Pileus Git - ~andy/linux/blob - net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/ipv6.h>
  74 #include <net/protocol.h>
  75 #include <net/ndisc.h>
  76 #include <net/ip6_route.h>
  77 #include <net/addrconf.h>
  78 #include <net/tcp.h>
  79 #include <net/ip.h>
  80 #include <net/netlink.h>
  81 #include <net/pkt_sched.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85
  86 #ifdef CONFIG_IPV6_PRIVACY
  87 #include <linux/random.h>
  88 #endif
  89
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 /* Set to 3 to get tracing... */
  98 #define ACONF_DEBUG 2
  99
 100 #if ACONF_DEBUG >= 3
 101 #define ADBG(x) printk x
 102 #else
 103 #define ADBG(x)
 104 #endif
 105
 106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 107
 108 static inline u32 cstamp_delta(unsigned long cstamp)
 109 {
 110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 111 }
 112
 113 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 114 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 115 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 116
 117 #ifdef CONFIG_SYSCTL
 118 static void addrconf_sysctl_register(struct inet6_dev *idev);
 119 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 120 #else
 121 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 122 {
 123 }
 124
 125 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 126 {
 127 }
 128 #endif
 129
 130 #ifdef CONFIG_IPV6_PRIVACY
 131 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 132 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 133 static void ipv6_regen_rndid(unsigned long data);
 134 #endif
 135
 136 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 137 static int ipv6_count_addresses(struct inet6_dev *idev);
 138
 139 /*
 140  *      Configured unicast address hash table
 141  */
 142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 143 static DEFINE_SPINLOCK(addrconf_hash_lock);
 144
 145 static void addrconf_verify(unsigned long);
 146
 147 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 148 static DEFINE_SPINLOCK(addrconf_verify_lock);
 149
 150 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 151 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 152
 153 static void addrconf_type_change(struct net_device *dev,
 154                                  unsigned long event);
 155 static int addrconf_ifdown(struct net_device *dev, int how);
 156
 157 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 158 static void addrconf_dad_timer(unsigned long data);
 159 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 160 static void addrconf_dad_run(struct inet6_dev *idev);
 161 static void addrconf_rs_timer(unsigned long data);
 162 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 163 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 164
 165 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 166                                 struct prefix_info *pinfo);
 167 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 168                                struct net_device *dev);
 169
 170 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 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
 247 /* Check if a valid qdisc is available */
 248 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 249 {
 250         return !qdisc_tx_is_noop(dev);
 251 }
 252
 253 /* Check if a route is valid prefix route */
 254 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 255 {
 256         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 257 }
 258
 259 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 260 {
 261         if (del_timer(&ifp->timer))
 262                 __in6_ifa_put(ifp);
 263 }
 264
 265 enum addrconf_timer_t {
 266         AC_NONE,
 267         AC_DAD,
 268         AC_RS,
 269 };
 270
 271 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 272                                enum addrconf_timer_t what,
 273                                unsigned long when)
 274 {
 275         if (!del_timer(&ifp->timer))
 276                 in6_ifa_hold(ifp);
 277
 278         switch (what) {
 279         case AC_DAD:
 280                 ifp->timer.function = addrconf_dad_timer;
 281                 break;
 282         case AC_RS:
 283                 ifp->timer.function = addrconf_rs_timer;
 284                 break;
 285         default:
 286                 break;
 287         }
 288         ifp->timer.expires = jiffies + when;
 289         add_timer(&ifp->timer);
 290 }
 291
 292 static int snmp6_alloc_dev(struct inet6_dev *idev)
 293 {
 294         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 295                           sizeof(struct ipstats_mib),
 296                           __alignof__(struct ipstats_mib)) < 0)
 297                 goto err_ip;
 298         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 299                                         GFP_KERNEL);
 300         if (!idev->stats.icmpv6dev)
 301                 goto err_icmp;
 302         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 303                                            GFP_KERNEL);
 304         if (!idev->stats.icmpv6msgdev)
 305                 goto err_icmpmsg;
 306
 307         return 0;
 308
 309 err_icmpmsg:
 310         kfree(idev->stats.icmpv6dev);
 311 err_icmp:
 312         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 313 err_ip:
 314         return -ENOMEM;
 315 }
 316
 317 static void snmp6_free_dev(struct inet6_dev *idev)
 318 {
 319         kfree(idev->stats.icmpv6msgdev);
 320         kfree(idev->stats.icmpv6dev);
 321         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 322 }
 323
 324 /* Nobody refers to this device, we may destroy it. */
 325
 326 void in6_dev_finish_destroy(struct inet6_dev *idev)
 327 {
 328         struct net_device *dev = idev->dev;
 329
 330         WARN_ON(!list_empty(&idev->addr_list));
 331         WARN_ON(idev->mc_list != NULL);
 332
 333 #ifdef NET_REFCNT_DEBUG
 334         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 335 #endif
 336         dev_put(dev);
 337         if (!idev->dead) {
 338                 pr_warn("Freeing alive inet6 device %p\n", idev);
 339                 return;
 340         }
 341         snmp6_free_dev(idev);
 342         kfree_rcu(idev, rcu);
 343 }
 344 EXPORT_SYMBOL(in6_dev_finish_destroy);
 345
 346 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 347 {
 348         struct inet6_dev *ndev;
 349
 350         ASSERT_RTNL();
 351
 352         if (dev->mtu < IPV6_MIN_MTU)
 353                 return NULL;
 354
 355         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 356
 357         if (ndev == NULL)
 358                 return NULL;
 359
 360         rwlock_init(&ndev->lock);
 361         ndev->dev = dev;
 362         INIT_LIST_HEAD(&ndev->addr_list);
 363
 364         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 365         ndev->cnf.mtu6 = dev->mtu;
 366         ndev->cnf.sysctl = NULL;
 367         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 368         if (ndev->nd_parms == NULL) {
 369                 kfree(ndev);
 370                 return NULL;
 371         }
 372         if (ndev->cnf.forwarding)
 373                 dev_disable_lro(dev);
 374         /* We refer to the device */
 375         dev_hold(dev);
 376
 377         if (snmp6_alloc_dev(ndev) < 0) {
 378                 ADBG((KERN_WARNING
 379                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 380                         __func__, dev->name));
 381                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 382                 dev_put(dev);
 383                 kfree(ndev);
 384                 return NULL;
 385         }
 386
 387         if (snmp6_register_dev(ndev) < 0) {
 388                 ADBG((KERN_WARNING
 389                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 390                         __func__, dev->name));
 391                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 392                 ndev->dead = 1;
 393                 in6_dev_finish_destroy(ndev);
 394                 return NULL;
 395         }
 396
 397         /* One reference from device.  We must do this before
 398          * we invoke __ipv6_regen_rndid().
 399          */
 400         in6_dev_hold(ndev);
 401
 402         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 403                 ndev->cnf.accept_dad = -1;
 404
 405 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 406         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 407                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 408                 ndev->cnf.rtr_solicits = 0;
 409         }
 410 #endif
 411
 412 #ifdef CONFIG_IPV6_PRIVACY
 413         INIT_LIST_HEAD(&ndev->tempaddr_list);
 414         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 415         if ((dev->flags&IFF_LOOPBACK) ||
 416             dev->type == ARPHRD_TUNNEL ||
 417             dev->type == ARPHRD_TUNNEL6 ||
 418             dev->type == ARPHRD_SIT ||
 419             dev->type == ARPHRD_NONE) {
 420                 ndev->cnf.use_tempaddr = -1;
 421         } else {
 422                 in6_dev_hold(ndev);
 423                 ipv6_regen_rndid((unsigned long) ndev);
 424         }
 425 #endif
 426
 427         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 428                 ndev->if_flags |= IF_READY;
 429
 430         ipv6_mc_init_dev(ndev);
 431         ndev->tstamp = jiffies;
 432         addrconf_sysctl_register(ndev);
 433         /* protected by rtnl_lock */
 434         rcu_assign_pointer(dev->ip6_ptr, ndev);
 435
 436         /* Join all-node multicast group */
 437         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 438
 439         /* Join all-router multicast group if forwarding is set */
 440         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 442
 443         return ndev;
 444 }
 445
 446 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 447 {
 448         struct inet6_dev *idev;
 449
 450         ASSERT_RTNL();
 451
 452         idev = __in6_dev_get(dev);
 453         if (!idev) {
 454                 idev = ipv6_add_dev(dev);
 455                 if (!idev)
 456                         return NULL;
 457         }
 458
 459         if (dev->flags&IFF_UP)
 460                 ipv6_mc_up(idev);
 461         return idev;
 462 }
 463
 464 static int inet6_netconf_msgsize_devconf(int type)
 465 {
 466         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 467                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 468
 469         if (type == NETCONFA_FORWARDING)
 470                 size += nla_total_size(4);
 471
 472         return size;
 473 }
 474
 475 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 476                                       struct ipv6_devconf *devconf, u32 portid,
 477                                       u32 seq, int event, unsigned int flags,
 478                                       int type)
 479 {
 480         struct nlmsghdr  *nlh;
 481         struct netconfmsg *ncm;
 482
 483         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 484                         flags);
 485         if (nlh == NULL)
 486                 return -EMSGSIZE;
 487
 488         ncm = nlmsg_data(nlh);
 489         ncm->ncm_family = AF_INET6;
 490
 491         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 492                 goto nla_put_failure;
 493
 494         if (type == NETCONFA_FORWARDING &&
 495             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 496                 goto nla_put_failure;
 497
 498         return nlmsg_end(skb, nlh);
 499
 500 nla_put_failure:
 501         nlmsg_cancel(skb, nlh);
 502         return -EMSGSIZE;
 503 }
 504
 505 static void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 506                                          struct ipv6_devconf *devconf)
 507 {
 508         struct sk_buff *skb;
 509         int err = -ENOBUFS;
 510
 511         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 512         if (skb == NULL)
 513                 goto errout;
 514
 515         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 516                                          RTM_NEWNETCONF, 0, type);
 517         if (err < 0) {
 518                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 519                 WARN_ON(err == -EMSGSIZE);
 520                 kfree_skb(skb);
 521                 goto errout;
 522         }
 523         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 524         return;
 525 errout:
 526         if (err < 0)
 527                 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 528 }
 529
 530 #ifdef CONFIG_SYSCTL
 531 static void dev_forward_change(struct inet6_dev *idev)
 532 {
 533         struct net_device *dev;
 534         struct inet6_ifaddr *ifa;
 535
 536         if (!idev)
 537                 return;
 538         dev = idev->dev;
 539         if (idev->cnf.forwarding)
 540                 dev_disable_lro(dev);
 541         if (dev && (dev->flags & IFF_MULTICAST)) {
 542                 if (idev->cnf.forwarding)
 543                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 544                 else
 545                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 546         }
 547
 548         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 549                 if (ifa->flags&IFA_F_TENTATIVE)
 550                         continue;
 551                 if (idev->cnf.forwarding)
 552                         addrconf_join_anycast(ifa);
 553                 else
 554                         addrconf_leave_anycast(ifa);
 555         }
 556         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 557                                      dev->ifindex, &idev->cnf);
 558 }
 559
 560
 561 static void addrconf_forward_change(struct net *net, __s32 newf)
 562 {
 563         struct net_device *dev;
 564         struct inet6_dev *idev;
 565
 566         for_each_netdev(net, dev) {
 567                 idev = __in6_dev_get(dev);
 568                 if (idev) {
 569                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 570                         idev->cnf.forwarding = newf;
 571                         if (changed)
 572                                 dev_forward_change(idev);
 573                 }
 574         }
 575 }
 576
 577 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 578 {
 579         struct net *net;
 580         int old;
 581
 582         if (!rtnl_trylock())
 583                 return restart_syscall();
 584
 585         net = (struct net *)table->extra2;
 586         old = *p;
 587         *p = newf;
 588
 589         if (p == &net->ipv6.devconf_dflt->forwarding) {
 590                 if ((!newf) ^ (!old))
 591                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 592                                                      NETCONFA_IFINDEX_DEFAULT,
 593                                                      net->ipv6.devconf_dflt);
 594                 rtnl_unlock();
 595                 return 0;
 596         }
 597
 598         if (p == &net->ipv6.devconf_all->forwarding) {
 599                 net->ipv6.devconf_dflt->forwarding = newf;
 600                 addrconf_forward_change(net, newf);
 601                 if ((!newf) ^ (!old))
 602                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 603                                                      NETCONFA_IFINDEX_ALL,
 604                                                      net->ipv6.devconf_all);
 605         } else if ((!newf) ^ (!old))
 606                 dev_forward_change((struct inet6_dev *)table->extra1);
 607         rtnl_unlock();
 608
 609         if (newf)
 610                 rt6_purge_dflt_routers(net);
 611         return 1;
 612 }
 613 #endif
 614
 615 /* Nobody refers to this ifaddr, destroy it */
 616 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 617 {
 618         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 619
 620 #ifdef NET_REFCNT_DEBUG
 621         pr_debug("%s\n", __func__);
 622 #endif
 623
 624         in6_dev_put(ifp->idev);
 625
 626         if (del_timer(&ifp->timer))
 627                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 628
 629         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 630                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 631                 return;
 632         }
 633         dst_release(&ifp->rt->dst);
 634
 635         kfree_rcu(ifp, rcu);
 636 }
 637
 638 static void
 639 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 640 {
 641         struct list_head *p;
 642         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 643
 644         /*
 645          * Each device address list is sorted in order of scope -
 646          * global before linklocal.
