev;
1540         struct inet6_ifaddr *ifa;
1541         int     onlink;
1542
1543         onlink = 0;
1544         rcu_read_lock();
1545         idev = __in6_dev_get(dev);
1546         if (idev) {
1547                 read_lock_bh(&idev->lock);
1548                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1549                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1550                                                    ifa->prefix_len);
1551                         if (onlink)
1552                                 break;
1553                 }
1554                 read_unlock_bh(&idev->lock);
1555         }
1556         rcu_read_unlock();
1557         return onlink;
1558 }
1559 EXPORT_SYMBOL(ipv6_chk_prefix);
1560
1561 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1562                                      struct net_device *dev, int strict)
1563 {
1564         struct inet6_ifaddr *ifp, *result = NULL;
1565         unsigned int hash = inet6_addr_hash(addr);
1566
1567         rcu_read_lock_bh();
1568         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1569                 if (!net_eq(dev_net(ifp->idev->dev), net))
1570                         continue;
1571                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1572                         if (dev == NULL || ifp->idev->dev == dev ||
1573                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1574                                 result = ifp;
1575                                 in6_ifa_hold(ifp);
1576                                 break;
1577                         }
1578                 }
1579         }
1580         rcu_read_unlock_bh();
1581
1582         return result;
1583 }
1584
1585 /* Gets referenced address, destroys ifaddr */
1586
1587 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1588 {
1589         if (ifp->flags&IFA_F_PERMANENT) {
1590                 spin_lock_bh(&ifp->lock);
1591                 addrconf_del_dad_timer(ifp);
1592                 ifp->flags |= IFA_F_TENTATIVE;
1593                 if (dad_failed)
1594                         ifp->flags |= IFA_F_DADFAILED;
1595                 spin_unlock_bh(&ifp->lock);
1596                 if (dad_failed)
1597                         ipv6_ifa_notify(0, ifp);
1598                 in6_ifa_put(ifp);
1599 #ifdef CONFIG_IPV6_PRIVACY
1600         } else if (ifp->flags&IFA_F_TEMPORARY) {
1601                 struct inet6_ifaddr *ifpub;
1602                 spin_lock_bh(&ifp->lock);
1603                 ifpub = ifp->ifpub;
1604                 if (ifpub) {
1605                         in6_ifa_hold(ifpub);
1606                         spin_unlock_bh(&ifp->lock);
1607                         ipv6_create_tempaddr(ifpub, ifp);
1608                         in6_ifa_put(ifpub);
1609                 } else {
1610                         spin_unlock_bh(&ifp->lock);
1611                 }
1612                 ipv6_del_addr(ifp);
1613 #endif
1614         } else
1615                 ipv6_del_addr(ifp);
1616 }
1617
1618 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1619 {
1620         int err = -ENOENT;
1621
1622         spin_lock(&ifp->state_lock);
1623         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1624                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1625                 err = 0;
1626         }
1627         spin_unlock(&ifp->state_lock);
1628
1629         return err;
1630 }
1631
1632 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1633 {
1634         struct inet6_dev *idev = ifp->idev;
1635
1636         if (addrconf_dad_end(ifp)) {
1637                 in6_ifa_put(ifp);
1638                 return;
1639         }
1640
1641         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1642                              ifp->idev->dev->name, &ifp->addr);
1643
1644         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1645                 struct in6_addr addr;
1646
1647                 addr.s6_addr32[0] = htonl(0xfe800000);
1648                 addr.s6_addr32[1] = 0;
1649
1650                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1651                     ipv6_addr_equal(&ifp->addr, &addr)) {
1652                         /* DAD failed for link-local based on MAC address */
1653                         idev->cnf.disable_ipv6 = 1;
1654
1655                         pr_info("%s: IPv6 being disabled!\n",
1656                                 ifp->idev->dev->name);
1657                 }
1658         }
1659
1660         addrconf_dad_stop(ifp, 1);
1661 }
1662
1663 /* Join to solicited addr multicast group. */
1664
1665 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1666 {
1667         struct in6_addr maddr;
1668
1669         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1670                 return;
1671
1672         addrconf_addr_solict_mult(addr, &maddr);
1673         ipv6_dev_mc_inc(dev, &maddr);
1674 }
1675
1676 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1677 {
1678         struct in6_addr maddr;
1679
1680         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1681                 return;
1682
1683         addrconf_addr_solict_mult(addr, &maddr);
1684         __ipv6_dev_mc_dec(idev, &maddr);
1685 }
1686
1687 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1688 {
1689         struct in6_addr addr;
1690         if (ifp->prefix_len == 127) /* RFC 6164 */
1691                 return;
1692         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1693         if (ipv6_addr_any(&addr))
1694                 return;
1695         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1696 }
1697
1698 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1699 {
1700         struct in6_addr addr;
1701         if (ifp->prefix_len == 127) /* RFC 6164 */
1702                 return;
1703         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1704         if (ipv6_addr_any(&addr))
1705                 return;
1706         __ipv6_dev_ac_dec(ifp->idev, &addr);
1707 }
1708
1709 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1710 {
1711         if (dev->addr_len != ETH_ALEN)
1712                 return -1;
1713         memcpy(eui, dev->dev_addr, 3);
1714         memcpy(eui + 5, dev->dev_addr + 3, 3);
1715
1716         /*
1717          * The zSeries OSA network cards can be shared among various
1718          * OS instances, but the OSA cards have only one MAC address.
1719          * This leads to duplicate address conflicts in conjunction
1720          * with IPv6 if more than one instance uses the same card.
1721          *
1722          * The driver for these cards can deliver a unique 16-bit
1723          * identifier for each instance sharing the same card.  It is
1724          * placed instead of 0xFFFE in the interface identifier.  The
1725          * "u" bit of the interface identifier is not inverted in this
1726          * case.  Hence the resulting interface identifier has local
1727          * scope according to RFC2373.
1728          */
1729         if (dev->dev_id) {
1730                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1731                 eui[4] = dev->dev_id & 0xFF;
1732         } else {
1733                 eui[3] = 0xFF;
1734                 eui[4] = 0xFE;
1735                 eui[0] ^= 2;
1736         }
1737         return 0;
1738 }
1739
1740 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1741 {
1742         if (dev->addr_len != IEEE802154_ADDR_LEN)
1743                 return -1;
1744         memcpy(eui, dev->dev_addr, 8);
1745         eui[0] ^= 2;
1746         return 0;
1747 }
1748
1749 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1750 {
1751         union fwnet_hwaddr *ha;
1752
1753         if (dev->addr_len != FWNET_ALEN)
1754                 return -1;
1755
1756         ha = (union fwnet_hwaddr *)dev->dev_addr;
1757
1758         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1759         eui[0] ^= 2;
1760         return 0;
1761 }
1762
1763 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1764 {
1765         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1766         if (dev->addr_len != ARCNET_ALEN)
1767                 return -1;
1768         memset(eui, 0, 7);
1769         eui[7] = *(u8 *)dev->dev_addr;
1770         return 0;
1771 }
1772
1773 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1774 {
1775         if (dev->addr_len != INFINIBAND_ALEN)
1776                 return -1;
1777         memcpy(eui, dev->dev_addr + 12, 8);
1778         eui[0] |= 2;
1779         return 0;
1780 }
1781
1782 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1783 {
1784         if (addr == 0)
1785                 return -1;
1786         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1787                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1788                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1789                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1790                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1791                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1792         eui[1] = 0;
1793         eui[2] = 0x5E;
1794         eui[3] = 0xFE;
1795         memcpy(eui + 4, &addr, 4);
1796         return 0;
1797 }
1798
1799 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1800 {
1801         if (dev->priv_flags & IFF_ISATAP)
1802                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1803         return -1;
1804 }
1805
1806 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1807 {
1808         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1809 }
1810
1811 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1812 {
1813         switch (dev->type) {
1814         case ARPHRD_ETHER:
1815         case ARPHRD_FDDI:
1816                 return addrconf_ifid_eui48(eui, dev);
1817         case ARPHRD_ARCNET:
1818                 return addrconf_ifid_arcnet(eui, dev);
1819         case ARPHRD_INFINIBAND:
1820                 return addrconf_ifid_infiniband(eui, dev);
1821         case ARPHRD_SIT:
1822                 return addrconf_ifid_sit(eui, dev);
1823         case ARPHRD_IPGRE:
1824                 return addrconf_ifid_gre(eui, dev);
1825         case ARPHRD_IEEE802154:
1826                 return addrconf_ifid_eui64(eui, dev);
1827         case ARPHRD_IEEE1394:
1828                 return addrconf_ifid_ieee1394(eui, dev);
1829         }
1830         return -1;
1831 }
1832
1833 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1834 {
1835         int err = -1;
1836         struct inet6_ifaddr *ifp;
1837
1838         read_lock_bh(&idev->lock);
1839         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1840                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1841                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1842                         err = 0;
1843                         break;
1844                 }
1845         }
1846         read_unlock_bh(&idev->lock);
1847         return err;
1848 }
1849
1850 #ifdef CONFIG_IPV6_PRIVACY
1851 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1852 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1853 {
1854 regen:
1855         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1856         idev->rndid[0] &= ~0x02;
1857
1858         /*
1859          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1860          * check if generated address is not inappropriate
1861          *
1862          *  - Reserved subnet anycast (RFC 2526)
1863          *      11111101 11....11 1xxxxxxx
1864          *  - ISATAP (RFC4214) 6.1
1865          *      00-00-5E-FE-xx-xx-xx-xx
1866          *  - value 0
1867          *  - XXX: already assigned to an address on the device
1868          */
1869         if (idev->rndid[0] == 0xfd &&
1870             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1871             (idev->rndid[7]&0x80))
1872                 goto regen;
1873         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1874                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1875                         goto regen;
1876                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1877                         goto regen;
1878         }
1879 }
1880
1881 static void ipv6_regen_rndid(unsigned long data)
1882 {
1883         struct inet6_dev *idev = (struct inet6_dev *) data;
1884         unsigned long expires;
1885
1886         rcu_read_lock_bh();
1887         write_lock_bh(&idev->lock);
1888
1889         if (idev->dead)
1890                 goto out;
1891
1892         __ipv6_regen_rndid(idev);
1893
1894         expires = jiffies +
1895                 idev->cnf.temp_prefered_lft * HZ -
1896                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1897                 idev->cnf.max_desync_factor * HZ;
1898         if (time_before(expires, jiffies)) {
1899                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1900                         __func__, idev->dev->name);
1901                 goto out;
1902         }
1903
1904         if (!mod_timer(&idev->regen_timer, expires))
1905                 in6_dev_hold(idev);
1906
1907 out:
1908         write_unlock_bh(&idev->lock);
1909         rcu_read_unlock_bh();
1910         in6_dev_put(idev);
1911 }
1912
1913 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1914 {
1915         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1916                 __ipv6_regen_rndid(idev);
1917 }
1918 #endif
1919
1920 /*
1921  *      Add prefix route.
1922  */
1923
1924 static void
1925 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1926                       unsigned long expires, u32 flags)
1927 {
1928         struct fib6_config cfg = {
1929                 .fc_table = RT6_TABLE_PREFIX,
1930                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1931                 .fc_ifindex = dev->ifindex,
1932                 .fc_expires = expires,
1933                 .fc_dst_len = plen,
1934                 .fc_flags = RTF_UP | flags,
1935                 .fc_nlinfo.nl_net = dev_net(dev),
1936                 .fc_protocol = RTPROT_KERNEL,
1937         };
1938
1939         cfg.fc_dst = *pfx;
1940
1941         /* Prevent useless cloning on PtP SIT.
1942            This thing is done here expecting that the whole
1943            class of non-broadcast devices need not cloning.
1944          */
1945 #if IS_ENABLED(CONFIG_IPV6_SIT)
1946         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1947                 cfg.fc_flags |= RTF_NONEXTHOP;
1948 #endif
1949
1950         ip6_route_add(&cfg);
1951 }
1952
1953
1954 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1955                                                   int plen,
1956                                                   const struct net_device *dev,
1957                                                   u32 flags, u32 noflags)
1958 {
1959         struct fib6_node *fn;
1960         struct rt6_info *rt = NULL;
1961         struct fib6_table *table;
1962
1963         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1964         if (table == NULL)
1965                 return NULL;
1966
1967         read_lock_bh(&table->tb6_lock);
1968         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1969         if (!fn)
1970                 goto out;
1971         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1972                 if (rt->dst.dev->ifindex != dev->ifindex)
1973                         continue;
1974                 if ((rt->rt6i_flags & flags) != flags)
1975                         continue;
1976                 if ((rt->rt6i_flags & noflags) != 0)
1977                         continue;
1978                 dst_hold(&rt->dst);
1979                 break;
1980         }
1981 out:
1982         read_unlock_bh(&table->tb6_lock);
1983         return rt;
1984 }
1985
1986
1987 /* Create "default" multicast route to the interface */
1988
1989 static void addrconf_add_mroute(struct net_device *dev)
1990 {
1991         struct fib6_config cfg = {
1992                 .fc_table = RT6_TABLE_LOCAL,
1993                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1994                 .fc_ifindex = dev->ifindex,
1995                 .fc_dst_len = 8,
1996                 .fc_flags = RTF_UP,
1997                 .fc_nlinfo.nl_net = dev_net(dev),
1998         };
1999
2000         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2001
2002         ip6_route_add(&cfg);
2003 }
2004
2005 #if IS_ENABLED(CONFIG_IPV6_SIT)
2006 static void sit_route_add(struct net_device *dev)
2007 {
2008         struct fib6_config cfg = {
2009                 .fc_table = RT6_TABLE_MAIN,
2010                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2011                 .fc_ifindex = dev->ifindex,
2012                 .fc_dst_len = 96,
2013                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2014                 .fc_nlinfo.nl_net = dev_net(dev),
2015         };
2016
2017         /* prefix length - 96 bits "::d.d.d.d" */
2018         ip6_route_add(&cfg);
2019 }
2020 #endif
2021
2022 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2023 {
2024         struct inet6_dev *idev;
2025
2026         ASSERT_RTNL();
2027
2028         idev = ipv6_find_idev(dev);
2029         if (!idev)
2030                 return ERR_PTR(-ENOBUFS);
2031
2032         if (idev->cnf.disable_ipv6)
2033                 return ERR_PTR(-EACCES);
2034
2035         /* Add default multicast route */
2036         if (!(dev->flags & IFF_LOOPBACK))
2037                 addrconf_add_mroute(dev);
2038
2039         return idev;
2040 }
2041
2042 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2043 {
2044         struct prefix_info *pinfo;
2045         __u32 valid_lft;
2046         __u32 prefered_lft;
2047         int addr_type;
2048         struct inet6_dev *in6_dev;
2049         struct net *net = dev_net(dev);
2050
2051         pinfo = (struct prefix_info *) opt;
2052
2053         if (len < sizeof(struct prefix_info)) {
2054                 ADBG(("addrconf: prefix option too short\n"));
2055                 return;
2056         }
2057
2058         /*
2059          *      Validation checks ([ADDRCONF], page 19)
2060          */
2061
2062         addr_type = ipv6_addr_type(&pinfo->prefix);
2063
2064         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2065                 return;
2066
2067         valid_lft = ntohl(pinfo->valid);
2068         prefered_lft = ntohl(pinfo->prefered);
2069
2070         if (prefered_lft > valid_lft) {
2071                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2072                 return;
2073         }
2074
2075         in6_dev = in6_dev_get(dev);
2076
2077         if (in6_dev == NULL) {
2078                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2079                                     dev->name);
2080                 return;
2081         }
2082
2083         /*
2084          *      Two things going on here:
2085          *      1) Add routes for on-link prefixes
2086          *      2) Configure prefixes with the auto flag set
2087          */
2088
2089         if (pinfo->onlink) {
2090                 struct rt6_info *rt;
2091                 unsigned long rt_expires;
2092
2093                 /* Avoid arithmetic overflow. Really, we could
2094                  * save rt_expires in seconds, likely valid_lft,
2095                  * but it would require division in fib gc, that it
2096                  * not good.
2097                  */
2098                 if (HZ > USER_HZ)
2099                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2100                 else
2101                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2102
2103                 if (addrconf_finite_timeout(rt_expires))
2104                         rt_expires *= HZ;
2105
2106                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2107                                                pinfo->prefix_len,
2108                                                dev,
2109                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2110                                                RTF_GATEWAY | RTF_DEFAULT);
2111
2112                 if (rt) {
2113                         /* Autoconf prefix route */
2114                         if (valid_lft == 0) {
2115                                 ip6_del_rt(rt);
2116                                 rt = NULL;
2117                         } else if (addrconf_finite_timeout(rt_expires)) {
2118                                 /* not infinity */
2119                                 rt6_set_expires(rt, jiffies + rt_expires);
2120                         } else {
2121                                 rt6_clean_expires(rt);
2122                         }
2123                 } else if (valid_lft) {
2124                         clock_t expires = 0;
2125                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2126                         if (addrconf_finite_timeout(rt_expires)) {
2127                                 /* not infinity */
2128                                 flags |= RTF_EXPIRES;
2129                                 expires = jiffies_to_clock_t(rt_expires);
2130                         }
2131                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2132                                               dev, expires, flags);
2133                 }
2134                 ip6_rt_put(rt);
2135         }
2136
2137         /* Try to figure out our local address for this prefix */
2138
2139         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2140                 struct inet6_ifaddr *ifp;
2141                 struct in6_addr addr;
2142                 int create = 0, update_lft = 0;
2143                 bool tokenized = false;
2144
2145                 if (pinfo->prefix_len == 64) {
2146                         memcpy(&addr, &pinfo->prefix, 8);
2147
2148                         if (!ipv6_addr_any(&in6_dev->token)) {
2149                                 read_lock_bh(&in6_dev->lock);
2150                                 memcpy(addr.s6_addr + 8,
2151                                        in6_dev->token.s6_addr + 8, 8);
2152                                 read_unlock_bh(&in6_dev->lock);
2153                                 tokenized = true;
2154                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2155                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2156                                 in6_dev_put(in6_dev);
2157                                 return;
2158                         }
2159                         goto ok;
2160                 }
2161                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2162                                     pinfo->prefix_len);
2163                 in6_dev_put(in6_dev);
2164                 return;
2165
2166 ok:
2167
2168                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2169
2170                 if (ifp == NULL && valid_lft) {
2171                         int max_addresses = in6_dev->cnf.max_addresses;
2172                         u32 addr_flags = 0;
2173
2174 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2175                         if (in6_dev->cnf.optimistic_dad &&
2176                             !net->ipv6.devconf_all->forwarding && sllao)
2177                                 addr_flags = IFA_F_OPTIMISTIC;
2178 #endif
2179
2180                         /* Do not allow to create too much of autoconfigured
2181                          * addresses; this would be too easy way to crash kernel.
2182                          */
2183                         if (!max_addresses ||
2184                             ipv6_count_addresses(in6_dev) < max_addresses)
2185                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2186                                                     pinfo->prefix_len,
2187                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2188                                                     addr_flags);
2189
2190                         if (IS_ERR_OR_NULL(ifp)) {
2191                                 in6_dev_put(in6_dev);
2192                                 return;
2193                         }
2194
2195                         update_lft = create = 1;
2196                         ifp->cstamp = jiffies;
2197                         ifp->tokenized = tokenized;
2198                         addrconf_dad_start(ifp);
2199                 }
2200
2201                 if (ifp) {
2202                         int flags;
2203                         unsigned long now;
2204 #ifdef CONFIG_IPV6_PRIVACY
2205                         struct inet6_ifaddr *ift;
2206 #endif
2207                         u32 stored_lft;
2208
2209                         /* update lifetime (RFC2462 5.5.3 e) */
2210                         spin_lock(&ifp->lock);
2211                         now = jiffies;
2212                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2213                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2214                         else
2215                                 stored_lft = 0;
2216                         if (!update_lft && stored_lft) {
2217                                 if (valid_lft > MIN_VALID_LIFETIME ||
2218                                     valid_lft > stored_lft)
2219                                         update_lft = 1;
2220                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2221                                         /* valid_lft <= stored_lft is always true */
2222                                         /*
2223                                          * RFC 4862 Section 5.5.3e:
2224                                          * "Note that the preferred lifetime of
2225                                          *  the corresponding address is always
2226                                          *  reset to the Preferred Lifetime in
2227                                          *  the received Prefix Information
2228                                          *  option, regardless of whether the
2229                                          *  valid lifetime is also reset or
2230                                          *  ignored."
2231                                          *
2232                                          *  So if the preferred lifetime in
2233                                          *  this advertisement is different
2234                                          *  than what we have stored, but the
2235                                          *  valid lifetime is invalid, just
2236                                          *  reset prefered_lft.
2237                                          *
2238                                          *  We must set the valid lifetime
2239                                          *  to the stored lifetime since we'll
2240                                          *  be updating the timestamp below,
2241                                          *  else we'll set it back to the
2242                                          *  minimum.
2243                                          */
2244                                         if (prefered_lft != ifp->prefered_lft) {
2245                                                 valid_lft = stored_lft;
2246                                                 update_lft = 1;
2247                                         }
2248                                 } else {
2249                                         valid_lft = MIN_VALID_LIFETIME;
2250                                         if (valid_lft < prefered_lft)
2251                                                 prefered_lft = valid_lft;
2252                                         update_lft = 1;
2253                                 }
2254                         }
2255
2256                         if (update_lft) {
2257                                 ifp->valid_lft = valid_lft;
2258                                 ifp->prefered_lft = prefered_lft;
2259                                 ifp->tstamp = now;
2260                                 flags = ifp->flags;
2261                                 ifp->flags &= ~IFA_F_DEPRECATED;
2262                                 spin_unlock(&ifp->lock);
2263
2264                                 if (!(flags&IFA_F_TENTATIVE))
2265                                         ipv6_ifa_notify(0, ifp);
2266                         } else
2267                                 spin_unlock(&ifp->lock);
2268
2269 #ifdef CONFIG_IPV6_PRIVACY
2270                         read_lock_bh(&in6_dev->lock);
2271                         /* update all temporary addresses in the list */
2272                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2273                                             tmp_list) {
2274                                 int age, max_valid, max_prefered;
2275
2276                                 if (ifp != ift->ifpub)
2277                                         continue;
2278
2279                                 /*
2280                                  * RFC 4941 section 3.3:
2281                                  * If a received option will extend the lifetime
2282                                  * of a public address, the lifetimes of
2283                                  * temporary addresses should be extended,
2284                                  * subject to the overall constraint that no
2285                                  * temporary addresses should ever remain
2286                                  * "valid" or "preferred" for a time longer than
2287                                  * (TEMP_VALID_LIFETIME) or
2288                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2289                                  * respectively.
2290                                  */
2291                                 age = (now - ift->cstamp) / HZ;
2292                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2293                                 if (max_valid < 0)
2294                                         max_valid = 0;
2295
2296                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2297                                                in6_dev->cnf.max_desync_factor -
2298                                                age;
2299                                 if (max_prefered < 0)
2300                                         max_prefered = 0;
2301
2302                                 if (valid_lft > max_valid)
2303                                         valid_lft = max_valid;
2304
2305                                 if (prefered_lft > max_prefered)
2306                                         prefered_lft = max_prefered;
2307
2308                                 spin_lock(&ift->lock);
2309                                 flags = ift->flags;
2310                                 ift->valid_lft = valid_lft;
2311                                 ift->prefered_lft = prefered_lft;
2312                                 ift->tstamp = now;
2313                                 if (prefered_lft > 0)
2314                                         ift->flags &= ~IFA_F_DEPRECATED;
2315
2316                                 spin_unlock(&ift->lock);
2317                                 if (!(flags&IFA_F_TENTATIVE))
2318                                         ipv6_ifa_notify(0, ift);
2319                         }
2320
2321                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2322                                 /*
2323                                  * When a new public address is created as
2324                                  * described in [ADDRCONF], also create a new
2325                                  * temporary address. Also create a temporary
2326                                  * address if it's enabled but no temporary
2327                                  * address currently exists.
2328                                  */
2329                                 read_unlock_bh(&in6_dev->lock);
2330                                 ipv6_create_tempaddr(ifp, NULL);
2331                         } else {
2332                                 read_unlock_bh(&in6_dev->lock);
2333                         }
2334 #endif
2335                         in6_ifa_put(ifp);
2336                         addrconf_verify(0);
2337                 }
2338         }
2339         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2340         in6_dev_put(in6_dev);
2341 }
2342
2343 /*
2344  *      Set destination address.
2345  *      Special case for SIT interfaces where we create a new "virtual"
2346  *      device.