 647          */
 648         list_for_each(p, &idev->addr_list) {
 649                 struct inet6_ifaddr *ifa
 650                         = list_entry(p, struct inet6_ifaddr, if_list);
 651                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 652                         break;
 653         }
 654
 655         list_add_tail(&ifp->if_list, p);
 656 }
 657
 658 static u32 inet6_addr_hash(const struct in6_addr *addr)
 659 {
 660         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 661 }
 662
 663 /* On success it returns ifp with increased reference count */
 664
 665 static struct inet6_ifaddr *
 666 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 667               int scope, u32 flags)
 668 {
 669         struct inet6_ifaddr *ifa = NULL;
 670         struct rt6_info *rt;
 671         unsigned int hash;
 672         int err = 0;
 673         int addr_type = ipv6_addr_type(addr);
 674
 675         if (addr_type == IPV6_ADDR_ANY ||
 676             addr_type & IPV6_ADDR_MULTICAST ||
 677             (!(idev->dev->flags & IFF_LOOPBACK) &&
 678              addr_type & IPV6_ADDR_LOOPBACK))
 679                 return ERR_PTR(-EADDRNOTAVAIL);
 680
 681         rcu_read_lock_bh();
 682         if (idev->dead) {
 683                 err = -ENODEV;                  /*XXX*/
 684                 goto out2;
 685         }
 686
 687         if (idev->cnf.disable_ipv6) {
 688                 err = -EACCES;
 689                 goto out2;
 690         }
 691
 692         spin_lock(&addrconf_hash_lock);
 693
 694         /* Ignore adding duplicate addresses on an interface */
 695         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 696                 ADBG(("ipv6_add_addr: already assigned\n"));
 697                 err = -EEXIST;
 698                 goto out;
 699         }
 700
 701         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 702
 703         if (ifa == NULL) {
 704                 ADBG(("ipv6_add_addr: malloc failed\n"));
 705                 err = -ENOBUFS;
 706                 goto out;
 707         }
 708
 709         rt = addrconf_dst_alloc(idev, addr, false);
 710         if (IS_ERR(rt)) {
 711                 err = PTR_ERR(rt);
 712                 goto out;
 713         }
 714
 715         ifa->addr = *addr;
 716
 717         spin_lock_init(&ifa->lock);
 718         spin_lock_init(&ifa->state_lock);
 719         init_timer(&ifa->timer);
 720         INIT_HLIST_NODE(&ifa->addr_lst);
 721         ifa->timer.data = (unsigned long) ifa;
 722         ifa->scope = scope;
 723         ifa->prefix_len = pfxlen;
 724         ifa->flags = flags | IFA_F_TENTATIVE;
 725         ifa->cstamp = ifa->tstamp = jiffies;
 726
 727         ifa->rt = rt;
 728
 729         ifa->idev = idev;
 730         in6_dev_hold(idev);
 731         /* For caller */
 732         in6_ifa_hold(ifa);
 733
 734         /* Add to big hash table */
 735         hash = inet6_addr_hash(addr);
 736
 737         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 738         spin_unlock(&addrconf_hash_lock);
 739
 740         write_lock(&idev->lock);
 741         /* Add to inet6_dev unicast addr list. */
 742         ipv6_link_dev_addr(idev, ifa);
 743
 744 #ifdef CONFIG_IPV6_PRIVACY
 745         if (ifa->flags&IFA_F_TEMPORARY) {
 746                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 747                 in6_ifa_hold(ifa);
 748         }
 749 #endif
 750
 751         in6_ifa_hold(ifa);
 752         write_unlock(&idev->lock);
 753 out2:
 754         rcu_read_unlock_bh();
 755
 756         if (likely(err == 0))
 757                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 758         else {
 759                 kfree(ifa);
 760                 ifa = ERR_PTR(err);
 761         }
 762
 763         return ifa;
 764 out:
 765         spin_unlock(&addrconf_hash_lock);
 766         goto out2;
 767 }
 768
 769 /* This function wants to get referenced ifp and releases it before return */
 770
 771 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 772 {
 773         struct inet6_ifaddr *ifa, *ifn;
 774         struct inet6_dev *idev = ifp->idev;
 775         int state;
 776         int deleted = 0, onlink = 0;
 777         unsigned long expires = jiffies;
 778
 779         spin_lock_bh(&ifp->state_lock);
 780         state = ifp->state;
 781         ifp->state = INET6_IFADDR_STATE_DEAD;
 782         spin_unlock_bh(&ifp->state_lock);
 783
 784         if (state == INET6_IFADDR_STATE_DEAD)
 785                 goto out;
 786
 787         spin_lock_bh(&addrconf_hash_lock);
 788         hlist_del_init_rcu(&ifp->addr_lst);
 789         spin_unlock_bh(&addrconf_hash_lock);
 790
 791         write_lock_bh(&idev->lock);
 792 #ifdef CONFIG_IPV6_PRIVACY
 793         if (ifp->flags&IFA_F_TEMPORARY) {
 794                 list_del(&ifp->tmp_list);
 795                 if (ifp->ifpub) {
 796                         in6_ifa_put(ifp->ifpub);
 797                         ifp->ifpub = NULL;
 798                 }
 799                 __in6_ifa_put(ifp);
 800         }
 801 #endif
 802
 803         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 804                 if (ifa == ifp) {
 805                         list_del_init(&ifp->if_list);
 806                         __in6_ifa_put(ifp);
 807
 808                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 809                                 break;
 810                         deleted = 1;
 811                         continue;
 812                 } else if (ifp->flags & IFA_F_PERMANENT) {
 813                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 814                                               ifp->prefix_len)) {
 815                                 if (ifa->flags & IFA_F_PERMANENT) {
 816                                         onlink = 1;
 817                                         if (deleted)
 818                                                 break;
 819                                 } else {
 820                                         unsigned long lifetime;
 821
 822                                         if (!onlink)
 823                                                 onlink = -1;
 824
 825                                         spin_lock(&ifa->lock);
 826
 827                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 828                                         /*
 829                                          * Note: Because this address is
 830                                          * not permanent, lifetime <
 831                                          * LONG_MAX / HZ here.
 832                                          */
 833                                         if (time_before(expires,
 834                                                         ifa->tstamp + lifetime * HZ))
 835                                                 expires = ifa->tstamp + lifetime * HZ;
 836                                         spin_unlock(&ifa->lock);
 837                                 }
 838                         }
 839                 }
 840         }
 841         write_unlock_bh(&idev->lock);
 842
 843         addrconf_del_timer(ifp);
 844
 845         ipv6_ifa_notify(RTM_DELADDR, ifp);
 846
 847         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 848
 849         /*
 850          * Purge or update corresponding prefix
 851          *
 852          * 1) we don't purge prefix here if address was not permanent.
 853          *    prefix is managed by its own lifetime.
 854          * 2) if there're no addresses, delete prefix.
 855          * 3) if there're still other permanent address(es),
 856          *    corresponding prefix is still permanent.
 857          * 4) otherwise, update prefix lifetime to the
 858          *    longest valid lifetime among the corresponding
 859          *    addresses on the device.
 860          *    Note: subsequent RA will update lifetime.
 861          *
 862          * --yoshfuji
 863          */
 864         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 865                 struct in6_addr prefix;
 866                 struct rt6_info *rt;
 867                 struct net *net = dev_net(ifp->idev->dev);
 868                 struct flowi6 fl6 = {};
 869
 870                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 871                 fl6.flowi6_oif = ifp->idev->dev->ifindex;
 872                 fl6.daddr = prefix;
 873                 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
 874                                                          RT6_LOOKUP_F_IFACE);
 875
 876                 if (rt != net->ipv6.ip6_null_entry &&
 877                     addrconf_is_prefix_route(rt)) {
 878                         if (onlink == 0) {
 879                                 ip6_del_rt(rt);
 880                                 rt = NULL;
 881                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 882                                 rt6_set_expires(rt, expires);
 883                         }
 884                 }
 885                 dst_release(&rt->dst);
 886         }
 887
 888         /* clean up prefsrc entries */
 889         rt6_remove_prefsrc(ifp);
 890 out:
 891         in6_ifa_put(ifp);
 892 }
 893
 894 #ifdef CONFIG_IPV6_PRIVACY
 895 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 896 {
 897         struct inet6_dev *idev = ifp->idev;
 898         struct in6_addr addr, *tmpaddr;
 899         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 900         unsigned long regen_advance;
 901         int tmp_plen;
 902         int ret = 0;
 903         int max_addresses;
 904         u32 addr_flags;
 905         unsigned long now = jiffies;
 906
 907         write_lock(&idev->lock);
 908         if (ift) {
 909                 spin_lock_bh(&ift->lock);
 910                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 911                 spin_unlock_bh(&ift->lock);
 912                 tmpaddr = &addr;
 913         } else {
 914                 tmpaddr = NULL;
 915         }
 916 retry:
 917         in6_dev_hold(idev);
 918         if (idev->cnf.use_tempaddr <= 0) {
 919                 write_unlock(&idev->lock);
 920                 pr_info("%s: use_tempaddr is disabled\n", __func__);
 921                 in6_dev_put(idev);
 922                 ret = -1;
 923                 goto out;
 924         }
 925         spin_lock_bh(&ifp->lock);
 926         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 927                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 928                 spin_unlock_bh(&ifp->lock);
 929                 write_unlock(&idev->lock);
 930                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
 931                         __func__);
 932                 in6_dev_put(idev);
 933                 ret = -1;
 934                 goto out;
 935         }
 936         in6_ifa_hold(ifp);
 937         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 938         __ipv6_try_regen_rndid(idev, tmpaddr);
 939         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 940         age = (now - ifp->tstamp) / HZ;
 941         tmp_valid_lft = min_t(__u32,
 942                               ifp->valid_lft,
 943                               idev->cnf.temp_valid_lft + age);
 944         tmp_prefered_lft = min_t(__u32,
 945                                  ifp->prefered_lft,
 946                                  idev->cnf.temp_prefered_lft + age -
 947                                  idev->cnf.max_desync_factor);
 948         tmp_plen = ifp->prefix_len;
 949         max_addresses = idev->cnf.max_addresses;
 950         tmp_tstamp = ifp->tstamp;
 951         spin_unlock_bh(&ifp->lock);
 952
 953         regen_advance = idev->cnf.regen_max_retry *
 954                         idev->cnf.dad_transmits *
 955                         idev->nd_parms->retrans_time / HZ;
 956         write_unlock(&idev->lock);
 957
 958         /* A temporary address is created only if this calculated Preferred
 959          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 960          * an implementation must not create a temporary address with a zero
 961          * Preferred Lifetime.
 962          */
 963         if (tmp_prefered_lft <= regen_advance) {
 964                 in6_ifa_put(ifp);
 965                 in6_dev_put(idev);
 966                 ret = -1;
 967                 goto out;
 968         }
 969
 970         addr_flags = IFA_F_TEMPORARY;
 971         /* set in addrconf_prefix_rcv() */
 972         if (ifp->flags & IFA_F_OPTIMISTIC)
 973                 addr_flags |= IFA_F_OPTIMISTIC;
 974
 975         ift = !max_addresses ||
 976               ipv6_count_addresses(idev) < max_addresses ?
 977                 ipv6_add_addr(idev, &addr, tmp_plen,
 978                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 979                               addr_flags) : NULL;
 980         if (!ift || IS_ERR(ift)) {
 981                 in6_ifa_put(ifp);
 982                 in6_dev_put(idev);
 983                 pr_info("%s: retry temporary address regeneration\n", __func__);
 984                 tmpaddr = &addr;
 985                 write_lock(&idev->lock);
 986                 goto retry;
 987         }
 988
 989         spin_lock_bh(&ift->lock);
 990         ift->ifpub = ifp;
 991         ift->valid_lft = tmp_valid_lft;
 992         ift->prefered_lft = tmp_prefered_lft;
 993         ift->cstamp = now;
 994         ift->tstamp = tmp_tstamp;
 995         spin_unlock_bh(&ift->lock);
 996
 997         addrconf_dad_start(ift);
 998         in6_ifa_put(ift);
 999         in6_dev_put(idev);
1000 out:
1001         return ret;
1002 }
1003 #endif
1004
1005 /*
1006  *      Choose an appropriate source address (RFC3484)
1007  */
1008 enum {
1009         IPV6_SADDR_RULE_INIT = 0,
1010         IPV6_SADDR_RULE_LOCAL,
1011         IPV6_SADDR_RULE_SCOPE,
1012         IPV6_SADDR_RULE_PREFERRED,
1013 #ifdef CONFIG_IPV6_MIP6
1014         IPV6_SADDR_RULE_HOA,
1015 #endif
1016         IPV6_SADDR_RULE_OIF,
1017         IPV6_SADDR_RULE_LABEL,
1018 #ifdef CONFIG_IPV6_PRIVACY
1019         IPV6_SADDR_RULE_PRIVACY,
1020 #endif
1021         IPV6_SADDR_RULE_ORCHID,
1022         IPV6_SADDR_RULE_PREFIX,
1023         IPV6_SADDR_RULE_MAX
1024 };
1025
1026 struct ipv6_saddr_score {
1027         int                     rule;
1028         int                     addr_type;
1029         struct inet6_ifaddr     *ifa;
1030         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1031         int                     scopedist;
1032         int                     matchlen;
1033 };
1034
1035 struct ipv6_saddr_dst {
1036         const struct in6_addr *addr;
1037         int ifindex;
1038         int scope;
1039         int label;
1040         unsigned int prefs;
1041 };
1042
1043 static inline int ipv6_saddr_preferred(int type)
1044 {
1045         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1046                 return 1;
1047         return 0;
1048 }
1049
1050 static int ipv6_get_saddr_eval(struct net *net,
1051                                struct ipv6_saddr_score *score,
1052                                struct ipv6_saddr_dst *dst,
1053                                int i)
1054 {
1055         int ret;
1056
1057         if (i <= score->rule) {
1058                 switch (i) {
1059                 case IPV6_SADDR_RULE_SCOPE:
1060                         ret = score->scopedist;
1061                         break;
1062                 case IPV6_SADDR_RULE_PREFIX:
1063                         ret = score->matchlen;
1064                         break;
1065                 default:
1066                         ret = !!test_bit(i, score->scorebits);
1067                 }
1068                 goto out;
1069         }
1070
1071         switch (i) {
1072         case IPV6_SADDR_RULE_INIT:
1073                 /* Rule 0: remember if hiscore is not ready yet */
1074                 ret = !!score->ifa;
1075                 break;
1076         case IPV6_SADDR_RULE_LOCAL:
1077                 /* Rule 1: Prefer same address */
1078                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1079                 break;
1080         case IPV6_SADDR_RULE_SCOPE:
1081                 /* Rule 2: Prefer appropriate scope
1082                  *
1083                  *      ret
1084                  *       ^
1085                  *    -1 |  d 15
1086                  *    ---+--+-+---> scope
1087                  *       |
1088                  *       |             d is scope of the destination.
1089                  *  B-d  |  \
1090                  *       |   \      <- smaller scope is better if
1091                  *  B-15 |    \        if scope is enough for destinaion.
1092                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1093                  * d-C-1 | /
1094                  *       |/         <- greater is better
1095                  *   -C  /             if scope is not enough for destination.
1096                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1097                  *
1098                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1099                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1100                  * Assume B = 0 and we get C > 29.
1101                  */
1102                 ret = __ipv6_addr_src_scope(score->addr_type);
1103                 if (ret >= dst->scope)
1104                         ret = -ret;
1105                 else
1106                         ret -= 128;     /* 30 is enough */
1107                 score->scopedist = ret;
1108                 break;
1109         case IPV6_SADDR_RULE_PREFERRED:
1110                 /* Rule 3: Avoid deprecated and optimistic addresses */
1111                 ret = ipv6_saddr_preferred(score->addr_type) ||
1112                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1113                 break;
1114 #ifdef CONFIG_IPV6_MIP6
1115         case IPV6_SADDR_RULE_HOA:
1116             {
1117                 /* Rule 4: Prefer home address */
1118                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1119                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1120                 break;
1121             }
1122 #endif
1123         case IPV6_SADDR_RULE_OIF:
1124                 /* Rule 5: Prefer outgoing interface */
1125                 ret = (!dst->ifindex ||
1126                        dst->ifindex == score->ifa->idev->dev->ifindex);
1127                 break;
1128         case IPV6_SADDR_RULE_LABEL:
1129                 /* Rule 6: Prefer matching label */
1130                 ret = ipv6_addr_label(net,
1131                                       &score->ifa->addr, score->addr_type,
1132                                       score->ifa->idev->dev->ifindex) == dst->label;
1133                 break;
1134 #ifdef CONFIG_IPV6_PRIVACY
1135         case IPV6_SADDR_RULE_PRIVACY:
1136             {
1137                 /* Rule 7: Prefer public address
1138                  * Note: prefer temporary address if use_tempaddr >= 2
1139                  */
1140                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1141                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1142                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1143                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1144                 break;
1145             }
1146 #endif
1147         case IPV6_SADDR_RULE_ORCHID:
1148                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1149                  *          non-ORCHID vs non-ORCHID
1150                  */
1151                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1152                         ipv6_addr_orchid(dst->addr));
1153                 break;
1154         case IPV6_SADDR_RULE_PREFIX:
1155                 /* Rule 8: Use longest matching prefix */
1156                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1157                 if (ret > score->ifa->prefix_len)
1158                         ret = score->ifa->prefix_len;
1159                 score->matchlen = ret;
1160                 break;
1161         default:
1162                 ret = 0;
1163         }
1164
1165         if (ret)
1166                 __set_bit(i, score->scorebits);
1167         score->rule = i;
1168 out:
1169         return ret;
1170 }
1171
1172 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1173                        const struct in6_addr *daddr, unsigned int prefs,
1174                        struct in6_addr *saddr)
1175 {
1176         struct ipv6_saddr_score scores[2],
1177                                 *score = &scores[0], *hiscore = &scores[1];
1178         struct ipv6_saddr_dst dst;
1179         struct net_device *dev;
1180         int dst_type;
1181
1182         dst_type = __ipv6_addr_type(daddr);
1183         dst.addr = daddr;
1184         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1185         dst.scope = __ipv6_addr_src_scope(dst_type);
1186         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1187         dst.prefs = prefs;
1188
1189         hiscore->rule = -1;
1190         hiscore->ifa = NULL;
1191
1192         rcu_read_lock();
1193
1194         for_each_netdev_rcu(net, dev) {
1195                 struct inet6_dev *idev;
1196
1197                 /* Candidate Source Address (section 4)
1198                  *  - multicast and link-local destination address,
1199                  *    the set of candidate source address MUST only
1200                  *    include addresses assigned to interfaces
1201                  *    belonging to the same link as the outgoing
1202                  *    interface.
1203                  * (- For site-local destination addresses, the
1204                  *    set of candidate source addresses MUST only
1205                  *    include addresses assigned to interfaces
1206                  *    belonging to the same site as the outgoing
1207                  *    interface.)
1208                  */
1209                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1210                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1211                     dst.ifindex && dev->ifindex != dst.ifindex)
1212                         continue;
1213
1214                 idev = __in6_dev_get(dev);
1215                 if (!idev)
1216                         continue;
1217
1218                 read_lock_bh(&idev->lock);
1219                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1220                         int i;
1221
1222                         /*
1223                          * - Tentative Address (RFC2462 section 5.4)
1224                          *  - A tentative address is not considered
1225                          *    "assigned to an interface" in the traditional
1226                          *    sense, unless it is also flagged as optimistic.
1227                          * - Candidate Source Address (section 4)
1228                          *  - In any case, anycast addresses, multicast
1229                          *    addresses, and the unspecified address MUST
1230                          *    NOT be included in a candidate set.
1231                          */
1232                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1233                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1234                                 continue;
1235
1236                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1237
1238                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1239                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1240                                 LIMIT_NETDEBUG(KERN_DEBUG
1241                                                "ADDRCONF: unspecified / multicast address "
1242                                                "assigned as unicast address on %s",
1243                                                dev->name);
1244                                 continue;
1245                         }
1246
1247                         score->rule = -1;
1248                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1249
1250                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1251                                 int minihiscore, miniscore;
1252
1253                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1254                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1255
1256                                 if (minihiscore > miniscore) {
1257                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1258                                             score->scopedist > 0) {
1259                                                 /*
1260                                                  * special case:
1261                                                  * each remaining entry
1262                                                  * has too small (not enough)
1263                                                  * scope, because ifa entries
1264                                                  * are sorted by their scope
1265                                                  * values.
1266                                                  */
1267                                                 goto try_nextdev;
1268                                         }
1269                                         break;
1270                                 } else if (minihiscore < miniscore) {
1271                                         if (hiscore->ifa)
1272                                                 in6_ifa_put(hiscore->ifa);
1273
1274                                         in6_ifa_hold(score->ifa);
1275
1276                                         swap(hiscore, score);
1277
1278                                         /* restore our iterator */
1279                                         score->ifa = hiscore->ifa;
1280
1281                                         break;
1282                                 }
1283                         }
1284                 }
1285 try_nextdev:
1286                 read_unlock_bh(&idev->lock);
1287         }
1288         rcu_read_unlock();
1289
1290         if (!hiscore->ifa)
1291                 return -EADDRNOTAVAIL;
1292
1293         *saddr = hiscore->ifa->addr;
1294         in6_ifa_put(hiscore->ifa);
1295         return 0;
1296 }
1297 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1298
1299 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1300                     unsigned char banned_flags)
1301 {
1302         struct inet6_dev *idev;
1303         int err = -EADDRNOTAVAIL;
1304
1305         rcu_read_lock();
1306         idev = __in6_dev_get(dev);
1307         if (idev) {
1308                 struct inet6_ifaddr *ifp;
1309
1310                 read_lock_bh(&idev->lock);
1311                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1312                         if (ifp->scope == IFA_LINK &&
1313                             !(ifp->flags & banned_flags)) {
1314                                 *addr = ifp->addr;
1315                                 err = 0;
1316                                 break;
1317                         }
1318                 }
1319                 read_unlock_bh(&idev->lock);
1320         }
1321         rcu_read_unlock();
1322         return err;
1323 }
1324
1325 static int ipv6_count_addresses(struct inet6_dev *idev)
1326 {
1327         int cnt = 0;
1328         struct inet6_ifaddr *ifp;
1329
1330         read_lock_bh(&idev->lock);
1331         list_for_each_entry(ifp, &idev->addr_list, if_list)
1332                 cnt++;
1333         read_unlock_bh(&idev->lock);
1334         return cnt;
1335 }
1336
1337 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1338                   struct net_device *dev, int strict)
1339 {
1340         struct inet6_ifaddr *ifp;
1341         struct hlist_node *node;
1342         unsigned int hash = inet6_addr_hash(addr);
1343
1344         rcu_read_lock_bh();
1345         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1346                 if (!net_eq(dev_net(ifp->idev->dev), net))
1347                         continue;
1348                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1349                     !(ifp->flags&IFA_F_TENTATIVE) &&
1350                     (dev == NULL || ifp->idev->dev == dev ||
1351                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1352                         rcu_read_unlock_bh();
1353                         return 1;
1354                 }
1355         }
1356
1357         rcu_read_unlock_bh();
1358         return 0;
1359 }
1360 EXPORT_SYMBOL(ipv6_chk_addr);
1361
1362 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1363                                struct net_device *dev)
1364 {
1365         unsigned int hash = inet6_addr_hash(addr);
1366         struct inet6_ifaddr *ifp;
1367         struct hlist_node *node;
1368
1369         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1370                 if (!net_eq(dev_net(ifp->idev->dev), net))
1371                         continue;
1372                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1373                         if (dev == NULL || ifp->idev->dev == dev)
1374                                 return true;
1375                 }
1376         }
1377         return false;
1378 }
1379
1380 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1381 {
1382         struct inet6_dev *idev;
1383         struct inet6_ifaddr *ifa;
1384         int     onlink;
1385
1386         onlink = 0;
1387         rcu_read_lock();
1388         idev = __in6_dev_get(dev);
1389         if (idev) {
1390                 read_lock_bh(&idev->lock);
1391                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1392                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1393                                                    ifa->prefix_len);
1394                         if (onlink)
1395                                 break;
1396                 }
1397                 read_unlock_bh(&idev->lock);
1398         }
1399         rcu_read_unlock();
1400         return onlink;
1401 }
1402 EXPORT_SYMBOL(ipv6_chk_prefix);
1403
1404 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1405                                      struct net_device *dev, int strict)
1406 {
1407         struct inet6_ifaddr *ifp, *result = NULL;
1408         unsigned int hash = inet6_addr_hash(addr);
1409         struct hlist_node *node;
1410
1411         rcu_read_lock_bh();
1412         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1413                 if (!net_eq(dev_net(ifp->idev->dev), net))
1414                         continue;
1415                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1416                         if (dev == NULL || ifp->idev->dev == dev ||
1417                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1418                                 result = ifp;
1419                                 in6_ifa_hold(ifp);
1420                                 break;
1421                         }
1422                 }
1423         }
1424         rcu_read_unlock_bh();
1425
1426         return result;
1427 }
1428
1429 /* Gets referenced address, destroys ifaddr */
1430
1431 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1432 {
1433         if (ifp->flags&IFA_F_PERMANENT) {
1434                 spin_lock_bh(&ifp->lock);
1435                 addrconf_del_timer(ifp);
1436                 ifp->flags |= IFA_F_TENTATIVE;
1437                 if (dad_failed)
1438                         ifp->flags |= IFA_F_DADFAILED;
1439                 spin_unlock_bh(&ifp->lock);
1440                 if (dad_failed)
1441                         ipv6_ifa_notify(0, ifp);
1442                 in6_ifa_put(ifp);
1443 #ifdef CONFIG_IPV6_PRIVACY
1444         } else if (ifp->flags&IFA_F_TEMPORARY) {
1445                 struct inet6_ifaddr *ifpub;
1446                 spin_lock_bh(&ifp->lock);
1447                 ifpub = ifp->ifpub;
1448                 if (ifpub) {
1449                         in6_ifa_hold(ifpub);
1450                         spin_unlock_bh(&ifp->lock);
1451                         ipv6_create_tempaddr(ifpub, ifp);
1452                         in6_ifa_put(ifpub);
1453                 } else {
1454                         spin_unlock_bh(&ifp->lock);
1455                 }
1456                 ipv6_del_addr(ifp);
1457 #endif
1458         } else
1459                 ipv6_del_addr(ifp);
1460 }
1461
1462 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1463 {
1464         int err = -ENOENT;
1465
1466         spin_lock(&ifp->state_lock);
1467         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1468                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1469                 err = 0;
1470         }
1471         spin_unlock(&ifp->state_lock);
1472
1473         return err;
1474 }
1475
1476 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1477 {
1478         struct inet6_dev *idev = ifp->idev;
1479
1480         if (addrconf_dad_end(ifp)) {
1481                 in6_ifa_put(ifp);
1482                 return;
1483         }
1484
1485         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1486                              ifp->idev->dev->name, &ifp->addr);
1487
1488         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1489                 struct in6_addr addr;
1490
1491                 addr.s6_addr32[0] = htonl(0xfe800000);
1492                 addr.s6_addr32[1] = 0;
1493
1494                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1495                     ipv6_addr_equal(&ifp->addr, &addr)) {
1496                         /* DAD failed for link-local based on MAC address */
1497                         idev->cnf.disable_ipv6 = 1;
1498
1499                         pr_info("%s: IPv6 being disabled!\n",
1500                                 ifp->idev->dev->name);
1501                 }
1502         }
1503
1504         addrconf_dad_stop(ifp, 1);
1505 }
1506
1507 /* Join to solicited addr multicast group. */
1508
1509 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1510 {
1511         struct in6_addr maddr;
1512
1513         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1514                 return;
1515
1516         addrconf_addr_solict_mult(addr, &maddr);
1517         ipv6_dev_mc_inc(dev, &maddr);
1518 }
1519
1520 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1521 {
1522         struct in6_addr maddr;
1523
1524         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1525                 return;
1526
1527         addrconf_addr_solict_mult(addr, &maddr);
1528         __ipv6_dev_mc_dec(idev, &maddr);
1529 }
1530
1531 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1532 {
1533         struct in6_addr addr;
1534         if (ifp->prefix_len == 127) /* RFC 6164 */
1535                 return;
1536         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1537         if (ipv6_addr_any(&addr))
1538                 return;
1539         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1540 }
1541
1542 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1543 {
1544         struct in6_addr addr;
1545         if (ifp->prefix_len == 127) /* RFC 6164 */
1546                 return;
1547         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1548         if (ipv6_addr_any(&addr))
1549                 return;
1550         __ipv6_dev_ac_dec(ifp->idev, &addr);
1551 }
1552
1553 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1554 {
1555         if (dev->addr_len != ETH_ALEN)
1556                 return -1;
1557         memcpy(eui, dev->dev_addr, 3);
1558         memcpy(eui + 5, dev->dev_addr + 3, 3);
1559
1560         /*
1561          * The zSeries OSA network cards can be shared among various
1562          * OS instances, but the OSA cards have only one MAC address.
1563          * This leads to duplicate address conflicts in conjunction
1564          * with IPv6 if more than one instance uses the same card.
1565          *
1566          * The driver for these cards can deliver a unique 16-bit
1567          * identifier for each instance sharing the same card.  It is
1568          * placed instead of 0xFFFE in the interface identifier.  The
1569          * "u" bit of the interface identifier is not inverted in this
1570          * case.  Hence the resulting interface identifier has local
1571          * scope according to RFC2373.
1572          */
1573         if (dev->dev_id) {
1574                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1575                 eui[4] = dev->dev_id & 0xFF;
1576         } else {
1577                 eui[3] = 0xFF;
1578                 eui[4] = 0xFE;
1579                 eui[0] ^= 2;
1580         }
1581         return 0;
1582 }
1583
1584 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1585 {
1586         if (dev->addr_len != IEEE802154_ADDR_LEN)
1587                 return -1;
1588         memcpy(eui, dev->dev_addr, 8);
1589         return 0;
1590 }
1591
1592 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1593 {
1594         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1595         if (dev->addr_len != ARCNET_ALEN)
1596                 return -1;
1597         memset(eui, 0, 7);
1598         eui[7] = *(u8 *)dev->dev_addr;
1599         return 0;
1600 }
1601
1602 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1603 {
1604         if (dev->addr_len != INFINIBAND_ALEN)
1605                 return -1;
1606         memcpy(eui, dev->dev_addr + 12, 8);
1607         eui[0] |= 2;
1608         return 0;
1609 }
1610
1611 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1612 {
1613         if (addr == 0)
1614                 return -1;
1615         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1616                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1617                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1618                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1619                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1620                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1621         eui[1] = 0;
1622         eui[2] = 0x5E;
1623         eui[3] = 0xFE;
1624         memcpy(eui + 4, &addr, 4);
1625         return 0;
1626 }
1627
1628 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1629 {
1630         if (dev->priv_flags & IFF_ISATAP)
1631                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1632         return -1;
1633 }
1634
1635 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1636 {
1637         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1638 }
1639
1640 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1641 {
1642         switch (dev->type) {
1643         case ARPHRD_ETHER:
1644         case ARPHRD_FDDI:
1645                 return addrconf_ifid_eui48(eui, dev);
1646         case ARPHRD_ARCNET:
1647                 return addrconf_ifid_arcnet(eui, dev);
1648         case ARPHRD_INFINIBAND:
1649                 return addrconf_ifid_infiniband(eui, dev);
1650         case ARPHRD_SIT:
1651                 return addrconf_ifid_sit(eui, dev);
1652         case ARPHRD_IPGRE:
1653                 return addrconf_ifid_gre(eui, dev);
1654         case ARPHRD_IEEE802154:
1655                 return addrconf_ifid_eui64(eui, dev);
1656         }
1657         return -1;
1658 }
1659
1660 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1661 {
1662         int err = -1;
1663         struct inet6_ifaddr *ifp;
1664
1665         read_lock_bh(&idev->lock);
1666         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1667                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1668                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1669                         err = 0;
1670                         break;
1671                 }
1672         }
1673         read_unlock_bh(&idev->lock);
1674         return err;
1675 }
1676
1677 #ifdef CONFIG_IPV6_PRIVACY
1678 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1679 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1680 {
1681 regen:
1682         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1683         idev->rndid[0] &= ~0x02;
1684
1685         /*
1686          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1687          * check if generated address is not inappropriate
1688          *
1689          *  - Reserved subnet anycast (RFC 2526)
1690          *      11111101 11....11 1xxxxxxx
1691          *  - ISATAP (RFC4214) 6.1
1692          *      00-00-5E-FE-xx-xx-xx-xx
1693          *  - value 0
1694          *  - XXX: already assigned to an address on the device
1695          */
1696         if (idev->rndid[0] == 0xfd &&
1697             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1698             (idev->rndid[7]&0x80))
1699                 goto regen;
1700         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1701                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1702                         goto regen;
1703                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1704                         goto regen;
1705         }
1706 }
1707
1708 static void ipv6_regen_rndid(unsigned long data)
1709 {
1710         struct inet6_dev *idev = (struct inet6_dev *) data;
1711         unsigned long expires;
1712
1713         rcu_read_lock_bh();
1714         write_lock_bh(&idev->lock);
1715
1716         if (idev->dead)
1717                 goto out;
1718
1719         __ipv6_regen_rndid(idev);
1720
1721         expires = jiffies +
1722                 idev->cnf.temp_prefered_lft * HZ -
1723                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1724                 idev->cnf.max_desync_factor * HZ;
1725         if (time_before(expires, jiffies)) {
1726                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1727                         __func__, idev->dev->name);
1728                 goto out;
1729         }
1730
1731         if (!mod_timer(&idev->regen_timer, expires))
1732                 in6_dev_hold(idev);
1733
1734 out:
1735         write_unlock_bh(&idev->lock);
1736         rcu_read_unlock_bh();
1737         in6_dev_put(idev);
1738 }
1739
1740 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1741 {
1742         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1743                 __ipv6_regen_rndid(idev);
1744 }
1745 #endif
1746
1747 /*
1748  *      Add prefix route.