2347  */
2348 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2349 {
2350         struct in6_ifreq ireq;
2351         struct net_device *dev;
2352         int err = -EINVAL;
2353
2354         rtnl_lock();
2355
2356         err = -EFAULT;
2357         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2358                 goto err_exit;
2359
2360         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2361
2362         err = -ENODEV;
2363         if (dev == NULL)
2364                 goto err_exit;
2365
2366 #if IS_ENABLED(CONFIG_IPV6_SIT)
2367         if (dev->type == ARPHRD_SIT) {
2368                 const struct net_device_ops *ops = dev->netdev_ops;
2369                 struct ifreq ifr;
2370                 struct ip_tunnel_parm p;
2371
2372                 err = -EADDRNOTAVAIL;
2373                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2374                         goto err_exit;
2375
2376                 memset(&p, 0, sizeof(p));
2377                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2378                 p.iph.saddr = 0;
2379                 p.iph.version = 4;
2380                 p.iph.ihl = 5;
2381                 p.iph.protocol = IPPROTO_IPV6;
2382                 p.iph.ttl = 64;
2383                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2384
2385                 if (ops->ndo_do_ioctl) {
2386                         mm_segment_t oldfs = get_fs();
2387
2388                         set_fs(KERNEL_DS);
2389                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2390                         set_fs(oldfs);
2391                 } else
2392                         err = -EOPNOTSUPP;
2393
2394                 if (err == 0) {
2395                         err = -ENOBUFS;
2396                         dev = __dev_get_by_name(net, p.name);
2397                         if (!dev)
2398                                 goto err_exit;
2399                         err = dev_open(dev);
2400                 }
2401         }
2402 #endif
2403
2404 err_exit:
2405         rtnl_unlock();
2406         return err;
2407 }
2408
2409 /*
2410  *      Manual configuration of address on an interface
2411  */
2412 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2413                           const struct in6_addr *peer_pfx,
2414                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2415                           __u32 valid_lft)
2416 {
2417         struct inet6_ifaddr *ifp;
2418         struct inet6_dev *idev;
2419         struct net_device *dev;
2420         int scope;
2421         u32 flags;
2422         clock_t expires;
2423         unsigned long timeout;
2424
2425         ASSERT_RTNL();
2426
2427         if (plen > 128)
2428                 return -EINVAL;
2429
2430         /* check the lifetime */
2431         if (!valid_lft || prefered_lft > valid_lft)
2432                 return -EINVAL;
2433
2434         dev = __dev_get_by_index(net, ifindex);
2435         if (!dev)
2436                 return -ENODEV;
2437
2438         idev = addrconf_add_dev(dev);
2439         if (IS_ERR(idev))
2440                 return PTR_ERR(idev);
2441
2442         scope = ipv6_addr_scope(pfx);
2443
2444         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2445         if (addrconf_finite_timeout(timeout)) {
2446                 expires = jiffies_to_clock_t(timeout * HZ);
2447                 valid_lft = timeout;
2448                 flags = RTF_EXPIRES;
2449         } else {
2450                 expires = 0;
2451                 flags = 0;
2452                 ifa_flags |= IFA_F_PERMANENT;
2453         }
2454
2455         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2456         if (addrconf_finite_timeout(timeout)) {
2457                 if (timeout == 0)
2458                         ifa_flags |= IFA_F_DEPRECATED;
2459                 prefered_lft = timeout;
2460         }
2461
2462         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags);
2463
2464         if (!IS_ERR(ifp)) {
2465                 spin_lock_bh(&ifp->lock);
2466                 ifp->valid_lft = valid_lft;
2467                 ifp->prefered_lft = prefered_lft;
2468                 ifp->tstamp = jiffies;
2469                 spin_unlock_bh(&ifp->lock);
2470
2471                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2472                                       expires, flags);
2473                 /*
2474                  * Note that section 3.1 of RFC 4429 indicates
2475                  * that the Optimistic flag should not be set for
2476                  * manually configured addresses
2477                  */
2478                 addrconf_dad_start(ifp);
2479                 in6_ifa_put(ifp);
2480                 addrconf_verify(0);
2481                 return 0;
2482         }
2483
2484         return PTR_ERR(ifp);
2485 }
2486
2487 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2488                           unsigned int plen)
2489 {
2490         struct inet6_ifaddr *ifp;
2491         struct inet6_dev *idev;
2492         struct net_device *dev;
2493
2494         if (plen > 128)
2495                 return -EINVAL;
2496
2497         dev = __dev_get_by_index(net, ifindex);
2498         if (!dev)
2499                 return -ENODEV;
2500
2501         if ((idev = __in6_dev_get(dev)) == NULL)
2502                 return -ENXIO;
2503
2504         read_lock_bh(&idev->lock);
2505         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2506                 if (ifp->prefix_len == plen &&
2507                     ipv6_addr_equal(pfx, &ifp->addr)) {
2508                         in6_ifa_hold(ifp);
2509                         read_unlock_bh(&idev->lock);
2510
2511                         ipv6_del_addr(ifp);
2512                         return 0;
2513                 }
2514         }
2515         read_unlock_bh(&idev->lock);
2516         return -EADDRNOTAVAIL;
2517 }
2518
2519
2520 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2521 {
2522         struct in6_ifreq ireq;
2523         int err;
2524
2525         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2526                 return -EPERM;
2527
2528         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2529                 return -EFAULT;
2530
2531         rtnl_lock();
2532         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2533                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2534                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2535         rtnl_unlock();
2536         return err;
2537 }
2538
2539 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2540 {
2541         struct in6_ifreq ireq;
2542         int err;
2543
2544         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2545                 return -EPERM;
2546
2547         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2548                 return -EFAULT;
2549
2550         rtnl_lock();
2551         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2552                              ireq.ifr6_prefixlen);
2553         rtnl_unlock();
2554         return err;
2555 }
2556
2557 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2558                      int plen, int scope)
2559 {
2560         struct inet6_ifaddr *ifp;
2561
2562         ifp = ipv6_add_addr(idev, addr, NULL, plen, scope, IFA_F_PERMANENT);
2563         if (!IS_ERR(ifp)) {
2564                 spin_lock_bh(&ifp->lock);
2565                 ifp->flags &= ~IFA_F_TENTATIVE;
2566                 spin_unlock_bh(&ifp->lock);
2567                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2568                 in6_ifa_put(ifp);
2569         }
2570 }
2571
2572 #if IS_ENABLED(CONFIG_IPV6_SIT)
2573 static void sit_add_v4_addrs(struct inet6_dev *idev)
2574 {
2575         struct in6_addr addr;
2576         struct net_device *dev;
2577         struct net *net = dev_net(idev->dev);
2578         int scope;
2579
2580         ASSERT_RTNL();
2581
2582         memset(&addr, 0, sizeof(struct in6_addr));
2583         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2584
2585         if (idev->dev->flags&IFF_POINTOPOINT) {
2586                 addr.s6_addr32[0] = htonl(0xfe800000);
2587                 scope = IFA_LINK;
2588         } else {
2589                 scope = IPV6_ADDR_COMPATv4;
2590         }
2591
2592         if (addr.s6_addr32[3]) {
2593                 add_addr(idev, &addr, 128, scope);
2594                 return;
2595         }
2596
2597         for_each_netdev(net, dev) {
2598                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2599                 if (in_dev && (dev->flags & IFF_UP)) {
2600                         struct in_ifaddr *ifa;
2601
2602                         int flag = scope;
2603
2604                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2605                                 int plen;
2606
2607                                 addr.s6_addr32[3] = ifa->ifa_local;
2608
2609                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2610                                         continue;
2611                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2612                                         if (idev->dev->flags&IFF_POINTOPOINT)
2613                                                 continue;
2614                                         flag |= IFA_HOST;
2615                                 }
2616                                 if (idev->dev->flags&IFF_POINTOPOINT)
2617                                         plen = 64;
2618                                 else
2619                                         plen = 96;
2620
2621                                 add_addr(idev, &addr, plen, flag);
2622                         }
2623                 }
2624         }
2625 }
2626 #endif
2627
2628 static void init_loopback(struct net_device *dev)
2629 {
2630         struct inet6_dev  *idev;
2631         struct net_device *sp_dev;
2632         struct inet6_ifaddr *sp_ifa;
2633         struct rt6_info *sp_rt;
2634
2635         /* ::1 */
2636
2637         ASSERT_RTNL();
2638
2639         if ((idev = ipv6_find_idev(dev)) == NULL) {
2640                 pr_debug("%s: add_dev failed\n", __func__);
2641                 return;
2642         }
2643
2644         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2645
2646         /* Add routes to other interface's IPv6 addresses */
2647         for_each_netdev(dev_net(dev), sp_dev) {
2648                 if (!strcmp(sp_dev->name, dev->name))
2649                         continue;
2650
2651                 idev = __in6_dev_get(sp_dev);
2652                 if (!idev)
2653                         continue;
2654
2655                 read_lock_bh(&idev->lock);
2656                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2657
2658                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2659                                 continue;
2660
2661                         if (sp_ifa->rt)
2662                                 continue;
2663
2664                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2665
2666                         /* Failure cases are ignored */
2667                         if (!IS_ERR(sp_rt)) {
2668                                 sp_ifa->rt = sp_rt;
2669                                 ip6_ins_rt(sp_rt);
2670                         }
2671                 }
2672                 read_unlock_bh(&idev->lock);
2673         }
2674 }
2675
2676 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2677 {
2678         struct inet6_ifaddr *ifp;
2679         u32 addr_flags = IFA_F_PERMANENT;
2680
2681 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2682         if (idev->cnf.optimistic_dad &&
2683             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2684                 addr_flags |= IFA_F_OPTIMISTIC;
2685 #endif
2686
2687
2688         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags);
2689         if (!IS_ERR(ifp)) {
2690                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2691                 addrconf_dad_start(ifp);
2692                 in6_ifa_put(ifp);
2693         }
2694 }
2695
2696 static void addrconf_dev_config(struct net_device *dev)
2697 {
2698         struct in6_addr addr;
2699         struct inet6_dev *idev;
2700
2701         ASSERT_RTNL();
2702
2703         if ((dev->type != ARPHRD_ETHER) &&
2704             (dev->type != ARPHRD_FDDI) &&
2705             (dev->type != ARPHRD_ARCNET) &&
2706             (dev->type != ARPHRD_INFINIBAND) &&
2707             (dev->type != ARPHRD_IEEE802154) &&
2708             (dev->type != ARPHRD_IEEE1394)) {
2709                 /* Alas, we support only Ethernet autoconfiguration. */
2710                 return;
2711         }
2712
2713         idev = addrconf_add_dev(dev);
2714         if (IS_ERR(idev))
2715                 return;
2716
2717         memset(&addr, 0, sizeof(struct in6_addr));
2718         addr.s6_addr32[0] = htonl(0xFE800000);
2719
2720         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2721                 addrconf_add_linklocal(idev, &addr);
2722 }
2723
2724 #if IS_ENABLED(CONFIG_IPV6_SIT)
2725 static void addrconf_sit_config(struct net_device *dev)
2726 {
2727         struct inet6_dev *idev;
2728
2729         ASSERT_RTNL();
2730
2731         /*
2732          * Configure the tunnel with one of our IPv4
2733          * addresses... we should configure all of
2734          * our v4 addrs in the tunnel
2735          */
2736
2737         if ((idev = ipv6_find_idev(dev)) == NULL) {
2738                 pr_debug("%s: add_dev failed\n", __func__);
2739                 return;
2740         }
2741
2742         if (dev->priv_flags & IFF_ISATAP) {
2743                 struct in6_addr addr;
2744
2745                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2746                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2747                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2748                         addrconf_add_linklocal(idev, &addr);
2749                 return;
2750         }
2751
2752         sit_add_v4_addrs(idev);
2753
2754         if (dev->flags&IFF_POINTOPOINT)
2755                 addrconf_add_mroute(dev);
2756         else
2757                 sit_route_add(dev);
2758 }
2759 #endif
2760
2761 #if IS_ENABLED(CONFIG_NET_IPGRE)
2762 static void addrconf_gre_config(struct net_device *dev)
2763 {
2764         struct inet6_dev *idev;
2765         struct in6_addr addr;
2766
2767         ASSERT_RTNL();
2768
2769         if ((idev = ipv6_find_idev(dev)) == NULL) {
2770                 pr_debug("%s: add_dev failed\n", __func__);
2771                 return;
2772         }
2773
2774         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2775         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2776
2777         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2778                 addrconf_add_linklocal(idev, &addr);
2779 }
2780 #endif
2781
2782 static inline int
2783 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2784 {
2785         struct in6_addr lladdr;
2786
2787         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2788                 addrconf_add_linklocal(idev, &lladdr);
2789                 return 0;
2790         }
2791         return -1;
2792 }
2793
2794 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2795 {
2796         struct net_device *link_dev;
2797         struct net *net = dev_net(idev->dev);
2798
2799         /* first try to inherit the link-local address from the link device */
2800         if (idev->dev->iflink &&
2801             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2802                 if (!ipv6_inherit_linklocal(idev, link_dev))
2803                         return;
2804         }
2805         /* then try to inherit it from any device */
2806         for_each_netdev(net, link_dev) {
2807                 if (!ipv6_inherit_linklocal(idev, link_dev))
2808                         return;
2809         }
2810         pr_debug("init ip6-ip6: add_linklocal failed\n");
2811 }
2812
2813 /*
2814  * Autoconfigure tunnel with a link-local address so routing protocols,
2815  * DHCPv6, MLD etc. can be run over the virtual link
2816  */
2817
2818 static void addrconf_ip6_tnl_config(struct net_device *dev)
2819 {
2820         struct inet6_dev *idev;
2821
2822         ASSERT_RTNL();
2823
2824         idev = addrconf_add_dev(dev);
2825         if (IS_ERR(idev)) {
2826                 pr_debug("init ip6-ip6: add_dev failed\n");
2827                 return;
2828         }
2829         ip6_tnl_add_linklocal(idev);
2830 }
2831
2832 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2833                            void *ptr)
2834 {
2835         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2836         struct inet6_dev *idev = __in6_dev_get(dev);
2837         int run_pending = 0;
2838         int err;
2839
2840         switch (event) {
2841         case NETDEV_REGISTER:
2842                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2843                         idev = ipv6_add_dev(dev);
2844                         if (!idev)
2845                                 return notifier_from_errno(-ENOMEM);
2846                 }
2847                 break;
2848
2849         case NETDEV_UP:
2850         case NETDEV_CHANGE:
2851                 if (dev->flags & IFF_SLAVE)
2852                         break;
2853
2854                 if (event == NETDEV_UP) {
2855                         if (!addrconf_qdisc_ok(dev)) {
2856                                 /* device is not ready yet. */
2857                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2858                                         dev->name);
2859                                 break;
2860                         }
2861
2862                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2863                                 idev = ipv6_add_dev(dev);
2864
2865                         if (idev) {
2866                                 idev->if_flags |= IF_READY;
2867                                 run_pending = 1;
2868                         }
2869                 } else {
2870                         if (!addrconf_qdisc_ok(dev)) {
2871                                 /* device is still not ready. */
2872                                 break;
2873                         }
2874
2875                         if (idev) {
2876                                 if (idev->if_flags & IF_READY)
2877                                         /* device is already configured. */
2878                                         break;
2879                                 idev->if_flags |= IF_READY;
2880                         }
2881
2882                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2883                                 dev->name);
2884
2885                         run_pending = 1;
2886                 }
2887
2888                 switch (dev->type) {
2889 #if IS_ENABLED(CONFIG_IPV6_SIT)
2890                 case ARPHRD_SIT:
2891                         addrconf_sit_config(dev);
2892                         break;
2893 #endif
2894 #if IS_ENABLED(CONFIG_NET_IPGRE)
2895                 case ARPHRD_IPGRE:
2896                         addrconf_gre_config(dev);
2897                         break;
2898 #endif
2899                 case ARPHRD_TUNNEL6:
2900                         addrconf_ip6_tnl_config(dev);
2901                         break;
2902                 case ARPHRD_LOOPBACK:
2903                         init_loopback(dev);
2904                         break;
2905
2906                 default:
2907                         addrconf_dev_config(dev);
2908                         break;
2909                 }
2910
2911                 if (idev) {
2912                         if (run_pending)
2913                                 addrconf_dad_run(idev);
2914
2915                         /*
2916                          * If the MTU changed during the interface down,
2917                          * when the interface up, the changed MTU must be
2918                          * reflected in the idev as well as routers.
2919                          */
2920                         if (idev->cnf.mtu6 != dev->mtu &&
2921                             dev->mtu >= IPV6_MIN_MTU) {
2922                                 rt6_mtu_change(dev, dev->mtu);
2923                                 idev->cnf.mtu6 = dev->mtu;
2924                         }
2925                         idev->tstamp = jiffies;
2926                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2927
2928                         /*
2929                          * If the changed mtu during down is lower than
2930                          * IPV6_MIN_MTU stop IPv6 on this interface.