1749  */
1750
1751 static void
1752 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1753                       unsigned long expires, u32 flags)
1754 {
1755         struct fib6_config cfg = {
1756                 .fc_table = RT6_TABLE_PREFIX,
1757                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1758                 .fc_ifindex = dev->ifindex,
1759                 .fc_expires = expires,
1760                 .fc_dst_len = plen,
1761                 .fc_flags = RTF_UP | flags,
1762                 .fc_nlinfo.nl_net = dev_net(dev),
1763                 .fc_protocol = RTPROT_KERNEL,
1764         };
1765
1766         cfg.fc_dst = *pfx;
1767
1768         /* Prevent useless cloning on PtP SIT.
1769            This thing is done here expecting that the whole
1770            class of non-broadcast devices need not cloning.
1771          */
1772 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1773         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1774                 cfg.fc_flags |= RTF_NONEXTHOP;
1775 #endif
1776
1777         ip6_route_add(&cfg);
1778 }
1779
1780
1781 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1782                                                   int plen,
1783                                                   const struct net_device *dev,
1784                                                   u32 flags, u32 noflags)
1785 {
1786         struct fib6_node *fn;
1787         struct rt6_info *rt = NULL;
1788         struct fib6_table *table;
1789
1790         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1791         if (table == NULL)
1792                 return NULL;
1793
1794         read_lock_bh(&table->tb6_lock);
1795         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1796         if (!fn)
1797                 goto out;
1798         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1799                 if (rt->dst.dev->ifindex != dev->ifindex)
1800                         continue;
1801                 if ((rt->rt6i_flags & flags) != flags)
1802                         continue;
1803                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1804                         continue;
1805                 dst_hold(&rt->dst);
1806                 break;
1807         }
1808 out:
1809         read_unlock_bh(&table->tb6_lock);
1810         return rt;
1811 }
1812
1813
1814 /* Create "default" multicast route to the interface */
1815
1816 static void addrconf_add_mroute(struct net_device *dev)
1817 {
1818         struct fib6_config cfg = {
1819                 .fc_table = RT6_TABLE_LOCAL,
1820                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1821                 .fc_ifindex = dev->ifindex,
1822                 .fc_dst_len = 8,
1823                 .fc_flags = RTF_UP,
1824                 .fc_nlinfo.nl_net = dev_net(dev),
1825         };
1826
1827         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1828
1829         ip6_route_add(&cfg);
1830 }
1831
1832 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1833 static void sit_route_add(struct net_device *dev)
1834 {
1835         struct fib6_config cfg = {
1836                 .fc_table = RT6_TABLE_MAIN,
1837                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1838                 .fc_ifindex = dev->ifindex,
1839                 .fc_dst_len = 96,
1840                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1841                 .fc_nlinfo.nl_net = dev_net(dev),
1842         };
1843
1844         /* prefix length - 96 bits "::d.d.d.d" */
1845         ip6_route_add(&cfg);
1846 }
1847 #endif
1848
1849 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1850 {
1851         struct inet6_dev *idev;
1852
1853         ASSERT_RTNL();
1854
1855         idev = ipv6_find_idev(dev);
1856         if (!idev)
1857                 return ERR_PTR(-ENOBUFS);
1858
1859         if (idev->cnf.disable_ipv6)
1860                 return ERR_PTR(-EACCES);
1861
1862         /* Add default multicast route */
1863         if (!(dev->flags & IFF_LOOPBACK))
1864                 addrconf_add_mroute(dev);
1865
1866         return idev;
1867 }
1868
1869 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1870 {
1871         struct prefix_info *pinfo;
1872         __u32 valid_lft;
1873         __u32 prefered_lft;
1874         int addr_type;
1875         struct inet6_dev *in6_dev;
1876         struct net *net = dev_net(dev);
1877
1878         pinfo = (struct prefix_info *) opt;
1879
1880         if (len < sizeof(struct prefix_info)) {
1881                 ADBG(("addrconf: prefix option too short\n"));
1882                 return;
1883         }
1884
1885         /*
1886          *      Validation checks ([ADDRCONF], page 19)
1887          */
1888
1889         addr_type = ipv6_addr_type(&pinfo->prefix);
1890
1891         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1892                 return;
1893
1894         valid_lft = ntohl(pinfo->valid);
1895         prefered_lft = ntohl(pinfo->prefered);
1896
1897         if (prefered_lft > valid_lft) {
1898                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1899                 return;
1900         }
1901
1902         in6_dev = in6_dev_get(dev);
1903
1904         if (in6_dev == NULL) {
1905                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1906                                     dev->name);
1907                 return;
1908         }
1909
1910         /*
1911          *      Two things going on here:
1912          *      1) Add routes for on-link prefixes
1913          *      2) Configure prefixes with the auto flag set
1914          */
1915
1916         if (pinfo->onlink) {
1917                 struct rt6_info *rt;
1918                 unsigned long rt_expires;
1919
1920                 /* Avoid arithmetic overflow. Really, we could
1921                  * save rt_expires in seconds, likely valid_lft,
1922                  * but it would require division in fib gc, that it
1923                  * not good.
1924                  */
1925                 if (HZ > USER_HZ)
1926                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1927                 else
1928                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1929
1930                 if (addrconf_finite_timeout(rt_expires))
1931                         rt_expires *= HZ;
1932
1933                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1934                                                pinfo->prefix_len,
1935                                                dev,
1936                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1937                                                RTF_GATEWAY | RTF_DEFAULT);
1938
1939                 if (rt) {
1940                         /* Autoconf prefix route */
1941                         if (valid_lft == 0) {
1942                                 ip6_del_rt(rt);
1943                                 rt = NULL;
1944                         } else if (addrconf_finite_timeout(rt_expires)) {
1945                                 /* not infinity */
1946                                 rt6_set_expires(rt, jiffies + rt_expires);
1947                         } else {
1948                                 rt6_clean_expires(rt);
1949                         }
1950                 } else if (valid_lft) {
1951                         clock_t expires = 0;
1952                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1953                         if (addrconf_finite_timeout(rt_expires)) {
1954                                 /* not infinity */
1955                                 flags |= RTF_EXPIRES;
1956                                 expires = jiffies_to_clock_t(rt_expires);
1957                         }
1958                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1959                                               dev, expires, flags);
1960                 }
1961                 if (rt)
1962                         dst_release(&rt->dst);
1963         }
1964
1965         /* Try to figure out our local address for this prefix */
1966
1967         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1968                 struct inet6_ifaddr *ifp;
1969                 struct in6_addr addr;
1970                 int create = 0, update_lft = 0;
1971
1972                 if (pinfo->prefix_len == 64) {
1973                         memcpy(&addr, &pinfo->prefix, 8);
1974                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1975                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1976                                 in6_dev_put(in6_dev);
1977                                 return;
1978                         }
1979                         goto ok;
1980                 }
1981                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
1982                                     pinfo->prefix_len);
1983                 in6_dev_put(in6_dev);
1984                 return;
1985
1986 ok:
1987
1988                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1989
1990                 if (ifp == NULL && valid_lft) {
1991                         int max_addresses = in6_dev->cnf.max_addresses;
1992                         u32 addr_flags = 0;
1993
1994 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1995                         if (in6_dev->cnf.optimistic_dad &&
1996                             !net->ipv6.devconf_all->forwarding && sllao)
1997                                 addr_flags = IFA_F_OPTIMISTIC;
1998 #endif
1999
2000                         /* Do not allow to create too much of autoconfigured
2001                          * addresses; this would be too easy way to crash kernel.
2002                          */
2003                         if (!max_addresses ||
2004                             ipv6_count_addresses(in6_dev) < max_addresses)
2005                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2006                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2007                                                     addr_flags);
2008
2009                         if (!ifp || IS_ERR(ifp)) {
2010                                 in6_dev_put(in6_dev);
2011                                 return;
2012                         }
2013
2014                         update_lft = create = 1;
2015                         ifp->cstamp = jiffies;
2016                         addrconf_dad_start(ifp);
2017                 }
2018
2019                 if (ifp) {
2020                         int flags;
2021                         unsigned long now;
2022 #ifdef CONFIG_IPV6_PRIVACY
2023                         struct inet6_ifaddr *ift;
2024 #endif
2025                         u32 stored_lft;
2026
2027                         /* update lifetime (RFC2462 5.5.3 e) */
2028                         spin_lock(&ifp->lock);
2029                         now = jiffies;
2030                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2031                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2032                         else
2033                                 stored_lft = 0;
2034                         if (!update_lft && stored_lft) {
2035                                 if (valid_lft > MIN_VALID_LIFETIME ||
2036                                     valid_lft > stored_lft)
2037                                         update_lft = 1;
2038                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2039                                         /* valid_lft <= stored_lft is always true */
2040                                         /*
2041                                          * RFC 4862 Section 5.5.3e:
2042                                          * "Note that the preferred lifetime of
2043                                          *  the corresponding address is always
2044                                          *  reset to the Preferred Lifetime in
2045                                          *  the received Prefix Information
2046                                          *  option, regardless of whether the
2047                                          *  valid lifetime is also reset or
2048                                          *  ignored."
2049                                          *
2050                                          *  So if the preferred lifetime in
2051                                          *  this advertisement is different
2052                                          *  than what we have stored, but the
2053                                          *  valid lifetime is invalid, just
2054                                          *  reset prefered_lft.
2055                                          *
2056                                          *  We must set the valid lifetime
2057                                          *  to the stored lifetime since we'll
2058                                          *  be updating the timestamp below,
2059                                          *  else we'll set it back to the
2060                                          *  minimum.
2061                                          */
2062                                         if (prefered_lft != ifp->prefered_lft) {
2063                                                 valid_lft = stored_lft;
2064                                                 update_lft = 1;
2065                                         }
2066                                 } else {
2067                                         valid_lft = MIN_VALID_LIFETIME;
2068                                         if (valid_lft < prefered_lft)
2069                                                 prefered_lft = valid_lft;
2070                                         update_lft = 1;
2071                                 }
2072                         }
2073
2074                         if (update_lft) {
2075                                 ifp->valid_lft = valid_lft;
2076                                 ifp->prefered_lft = prefered_lft;
2077                                 ifp->tstamp = now;
2078                                 flags = ifp->flags;
2079                                 ifp->flags &= ~IFA_F_DEPRECATED;
2080                                 spin_unlock(&ifp->lock);
2081
2082                                 if (!(flags&IFA_F_TENTATIVE))
2083                                         ipv6_ifa_notify(0, ifp);
2084                         } else
2085                                 spin_unlock(&ifp->lock);
2086
2087 #ifdef CONFIG_IPV6_PRIVACY
2088                         read_lock_bh(&in6_dev->lock);
2089                         /* update all temporary addresses in the list */
2090                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2091                                             tmp_list) {
2092                                 int age, max_valid, max_prefered;
2093
2094                                 if (ifp != ift->ifpub)
2095                                         continue;
2096
2097                                 /*
2098                                  * RFC 4941 section 3.3:
2099                                  * If a received option will extend the lifetime
2100                                  * of a public address, the lifetimes of
2101                                  * temporary addresses should be extended,
2102                                  * subject to the overall constraint that no
2103                                  * temporary addresses should ever remain
2104                                  * "valid" or "preferred" for a time longer than
2105                                  * (TEMP_VALID_LIFETIME) or
2106                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2107                                  * respectively.
2108                                  */
2109                                 age = (now - ift->cstamp) / HZ;
2110                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2111                                 if (max_valid < 0)
2112                                         max_valid = 0;
2113
2114                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2115                                                in6_dev->cnf.max_desync_factor -
2116                                                age;
2117                                 if (max_prefered < 0)
2118                                         max_prefered = 0;
2119
2120                                 if (valid_lft > max_valid)
2121                                         valid_lft = max_valid;
2122
2123                                 if (prefered_lft > max_prefered)
2124                                         prefered_lft = max_prefered;
2125
2126                                 spin_lock(&ift->lock);
2127                                 flags = ift->flags;
2128                                 ift->valid_lft = valid_lft;
2129                                 ift->prefered_lft = prefered_lft;
2130                                 ift->tstamp = now;
2131                                 if (prefered_lft > 0)
2132                                         ift->flags &= ~IFA_F_DEPRECATED;
2133
2134                                 spin_unlock(&ift->lock);
2135                                 if (!(flags&IFA_F_TENTATIVE))
2136                                         ipv6_ifa_notify(0, ift);
2137                         }
2138
2139                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2140                                 /*
2141                                  * When a new public address is created as
2142                                  * described in [ADDRCONF], also create a new
2143                                  * temporary address. Also create a temporary
2144                                  * address if it's enabled but no temporary
2145                                  * address currently exists.
2146                                  */
2147                                 read_unlock_bh(&in6_dev->lock);
2148                                 ipv6_create_tempaddr(ifp, NULL);
2149                         } else {
2150                                 read_unlock_bh(&in6_dev->lock);
2151                         }
2152 #endif
2153                         in6_ifa_put(ifp);
2154                         addrconf_verify(0);
2155                 }
2156         }
2157         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2158         in6_dev_put(in6_dev);
2159 }
2160
2161 /*
2162  *      Set destination address.
2163  *      Special case for SIT interfaces where we create a new "virtual"
2164  *      device.
2165  */
2166 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2167 {
2168         struct in6_ifreq ireq;
2169         struct net_device *dev;
2170         int err = -EINVAL;
2171
2172         rtnl_lock();
2173
2174         err = -EFAULT;
2175         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2176                 goto err_exit;
2177
2178         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2179
2180         err = -ENODEV;
2181         if (dev == NULL)
2182                 goto err_exit;
2183
2184 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2185         if (dev->type == ARPHRD_SIT) {
2186                 const struct net_device_ops *ops = dev->netdev_ops;
2187                 struct ifreq ifr;
2188                 struct ip_tunnel_parm p;
2189
2190                 err = -EADDRNOTAVAIL;
2191                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2192                         goto err_exit;
2193
2194                 memset(&p, 0, sizeof(p));
2195                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2196                 p.iph.saddr = 0;
2197                 p.iph.version = 4;
2198                 p.iph.ihl = 5;
2199                 p.iph.protocol = IPPROTO_IPV6;
2200                 p.iph.ttl = 64;
2201                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2202
2203                 if (ops->ndo_do_ioctl) {
2204                         mm_segment_t oldfs = get_fs();
2205
2206                         set_fs(KERNEL_DS);
2207                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2208                         set_fs(oldfs);
2209                 } else
2210                         err = -EOPNOTSUPP;
2211
2212                 if (err == 0) {
2213                         err = -ENOBUFS;
2214                         dev = __dev_get_by_name(net, p.name);
2215                         if (!dev)
2216                                 goto err_exit;
2217                         err = dev_open(dev);
2218                 }
2219         }
2220 #endif
2221
2222 err_exit:
2223         rtnl_unlock();
2224         return err;
2225 }
2226
2227 /*
2228  *      Manual configuration of address on an interface
2229  */
2230 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2231                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2232                           __u32 valid_lft)
2233 {
2234         struct inet6_ifaddr *ifp;
2235         struct inet6_dev *idev;
2236         struct net_device *dev;
2237         int scope;
2238         u32 flags;
2239         clock_t expires;
2240         unsigned long timeout;
2241
2242         ASSERT_RTNL();
2243
2244         if (plen > 128)
2245                 return -EINVAL;
2246
2247         /* check the lifetime */
2248         if (!valid_lft || prefered_lft > valid_lft)
2249                 return -EINVAL;
2250
2251         dev = __dev_get_by_index(net, ifindex);
2252         if (!dev)
2253                 return -ENODEV;
2254
2255         idev = addrconf_add_dev(dev);
2256         if (IS_ERR(idev))
2257                 return PTR_ERR(idev);
2258
2259         scope = ipv6_addr_scope(pfx);
2260
2261         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2262         if (addrconf_finite_timeout(timeout)) {
2263                 expires = jiffies_to_clock_t(timeout * HZ);
2264                 valid_lft = timeout;
2265                 flags = RTF_EXPIRES;
2266         } else {
2267                 expires = 0;
2268                 flags = 0;
2269                 ifa_flags |= IFA_F_PERMANENT;
2270         }
2271
2272         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2273         if (addrconf_finite_timeout(timeout)) {
2274                 if (timeout == 0)
2275                         ifa_flags |= IFA_F_DEPRECATED;
2276                 prefered_lft = timeout;
2277         }
2278
2279         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2280
2281         if (!IS_ERR(ifp)) {
2282                 spin_lock_bh(&ifp->lock);
2283                 ifp->valid_lft = valid_lft;
2284                 ifp->prefered_lft = prefered_lft;
2285                 ifp->tstamp = jiffies;
2286                 spin_unlock_bh(&ifp->lock);
2287
2288                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2289                                       expires, flags);
2290                 /*
2291                  * Note that section 3.1 of RFC 4429 indicates
2292                  * that the Optimistic flag should not be set for
2293                  * manually configured addresses
2294                  */
2295                 addrconf_dad_start(ifp);
2296                 in6_ifa_put(ifp);
2297                 addrconf_verify(0);
2298                 return 0;
2299         }
2300
2301         return PTR_ERR(ifp);
2302 }
2303
2304 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2305                           unsigned int plen)
2306 {
2307         struct inet6_ifaddr *ifp;
2308         struct inet6_dev *idev;
2309         struct net_device *dev;
2310
2311         if (plen > 128)
2312                 return -EINVAL;
2313
2314         dev = __dev_get_by_index(net, ifindex);
2315         if (!dev)
2316                 return -ENODEV;
2317
2318         if ((idev = __in6_dev_get(dev)) == NULL)
2319                 return -ENXIO;
2320
2321         read_lock_bh(&idev->lock);
2322         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2323                 if (ifp->prefix_len == plen &&
2324                     ipv6_addr_equal(pfx, &ifp->addr)) {
2325                         in6_ifa_hold(ifp);
2326                         read_unlock_bh(&idev->lock);
2327
2328                         ipv6_del_addr(ifp);
2329
2330                         /* If the last address is deleted administratively,
2331                            disable IPv6 on this interface.
2332                          */
2333                         if (list_empty(&idev->addr_list))
2334                                 addrconf_ifdown(idev->dev, 1);
2335                         return 0;
2336                 }
2337         }
2338         read_unlock_bh(&idev->lock);
2339         return -EADDRNOTAVAIL;
2340 }
2341
2342
2343 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2344 {
2345         struct in6_ifreq ireq;
2346         int err;
2347
2348         if (!capable(CAP_NET_ADMIN))
2349                 return -EPERM;
2350
2351         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2352                 return -EFAULT;
2353
2354         rtnl_lock();
2355         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2356                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2357                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2358         rtnl_unlock();
2359         return err;
2360 }
2361
2362 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2363 {
2364         struct in6_ifreq ireq;
2365         int err;
2366
2367         if (!capable(CAP_NET_ADMIN))
2368                 return -EPERM;
2369
2370         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2371                 return -EFAULT;
2372
2373         rtnl_lock();
2374         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2375                              ireq.ifr6_prefixlen);
2376         rtnl_unlock();
2377         return err;
2378 }
2379
2380 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2381                      int plen, int scope)
2382 {
2383         struct inet6_ifaddr *ifp;
2384
2385         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2386         if (!IS_ERR(ifp)) {
2387                 spin_lock_bh(&ifp->lock);
2388                 ifp->flags &= ~IFA_F_TENTATIVE;
2389                 spin_unlock_bh(&ifp->lock);
2390                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2391                 in6_ifa_put(ifp);
2392         }
2393 }
2394
2395 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2396 static void sit_add_v4_addrs(struct inet6_dev *idev)
2397 {
2398         struct in6_addr addr;
2399         struct net_device *dev;
2400         struct net *net = dev_net(idev->dev);
2401         int scope;
2402
2403         ASSERT_RTNL();
2404
2405         memset(&addr, 0, sizeof(struct in6_addr));
2406         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2407
2408         if (idev->dev->flags&IFF_POINTOPOINT) {
2409                 addr.s6_addr32[0] = htonl(0xfe800000);
2410                 scope = IFA_LINK;
2411         } else {
2412                 scope = IPV6_ADDR_COMPATv4;
2413         }
2414
2415         if (addr.s6_addr32[3]) {
2416                 add_addr(idev, &addr, 128, scope);
2417                 return;
2418         }
2419
2420         for_each_netdev(net, dev) {
2421                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2422                 if (in_dev && (dev->flags & IFF_UP)) {
2423                         struct in_ifaddr *ifa;
2424
2425                         int flag = scope;
2426
2427                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2428                                 int plen;
2429
2430                                 addr.s6_addr32[3] = ifa->ifa_local;
2431
2432                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2433                                         continue;
2434                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2435                                         if (idev->dev->flags&IFF_POINTOPOINT)
2436                                                 continue;
2437                                         flag |= IFA_HOST;
2438                                 }
2439                                 if (idev->dev->flags&IFF_POINTOPOINT)
2440                                         plen = 64;
2441                                 else
2442                                         plen = 96;
2443
2444                                 add_addr(idev, &addr, plen, flag);
2445                         }
2446                 }
2447         }
2448 }
2449 #endif
2450
2451 static void init_loopback(struct net_device *dev)
2452 {
2453         struct inet6_dev  *idev;
2454
2455         /* ::1 */
2456
2457         ASSERT_RTNL();
2458
2459         if ((idev = ipv6_find_idev(dev)) == NULL) {
2460                 pr_debug("%s: add_dev failed\n", __func__);
2461                 return;
2462         }
2463
2464         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2465 }
2466
2467 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2468 {
2469         struct inet6_ifaddr *ifp;
2470         u32 addr_flags = IFA_F_PERMANENT;
2471
2472 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2473         if (idev->cnf.optimistic_dad &&
2474             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2475                 addr_flags |= IFA_F_OPTIMISTIC;
2476 #endif
2477
2478
2479         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2480         if (!IS_ERR(ifp)) {
2481                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2482                 addrconf_dad_start(ifp);
2483                 in6_ifa_put(ifp);
2484         }
2485 }
2486
2487 static void addrconf_dev_config(struct net_device *dev)
2488 {
2489         struct in6_addr addr;
2490         struct inet6_dev *idev;
2491
2492         ASSERT_RTNL();
2493
2494         if ((dev->type != ARPHRD_ETHER) &&
2495             (dev->type != ARPHRD_FDDI) &&
2496             (dev->type != ARPHRD_ARCNET) &&
2497             (dev->type != ARPHRD_INFINIBAND) &&
2498             (dev->type != ARPHRD_IEEE802154)) {
2499                 /* Alas, we support only Ethernet autoconfiguration. */
2500                 return;
2501         }
2502
2503         idev = addrconf_add_dev(dev);
2504         if (IS_ERR(idev))
2505                 return;
2506
2507         memset(&addr, 0, sizeof(struct in6_addr));
2508         addr.s6_addr32[0] = htonl(0xFE800000);
2509
2510         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2511                 addrconf_add_linklocal(idev, &addr);
2512 }
2513
2514 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2515 static void addrconf_sit_config(struct net_device *dev)
2516 {
2517         struct inet6_dev *idev;
2518
2519         ASSERT_RTNL();
2520
2521         /*
2522          * Configure the tunnel with one of our IPv4
2523          * addresses... we should configure all of
2524          * our v4 addrs in the tunnel
2525          */
2526
2527         if ((idev = ipv6_find_idev(dev)) == NULL) {
2528                 pr_debug("%s: add_dev failed\n", __func__);
2529                 return;
2530         }
2531
2532         if (dev->priv_flags & IFF_ISATAP) {
2533                 struct in6_addr addr;
2534
2535                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2536                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2537                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2538                         addrconf_add_linklocal(idev, &addr);
2539                 return;
2540         }
2541
2542         sit_add_v4_addrs(idev);
2543
2544         if (dev->flags&IFF_POINTOPOINT)
2545                 addrconf_add_mroute(dev);
2546         else
2547                 sit_route_add(dev);
2548 }
2549 #endif
2550
2551 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2552 static void addrconf_gre_config(struct net_device *dev)
2553 {
2554         struct inet6_dev *idev;
2555         struct in6_addr addr;
2556
2557         pr_info("%s(%s)\n", __func__, dev->name);
2558
2559         ASSERT_RTNL();
2560
2561         if ((idev = ipv6_find_idev(dev)) == NULL) {
2562                 pr_debug("%s: add_dev failed\n", __func__);
2563                 return;
2564         }
2565
2566         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2567         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2568
2569         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2570                 addrconf_add_linklocal(idev, &addr);
2571 }
2572 #endif
2573
2574 static inline int
2575 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2576 {
2577         struct in6_addr lladdr;
2578
2579         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2580                 addrconf_add_linklocal(idev, &lladdr);
2581                 return 0;
2582         }
2583         return -1;
2584 }
2585
2586 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2587 {
2588         struct net_device *link_dev;
2589         struct net *net = dev_net(idev->dev);
2590
2591         /* first try to inherit the link-local address from the link device */
2592         if (idev->dev->iflink &&
2593             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2594                 if (!ipv6_inherit_linklocal(idev, link_dev))
2595                         return;
2596         }
2597         /* then try to inherit it from any device */
2598         for_each_netdev(net, link_dev) {
2599                 if (!ipv6_inherit_linklocal(idev, link_dev))
2600                         return;
2601         }
2602         pr_debug("init ip6-ip6: add_linklocal failed\n");
2603 }
2604
2605 /*
2606  * Autoconfigure tunnel with a link-local address so routing protocols,
2607  * DHCPv6, MLD etc. can be run over the virtual link
2608  */
2609
2610 static void addrconf_ip6_tnl_config(struct net_device *dev)
2611 {
2612         struct inet6_dev *idev;
2613
2614         ASSERT_RTNL();
2615
2616         idev = addrconf_add_dev(dev);
2617         if (IS_ERR(idev)) {
2618                 pr_debug("init ip6-ip6: add_dev failed\n");
2619                 return;
2620         }
2621         ip6_tnl_add_linklocal(idev);
2622 }
2623
2624 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2625                            void *data)
2626 {
2627         struct net_device *dev = (struct net_device *) data;
2628         struct inet6_dev *idev = __in6_dev_get(dev);
2629         int run_pending = 0;
2630         int err;
2631
2632         switch (event) {
2633         case NETDEV_REGISTER:
2634                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2635                         idev = ipv6_add_dev(dev);
2636                         if (!idev)
2637                                 return notifier_from_errno(-ENOMEM);
2638                 }
2639                 break;
2640
2641         case NETDEV_UP:
2642         case NETDEV_CHANGE:
2643                 if (dev->flags & IFF_SLAVE)
2644                         break;
2645
2646                 if (event == NETDEV_UP) {
2647                         if (!addrconf_qdisc_ok(dev)) {
2648                                 /* device is not ready yet. */
2649                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2650                                         dev->name);
2651                                 break;
2652                         }
2653
2654                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2655                                 idev = ipv6_add_dev(dev);
2656
2657                         if (idev) {
2658                                 idev->if_flags |= IF_READY;
2659                                 run_pending = 1;
2660                         }
2661                 } else {
2662                         if (!addrconf_qdisc_ok(dev)) {
2663                                 /* device is still not ready. */
2664                                 break;
2665                         }
2666
2667                         if (idev) {
2668                                 if (idev->if_flags & IF_READY)
2669                                         /* device is already configured. */
2670                                         break;
2671                                 idev->if_flags |= IF_READY;
2672                         }
2673
2674                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2675                                 dev->name);
2676
2677                         run_pending = 1;
2678                 }
2679
2680                 switch (dev->type) {
2681 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2682                 case ARPHRD_SIT:
2683                         addrconf_sit_config(dev);
2684                         break;
2685 #endif
2686 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2687                 case ARPHRD_IPGRE:
2688                         addrconf_gre_config(dev);
2689                         break;
2690 #endif
2691                 case ARPHRD_TUNNEL6:
2692                         addrconf_ip6_tnl_config(dev);
2693                         break;
2694                 case ARPHRD_LOOPBACK:
2695                         init_loopback(dev);
2696                         break;
2697
2698                 default:
2699                         addrconf_dev_config(dev);
2700                         break;
2701                 }
2702
2703                 if (idev) {
2704                         if (run_pending)
2705                                 addrconf_dad_run(idev);
2706
2707                         /*
2708                          * If the MTU changed during the interface down,
2709                          * when the interface up, the changed MTU must be
2710                          * reflected in the idev as well as routers.
2711                          */
2712                         if (idev->cnf.mtu6 != dev->mtu &&
2713                             dev->mtu >= IPV6_MIN_MTU) {
2714                                 rt6_mtu_change(dev, dev->mtu);
2715                                 idev->cnf.mtu6 = dev->mtu;
2716                         }
2717                         idev->tstamp = jiffies;
2718                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2719
2720                         /*
2721                          * If the changed mtu during down is lower than
2722                          * IPV6_MIN_MTU stop IPv6 on this interface.
2723                          */
2724                         if (dev->mtu < IPV6_MIN_MTU)
2725                                 addrconf_ifdown(dev, 1);
2726                 }
2727                 break;
2728
2729         case NETDEV_CHANGEMTU:
2730                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2731                         rt6_mtu_change(dev, dev->mtu);
2732                         idev->cnf.mtu6 = dev->mtu;
2733                         break;
2734                 }
2735
2736                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2737                         idev = ipv6_add_dev(dev);
2738                         if (idev)
2739                                 break;
2740                 }
2741
2742                 /*
2743                  * MTU falled under IPV6_MIN_MTU.
2744                  * Stop IPv6 on this interface.
2745                  */
2746
2747         case NETDEV_DOWN:
2748         case NETDEV_UNREGISTER:
2749                 /*
2750                  *      Remove all addresses from this interface.