2931                          */
2932                         if (dev->mtu < IPV6_MIN_MTU)
2933                                 addrconf_ifdown(dev, 1);
2934                 }
2935                 break;
2936
2937         case NETDEV_CHANGEMTU:
2938                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2939                         rt6_mtu_change(dev, dev->mtu);
2940                         idev->cnf.mtu6 = dev->mtu;
2941                         break;
2942                 }
2943
2944                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2945                         idev = ipv6_add_dev(dev);
2946                         if (idev)
2947                                 break;
2948                 }
2949
2950                 /*
2951                  * MTU falled under IPV6_MIN_MTU.
2952                  * Stop IPv6 on this interface.
2953                  */
2954
2955         case NETDEV_DOWN:
2956         case NETDEV_UNREGISTER:
2957                 /*
2958                  *      Remove all addresses from this interface.
2959                  */
2960                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2961                 break;
2962
2963         case NETDEV_CHANGENAME:
2964                 if (idev) {
2965                         snmp6_unregister_dev(idev);
2966                         addrconf_sysctl_unregister(idev);
2967                         addrconf_sysctl_register(idev);
2968                         err = snmp6_register_dev(idev);
2969                         if (err)
2970                                 return notifier_from_errno(err);
2971                 }
2972                 break;
2973
2974         case NETDEV_PRE_TYPE_CHANGE:
2975         case NETDEV_POST_TYPE_CHANGE:
2976                 addrconf_type_change(dev, event);
2977                 break;
2978         }
2979
2980         return NOTIFY_OK;
2981 }
2982
2983 /*
2984  *      addrconf module should be notified of a device going up
2985  */
2986 static struct notifier_block ipv6_dev_notf = {
2987         .notifier_call = addrconf_notify,
2988 };
2989
2990 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2991 {
2992         struct inet6_dev *idev;
2993         ASSERT_RTNL();
2994
2995         idev = __in6_dev_get(dev);
2996
2997         if (event == NETDEV_POST_TYPE_CHANGE)
2998                 ipv6_mc_remap(idev);
2999         else if (event == NETDEV_PRE_TYPE_CHANGE)
3000                 ipv6_mc_unmap(idev);
3001 }
3002
3003 static int addrconf_ifdown(struct net_device *dev, int how)
3004 {
3005         struct net *net = dev_net(dev);
3006         struct inet6_dev *idev;
3007         struct inet6_ifaddr *ifa;
3008         int state, i;
3009
3010         ASSERT_RTNL();
3011
3012         rt6_ifdown(net, dev);
3013         neigh_ifdown(&nd_tbl, dev);
3014
3015         idev = __in6_dev_get(dev);
3016         if (idev == NULL)
3017                 return -ENODEV;
3018
3019         /*
3020          * Step 1: remove reference to ipv6 device from parent device.
3021          *         Do not dev_put!
3022          */
3023         if (how) {
3024                 idev->dead = 1;
3025
3026                 /* protected by rtnl_lock */
3027                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3028
3029                 /* Step 1.5: remove snmp6 entry */
3030                 snmp6_unregister_dev(idev);
3031
3032         }
3033
3034         /* Step 2: clear hash table */
3035         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3036                 struct hlist_head *h = &inet6_addr_lst[i];
3037
3038                 spin_lock_bh(&addrconf_hash_lock);
3039         restart:
3040                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3041                         if (ifa->idev == idev) {
3042                                 hlist_del_init_rcu(&ifa->addr_lst);
3043                                 addrconf_del_dad_timer(ifa);
3044                                 goto restart;
3045                         }
3046                 }
3047                 spin_unlock_bh(&addrconf_hash_lock);
3048         }
3049
3050         write_lock_bh(&idev->lock);
3051
3052         addrconf_del_rs_timer(idev);
3053
3054         /* Step 2: clear flags for stateless addrconf */
3055         if (!how)
3056                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3057
3058 #ifdef CONFIG_IPV6_PRIVACY
3059         if (how && del_timer(&idev->regen_timer))
3060                 in6_dev_put(idev);
3061
3062         /* Step 3: clear tempaddr list */
3063         while (!list_empty(&idev->tempaddr_list)) {
3064                 ifa = list_first_entry(&idev->tempaddr_list,
3065                                        struct inet6_ifaddr, tmp_list);
3066                 list_del(&ifa->tmp_list);
3067                 write_unlock_bh(&idev->lock);
3068                 spin_lock_bh(&ifa->lock);
3069
3070                 if (ifa->ifpub) {
3071                         in6_ifa_put(ifa->ifpub);
3072                         ifa->ifpub = NULL;
3073                 }
3074                 spin_unlock_bh(&ifa->lock);
3075                 in6_ifa_put(ifa);
3076                 write_lock_bh(&idev->lock);
3077         }
3078 #endif
3079
3080         while (!list_empty(&idev->addr_list)) {
3081                 ifa = list_first_entry(&idev->addr_list,
3082                                        struct inet6_ifaddr, if_list);
3083                 addrconf_del_dad_timer(ifa);
3084
3085                 list_del(&ifa->if_list);
3086
3087                 write_unlock_bh(&idev->lock);
3088
3089                 spin_lock_bh(&ifa->state_lock);
3090                 state = ifa->state;
3091                 ifa->state = INET6_IFADDR_STATE_DEAD;
3092                 spin_unlock_bh(&ifa->state_lock);
3093
3094                 if (state != INET6_IFADDR_STATE_DEAD) {
3095                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3096                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3097                 }
3098                 in6_ifa_put(ifa);
3099
3100                 write_lock_bh(&idev->lock);
3101         }
3102
3103         write_unlock_bh(&idev->lock);
3104
3105         /* Step 5: Discard multicast list */
3106         if (how)
3107                 ipv6_mc_destroy_dev(idev);
3108         else
3109                 ipv6_mc_down(idev);
3110
3111         idev->tstamp = jiffies;
3112
3113         /* Last: Shot the device (if unregistered) */
3114         if (how) {
3115                 addrconf_sysctl_unregister(idev);
3116                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3117                 neigh_ifdown(&nd_tbl, dev);
3118                 in6_dev_put(idev);
3119         }
3120         return 0;
3121 }
3122
3123 static void addrconf_rs_timer(unsigned long data)
3124 {
3125         struct inet6_dev *idev = (struct inet6_dev *)data;
3126         struct in6_addr lladdr;
3127
3128         write_lock(&idev->lock);
3129         if (idev->dead || !(idev->if_flags & IF_READY))
3130                 goto out;
3131
3132         if (!ipv6_accept_ra(idev))
3133                 goto out;
3134
3135         /* Announcement received after solicitation was sent */
3136         if (idev->if_flags & IF_RA_RCVD)
3137                 goto out;
3138
3139         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3140                 if (!__ipv6_get_lladdr(idev, &lladdr, IFA_F_TENTATIVE))
3141                         ndisc_send_rs(idev->dev, &lladdr,
3142                                       &in6addr_linklocal_allrouters);
3143                 else
3144                         goto out;
3145
3146                 /* The wait after the last probe can be shorter */
3147                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3148                                              idev->cnf.rtr_solicits) ?
3149                                       idev->cnf.rtr_solicit_delay :
3150                                       idev->cnf.rtr_solicit_interval);
3151         } else {
3152                 /*
3153                  * Note: we do not support deprecated "all on-link"
3154                  * assumption any longer.
3155                  */
3156                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3157         }
3158
3159 out:
3160         write_unlock(&idev->lock);
3161         in6_dev_put(idev);
3162 }
3163
3164 /*
3165  *      Duplicate Address Detection
3166  */
3167 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3168 {
3169         unsigned long rand_num;
3170         struct inet6_dev *idev = ifp->idev;
3171
3172         if (ifp->flags & IFA_F_OPTIMISTIC)
3173                 rand_num = 0;
3174         else
3175                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3176
3177         ifp->dad_probes = idev->cnf.dad_transmits;
3178         addrconf_mod_dad_timer(ifp, rand_num);
3179 }
3180
3181 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3182 {
3183         struct inet6_dev *idev = ifp->idev;
3184         struct net_device *dev = idev->dev;
3185
3186         addrconf_join_solict(dev, &ifp->addr);
3187
3188         net_srandom(ifp->addr.s6_addr32[3]);
3189
3190         read_lock_bh(&idev->lock);
3191         spin_lock(&ifp->lock);
3192         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3193                 goto out;
3194
3195         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3196             idev->cnf.accept_dad < 1 ||
3197             !(ifp->flags&IFA_F_TENTATIVE) ||
3198             ifp->flags & IFA_F_NODAD) {
3199                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3200                 spin_unlock(&ifp->lock);
3201                 read_unlock_bh(&idev->lock);
3202
3203                 addrconf_dad_completed(ifp);
3204                 return;
3205         }
3206
3207         if (!(idev->if_flags & IF_READY)) {
3208                 spin_unlock(&ifp->lock);
3209                 read_unlock_bh(&idev->lock);
3210                 /*
3211                  * If the device is not ready:
3212                  * - keep it tentative if it is a permanent address.
3213                  * - otherwise, kill it.
3214                  */
3215                 in6_ifa_hold(ifp);
3216                 addrconf_dad_stop(ifp, 0);
3217                 return;
3218         }
3219
3220         /*
3221          * Optimistic nodes can start receiving
3222          * Frames right away
3223          */
3224         if (ifp->flags & IFA_F_OPTIMISTIC)
3225                 ip6_ins_rt(ifp->rt);
3226
3227         addrconf_dad_kick(ifp);
3228 out:
3229         spin_unlock(&ifp->lock);
3230         read_unlock_bh(&idev->lock);
3231 }
3232
3233 static void addrconf_dad_timer(unsigned long data)
3234 {
3235         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3236         struct inet6_dev *idev = ifp->idev;
3237         struct in6_addr mcaddr;
3238
3239         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3240                 goto out;
3241
3242         write_lock(&idev->lock);
3243         if (idev->dead || !(idev->if_flags & IF_READY)) {
3244                 write_unlock(&idev->lock);
3245                 goto out;
3246         }
3247
3248         spin_lock(&ifp->lock);
3249         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3250                 spin_unlock(&ifp->lock);
3251                 write_unlock(&idev->lock);
3252                 goto out;
3253         }
3254
3255         if (ifp->dad_probes == 0) {
3256                 /*
3257                  * DAD was successful
3258                  */
3259
3260                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3261                 spin_unlock(&ifp->lock);
3262                 write_unlock(&idev->lock);
3263
3264                 addrconf_dad_completed(ifp);
3265
3266                 goto out;
3267         }
3268
3269         ifp->dad_probes--;
3270         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3271         spin_unlock(&ifp->lock);
3272         write_unlock(&idev->lock);
3273
3274         /* send a neighbour solicitation for our addr */
3275         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3276         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3277 out:
3278         in6_ifa_put(ifp);
3279 }
3280
3281 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3282 {
3283         struct net_device *dev = ifp->idev->dev;
3284         struct in6_addr lladdr;
3285         bool send_rs, send_mld;
3286
3287         addrconf_del_dad_timer(ifp);
3288
3289         /*
3290          *      Configure the address for reception. Now it is valid.
3291          */
3292
3293         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3294
3295         /* If added prefix is link local and we are prepared to process
3296            router advertisements, start sending router solicitations.
3297          */
3298
3299         read_lock_bh(&ifp->idev->lock);
3300         spin_lock(&ifp->lock);
3301         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3302                    ifp->idev->valid_ll_addr_cnt == 1;
3303         send_rs = send_mld &&
3304                   ipv6_accept_ra(ifp->idev) &&
3305                   ifp->idev->cnf.rtr_solicits > 0 &&
3306                   (dev->flags&IFF_LOOPBACK) == 0;
3307         spin_unlock(&ifp->lock);
3308         read_unlock_bh(&ifp->idev->lock);
3309
3310         /* While dad is in progress mld report's source address is in6_addrany.
3311          * Resend with proper ll now.