2751                  */
2752                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2753                 break;
2754
2755         case NETDEV_CHANGENAME:
2756                 if (idev) {
2757                         snmp6_unregister_dev(idev);
2758                         addrconf_sysctl_unregister(idev);
2759                         addrconf_sysctl_register(idev);
2760                         err = snmp6_register_dev(idev);
2761                         if (err)
2762                                 return notifier_from_errno(err);
2763                 }
2764                 break;
2765
2766         case NETDEV_PRE_TYPE_CHANGE:
2767         case NETDEV_POST_TYPE_CHANGE:
2768                 addrconf_type_change(dev, event);
2769                 break;
2770         }
2771
2772         return NOTIFY_OK;
2773 }
2774
2775 /*
2776  *      addrconf module should be notified of a device going up
2777  */
2778 static struct notifier_block ipv6_dev_notf = {
2779         .notifier_call = addrconf_notify,
2780 };
2781
2782 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2783 {
2784         struct inet6_dev *idev;
2785         ASSERT_RTNL();
2786
2787         idev = __in6_dev_get(dev);
2788
2789         if (event == NETDEV_POST_TYPE_CHANGE)
2790                 ipv6_mc_remap(idev);
2791         else if (event == NETDEV_PRE_TYPE_CHANGE)
2792                 ipv6_mc_unmap(idev);
2793 }
2794
2795 static int addrconf_ifdown(struct net_device *dev, int how)
2796 {
2797         struct net *net = dev_net(dev);
2798         struct inet6_dev *idev;
2799         struct inet6_ifaddr *ifa;
2800         int state, i;
2801
2802         ASSERT_RTNL();
2803
2804         rt6_ifdown(net, dev);
2805         neigh_ifdown(&nd_tbl, dev);
2806
2807         idev = __in6_dev_get(dev);
2808         if (idev == NULL)
2809                 return -ENODEV;
2810
2811         /*
2812          * Step 1: remove reference to ipv6 device from parent device.
2813          *         Do not dev_put!
2814          */
2815         if (how) {
2816                 idev->dead = 1;
2817
2818                 /* protected by rtnl_lock */
2819                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2820
2821                 /* Step 1.5: remove snmp6 entry */
2822                 snmp6_unregister_dev(idev);
2823
2824         }
2825
2826         /* Step 2: clear hash table */
2827         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2828                 struct hlist_head *h = &inet6_addr_lst[i];
2829                 struct hlist_node *n;
2830
2831                 spin_lock_bh(&addrconf_hash_lock);
2832         restart:
2833                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2834                         if (ifa->idev == idev) {
2835                                 hlist_del_init_rcu(&ifa->addr_lst);
2836                                 addrconf_del_timer(ifa);
2837                                 goto restart;
2838                         }
2839                 }
2840                 spin_unlock_bh(&addrconf_hash_lock);
2841         }
2842
2843         write_lock_bh(&idev->lock);
2844
2845         /* Step 2: clear flags for stateless addrconf */
2846         if (!how)
2847                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2848
2849 #ifdef CONFIG_IPV6_PRIVACY
2850         if (how && del_timer(&idev->regen_timer))
2851                 in6_dev_put(idev);
2852
2853         /* Step 3: clear tempaddr list */
2854         while (!list_empty(&idev->tempaddr_list)) {
2855                 ifa = list_first_entry(&idev->tempaddr_list,
2856                                        struct inet6_ifaddr, tmp_list);
2857                 list_del(&ifa->tmp_list);
2858                 write_unlock_bh(&idev->lock);
2859                 spin_lock_bh(&ifa->lock);
2860
2861                 if (ifa->ifpub) {
2862                         in6_ifa_put(ifa->ifpub);
2863                         ifa->ifpub = NULL;
2864                 }
2865                 spin_unlock_bh(&ifa->lock);
2866                 in6_ifa_put(ifa);
2867                 write_lock_bh(&idev->lock);
2868         }
2869 #endif
2870
2871         while (!list_empty(&idev->addr_list)) {
2872                 ifa = list_first_entry(&idev->addr_list,
2873                                        struct inet6_ifaddr, if_list);
2874                 addrconf_del_timer(ifa);
2875
2876                 list_del(&ifa->if_list);
2877
2878                 write_unlock_bh(&idev->lock);
2879
2880                 spin_lock_bh(&ifa->state_lock);
2881                 state = ifa->state;
2882                 ifa->state = INET6_IFADDR_STATE_DEAD;
2883                 spin_unlock_bh(&ifa->state_lock);
2884
2885                 if (state != INET6_IFADDR_STATE_DEAD) {
2886                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2887                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2888                 }
2889                 in6_ifa_put(ifa);
2890
2891                 write_lock_bh(&idev->lock);
2892         }
2893
2894         write_unlock_bh(&idev->lock);
2895
2896         /* Step 5: Discard multicast list */
2897         if (how)
2898                 ipv6_mc_destroy_dev(idev);
2899         else
2900                 ipv6_mc_down(idev);
2901
2902         idev->tstamp = jiffies;
2903
2904         /* Last: Shot the device (if unregistered) */
2905         if (how) {
2906                 addrconf_sysctl_unregister(idev);
2907                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2908                 neigh_ifdown(&nd_tbl, dev);
2909                 in6_dev_put(idev);
2910         }
2911         return 0;
2912 }
2913
2914 static void addrconf_rs_timer(unsigned long data)
2915 {
2916         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2917         struct inet6_dev *idev = ifp->idev;
2918
2919         read_lock(&idev->lock);
2920         if (idev->dead || !(idev->if_flags & IF_READY))
2921                 goto out;
2922
2923         if (idev->cnf.forwarding)
2924                 goto out;
2925
2926         /* Announcement received after solicitation was sent */
2927         if (idev->if_flags & IF_RA_RCVD)
2928                 goto out;
2929
2930         spin_lock(&ifp->lock);
2931         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2932                 /* The wait after the last probe can be shorter */
2933                 addrconf_mod_timer(ifp, AC_RS,
2934                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2935                                    idev->cnf.rtr_solicit_delay :
2936                                    idev->cnf.rtr_solicit_interval);
2937                 spin_unlock(&ifp->lock);
2938
2939                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2940         } else {
2941                 spin_unlock(&ifp->lock);
2942                 /*
2943                  * Note: we do not support deprecated "all on-link"
2944                  * assumption any longer.
2945                  */
2946                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
2947         }
2948
2949 out:
2950         read_unlock(&idev->lock);
2951         in6_ifa_put(ifp);
2952 }
2953
2954 /*
2955  *      Duplicate Address Detection
2956  */
2957 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2958 {
2959         unsigned long rand_num;
2960         struct inet6_dev *idev = ifp->idev;
2961
2962         if (ifp->flags & IFA_F_OPTIMISTIC)
2963                 rand_num = 0;
2964         else
2965                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2966
2967         ifp->probes = idev->cnf.dad_transmits;
2968         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2969 }
2970
2971 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
2972 {
2973         struct inet6_dev *idev = ifp->idev;
2974         struct net_device *dev = idev->dev;
2975
2976         addrconf_join_solict(dev, &ifp->addr);
2977
2978         net_srandom(ifp->addr.s6_addr32[3]);
2979
2980         read_lock_bh(&idev->lock);
2981         spin_lock(&ifp->lock);
2982         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2983                 goto out;
2984
2985         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2986             idev->cnf.accept_dad < 1 ||
2987             !(ifp->flags&IFA_F_TENTATIVE) ||
2988             ifp->flags & IFA_F_NODAD) {
2989                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2990                 spin_unlock(&ifp->lock);
2991                 read_unlock_bh(&idev->lock);
2992
2993                 addrconf_dad_completed(ifp);
2994                 return;
2995         }
2996
2997         if (!(idev->if_flags & IF_READY)) {
2998                 spin_unlock(&ifp->lock);
2999                 read_unlock_bh(&idev->lock);
3000                 /*
3001                  * If the device is not ready:
3002                  * - keep it tentative if it is a permanent address.
3003                  * - otherwise, kill it.
3004                  */
3005                 in6_ifa_hold(ifp);
3006                 addrconf_dad_stop(ifp, 0);
3007                 return;
3008         }
3009
3010         /*
3011          * Optimistic nodes can start receiving
3012          * Frames right away
3013          */
3014         if (ifp->flags & IFA_F_OPTIMISTIC)
3015                 ip6_ins_rt(ifp->rt);
3016
3017         addrconf_dad_kick(ifp);
3018 out:
3019         spin_unlock(&ifp->lock);
3020         read_unlock_bh(&idev->lock);
3021 }
3022
3023 static void addrconf_dad_timer(unsigned long data)
3024 {
3025         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3026         struct inet6_dev *idev = ifp->idev;
3027         struct in6_addr mcaddr;
3028
3029         if (!ifp->probes && addrconf_dad_end(ifp))
3030                 goto out;
3031
3032         read_lock(&idev->lock);
3033         if (idev->dead || !(idev->if_flags & IF_READY)) {
3034                 read_unlock(&idev->lock);
3035                 goto out;
3036         }
3037
3038         spin_lock(&ifp->lock);
3039         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3040                 spin_unlock(&ifp->lock);
3041                 read_unlock(&idev->lock);
3042                 goto out;
3043         }
3044
3045         if (ifp->probes == 0) {
3046                 /*
3047                  * DAD was successful
3048                  */
3049
3050                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3051                 spin_unlock(&ifp->lock);
3052                 read_unlock(&idev->lock);
3053
3054                 addrconf_dad_completed(ifp);
3055
3056                 goto out;
3057         }
3058
3059         ifp->probes--;
3060         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3061         spin_unlock(&ifp->lock);
3062         read_unlock(&idev->lock);
3063
3064         /* send a neighbour solicitation for our addr */
3065         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3066         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3067 out:
3068         in6_ifa_put(ifp);
3069 }
3070
3071 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3072 {
3073         struct net_device *dev = ifp->idev->dev;
3074
3075         /*
3076          *      Configure the address for reception. Now it is valid.
3077          */
3078
3079         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3080
3081         /* If added prefix is link local and we are prepared to process
3082            router advertisements, start sending router solicitations.
3083          */
3084
3085         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3086              ifp->idev->cnf.accept_ra == 2) &&
3087             ifp->idev->cnf.rtr_solicits > 0 &&
3088             (dev->flags&IFF_LOOPBACK) == 0 &&
3089             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3090                 /*
3091                  *      If a host as already performed a random delay
3092                  *      [...] as part of DAD [...] there is no need
3093                  *      to delay again before sending the first RS
3094                  */
3095                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3096
3097                 spin_lock_bh(&ifp->lock);
3098                 ifp->probes = 1;
3099                 ifp->idev->if_flags |= IF_RS_SENT;
3100                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3101                 spin_unlock_bh(&ifp->lock);
3102         }
3103 }
3104
3105 static void addrconf_dad_run(struct inet6_dev *idev)
3106 {
3107         struct inet6_ifaddr *ifp;
3108
3109         read_lock_bh(&idev->lock);
3110         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3111                 spin_lock(&ifp->lock);
3112                 if (ifp->flags & IFA_F_TENTATIVE &&
3113                     ifp->state == INET6_IFADDR_STATE_DAD)
3114                         addrconf_dad_kick(ifp);
3115                 spin_unlock(&ifp->lock);
3116         }
3117         read_unlock_bh(&idev->lock);
3118 }
3119
3120 #ifdef CONFIG_PROC_FS
3121 struct if6_iter_state {
3122         struct seq_net_private p;
3123         int bucket;
3124         int offset;
3125 };
3126
3127 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3128 {
3129         struct inet6_ifaddr *ifa = NULL;
3130         struct if6_iter_state *state = seq->private;
3131         struct net *net = seq_file_net(seq);
3132         int p = 0;
3133
3134         /* initial bucket if pos is 0 */
3135         if (pos == 0) {
3136                 state->bucket = 0;
3137                 state->offset = 0;
3138         }
3139
3140         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3141                 struct hlist_node *n;
3142                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3143                                          addr_lst) {
3144                         if (!net_eq(dev_net(ifa->idev->dev), net))
3145                                 continue;
3146                         /* sync with offset */
3147                         if (p < state->offset) {
3148                                 p++;
3149                                 continue;
3150                         }
3151                         state->offset++;
3152                         return ifa;
3153                 }
3154
3155                 /* prepare for next bucket */
3156                 state->offset = 0;
3157                 p = 0;
3158         }
3159         return NULL;
3160 }
3161
3162 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3163                                          struct inet6_ifaddr *ifa)
3164 {
3165         struct if6_iter_state *state = seq->private;
3166         struct net *net = seq_file_net(seq);
3167         struct hlist_node *n = &ifa->addr_lst;
3168
3169         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3170                 if (!net_eq(dev_net(ifa->idev->dev), net))
3171                         continue;
3172                 state->offset++;
3173                 return ifa;
3174         }
3175
3176         while (++state->bucket < IN6_ADDR_HSIZE) {
3177                 state->offset = 0;
3178                 hlist_for_each_entry_rcu_bh(ifa, n,
3179                                      &inet6_addr_lst[state->bucket], addr_lst) {
3180                         if (!net_eq(dev_net(ifa->idev->dev), net))
3181                                 continue;
3182                         state->offset++;
3183                         return ifa;
3184                 }
3185         }
3186
3187         return NULL;
3188 }
3189
3190 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3191         __acquires(rcu_bh)
3192 {
3193         rcu_read_lock_bh();
3194         return if6_get_first(seq, *pos);
3195 }
3196
3197 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3198 {
3199         struct inet6_ifaddr *ifa;
3200
3201         ifa = if6_get_next(seq, v);
3202         ++*pos;
3203         return ifa;
3204 }
3205
3206 static void if6_seq_stop(struct seq_file *seq, void *v)
3207         __releases(rcu_bh)
3208 {
3209         rcu_read_unlock_bh();
3210 }
3211
3212 static int if6_seq_show(struct seq_file *seq, void *v)
3213 {
3214         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3215         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3216                    &ifp->addr,
3217                    ifp->idev->dev->ifindex,
3218                    ifp->prefix_len,
3219                    ifp->scope,
3220                    ifp->flags,
3221                    ifp->idev->dev->name);
3222         return 0;
3223 }
3224
3225 static const struct seq_operations if6_seq_ops = {
3226         .start  = if6_seq_start,
3227         .next   = if6_seq_next,
3228         .show   = if6_seq_show,
3229         .stop   = if6_seq_stop,
3230 };
3231
3232 static int if6_seq_open(struct inode *inode, struct file *file)
3233 {
3234         return seq_open_net(inode, file, &if6_seq_ops,
3235                             sizeof(struct if6_iter_state));
3236 }
3237
3238 static const struct file_operations if6_fops = {
3239         .owner          = THIS_MODULE,
3240         .open           = if6_seq_open,
3241         .read           = seq_read,
3242         .llseek         = seq_lseek,
3243         .release        = seq_release_net,
3244 };
3245
3246 static int __net_init if6_proc_net_init(struct net *net)
3247 {
3248         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3249                 return -ENOMEM;
3250         return 0;
3251 }
3252
3253 static void __net_exit if6_proc_net_exit(struct net *net)
3254 {
3255        proc_net_remove(net, "if_inet6");
3256 }
3257
3258 static struct pernet_operations if6_proc_net_ops = {
3259        .init = if6_proc_net_init,
3260        .exit = if6_proc_net_exit,
3261 };
3262
3263 int __init if6_proc_init(void)
3264 {
3265         return register_pernet_subsys(&if6_proc_net_ops);
3266 }
3267
3268 void if6_proc_exit(void)
3269 {
3270         unregister_pernet_subsys(&if6_proc_net_ops);
3271 }
3272 #endif  /* CONFIG_PROC_FS */
3273
3274 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3275 /* Check if address is a home address configured on any interface. */
3276 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3277 {
3278         int ret = 0;
3279         struct inet6_ifaddr *ifp = NULL;
3280         struct hlist_node *n;
3281         unsigned int hash = inet6_addr_hash(addr);
3282
3283         rcu_read_lock_bh();
3284         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3285                 if (!net_eq(dev_net(ifp->idev->dev), net))
3286                         continue;
3287                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3288                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3289                         ret = 1;
3290                         break;
3291                 }
3292         }
3293         rcu_read_unlock_bh();
3294         return ret;
3295 }
3296 #endif
3297
3298 /*
3299  *      Periodic address status verification
3300  */
3301
3302 static void addrconf_verify(unsigned long foo)
3303 {
3304         unsigned long now, next, next_sec, next_sched;
3305         struct inet6_ifaddr *ifp;
3306         struct hlist_node *node;
3307         int i;
3308
3309         rcu_read_lock_bh();
3310         spin_lock(&addrconf_verify_lock);
3311         now = jiffies;
3312         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3313
3314         del_timer(&addr_chk_timer);
3315
3316         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3317 restart:
3318                 hlist_for_each_entry_rcu_bh(ifp, node,
3319                                          &inet6_addr_lst[i], addr_lst) {
3320                         unsigned long age;
3321
3322                         if (ifp->flags & IFA_F_PERMANENT)
3323                                 continue;
3324
3325                         spin_lock(&ifp->lock);
3326                         /* We try to batch several events at once. */
3327                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3328