3312          */
3313         if (send_mld)
3314                 ipv6_mc_dad_complete(ifp->idev);
3315
3316         if (send_rs) {
3317                 /*
3318                  *      If a host as already performed a random delay
3319                  *      [...] as part of DAD [...] there is no need
3320                  *      to delay again before sending the first RS
3321                  */
3322                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3323                         return;
3324                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3325
3326                 write_lock_bh(&ifp->idev->lock);
3327                 spin_lock(&ifp->lock);
3328                 ifp->idev->rs_probes = 1;
3329                 ifp->idev->if_flags |= IF_RS_SENT;
3330                 addrconf_mod_rs_timer(ifp->idev,
3331                                       ifp->idev->cnf.rtr_solicit_interval);
3332                 spin_unlock(&ifp->lock);
3333                 write_unlock_bh(&ifp->idev->lock);
3334         }
3335 }
3336
3337 static void addrconf_dad_run(struct inet6_dev *idev)
3338 {
3339         struct inet6_ifaddr *ifp;
3340
3341         read_lock_bh(&idev->lock);
3342         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3343                 spin_lock(&ifp->lock);
3344                 if (ifp->flags & IFA_F_TENTATIVE &&
3345                     ifp->state == INET6_IFADDR_STATE_DAD)
3346                         addrconf_dad_kick(ifp);
3347                 spin_unlock(&ifp->lock);
3348         }
3349         read_unlock_bh(&idev->lock);
3350 }
3351
3352 #ifdef CONFIG_PROC_FS
3353 struct if6_iter_state {
3354         struct seq_net_private p;
3355         int bucket;
3356         int offset;
3357 };
3358
3359 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3360 {
3361         struct inet6_ifaddr *ifa = NULL;
3362         struct if6_iter_state *state = seq->private;
3363         struct net *net = seq_file_net(seq);
3364         int p = 0;
3365
3366         /* initial bucket if pos is 0 */
3367         if (pos == 0) {
3368                 state->bucket = 0;
3369                 state->offset = 0;
3370         }
3371
3372         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3373                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3374                                          addr_lst) {
3375                         if (!net_eq(dev_net(ifa->idev->dev), net))
3376                                 continue;
3377                         /* sync with offset */
3378                         if (p < state->offset) {
3379                                 p++;
3380                                 continue;
3381                         }
3382                         state->offset++;
3383                         return ifa;
3384                 }
3385
3386                 /* prepare for next bucket */
3387                 state->offset = 0;
3388                 p = 0;
3389         }
3390         return NULL;
3391 }
3392
3393 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3394                                          struct inet6_ifaddr *ifa)
3395 {
3396         struct if6_iter_state *state = seq->private;
3397         struct net *net = seq_file_net(seq);
3398
3399         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3400                 if (!net_eq(dev_net(ifa->idev->dev), net))
3401                         continue;
3402                 state->offset++;
3403                 return ifa;
3404         }
3405
3406         while (++state->bucket < IN6_ADDR_HSIZE) {
3407                 state->offset = 0;
3408                 hlist_for_each_entry_rcu_bh(ifa,
3409                                      &inet6_addr_lst[state->bucket], addr_lst) {
3410                         if (!net_eq(dev_net(ifa->idev->dev), net))
3411                                 continue;
3412                         state->offset++;
3413                         return ifa;
3414                 }
3415         }
3416
3417         return NULL;
3418 }
3419
3420 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3421         __acquires(rcu_bh)
3422 {
3423         rcu_read_lock_bh();
3424         return if6_get_first(seq, *pos);
3425 }
3426
3427 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3428 {
3429         struct inet6_ifaddr *ifa;
3430
3431         ifa = if6_get_next(seq, v);
3432         ++*pos;
3433         return ifa;
3434 }
3435
3436 static void if6_seq_stop(struct seq_file *seq, void *v)
3437         __releases(rcu_bh)
3438 {
3439         rcu_read_unlock_bh();
3440 }
3441
3442 static int if6_seq_show(struct seq_file *seq, void *v)
3443 {
3444         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3445         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3446                    &ifp->addr,
3447                    ifp->idev->dev->ifindex,
3448                    ifp->prefix_len,
3449                    ifp->scope,
3450                    ifp->flags,
3451                    ifp->idev->dev->name);
3452         return 0;
3453 }
3454
3455 static const struct seq_operations if6_seq_ops = {
3456         .start  = if6_seq_start,
3457         .next   = if6_seq_next,
3458         .show   = if6_seq_show,
3459         .stop   = if6_seq_stop,
3460 };
3461
3462 static int if6_seq_open(struct inode *inode, struct file *file)
3463 {
3464         return seq_open_net(inode, file, &if6_seq_ops,
3465                             sizeof(struct if6_iter_state));
3466 }
3467
3468 static const struct file_operations if6_fops = {
3469         .owner          = THIS_MODULE,
3470         .open           = if6_seq_open,
3471         .read           = seq_read,
3472         .llseek         = seq_lseek,
3473         .release        = seq_release_net,
3474 };
3475
3476 static int __net_init if6_proc_net_init(struct net *net)
3477 {
3478         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3479                 return -ENOMEM;
3480         return 0;
3481 }
3482
3483 static void __net_exit if6_proc_net_exit(struct net *net)
3484 {
3485         remove_proc_entry("if_inet6", net->proc_net);
3486 }
3487
3488 static struct pernet_operations if6_proc_net_ops = {
3489        .init = if6_proc_net_init,
3490        .exit = if6_proc_net_exit,
3491 };
3492
3493 int __init if6_proc_init(void)
3494 {
3495         return register_pernet_subsys(&if6_proc_net_ops);
3496 }
3497
3498 void if6_proc_exit(void)
3499 {
3500         unregister_pernet_subsys(&if6_proc_net_ops);
3501 }
3502 #endif  /* CONFIG_PROC_FS */
3503
3504 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3505 /* Check if address is a home address configured on any interface. */
3506 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3507 {
3508         int ret = 0;
3509         struct inet6_ifaddr *ifp = NULL;
3510         unsigned int hash = inet6_addr_hash(addr);
3511
3512         rcu_read_lock_bh();
3513         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3514                 if (!net_eq(dev_net(ifp->idev->dev), net))
3515                         continue;
3516                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3517                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3518                         ret = 1;
3519                         break;
3520                 }
3521         }
3522         rcu_read_unlock_bh();
3523         return ret;
3524 }
3525 #endif
3526
3527 /*
3528  *      Periodic address status verification
3529  */
3530
3531 static void addrconf_verify(unsigned long foo)
3532 {
3533         unsigned long now, next, next_sec, next_sched;
3534         struct inet6_ifaddr *ifp;
3535         int i;
3536
3537         rcu_read_lock_bh();
3538         spin_lock(&addrconf_verify_lock);
3539         now = jiffies;
3540         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3541
3542         del_timer(&addr_chk_timer);
3543
3544         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3545 restart:
3546                 hlist_for_each_entry_rcu_bh(ifp,
3547                                          &inet6_addr_lst[i], addr_lst) {
3548                         unsigned long age;
3549
3550                         if (ifp->flags & IFA_F_PERMANENT)
3551                                 continue;
3552
3553                         spin_lock(&ifp->lock);
3554                         /* We try to batch several events at once. */
3555                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3556
3557                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3558                             age >= ifp->valid_lft) {
3559                                 spin_unlock(&ifp->lock);
3560                                 in6_ifa_hold(ifp);
3561                                 ipv6_del_addr(ifp);
3562                                 goto restart;
3563                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3564                                 spin_unlock(&ifp->lock);
3565                                 continue;
3566                         } else if (age >= ifp->prefered_lft) {
3567                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3568                                 int deprecate = 0;
3569
3570                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3571                                         deprecate = 1;
3572                                         ifp->flags |= IFA_F_DEPRECATED;
3573                                 }
3574
3575                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3576                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3577
3578                                 spin_unlock(&ifp->lock);
3579
3580                                 if (deprecate) {
3581                                         in6_ifa_hold(ifp);
3582
3583                                         ipv6_ifa_notify(0, ifp);
3584                                         in6_ifa_put(ifp);
3585                                         goto restart;
3586                                 }
3587 #ifdef CONFIG_IPV6_PRIVACY
3588                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3589                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3590                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3591                                         ifp->idev->cnf.dad_transmits *
3592                                         ifp->idev->nd_parms->retrans_time / HZ;
3593
3594                                 if (age >= ifp->prefered_lft - regen_advance) {
3595                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3596                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3597                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3598                                         if (!ifp->regen_count && ifpub) {
3599                                                 ifp->regen_count++;
3600                                                 in6_ifa_hold(ifp);
3601                                                 in6_ifa_hold(ifpub);
3602                                                 spin_unlock(&ifp->lock);
3603
3604                                                 spin_lock(&ifpub->lock);
3605                                                 ifpub->regen_count = 0;
3606                                                 spin_unlock(&ifpub->lock);
3607                                                 ipv6_create_tempaddr(ifpub, ifp);
3608                                                 in6_ifa_put(ifpub);
3609                                                 in6_ifa_put(ifp);
3610                                                 goto restart;
3611                                         }
3612                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3613                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3614                                 spin_unlock(&ifp->lock);
3615 #endif
3616                         } else {
3617                                 /* ifp->prefered_lft <= ifp->valid_lft */
3618                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3619                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3620                                 spin_unlock(&ifp->lock);
3621                         }
3622                 }
3623         }
3624
3625         next_sec = round_jiffies_up(next);
3626         next_sched = next;
3627
3628         /* If rounded timeout is accurate enough, accept it. */
3629         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3630                 next_sched = next_sec;
3631
3632         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3633         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3634                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3635
3636         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3637               now, next, next_sec, next_sched));
3638
3639         addr_chk_timer.expires = next_sched;
3640         add_timer(&addr_chk_timer);
3641         spin_unlock(&addrconf_verify_lock);
3642         rcu_read_unlock_bh();
3643 }
3644
3645 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3646                                      struct in6_addr **peer_pfx)
3647 {
3648         struct in6_addr *pfx = NULL;
3649
3650         *peer_pfx = NULL;
3651
3652         if (addr)
3653                 pfx = nla_data(addr);
3654
3655         if (local) {
3656                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3657                         *peer_pfx = pfx;
3658                 pfx = nla_data(local);
3659         }
3660
3661         return pfx;
3662 }
3663
3664 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3665         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3666         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3667         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3668 };
3669
3670 static int
3671 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3672 {
3673         struct net *net = sock_net(skb->sk);
3674         struct ifaddrmsg *ifm;
3675         struct nlattr *tb[IFA_MAX+1];
3676         struct in6_addr *pfx, *peer_pfx;
3677         int err;
3678
3679         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3680         if (err < 0)
3681                 return err;
3682
3683         ifm = nlmsg_data(nlh);
3684         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3685         if (pfx == NULL)
3686                 return -EINVAL;
3687
3688         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3689 }
3690
3691 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3692                              u32 prefered_lft, u32 valid_lft)
3693 {
3694         u32 flags;
3695         clock_t expires;
3696         unsigned long timeout;
3697
3698         if (!valid_lft || (prefered_lft > valid_lft))
3699                 return -EINVAL;
3700
3701         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3702         if (addrconf_finite_timeout(timeout)) {
3703                 expires = jiffies_to_clock_t(timeout * HZ);
3704                 valid_lft = timeout;
3705                 flags = RTF_EXPIRES;
3706         } else {
3707                 expires = 0;
3708                 flags = 0;
3709                 ifa_flags |= IFA_F_PERMANENT;
3710         }
3711
3712         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3713         if (addrconf_finite_timeout(timeout)) {
3714                 if (timeout == 0)
3715                         ifa_flags |= IFA_F_DEPRECATED;
3716                 prefered_lft = timeout;
3717         }
3718
3719         spin_lock_bh(&ifp->lock);
3720         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3721         ifp->tstamp = jiffies;
3722         ifp->valid_lft = valid_lft;
3723         ifp->prefered_lft = prefered_lft;
3724
3725         spin_unlock_bh(&ifp->lock);
3726         if (!(ifp->flags&IFA_F_TENTATIVE))
3727                 ipv6_ifa_notify(0, ifp);
3728
3729         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3730                               expires, flags);
3731         addrconf_verify(0);
3732
3733         return 0;
3734 }
3735
3736 static int
3737 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3738 {
3739         struct net *net = sock_net(skb->sk);
3740         struct ifaddrmsg *ifm;
3741         struct nlattr *tb[IFA_MAX+1];
3742         struct in6_addr *pfx, *peer_pfx;
3743         struct inet6_ifaddr *ifa;
3744         struct net_device *dev;
3745         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3746         u8 ifa_flags;
3747         int err;
3748
3749         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3750         if (err < 0)
3751                 return err;
3752
3753         ifm = nlmsg_data(nlh);
3754         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3755         if (pfx == NULL)
3756                 return -EINVAL;
3757
3758         if (tb[IFA_CACHEINFO]) {
3759                 struct ifa_cacheinfo *ci;
3760
3761                 ci = nla_data(tb[IFA_CACHEINFO]);
3762                 valid_lft = ci->ifa_valid;
3763                 preferred_lft = ci->ifa_prefered;
3764         } else {
3765                 preferred_lft = INFINITY_LIFE_TIME;
3766                 valid_lft = INFINITY_LIFE_TIME;
3767         }
3768
3769         dev =  __dev_get_by_index(net, ifm->ifa_index);
3770         if (dev == NULL)
3771                 return -ENODEV;
3772
3773         /* We ignore other flags so far. */
3774         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3775
3776         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3777         if (ifa == NULL) {
3778                 /*
3779                  * It would be best to check for !NLM_F_CREATE here but
3780                  * userspace alreay relies on not having to provide this.
3781                  */
3782                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3783                                       ifm->ifa_prefixlen, ifa_flags,
3784                                       preferred_lft, valid_lft);
3785         }
3786
3787         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3788             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3789                 err = -EEXIST;
3790         else
3791                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3792
3793         in6_ifa_put(ifa);
3794
3795         return err;
3796 }
3797
3798 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3799                           u8 scope, int ifindex)
3800 {
3801         struct ifaddrmsg *ifm;
3802
3803         ifm = nlmsg_data(nlh);
3804         ifm->ifa_family = AF_INET6;
3805         ifm->ifa_prefixlen = prefixlen;
3806         ifm->ifa_flags = flags;
3807         ifm->ifa_scope = scope;
3808         ifm->ifa_index = ifindex;
3809 }
3810
3811 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3812                          unsigned long tstamp, u32 preferred, u32 valid)
3813 {
3814         struct ifa_cacheinfo ci;
3815
3816         ci.cstamp = cstamp_delta(cstamp);
3817         ci.tstamp = cstamp_delta(tstamp);
3818         ci.ifa_prefered = preferred;
3819         ci.ifa_valid = valid;
3820
3821         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3822 }
3823
3824 static inline int rt_scope(int ifa_scope)
3825 {
3826         if (ifa_scope & IFA_HOST)
3827                 return RT_SCOPE_HOST;
3828         else if (ifa_scope & IFA_LINK)
3829                 return RT_SCOPE_LINK;
3830         else if (ifa_scope & IFA_SITE)
3831                 return RT_SCOPE_SITE;
3832         else
3833                 return RT_SCOPE_UNIVERSE;
3834 }
3835
3836 static inline int inet6_ifaddr_msgsize(void)
3837 {
3838         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3839                + nla_total_size(16) /* IFA_LOCAL */
3840                + nla_total_size(16) /* IFA_ADDRESS */
3841                + nla_total_size(sizeof(struct ifa_cacheinfo));
3842 }
3843
3844 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3845                              u32 portid, u32 seq, int event, unsigned int flags)
3846 {
3847         struct nlmsghdr  *nlh;
3848         u32 preferred, valid;
3849
3850         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3851         if (nlh == NULL)
3852                 return -EMSGSIZE;
3853
3854         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3855                       ifa->idev->dev->ifindex);
3856
3857         if (!(ifa->flags&IFA_F_PERMANENT)) {
3858                 preferred = ifa->prefered_lft;
3859                 valid = ifa->valid_lft;
3860                 if (preferred != INFINITY_LIFE_TIME) {
3861                         long tval = (jiffies - ifa->tstamp)/HZ;
3862                         if (preferred > tval)
3863                                 preferred -= tval;
3864                         else
3865                                 preferred = 0;
3866                         if (valid != INFINITY_LIFE_TIME) {
3867                                 if (valid > tval)
3868                                         valid -= tval;
3869                                 else
3870                                         valid = 0;
3871                         }
3872                 }
3873         } else {
3874                 preferred = INFINITY_LIFE_TIME;
3875                 valid = INFINITY_LIFE_TIME;
3876         }
3877
3878         if (!ipv6_addr_any(&ifa->peer_addr)) {
3879                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3880                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3881                         goto error;
3882         } else
3883                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3884                         goto error;
3885
3886         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3887                 goto error;
3888
3889         return nlmsg_end(skb, nlh);
3890
3891 error:
3892         nlmsg_cancel(skb, nlh);
3893         return -EMSGSIZE;
3894 }
3895
3896 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3897                                 u32 portid, u32 seq, int event, u16 flags)
3898 {
3899         struct nlmsghdr  *nlh;
3900         u8 scope = RT_SCOPE_UNIVERSE;
3901         int ifindex = ifmca->idev->dev->ifindex;
3902
3903         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3904                 scope = RT_SCOPE_SITE;
3905
3906         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3907         if (nlh == NULL)
3908                 return -EMSGSIZE;
3909
3910         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3911         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3912             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3913                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3914                 nlmsg_cancel(skb, nlh);
3915                 return -EMSGSIZE;
3916         }
3917
3918         return nlmsg_end(skb, nlh);
3919 }
3920
3921 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3922                                 u32 portid, u32 seq, int event, unsigned int flags)
3923 {
3924         struct nlmsghdr  *nlh;
3925         u8 scope = RT_SCOPE_UNIVERSE;
3926         int ifindex = ifaca->aca_idev->dev->ifindex;
3927
3928         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3929                 scope = RT_SCOPE_SITE;
3930
3931         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3932         if (nlh == NULL)
3933                 return -EMSGSIZE;
3934
3935         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3936         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3937             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3938                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3939                 nlmsg_cancel(skb, nlh);
3940                 return -EMSGSIZE;
3941         }
3942
3943         return nlmsg_end(skb, nlh);
3944 }
3945
3946 enum addr_type_t {
3947         UNICAST_ADDR,
3948         MULTICAST_ADDR,
3949         ANYCAST_ADDR,
3950 };
3951
3952 /* called with rcu_read_lock() */
3953 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3954                           struct netlink_callback *cb, enum addr_type_t type,
3955                           int s_ip_idx, int *p_ip_idx)
3956 {
3957         struct ifmcaddr6 *ifmca;
3958         struct ifacaddr6 *ifaca;
3959         int err = 1;
3960         int ip_idx = *p_ip_idx;
3961
3962         read_lock_bh(&idev->lock);
3963         switch (type) {
3964         case UNICAST_ADDR: {
3965                 struct inet6_ifaddr *ifa;
3966
3967                 /* unicast address incl. temp addr */
3968                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3969                         if (++ip_idx < s_ip_idx)
3970                                 continue;
3971                         err = inet6_fill_ifaddr(skb, ifa,
3972                                                 NETLINK_CB(cb->skb).portid,
3973                                                 cb->nlh->nlmsg_seq,
3974                                                 RTM_NEWADDR,
3975                                                 NLM_F_MULTI);
3976                         if (err <= 0)
3977                                 break;
3978                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3979                 }
3980                 break;
3981         }
3982         case MULTICAST_ADDR:
3983                 /* multicast address */
3984                 for (ifmca = idev->mc_list; ifmca;
3985                      ifmca = ifmca->next, ip_idx++) {
3986                         if (ip_idx < s_ip_idx)
3987                                 continue;
3988                         err = inet6_fill_ifmcaddr(skb, ifmca,
3989                                                   NETLINK_CB(cb->skb).portid,
3990                                                   cb->nlh->nlmsg_seq,
3991                                                   RTM_GETMULTICAST,
3992                                                   NLM_F_MULTI);
3993                         if (err <= 0)
3994                                 break;
3995                 }
3996                 break;
3997         case ANYCAST_ADDR:
3998                 /* anycast address */
3999                 for (ifaca = idev->ac_list; ifaca;
4000                      ifaca = ifaca->aca_next, ip_idx++) {
4001                         if (ip_idx < s_ip_idx)
4002                                 continue;
4003                         err = inet6_fill_ifacaddr(skb, ifaca,
4004                                                   NETLINK_CB(cb->skb).portid,
4005                                                   cb->nlh->nlmsg_seq,
4006                                                   RTM_GETANYCAST,
4007                                                   NLM_F_MULTI);
4008                         if (err <= 0)
4009                                 break;
4010                 }
4011                 break;
4012         default:
4013                 break;
4014         }
4015         read_unlock_bh(&idev->lock);
4016         *p_ip_idx = ip_idx;
4017         return err;
4018 }
4019
4020 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4021                            enum addr_type_t type)
4022 {
4023         struct net *net = sock_net(skb->sk);
4024         int h, s_h;
4025         int idx, ip_idx;
4026         int s_idx, s_ip_idx;
4027         struct net_device *dev;
4028         struct inet6_dev *idev;
4029         struct hlist_head *head;
4030
4031         s_h = cb->args[0];
4032         s_idx = idx = cb->args[1];
4033         s_ip_idx = ip_idx = cb->args[2];
4034
4035         rcu_read_lock();
4036         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4037         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4038                 idx = 0;
4039                 head = &net->dev_index_head[h];
4040                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4041                         if (idx < s_idx)
4042                                 goto cont;
4043                         if (h > s_h || idx > s_idx)
4044                                 s_ip_idx = 0;
4045                         ip_idx = 0;
4046                         idev = __in6_dev_get(dev);
4047                         if (!idev)
4048                                 goto cont;
4049
4050                         if (in6_dump_addrs(idev, skb, cb, type,
4051                                            s_ip_idx, &ip_idx) <= 0)
4052                                 goto done;
4053 cont:
4054                         idx++;
4055                 }
4056         }
4057 done:
4058         rcu_read_unlock();
4059         cb->args[0] = h;
4060         cb->args[1] = idx;
4061         cb->args[2] = ip_idx;
4062
4063         return skb->len;
4064 }
4065
4066 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4067 {
4068         enum addr_type_t type = UNICAST_ADDR;
4069
4070         return inet6_dump_addr(skb, cb, type);
4071 }
4072
4073 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4074 {
4075         enum addr_type_t type = MULTICAST_ADDR;
4076
4077         return inet6_dump_addr(skb, cb, type);
4078 }
4079
4080
4081 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4082 {
4083         enum addr_type_t type = ANYCAST_ADDR;
4084
4085         return inet6_dump_addr(skb, cb, type);
4086 }
4087
4088 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4089 {
4090         struct net *net = sock_net(in_skb->sk);
4091         struct ifaddrmsg *ifm;
4092         struct nlattr *tb[IFA_MAX+1];
4093         struct in6_addr *addr = NULL, *peer;
4094         struct net_device *dev = NULL;
4095         struct inet6_ifaddr *ifa;
4096         struct sk_buff *skb;
4097         int err;
4098
4099         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4100         if (err < 0)
4101                 goto errout;
4102
4103         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4104         if (addr == NULL) {
4105                 err = -EINVAL;
4106                 goto errout;
4107         }
4108
4109         ifm = nlmsg_data(nlh);
4110         if (ifm->ifa_index)
4111                 dev = __dev_get_by_index(net, ifm->ifa_index);
4112
4113         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4114         if (!ifa) {
4115                 err = -EADDRNOTAVAIL;
4116                 goto errout;
4117         }
4118
4119         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4120         if (!skb) {
4121                 err = -ENOBUFS;
4122                 goto errout_ifa;
4123         }
4124
4125         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4126                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4127         if (err < 0) {
4128                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4129                 WARN_ON(err == -EMSGSIZE);
4130                 kfree_skb(skb);
4131                 goto errout_ifa;
4132         }
4133         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4134 errout_ifa:
4135         in6_ifa_put(ifa);
4136 errout:
4137         return err;
4138 }
4139
4140 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4141 {
4142         struct sk_buff *skb;
4143         struct net *net = dev_net(ifa->idev->dev);
4144         int err = -ENOBUFS;
4145
4146         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4147         if (skb == NULL)
4148                 goto errout;
4149
4150         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4151         if (err < 0) {
4152                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4153                 WARN_ON(err == -EMSGSIZE);
4154                 kfree_skb(skb);
4155                 goto errout;
4156         }
4157         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4158         return;
4159 errout:
4160         if (err < 0)
4161                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4162 }
4163
4164 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4165                                 __s32 *array, int bytes)
4166 {
4167         BUG_ON(bytes < (DEVCONF_MAX * 4));
4168
4169         memset(array, 0, bytes);
4170         array[DEVCONF_FORWARDING] = cnf->forwarding;
4171         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4172         array[DEVCONF_MTU6] = cnf->mtu6;
4173         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4174         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4175         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4176         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4177         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4178         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4179                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4180         array[DEVCONF_RTR_SOLICIT_DELAY] =
4181                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4182         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4183 #ifdef CONFIG_IPV6_PRIVACY
4184         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4185         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4186         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4187         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4188         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4189 #endif
4190         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4191         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4192         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4193 #ifdef CONFIG_IPV6_ROUTER_PREF
4194         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4195         array[DEVCONF_RTR_PROBE_INTERVAL] =
4196                 jiffies_to_msecs(cnf->rtr_probe_interval);
4197 #ifdef CONFIG_IPV6_ROUTE_INFO
4198         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4199 #endif
4200 #endif
4201         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4202         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4203 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4204         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4205 #endif
4206 #ifdef CONFIG_IPV6_MROUTE
4207         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4208 #endif
4209         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4210         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4211         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4212         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4213 }
4214
4215 static inline size_t inet6_ifla6_size(void)
4216 {
4217         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4218              + nla_total_size(sizeof(struct ifla_cacheinfo))
4219              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4220              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4221              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4222              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4223 }
4224
4225 static inline size_t inet6_if_nlmsg_size(void)
4226 {
4227         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4228                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4229                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4230                + nla_total_size(4) /* IFLA_MTU */
4231                + nla_total_size(4) /* IFLA_LINK */
4232                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4233 }
4234
4235 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4236                                       int items, int bytes)
4237 {
4238         int i;
4239         int pad = bytes - sizeof(u64) * items;
4240         BUG_ON(pad < 0);
4241
4242         /* Use put_unaligned() because stats may not be aligned for u64. */
4243         put_unaligned(items, &stats[0]);
4244         for (i = 1; i < items; i++)
4245                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4246
4247         memset(&stats[items], 0, pad);
4248 }
4249
4250 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4251                                       int items, int bytes, size_t syncpoff)
4252 {
4253         int i;
4254         int pad = bytes - sizeof(u64) * items;
4255         BUG_ON(pad < 0);
4256
4257         /* Use put_unaligned() because stats may not be aligned for u64. */
4258         put_unaligned(items, &stats[0]);
4259         for (i = 1; i < items; i++)
4260                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4261
4262         memset(&stats[items], 0, pad);
4263 }
4264
4265 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4266                              int bytes)
4267 {
4268         switch (attrtype) {
4269         case IFLA_INET6_STATS:
4270                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4271                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4272                 break;
4273         case IFLA_INET6_ICMP6STATS:
4274                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4275                 break;
4276         }
4277 }
4278
4279 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4280 {
4281         struct nlattr *nla;
4282         struct ifla_cacheinfo ci;
4283
4284         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4285                 goto nla_put_failure;
4286         ci.max_reasm_len = IPV6_MAXPLEN;
4287         ci.tstamp = cstamp_delta(idev->tstamp);
4288         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4289         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4290         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4291                 goto nla_put_failure;
4292         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4293         if (nla == NULL)
4294                 goto nla_put_failure;
4295         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4296
4297         /* XXX - MC not implemented */
4298
4299         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4300         if (nla == NULL)
4301                 goto nla_put_failure;
4302         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4303
4304         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4305         if (nla == NULL)
4306                 goto nla_put_failure;
4307         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4308
4309         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4310         if (nla == NULL)
4311                 goto nla_put_failure;
4312         read_lock_bh(&idev->lock);
4313         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4314         read_unlock_bh(&idev->lock);
4315
4316         return 0;
4317
4318 nla_put_failure:
4319         return -EMSGSIZE;
4320 }
4321
4322 static size_t inet6_get_link_af_size(const struct net_device *dev)
4323 {
4324         if (!__in6_dev_get(dev))
4325                 return 0;
4326
4327         return inet6_ifla6_size();
4328 }
4329
4330 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4331 {
4332         struct inet6_dev *idev = __in6_dev_get(dev);
4333
4334         if (!idev)
4335                 return -ENODATA;
4336
4337         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4338                 return -EMSGSIZE;
4339
4340         return 0;
4341 }
4342
4343 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4344 {
4345         struct inet6_ifaddr *ifp;
4346         struct net_device *dev = idev->dev;
4347         bool update_rs = false;
4348         struct in6_addr ll_addr;
4349
4350         if (token == NULL)
4351                 return -EINVAL;
4352         if (ipv6_addr_any(token))
4353                 return -EINVAL;
4354         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4355                 return -EINVAL;
4356         if (!ipv6_accept_ra(idev))
4357                 return -EINVAL;
4358         if (idev->cnf.rtr_solicits <= 0)
4359                 return -EINVAL;
4360
4361         write_lock_bh(&idev->lock);
4362
4363         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4364         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4365
4366         write_unlock_bh(&idev->lock);
4367
4368         if (!idev->dead && (idev->if_flags & IF_READY) &&
4369             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4370                              IFA_F_OPTIMISTIC)) {
4371
4372                 /* If we're not ready, then normal ifup will take care
4373                  * of this. Otherwise, we need to request our rs here.