3329                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3330                             age >= ifp->valid_lft) {
3331                                 spin_unlock(&ifp->lock);
3332                                 in6_ifa_hold(ifp);
3333                                 ipv6_del_addr(ifp);
3334                                 goto restart;
3335                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3336                                 spin_unlock(&ifp->lock);
3337                                 continue;
3338                         } else if (age >= ifp->prefered_lft) {
3339                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3340                                 int deprecate = 0;
3341
3342                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3343                                         deprecate = 1;
3344                                         ifp->flags |= IFA_F_DEPRECATED;
3345                                 }
3346
3347                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3348                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3349
3350                                 spin_unlock(&ifp->lock);
3351
3352                                 if (deprecate) {
3353                                         in6_ifa_hold(ifp);
3354
3355                                         ipv6_ifa_notify(0, ifp);
3356                                         in6_ifa_put(ifp);
3357                                         goto restart;
3358                                 }
3359 #ifdef CONFIG_IPV6_PRIVACY
3360                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3361                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3362                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3363                                         ifp->idev->cnf.dad_transmits *
3364                                         ifp->idev->nd_parms->retrans_time / HZ;
3365
3366                                 if (age >= ifp->prefered_lft - regen_advance) {
3367                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3368                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3369                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3370                                         if (!ifp->regen_count && ifpub) {
3371                                                 ifp->regen_count++;
3372                                                 in6_ifa_hold(ifp);
3373                                                 in6_ifa_hold(ifpub);
3374                                                 spin_unlock(&ifp->lock);
3375
3376                                                 spin_lock(&ifpub->lock);
3377                                                 ifpub->regen_count = 0;
3378                                                 spin_unlock(&ifpub->lock);
3379                                                 ipv6_create_tempaddr(ifpub, ifp);
3380                                                 in6_ifa_put(ifpub);
3381                                                 in6_ifa_put(ifp);
3382                                                 goto restart;
3383                                         }
3384                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3385                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3386                                 spin_unlock(&ifp->lock);
3387 #endif
3388                         } else {
3389                                 /* ifp->prefered_lft <= ifp->valid_lft */
3390                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3391                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3392                                 spin_unlock(&ifp->lock);
3393                         }
3394                 }
3395         }
3396
3397         next_sec = round_jiffies_up(next);
3398         next_sched = next;
3399
3400         /* If rounded timeout is accurate enough, accept it. */
3401         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3402                 next_sched = next_sec;
3403
3404         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3405         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3406                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3407
3408         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3409               now, next, next_sec, next_sched));
3410
3411         addr_chk_timer.expires = next_sched;
3412         add_timer(&addr_chk_timer);
3413         spin_unlock(&addrconf_verify_lock);
3414         rcu_read_unlock_bh();
3415 }
3416
3417 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3418 {
3419         struct in6_addr *pfx = NULL;
3420
3421         if (addr)
3422                 pfx = nla_data(addr);
3423
3424         if (local) {
3425                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3426                         pfx = NULL;
3427                 else
3428                         pfx = nla_data(local);
3429         }
3430
3431         return pfx;
3432 }
3433
3434 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3435         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3436         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3437         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3438 };
3439
3440 static int
3441 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3442 {
3443         struct net *net = sock_net(skb->sk);
3444         struct ifaddrmsg *ifm;
3445         struct nlattr *tb[IFA_MAX+1];
3446         struct in6_addr *pfx;
3447         int err;
3448
3449         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3450         if (err < 0)
3451                 return err;
3452
3453         ifm = nlmsg_data(nlh);
3454         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3455         if (pfx == NULL)
3456                 return -EINVAL;
3457
3458         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3459 }
3460
3461 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3462                              u32 prefered_lft, u32 valid_lft)
3463 {
3464         u32 flags;
3465         clock_t expires;
3466         unsigned long timeout;
3467
3468         if (!valid_lft || (prefered_lft > valid_lft))
3469                 return -EINVAL;
3470
3471         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3472         if (addrconf_finite_timeout(timeout)) {
3473                 expires = jiffies_to_clock_t(timeout * HZ);
3474                 valid_lft = timeout;
3475                 flags = RTF_EXPIRES;
3476         } else {
3477                 expires = 0;
3478                 flags = 0;
3479                 ifa_flags |= IFA_F_PERMANENT;
3480         }
3481
3482         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3483         if (addrconf_finite_timeout(timeout)) {
3484                 if (timeout == 0)
3485                         ifa_flags |= IFA_F_DEPRECATED;
3486                 prefered_lft = timeout;
3487         }
3488
3489         spin_lock_bh(&ifp->lock);
3490         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3491         ifp->tstamp = jiffies;
3492         ifp->valid_lft = valid_lft;
3493         ifp->prefered_lft = prefered_lft;
3494
3495         spin_unlock_bh(&ifp->lock);
3496         if (!(ifp->flags&IFA_F_TENTATIVE))
3497                 ipv6_ifa_notify(0, ifp);
3498
3499         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3500                               expires, flags);
3501         addrconf_verify(0);
3502
3503         return 0;
3504 }
3505
3506 static int
3507 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3508 {
3509         struct net *net = sock_net(skb->sk);
3510         struct ifaddrmsg *ifm;
3511         struct nlattr *tb[IFA_MAX+1];
3512         struct in6_addr *pfx;
3513         struct inet6_ifaddr *ifa;
3514         struct net_device *dev;
3515         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3516         u8 ifa_flags;
3517         int err;
3518
3519         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3520         if (err < 0)
3521                 return err;
3522
3523         ifm = nlmsg_data(nlh);
3524         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3525         if (pfx == NULL)
3526                 return -EINVAL;
3527
3528         if (tb[IFA_CACHEINFO]) {
3529                 struct ifa_cacheinfo *ci;
3530
3531                 ci = nla_data(tb[IFA_CACHEINFO]);
3532                 valid_lft = ci->ifa_valid;
3533                 preferred_lft = ci->ifa_prefered;
3534         } else {
3535                 preferred_lft = INFINITY_LIFE_TIME;
3536                 valid_lft = INFINITY_LIFE_TIME;
3537         }
3538
3539         dev =  __dev_get_by_index(net, ifm->ifa_index);
3540         if (dev == NULL)
3541                 return -ENODEV;
3542
3543         /* We ignore other flags so far. */
3544         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3545
3546         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3547         if (ifa == NULL) {
3548                 /*
3549                  * It would be best to check for !NLM_F_CREATE here but
3550                  * userspace alreay relies on not having to provide this.
3551                  */
3552                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3553                                       ifm->ifa_prefixlen, ifa_flags,
3554                                       preferred_lft, valid_lft);
3555         }
3556
3557         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3558             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3559                 err = -EEXIST;
3560         else
3561                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3562
3563         in6_ifa_put(ifa);
3564
3565         return err;
3566 }
3567
3568 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3569                           u8 scope, int ifindex)
3570 {
3571         struct ifaddrmsg *ifm;
3572
3573         ifm = nlmsg_data(nlh);
3574         ifm->ifa_family = AF_INET6;
3575         ifm->ifa_prefixlen = prefixlen;
3576         ifm->ifa_flags = flags;
3577         ifm->ifa_scope = scope;
3578         ifm->ifa_index = ifindex;
3579 }
3580
3581 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3582                          unsigned long tstamp, u32 preferred, u32 valid)
3583 {
3584         struct ifa_cacheinfo ci;
3585
3586         ci.cstamp = cstamp_delta(cstamp);
3587         ci.tstamp = cstamp_delta(tstamp);
3588         ci.ifa_prefered = preferred;
3589         ci.ifa_valid = valid;
3590
3591         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3592 }
3593
3594 static inline int rt_scope(int ifa_scope)
3595 {
3596         if (ifa_scope & IFA_HOST)
3597                 return RT_SCOPE_HOST;
3598         else if (ifa_scope & IFA_LINK)
3599                 return RT_SCOPE_LINK;
3600         else if (ifa_scope & IFA_SITE)
3601                 return RT_SCOPE_SITE;
3602         else
3603                 return RT_SCOPE_UNIVERSE;
3604 }
3605
3606 static inline int inet6_ifaddr_msgsize(void)
3607 {
3608         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3609                + nla_total_size(16) /* IFA_ADDRESS */
3610                + nla_total_size(sizeof(struct ifa_cacheinfo));
3611 }
3612
3613 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3614                              u32 portid, u32 seq, int event, unsigned int flags)
3615 {
3616         struct nlmsghdr  *nlh;
3617         u32 preferred, valid;
3618
3619         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3620         if (nlh == NULL)
3621                 return -EMSGSIZE;
3622
3623         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3624                       ifa->idev->dev->ifindex);
3625
3626         if (!(ifa->flags&IFA_F_PERMANENT)) {
3627                 preferred = ifa->prefered_lft;
3628                 valid = ifa->valid_lft;
3629                 if (preferred != INFINITY_LIFE_TIME) {
3630                         long tval = (jiffies - ifa->tstamp)/HZ;
3631                         if (preferred > tval)
3632                                 preferred -= tval;
3633                         else
3634                                 preferred = 0;
3635                         if (valid != INFINITY_LIFE_TIME) {
3636                                 if (valid > tval)
3637                                         valid -= tval;
3638                                 else
3639                                         valid = 0;
3640                         }
3641                 }
3642         } else {
3643                 preferred = INFINITY_LIFE_TIME;
3644                 valid = INFINITY_LIFE_TIME;
3645         }
3646
3647         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3648             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3649                 nlmsg_cancel(skb, nlh);
3650                 return -EMSGSIZE;
3651         }
3652
3653         return nlmsg_end(skb, nlh);
3654 }
3655
3656 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3657                                 u32 portid, u32 seq, int event, u16 flags)
3658 {
3659         struct nlmsghdr  *nlh;
3660         u8 scope = RT_SCOPE_UNIVERSE;
3661         int ifindex = ifmca->idev->dev->ifindex;
3662
3663         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3664                 scope = RT_SCOPE_SITE;
3665
3666         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3667         if (nlh == NULL)
3668                 return -EMSGSIZE;
3669
3670         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3671         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3672             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3673                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3674                 nlmsg_cancel(skb, nlh);
3675                 return -EMSGSIZE;
3676         }
3677
3678         return nlmsg_end(skb, nlh);
3679 }
3680
3681 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3682                                 u32 portid, u32 seq, int event, unsigned int flags)
3683 {
3684         struct nlmsghdr  *nlh;
3685         u8 scope = RT_SCOPE_UNIVERSE;
3686         int ifindex = ifaca->aca_idev->dev->ifindex;
3687
3688         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3689                 scope = RT_SCOPE_SITE;
3690
3691         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3692         if (nlh == NULL)
3693                 return -EMSGSIZE;
3694
3695         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3696         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3697             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3698                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3699                 nlmsg_cancel(skb, nlh);
3700                 return -EMSGSIZE;
3701         }
3702
3703         return nlmsg_end(skb, nlh);
3704 }
3705
3706 enum addr_type_t {
3707         UNICAST_ADDR,
3708         MULTICAST_ADDR,
3709         ANYCAST_ADDR,
3710 };
3711
3712 /* called with rcu_read_lock() */
3713 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3714                           struct netlink_callback *cb, enum addr_type_t type,
3715                           int s_ip_idx, int *p_ip_idx)
3716 {
3717         struct ifmcaddr6 *ifmca;
3718         struct ifacaddr6 *ifaca;
3719         int err = 1;
3720         int ip_idx = *p_ip_idx;
3721
3722         read_lock_bh(&idev->lock);
3723         switch (type) {
3724         case UNICAST_ADDR: {
3725                 struct inet6_ifaddr *ifa;
3726
3727                 /* unicast address incl. temp addr */
3728                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3729                         if (++ip_idx < s_ip_idx)
3730                                 continue;
3731                         err = inet6_fill_ifaddr(skb, ifa,
3732                                                 NETLINK_CB(cb->skb).portid,
3733                                                 cb->nlh->nlmsg_seq,
3734                                                 RTM_NEWADDR,
3735                                                 NLM_F_MULTI);
3736                         if (err <= 0)
3737                                 break;
3738                 }
3739                 break;
3740         }
3741         case MULTICAST_ADDR:
3742                 /* multicast address */
3743                 for (ifmca = idev->mc_list; ifmca;
3744                      ifmca = ifmca->next, ip_idx++) {
3745                         if (ip_idx < s_ip_idx)
3746                                 continue;
3747                         err = inet6_fill_ifmcaddr(skb, ifmca,
3748                                                   NETLINK_CB(cb->skb).portid,
3749                                                   cb->nlh->nlmsg_seq,
3750                                                   RTM_GETMULTICAST,
3751                                                   NLM_F_MULTI);
3752                         if (err <= 0)
3753                                 break;
3754                 }
3755                 break;
3756         case ANYCAST_ADDR:
3757                 /* anycast address */
3758                 for (ifaca = idev->ac_list; ifaca;
3759                      ifaca = ifaca->aca_next, ip_idx++) {
3760                         if (ip_idx < s_ip_idx)
3761                                 continue;
3762                         err = inet6_fill_ifacaddr(skb, ifaca,
3763                                                   NETLINK_CB(cb->skb).portid,
3764                                                   cb->nlh->nlmsg_seq,
3765                                                   RTM_GETANYCAST,
3766                                                   NLM_F_MULTI);
3767                         if (err <= 0)
3768                                 break;
3769                 }
3770                 break;
3771         default:
3772                 break;
3773         }
3774         read_unlock_bh(&idev->lock);
3775         *p_ip_idx = ip_idx;
3776         return err;
3777 }
3778
3779 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3780                            enum addr_type_t type)
3781 {
3782         struct net *net = sock_net(skb->sk);
3783         int h, s_h;
3784         int idx, ip_idx;
3785         int s_idx, s_ip_idx;
3786         struct net_device *dev;
3787         struct inet6_dev *idev;
3788         struct hlist_head *head;
3789         struct hlist_node *node;
3790
3791         s_h = cb->args[0];
3792         s_idx = idx = cb->args[1];
3793         s_ip_idx = ip_idx = cb->args[2];
3794
3795         rcu_read_lock();
3796         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3797                 idx = 0;
3798                 head = &net->dev_index_head[h];
3799                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3800                         if (idx < s_idx)
3801                                 goto cont;
3802                         if (h > s_h || idx > s_idx)
3803                                 s_ip_idx = 0;
3804                         ip_idx = 0;
3805                         idev = __in6_dev_get(dev);
3806                         if (!idev)
3807                                 goto cont;
3808
3809                         if (in6_dump_addrs(idev, skb, cb, type,
3810                                            s_ip_idx, &ip_idx) <= 0)
3811                                 goto done;
3812 cont:
3813                         idx++;
3814                 }
3815         }
3816 done:
3817         rcu_read_unlock();
3818         cb->args[0] = h;
3819         cb->args[1] = idx;
3820         cb->args[2] = ip_idx;
3821
3822         return skb->len;
3823 }
3824
3825 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3826 {
3827         enum addr_type_t type = UNICAST_ADDR;
3828
3829         return inet6_dump_addr(skb, cb, type);
3830 }
3831