4374                  */
4375                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4376                 update_rs = true;
4377         }
4378
4379         write_lock_bh(&idev->lock);
4380
4381         if (update_rs) {
4382                 idev->if_flags |= IF_RS_SENT;
4383                 idev->rs_probes = 1;
4384                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4385         }
4386
4387         /* Well, that's kinda nasty ... */
4388         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4389                 spin_lock(&ifp->lock);
4390                 if (ifp->tokenized) {
4391                         ifp->valid_lft = 0;
4392                         ifp->prefered_lft = 0;
4393                 }
4394                 spin_unlock(&ifp->lock);
4395         }
4396
4397         write_unlock_bh(&idev->lock);
4398         addrconf_verify(0);
4399         return 0;
4400 }
4401
4402 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4403 {
4404         int err = -EINVAL;
4405         struct inet6_dev *idev = __in6_dev_get(dev);
4406         struct nlattr *tb[IFLA_INET6_MAX + 1];
4407
4408         if (!idev)
4409                 return -EAFNOSUPPORT;
4410
4411         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4412                 BUG();
4413
4414         if (tb[IFLA_INET6_TOKEN])
4415                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4416
4417         return err;
4418 }
4419
4420 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4421                              u32 portid, u32 seq, int event, unsigned int flags)
4422 {
4423         struct net_device *dev = idev->dev;
4424         struct ifinfomsg *hdr;
4425         struct nlmsghdr *nlh;
4426         void *protoinfo;
4427
4428         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4429         if (nlh == NULL)
4430                 return -EMSGSIZE;
4431
4432         hdr = nlmsg_data(nlh);
4433         hdr->ifi_family = AF_INET6;
4434         hdr->__ifi_pad = 0;
4435         hdr->ifi_type = dev->type;
4436         hdr->ifi_index = dev->ifindex;
4437         hdr->ifi_flags = dev_get_flags(dev);
4438         hdr->ifi_change = 0;
4439
4440         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4441             (dev->addr_len &&
4442              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4443             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4444             (dev->ifindex != dev->iflink &&
4445              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4446                 goto nla_put_failure;
4447         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4448         if (protoinfo == NULL)
4449                 goto nla_put_failure;
4450
4451         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4452                 goto nla_put_failure;
4453
4454         nla_nest_end(skb, protoinfo);
4455         return nlmsg_end(skb, nlh);
4456
4457 nla_put_failure:
4458         nlmsg_cancel(skb, nlh);
4459         return -EMSGSIZE;
4460 }
4461
4462 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4463 {
4464         struct net *net = sock_net(skb->sk);
4465         int h, s_h;
4466         int idx = 0, s_idx;
4467         struct net_device *dev;
4468         struct inet6_dev *idev;
4469         struct hlist_head *head;
4470
4471         s_h = cb->args[0];
4472         s_idx = cb->args[1];
4473
4474         rcu_read_lock();
4475         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4476                 idx = 0;
4477                 head = &net->dev_index_head[h];
4478                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4479                         if (idx < s_idx)
4480                                 goto cont;
4481                         idev = __in6_dev_get(dev);
4482                         if (!idev)
4483                                 goto cont;
4484                         if (inet6_fill_ifinfo(skb, idev,
4485                                               NETLINK_CB(cb->skb).portid,
4486                                               cb->nlh->nlmsg_seq,
4487                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4488                                 goto out;
4489 cont:
4490                         idx++;
4491                 }
4492         }
4493 out:
4494         rcu_read_unlock();
4495         cb->args[1] = idx;
4496         cb->args[0] = h;
4497
4498         return skb->len;
4499 }
4500
4501 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4502 {
4503         struct sk_buff *skb;
4504         struct net *net = dev_net(idev->dev);
4505         int err = -ENOBUFS;
4506
4507         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4508         if (skb == NULL)
4509                 goto errout;
4510
4511         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4512         if (err < 0) {
4513                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4514                 WARN_ON(err == -EMSGSIZE);
4515                 kfree_skb(skb);
4516                 goto errout;
4517         }
4518         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4519         return;
4520 errout:
4521         if (err < 0)
4522                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4523 }
4524
4525 static inline size_t inet6_prefix_nlmsg_size(void)
4526 {
4527         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4528                + nla_total_size(sizeof(struct in6_addr))
4529                + nla_total_size(sizeof(struct prefix_cacheinfo));
4530 }
4531
4532 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4533                              struct prefix_info *pinfo, u32 portid, u32 seq,
4534                              int event, unsigned int flags)
4535 {
4536         struct prefixmsg *pmsg;
4537         struct nlmsghdr *nlh;
4538         struct prefix_cacheinfo ci;
4539
4540         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4541         if (nlh == NULL)
4542                 return -EMSGSIZE;
4543
4544         pmsg = nlmsg_data(nlh);
4545         pmsg->prefix_family = AF_INET6;
4546         pmsg->prefix_pad1 = 0;
4547         pmsg->prefix_pad2 = 0;
4548         pmsg->prefix_ifindex = idev->dev->ifindex;
4549         pmsg->prefix_len = pinfo->prefix_len;
4550         pmsg->prefix_type = pinfo->type;
4551         pmsg->prefix_pad3 = 0;
4552         pmsg->prefix_flags = 0;
4553         if (pinfo->onlink)
4554                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4555         if (pinfo->autoconf)
4556                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4557
4558         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4559                 goto nla_put_failure;
4560         ci.preferred_time = ntohl(pinfo->prefered);
4561         ci.valid_time = ntohl(pinfo->valid);
4562         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4563                 goto nla_put_failure;
4564         return nlmsg_end(skb, nlh);
4565
4566 nla_put_failure:
4567         nlmsg_cancel(skb, nlh);
4568         return -EMSGSIZE;
4569 }
4570
4571 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4572                          struct prefix_info *pinfo)
4573 {
4574         struct sk_buff *skb;
4575         struct net *net = dev_net(idev->dev);
4576         int err = -ENOBUFS;
4577
4578         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4579         if (skb == NULL)
4580                 goto errout;
4581
4582         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4583         if (err < 0) {
4584                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4585                 WARN_ON(err == -EMSGSIZE);
4586                 kfree_skb(skb);
4587                 goto errout;
4588         }
4589         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4590         return;
4591 errout:
4592         if (err < 0)
4593                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4594 }
4595
4596 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4597 {
4598         write_lock_bh(&ifp->idev->lock);
4599         spin_lock(&ifp->lock);
4600         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4601                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4602             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4603                 ifp->idev->valid_ll_addr_cnt += count;
4604         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4605         spin_unlock(&ifp->lock);
4606         write_unlock_bh(&ifp->idev->lock);
4607 }
4608
4609 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4610 {
4611         struct net *net = dev_net(ifp->idev->dev);
4612
4613         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4614
4615         switch (event) {
4616         case RTM_NEWADDR:
4617                 update_valid_ll_addr_cnt(ifp, 1);
4618
4619                 /*
4620                  * If the address was optimistic
4621                  * we inserted the route at the start of
4622                  * our DAD process, so we don't need
4623                  * to do it again
4624                  */
4625                 if (!(ifp->rt->rt6i_node))
4626                         ip6_ins_rt(ifp->rt);
4627                 if (ifp->idev->cnf.forwarding)
4628                         addrconf_join_anycast(ifp);
4629                 if (!ipv6_addr_any(&ifp->peer_addr))
4630                         addrconf_prefix_route(&ifp->peer_addr, 128,
4631                                               ifp->idev->dev, 0, 0);
4632                 break;
4633         case RTM_DELADDR:
4634                 update_valid_ll_addr_cnt(ifp, -1);
4635
4636                 if (ifp->idev->cnf.forwarding)
4637                         addrconf_leave_anycast(ifp);
4638                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4639                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4640                         struct rt6_info *rt;
4641                         struct net_device *dev = ifp->idev->dev;
4642
4643                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4644                                         dev->ifindex, 1);
4645                         if (rt) {
4646                                 dst_hold(&rt->dst);
4647                                 if (ip6_del_rt(rt))
4648                                         dst_free(&rt->dst);
4649                         }
4650                 }
4651                 dst_hold(&ifp->rt->dst);
4652
4653                 if (ip6_del_rt(ifp->rt))
4654                         dst_free(&ifp->rt->dst);
4655                 break;
4656         }
4657         atomic_inc(&net->ipv6.dev_addr_genid);
4658 }
4659
4660 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4661 {
4662         rcu_read_lock_bh();
4663         if (likely(ifp->idev->dead == 0))
4664                 __ipv6_ifa_notify(event, ifp);
4665         rcu_read_unlock_bh();
4666 }
4667
4668 #ifdef CONFIG_SYSCTL
4669
4670 static
4671 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4672                            void __user *buffer, size_t *lenp, loff_t *ppos)
4673 {
4674         int *valp = ctl->data;
4675         int val = *valp;
4676         loff_t pos = *ppos;
4677         struct ctl_table lctl;
4678         int ret;
4679
4680         /*
4681          * ctl->data points to idev->cnf.forwarding, we should
4682          * not modify it until we get the rtnl lock.
4683          */
4684         lctl = *ctl;
4685         lctl.data = &val;
4686
4687         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4688
4689         if (write)
4690                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4691         if (ret)
4692                 *ppos = pos;
4693         return ret;
4694 }
4695
4696 static void dev_disable_change(struct inet6_dev *idev)
4697 {
4698         struct netdev_notifier_info info;
4699
4700         if (!idev || !idev->dev)
4701                 return;
4702
4703         netdev_notifier_info_init(&info, idev->dev);
4704         if (idev->cnf.disable_ipv6)
4705                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4706         else
4707                 addrconf_notify(NULL, NETDEV_UP, &info);
4708 }
4709
4710 static void addrconf_disable_change(struct net *net, __s32 newf)
4711 {
4712         struct net_device *dev;
4713         struct inet6_dev *idev;
4714
4715         rcu_read_lock();
4716         for_each_netdev_rcu(net, dev) {
4717                 idev = __in6_dev_get(dev);
4718                 if (idev) {
4719                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4720                         idev->cnf.disable_ipv6 = newf;
4721                         if (changed)
4722                                 dev_disable_change(idev);
4723                 }
4724         }
4725         rcu_read_unlock();
4726 }
4727
4728 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4729 {
4730         struct net *net;
4731         int old;
4732
4733         if (!rtnl_trylock())
4734                 return restart_syscall();
4735
4736         net = (struct net *)table->extra2;
4737         old = *p;
4738         *p = newf;
4739
4740         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4741                 rtnl_unlock();
4742                 return 0;
4743         }
4744
4745         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4746                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4747                 addrconf_disable_change(net, newf);
4748         } else if ((!newf) ^ (!old))
4749                 dev_disable_change((struct inet6_dev *)table->extra1);
4750
4751         rtnl_unlock();
4752         return 0;
4753 }
4754
4755 static
4756 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4757                             void __user *buffer, size_t *lenp, loff_t *ppos)
4758 {
4759         int *valp = ctl->data;
4760         int val = *valp;
4761         loff_t pos = *ppos;
4762         struct ctl_table lctl;
4763         int ret;
4764
4765         /*
4766          * ctl->data points to idev->cnf.disable_ipv6, we should
4767          * not modify it until we get the rtnl lock.
4768          */
4769         lctl = *ctl;
4770         lctl.data = &val;
4771
4772         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4773
4774         if (write)
4775                 ret = addrconf_disable_ipv6(ctl, valp, val);
4776         if (ret)
4777                 *ppos = pos;
4778         return ret;
4779 }
4780
4781 static struct addrconf_sysctl_table
4782 {
4783         struct ctl_table_header *sysctl_header;
4784         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4785 } addrconf_sysctl __read_mostly = {
4786         .sysctl_header = NULL,
4787         .addrconf_vars = {
4788                 {
4789                         .procname       = "forwarding",
4790                         .data           = &ipv6_devconf.forwarding,
4791                         .maxlen         = sizeof(int),
4792                         .mode           = 0644,
4793                         .proc_handler   = addrconf_sysctl_forward,
4794                 },
4795                 {
4796                         .procname       = "hop_limit",
4797                         .data           = &ipv6_devconf.hop_limit,
4798                         .maxlen         = sizeof(int),
4799                         .mode           = 0644,
4800                         .proc_handler   = proc_dointvec,
4801                 },
4802                 {
4803                         .procname       = "mtu",
4804                         .data           = &ipv6_devconf.mtu6,
4805                         .maxlen         = sizeof(int),
4806                         .mode           = 0644,
4807                         .proc_handler   = proc_dointvec,
4808                 },
4809                 {
4810                         .procname       = "accept_ra",
4811                         .data           = &ipv6_devconf.accept_ra,
4812                         .maxlen         = sizeof(int),
4813                         .mode           = 0644,
4814                         .proc_handler   = proc_dointvec,
4815                 },
4816                 {
4817                         .procname       = "accept_redirects",
4818                         .data           = &ipv6_devconf.accept_redirects,
4819                         .maxlen         = sizeof(int),
4820                         .mode           = 0644,
4821                         .proc_handler   = proc_dointvec,
4822                 },
4823                 {
4824                         .procname       = "autoconf",
4825                         .data           = &ipv6_devconf.autoconf,
4826                         .maxlen         = sizeof(int),
4827                         .mode           = 0644,
4828                         .proc_handler   = proc_dointvec,
4829                 },
4830                 {
4831                         .procname       = "dad_transmits",
4832                         .data           = &ipv6_devconf.dad_transmits,
4833                         .maxlen         = sizeof(int),
4834                         .mode           = 0644,
4835                         .proc_handler   = proc_dointvec,
4836                 },
4837                 {
4838                         .procname       = "router_solicitations",
4839                         .data           = &ipv6_devconf.rtr_solicits,
4840                         .maxlen         = sizeof(int),
4841                         .mode           = 0644,
4842                         .proc_handler   = proc_dointvec,
4843                 },
4844                 {
4845                         .procname       = "router_solicitation_interval",
4846                         .data           = &ipv6_devconf.rtr_solicit_interval,
4847                         .maxlen         = sizeof(int),
4848                         .mode           = 0644,
4849                         .proc_handler   = proc_dointvec_jiffies,
4850                 },
4851                 {
4852                         .procname       = "router_solicitation_delay",
4853                         .data           = &ipv6_devconf.rtr_solicit_delay,
4854                         .maxlen         = sizeof(int),
4855                         .mode           = 0644,
4856                         .proc_handler   = proc_dointvec_jiffies,
4857                 },
4858                 {
4859                         .procname       = "force_mld_version",
4860                         .data           = &ipv6_devconf.force_mld_version,
4861                         .maxlen         = sizeof(int),
4862                         .mode           = 0644,
4863                         .proc_handler   = proc_dointvec,
4864                 },
4865 #ifdef CONFIG_IPV6_PRIVACY
4866                 {
4867                         .procname       = "use_tempaddr",
4868                         .data           = &ipv6_devconf.use_tempaddr,
4869                         .maxlen         = sizeof(int),
4870                         .mode           = 0644,
4871                         .proc_handler   = proc_dointvec,
4872                 },
4873                 {
4874                         .procname       = "temp_valid_lft",
4875                         .data           = &ipv6_devconf.temp_valid_lft,
4876                         .maxlen         = sizeof(int),
4877                         .mode           = 0644,
4878                         .proc_handler   = proc_dointvec,
4879                 },
4880                 {
4881                         .procname       = "temp_prefered_lft",
4882                         .data           = &ipv6_devconf.temp_prefered_lft,
4883                         .maxlen         = sizeof(int),
4884                         .mode           = 0644,
4885                         .proc_handler   = proc_dointvec,
4886                 },
4887                 {
4888                         .procname       = "regen_max_retry",
4889                         .data           = &ipv6_devconf.regen_max_retry,
4890                         .maxlen         = sizeof(int),
4891                         .mode           = 0644,
4892                         .proc_handler   = proc_dointvec,
4893                 },
4894                 {
4895                         .procname       = "max_desync_factor",
4896                         .data           = &ipv6_devconf.max_desync_factor,
4897                         .maxlen         = sizeof(int),
4898                         .mode           = 0644,
4899                         .proc_handler   = proc_dointvec,
4900                 },
4901 #endif
4902                 {
4903                         .procname       = "max_addresses",
4904                         .data           = &ipv6_devconf.max_addresses,
4905                         .maxlen         = sizeof(int),
4906                         .mode           = 0644,
4907                         .proc_handler   = proc_dointvec,
4908                 },
4909                 {
4910                         .procname       = "accept_ra_defrtr",
4911                         .data           = &ipv6_devconf.accept_ra_defrtr,
4912                         .maxlen         = sizeof(int),
4913                         .mode           = 0644,
4914                         .proc_handler   = proc_dointvec,
4915                 },
4916                 {
4917                         .procname       = "accept_ra_pinfo",
4918                         .data           = &ipv6_devconf.accept_ra_pinfo,
4919                         .maxlen         = sizeof(int),
4920                         .mode           = 0644,
4921                         .proc_handler   = proc_dointvec,
4922                 },
4923 #ifdef CONFIG_IPV6_ROUTER_PREF
4924                 {
4925                         .procname       = "accept_ra_rtr_pref",
4926                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4927                         .maxlen         = sizeof(int),
4928                         .mode           = 0644,
4929                         .proc_handler   = proc_dointvec,
4930                 },
4931                 {
4932                         .procname       = "router_probe_interval",
4933                         .data           = &ipv6_devconf.rtr_probe_interval,
4934                         .maxlen         = sizeof(int),
4935                         .mode           = 0644,
4936                         .proc_handler   = proc_dointvec_jiffies,
4937                 },
4938 #ifdef CONFIG_IPV6_ROUTE_INFO
4939                 {
4940                         .procname       = "accept_ra_rt_info_max_plen",
4941                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4942                         .maxlen         = sizeof(int),
4943                         .mode           = 0644,
4944                         .proc_handler   = proc_dointvec,
4945                 },
4946 #endif
4947 #endif
4948                 {
4949                         .procname       = "proxy_ndp",
4950                         .data           = &ipv6_devconf.proxy_ndp,
4951                         .maxlen         = sizeof(int),
4952                         .mode           = 0644,
4953                         .proc_handler   = proc_dointvec,
4954                 },
4955                 {
4956                         .procname       = "accept_source_route",
4957                         .data           = &ipv6_devconf.accept_source_route,
4958                         .maxlen         = sizeof(int),
4959                         .mode           = 0644,
4960                         .proc_handler   = proc_dointvec,
4961                 },
4962 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4963                 {
4964                         .procname       = "optimistic_dad",
4965                         .data           = &ipv6_devconf.optimistic_dad,
4966                         .maxlen         = sizeof(int),
4967                         .mode           = 0644,
4968                         .proc_handler   = proc_dointvec,
4969
4970                 },
4971 #endif
4972 #ifdef CONFIG_IPV6_MROUTE
4973                 {
4974                         .procname       = "mc_forwarding",
4975                         .data           = &ipv6_devconf.mc_forwarding,
4976                         .maxlen         = sizeof(int),
4977                         .mode           = 0444,
4978                         .proc_handler   = proc_dointvec,
4979                 },
4980 #endif
4981                 {
4982                         .procname       = "disable_ipv6",
4983                         .data           = &ipv6_devconf.disable_ipv6,
4984                         .maxlen         = sizeof(int),
4985                         .mode           = 0644,
4986                         .proc_handler   = addrconf_sysctl_disable,
4987                 },
4988                 {
4989                         .procname       = "accept_dad",
4990                         .data           = &ipv6_devconf.accept_dad,
4991                         .maxlen         = sizeof(int),
4992                         .mode           = 0644,
4993                         .proc_handler   = proc_dointvec,
4994                 },
4995                 {
4996                         .procname       = "force_tllao",
4997                         .data           = &ipv6_devconf.force_tllao,
4998                         .maxlen         = sizeof(int),
4999                         .mode           = 0644,
5000                         .proc_handler   = proc_dointvec
5001                 },
5002                 {
5003                         .procname       = "ndisc_notify",
5004                         .data           = &ipv6_devconf.ndisc_notify,
5005                         .maxlen         = sizeof(int),
5006                         .mode           = 0644,
5007                         .proc_handler   = proc_dointvec
5008                 },
5009                 {
5010                         /* sentinel */
5011                 }
5012         },
5013 };
5014
5015 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5016                 struct inet6_dev *idev, struct ipv6_devconf *p)
5017 {
5018         int i;
5019         struct addrconf_sysctl_table *t;
5020         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5021
5022         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5023         if (t == NULL)
5024                 goto out;
5025
5026         for (i = 0; t->addrconf_vars[i].data; i++) {
5027                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5028                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5029                 t->addrconf_vars[i].extra2 = net;
5030         }
5031
5032         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5033
5034         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5035         if (t->sysctl_header == NULL)
5036                 goto free;
5037
5038         p->sysctl = t;
5039         return 0;
5040
5041 free:
5042         kfree(t);
5043 out:
5044         return -ENOBUFS;
5045 }
5046
5047 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5048 {
5049         struct addrconf_sysctl_table *t;
5050
5051         if (p->sysctl == NULL)
5052                 return;
5053
5054         t = p->sysctl;
5055         p->sysctl = NULL;
5056         unregister_net_sysctl_table(t->sysctl_header);
5057         kfree(t);
5058 }
5059
5060 static void addrconf_sysctl_register(struct inet6_dev *idev)
5061 {
5062         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5063                               &ndisc_ifinfo_sysctl_change);
5064         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5065                                         idev, &idev->cnf);
5066 }
5067
5068 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5069 {
5070         __addrconf_sysctl_unregister(&idev->cnf);
5071         neigh_sysctl_unregister(idev->nd_parms);
5072 }
5073
5074
5075 #endif
5076
5077 static int __net_init addrconf_init_net(struct net *net)
5078 {
5079         int err = -ENOMEM;
5080         struct ipv6_devconf *all, *dflt;
5081
5082         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5083         if (all == NULL)
5084                 goto err_alloc_all;
5085
5086         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5087         if (dflt == NULL)
5088                 goto err_alloc_dflt;
5089
5090         /* these will be inherited by all namespaces */
5091         dflt->autoconf = ipv6_defaults.autoconf;
5092         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5093
5094         net->ipv6.devconf_all = all;
5095         net->ipv6.devconf_dflt = dflt;
5096
5097 #ifdef CONFIG_SYSCTL
5098         err = __addrconf_sysctl_register(net, "all", NULL, all);
5099         if (err < 0)
5100                 goto err_reg_all;
5101
5102         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5103         if (err < 0)
5104                 goto err_reg_dflt;
5105 #endif
5106         return 0;
5107
5108 #ifdef CONFIG_SYSCTL
5109 err_reg_dflt:
5110         __addrconf_sysctl_unregister(all);
5111 err_reg_all:
5112         kfree(dflt);
5113 #endif
5114 err_alloc_dflt:
5115         kfree(all);
5116 err_alloc_all:
5117         return err;
5118 }
5119
5120 static void __net_exit addrconf_exit_net(struct net *net)
5121 {
5122 #ifdef CONFIG_SYSCTL
5123         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5124         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5125 #endif
5126         if (!net_eq(net, &init_net)) {
5127                 kfree(net->ipv6.devconf_dflt);
5128                 kfree(net->ipv6.devconf_all);
5129         }
5130 }
5131
5132 static struct pernet_operations addrconf_ops = {
5133         .init = addrconf_init_net,
5134         .exit = addrconf_exit_net,
5135 };
5136
5137 static struct rtnl_af_ops inet6_ops = {
5138         .family           = AF_INET6,
5139         .fill_link_af     = inet6_fill_link_af,
5140         .get_link_af_size = inet6_get_link_af_size,
5141         .set_link_af      = inet6_set_link_af,
5142 };
5143
5144 /*
5145  *      Init / cleanup code
5146  */
5147
5148 int __init addrconf_init(void)
5149 {
5150         int i, err;
5151
5152         err = ipv6_addr_label_init();
5153         if (err < 0) {
5154                 pr_crit("%s: cannot initialize default policy table: %d\n",
5155                         __func__, err);
5156                 goto out;
5157         }
5158
5159         err = register_pernet_subsys(&addrconf_ops);
5160         if (err < 0)
5161                 goto out_addrlabel;
5162
5163         /* The addrconf netdev notifier requires that loopback_dev
5164          * has it's ipv6 private information allocated and setup
5165          * before it can bring up and give link-local addresses
5166          * to other devices which are up.
5167          *
5168          * Unfortunately, loopback_dev is not necessarily the first
5169          * entry in the global dev_base list of net devices.  In fact,
5170          * it is likely to be the very last entry on that list.
5171          * So this causes the notifier registry below to try and
5172          * give link-local addresses to all devices besides loopback_dev
5173          * first, then loopback_dev, which cases all the non-loopback_dev
5174          * devices to fail to get a link-local address.
5175          *
5176          * So, as a temporary fix, allocate the ipv6 structure for
5177          * loopback_dev first by hand.
5178          * Longer term, all of the dependencies ipv6 has upon the loopback
5179          * device and it being up should be removed.
5180          */
5181         rtnl_lock();
5182         if (!ipv6_add_dev(init_net.loopback_dev))
5183                 err = -ENOMEM;
5184         rtnl_unlock();
5185         if (err)
5186                 goto errlo;
5187
5188         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5189                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5190
5191         register_netdevice_notifier(&ipv6_dev_notf);
5192
5193         addrconf_verify(0);
5194
5195         err = rtnl_af_register(&inet6_ops);
5196         if (err < 0)
5197                 goto errout_af;
5198
5199         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5200                               NULL);
5201         if (err < 0)
5202                 goto errout;
5203
5204         /* Only the first call to __rtnl_register can fail */
5205         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5206         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5207         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5208                         inet6_dump_ifaddr, NULL);
5209         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5210                         inet6_dump_ifmcaddr, NULL);
5211         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5212                         inet6_dump_ifacaddr, NULL);
5213         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5214                         inet6_netconf_dump_devconf, NULL);
5215
5216         ipv6_addr_label_rtnl_register();
5217
5218         return 0;
5219 errout:
5220         rtnl_af_unregister(&inet6_ops);
5221 errout_af:
5222         unregister_netdevice_notifier(&ipv6_dev_notf);
5223 errlo:
5224         unregister_pernet_subsys(&addrconf_ops);
5225 out_addrlabel:
5226         ipv6_addr_label_cleanup();
5227 out:
5228         return err;
5229 }
5230
5231 void addrconf_cleanup(void)
5232 {
5233         struct net_device *dev;
5234         int i;
5235
5236         unregister_netdevice_notifier(&ipv6_dev_notf);
5237         unregister_pernet_subsys(&addrconf_ops);
5238         ipv6_addr_label_cleanup();
5239
5240         rtnl_lock();
5241
5242         __rtnl_af_unregister(&inet6_ops);
5243
5244         /* clean dev list */
5245         for_each_netdev(&init_net, dev) {
5246                 if (__in6_dev_get(dev) == NULL)
5247                         continue;
5248                 addrconf_ifdown(dev, 1);
5249         }
5250         addrconf_ifdown(init_net.loopback_dev, 2);
5251
5252         /*
5253          *      Check hash table.
5254          */
5255         spin_lock_bh(&addrconf_hash_lock);
5256         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5257                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5258         spin_unlock_bh(&addrconf_hash_lock);
5259
5260         del_timer(&addr_chk_timer);
5261         rtnl_unlock();
5262 }