3832 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3833 {
3834         enum addr_type_t type = MULTICAST_ADDR;
3835
3836         return inet6_dump_addr(skb, cb, type);
3837 }
3838
3839
3840 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3841 {
3842         enum addr_type_t type = ANYCAST_ADDR;
3843
3844         return inet6_dump_addr(skb, cb, type);
3845 }
3846
3847 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3848                              void *arg)
3849 {
3850         struct net *net = sock_net(in_skb->sk);
3851         struct ifaddrmsg *ifm;
3852         struct nlattr *tb[IFA_MAX+1];
3853         struct in6_addr *addr = NULL;
3854         struct net_device *dev = NULL;
3855         struct inet6_ifaddr *ifa;
3856         struct sk_buff *skb;
3857         int err;
3858
3859         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3860         if (err < 0)
3861                 goto errout;
3862
3863         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3864         if (addr == NULL) {
3865                 err = -EINVAL;
3866                 goto errout;
3867         }
3868
3869         ifm = nlmsg_data(nlh);
3870         if (ifm->ifa_index)
3871                 dev = __dev_get_by_index(net, ifm->ifa_index);
3872
3873         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3874         if (!ifa) {
3875                 err = -EADDRNOTAVAIL;
3876                 goto errout;
3877         }
3878
3879         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3880         if (!skb) {
3881                 err = -ENOBUFS;
3882                 goto errout_ifa;
3883         }
3884
3885         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3886                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3887         if (err < 0) {
3888                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3889                 WARN_ON(err == -EMSGSIZE);
3890                 kfree_skb(skb);
3891                 goto errout_ifa;
3892         }
3893         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3894 errout_ifa:
3895         in6_ifa_put(ifa);
3896 errout:
3897         return err;
3898 }
3899
3900 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3901 {
3902         struct sk_buff *skb;
3903         struct net *net = dev_net(ifa->idev->dev);
3904         int err = -ENOBUFS;
3905
3906         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3907         if (skb == NULL)
3908                 goto errout;
3909
3910         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3911         if (err < 0) {
3912                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3913                 WARN_ON(err == -EMSGSIZE);
3914                 kfree_skb(skb);
3915                 goto errout;
3916         }
3917         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3918         return;
3919 errout:
3920         if (err < 0)
3921                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3922 }
3923
3924 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3925                                 __s32 *array, int bytes)
3926 {
3927         BUG_ON(bytes < (DEVCONF_MAX * 4));
3928
3929         memset(array, 0, bytes);
3930         array[DEVCONF_FORWARDING] = cnf->forwarding;
3931         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3932         array[DEVCONF_MTU6] = cnf->mtu6;
3933         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3934         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3935         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3936         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3937         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3938         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3939                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3940         array[DEVCONF_RTR_SOLICIT_DELAY] =
3941                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3942         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3943 #ifdef CONFIG_IPV6_PRIVACY
3944         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3945         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3946         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3947         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3948         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3949 #endif
3950         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3951         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3952         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3953 #ifdef CONFIG_IPV6_ROUTER_PREF
3954         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3955         array[DEVCONF_RTR_PROBE_INTERVAL] =
3956                 jiffies_to_msecs(cnf->rtr_probe_interval);
3957 #ifdef CONFIG_IPV6_ROUTE_INFO
3958         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3959 #endif
3960 #endif
3961         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3962         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3963 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3964         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3965 #endif
3966 #ifdef CONFIG_IPV6_MROUTE
3967         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3968 #endif
3969         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3970         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3971         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3972 }
3973
3974 static inline size_t inet6_ifla6_size(void)
3975 {
3976         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3977              + nla_total_size(sizeof(struct ifla_cacheinfo))
3978              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3979              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3980              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3981 }
3982
3983 static inline size_t inet6_if_nlmsg_size(void)
3984 {
3985         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3986                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3987                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3988                + nla_total_size(4) /* IFLA_MTU */
3989                + nla_total_size(4) /* IFLA_LINK */
3990                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3991 }
3992
3993 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3994                                       int items, int bytes)
3995 {
3996         int i;
3997         int pad = bytes - sizeof(u64) * items;
3998         BUG_ON(pad < 0);
3999
4000         /* Use put_unaligned() because stats may not be aligned for u64. */
4001         put_unaligned(items, &stats[0]);
4002         for (i = 1; i < items; i++)
4003                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4004
4005         memset(&stats[items], 0, pad);
4006 }
4007
4008 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4009                                       int items, int bytes, size_t syncpoff)
4010 {
4011         int i;
4012         int pad = bytes - sizeof(u64) * items;
4013         BUG_ON(pad < 0);
4014
4015         /* Use put_unaligned() because stats may not be aligned for u64. */
4016         put_unaligned(items, &stats[0]);
4017         for (i = 1; i < items; i++)
4018                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4019
4020         memset(&stats[items], 0, pad);
4021 }
4022
4023 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4024                              int bytes)
4025 {
4026         switch (attrtype) {
4027         case IFLA_INET6_STATS:
4028                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4029                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4030                 break;
4031         case IFLA_INET6_ICMP6STATS:
4032                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4033                 break;
4034         }
4035 }
4036
4037 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4038 {
4039         struct nlattr *nla;
4040         struct ifla_cacheinfo ci;
4041
4042         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4043                 goto nla_put_failure;
4044         ci.max_reasm_len = IPV6_MAXPLEN;
4045         ci.tstamp = cstamp_delta(idev->tstamp);
4046         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4047         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4048         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4049                 goto nla_put_failure;
4050         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4051         if (nla == NULL)
4052                 goto nla_put_failure;
4053         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4054
4055         /* XXX - MC not implemented */
4056
4057         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4058         if (nla == NULL)
4059                 goto nla_put_failure;
4060         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4061
4062         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4063         if (nla == NULL)
4064                 goto nla_put_failure;
4065         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4066
4067         return 0;
4068
4069 nla_put_failure:
4070         return -EMSGSIZE;
4071 }
4072
4073 static size_t inet6_get_link_af_size(const struct net_device *dev)
4074 {
4075         if (!__in6_dev_get(dev))
4076                 return 0;
4077
4078         return inet6_ifla6_size();
4079 }
4080
4081 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4082 {
4083         struct inet6_dev *idev = __in6_dev_get(dev);
4084
4085         if (!idev)
4086                 return -ENODATA;
4087
4088         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4089                 return -EMSGSIZE;
4090
4091         return 0;
4092 }
4093
4094 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4095                              u32 portid, u32 seq, int event, unsigned int flags)
4096 {
4097         struct net_device *dev = idev->dev;
4098         struct ifinfomsg *hdr;
4099         struct nlmsghdr *nlh;
4100         void *protoinfo;
4101
4102         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4103         if (nlh == NULL)
4104                 return -EMSGSIZE;
4105
4106         hdr = nlmsg_data(nlh);
4107         hdr->ifi_family = AF_INET6;
4108         hdr->__ifi_pad = 0;
4109         hdr->ifi_type = dev->type;
4110         hdr->ifi_index = dev->ifindex;
4111         hdr->ifi_flags = dev_get_flags(dev);
4112         hdr->ifi_change = 0;
4113
4114         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4115             (dev->addr_len &&
4116              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4117             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4118             (dev->ifindex != dev->iflink &&
4119              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4120                 goto nla_put_failure;
4121         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4122         if (protoinfo == NULL)
4123                 goto nla_put_failure;
4124
4125         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4126                 goto nla_put_failure;
4127
4128         nla_nest_end(skb, protoinfo);
4129         return nlmsg_end(skb, nlh);
4130
4131 nla_put_failure:
4132         nlmsg_cancel(skb, nlh);
4133         return -EMSGSIZE;
4134 }
4135
4136 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4137 {
4138         struct net *net = sock_net(skb->sk);
4139         int h, s_h;
4140         int idx = 0, s_idx;
4141         struct net_device *dev;
4142         struct inet6_dev *idev;
4143         struct hlist_head *head;
4144         struct hlist_node *node;
4145
4146         s_h = cb->args[0];
4147         s_idx = cb->args[1];
4148
4149         rcu_read_lock();
4150         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4151                 idx = 0;
4152                 head = &net->dev_index_head[h];
4153                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4154                         if (idx < s_idx)
4155                                 goto cont;
4156                         idev = __in6_dev_get(dev);
4157                         if (!idev)
4158                                 goto cont;
4159                         if (inet6_fill_ifinfo(skb, idev,
4160                                               NETLINK_CB(cb->skb).portid,
4161                                               cb->nlh->nlmsg_seq,
4162                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4163                                 goto out;
4164 cont:
4165                         idx++;
4166                 }
4167         }
4168 out:
4169         rcu_read_unlock();
4170         cb->args[1] = idx;
4171         cb->args[0] = h;
4172
4173         return skb->len;
4174 }
4175
4176 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4177 {
4178         struct sk_buff *skb;
4179         struct net *net = dev_net(idev->dev);
4180         int err = -ENOBUFS;
4181
4182         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4183         if (skb == NULL)
4184                 goto errout;
4185
4186         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4187         if (err < 0) {
4188                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4189                 WARN_ON(err == -EMSGSIZE);
4190                 kfree_skb(skb);
4191                 goto errout;
4192         }
4193         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4194         return;
4195 errout:
4196         if (err < 0)
4197                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4198 }
4199
4200 static inline size_t inet6_prefix_nlmsg_size(void)
4201 {
4202         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4203                + nla_total_size(sizeof(struct in6_addr))
4204                + nla_total_size(sizeof(struct prefix_cacheinfo));
4205 }
4206
4207 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4208                              struct prefix_info *pinfo, u32 portid, u32 seq,
4209                              int event, unsigned int flags)
4210 {
4211         struct prefixmsg *pmsg;
4212         struct nlmsghdr *nlh;
4213         struct prefix_cacheinfo ci;
4214
4215         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4216         if (nlh == NULL)
4217                 return -EMSGSIZE;
4218
4219         pmsg = nlmsg_data(nlh);
4220         pmsg->prefix_family = AF_INET6;
4221         pmsg->prefix_pad1 = 0;
4222         pmsg->prefix_pad2 = 0;
4223         pmsg->prefix_ifindex = idev->dev->ifindex;
4224         pmsg->prefix_len = pinfo->prefix_len;
4225         pmsg->prefix_type = pinfo->type;
4226         pmsg->prefix_pad3 = 0;
4227         pmsg->prefix_flags = 0;
4228         if (pinfo->onlink)
4229                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4230         if (pinfo->autoconf)
4231                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4232
4233         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4234                 goto nla_put_failure;
4235         ci.preferred_time = ntohl(pinfo->prefered);
4236         ci.valid_time = ntohl(pinfo->valid);
4237         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4238                 goto nla_put_failure;
4239         return nlmsg_end(skb, nlh);
4240
4241 nla_put_failure:
4242         nlmsg_cancel(skb, nlh);
4243         return -EMSGSIZE;
4244 }
4245
4246 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4247                          struct prefix_info *pinfo)
4248 {
4249         struct sk_buff *skb;
4250         struct net *net = dev_net(idev->dev);
4251         int err = -ENOBUFS;
4252
4253         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4254         if (skb == NULL)
4255                 goto errout;
4256
4257         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4258         if (err < 0) {
4259                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4260                 WARN_ON(err == -EMSGSIZE);
4261                 kfree_skb(skb);
4262                 goto errout;
4263         }
4264         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4265         return;
4266 errout:
4267         if (err < 0)
4268                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4269 }
4270
4271 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4272 {
4273         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4274
4275         switch (event) {
4276         case RTM_NEWADDR:
4277                 /*
4278                  * If the address was optimistic
4279                  * we inserted the route at the start of
4280                  * our DAD process, so we don't need
4281                  * to do it again
4282                  */
4283                 if (!(ifp->rt->rt6i_node))
4284                         ip6_ins_rt(ifp->rt);
4285                 if (ifp->idev->cnf.forwarding)
4286                         addrconf_join_anycast(ifp);
4287                 break;
4288         case RTM_DELADDR:
4289                 if (ifp->idev->cnf.forwarding)
4290                         addrconf_leave_anycast(ifp);
4291                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4292                 dst_hold(&ifp->rt->dst);
4293
4294                 if (ip6_del_rt(ifp->rt))
4295                         dst_free(&ifp->rt->dst);
4296                 break;
4297         }
4298 }
4299
4300 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4301 {
4302         rcu_read_lock_bh();
4303         if (likely(ifp->idev->dead == 0))
4304                 __ipv6_ifa_notify(event, ifp);
4305         rcu_read_unlock_bh();
4306 }
4307
4308 #ifdef CONFIG_SYSCTL
4309
4310 static
4311 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4312                            void __user *buffer, size_t *lenp, loff_t *ppos)
4313 {
4314         int *valp = ctl->data;
4315         int val = *valp;
4316         loff_t pos = *ppos;
4317         ctl_table lctl;
4318         int ret;
4319
4320         /*
4321          * ctl->data points to idev->cnf.forwarding, we should
4322          * not modify it until we get the rtnl lock.
4323          */
4324         lctl = *ctl;
4325         lctl.data = &val;
4326
4327         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4328
4329         if (write)
4330                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4331         if (ret)
4332                 *ppos = pos;
4333         return ret;
4334 }
4335
4336 static void dev_disable_change(struct inet6_dev *idev)
4337 {
4338         if (!idev || !idev->dev)
4339                 return;
4340
4341         if (idev->cnf.disable_ipv6)
4342                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4343         else
4344                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4345 }
4346
4347 static void addrconf_disable_change(struct net *net, __s32 newf)
4348 {
4349         struct net_device *dev;
4350         struct inet6_dev *idev;
4351
4352         rcu_read_lock();
4353         for_each_netdev_rcu(net, dev) {
4354                 idev = __in6_dev_get(dev);
4355                 if (idev) {
4356                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4357                         idev->cnf.disable_ipv6 = newf;
4358                         if (changed)
4359                                 dev_disable_change(idev);
4360                 }
4361         }
4362         rcu_read_unlock();
4363 }
4364
4365 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4366 {
4367         struct net *net;
4368         int old;
4369
4370         if (!rtnl_trylock())
4371                 return restart_syscall();
4372
4373         net = (struct net *)table->extra2;
4374         old = *p;
4375         *p = newf;
4376
4377         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4378                 rtnl_unlock();
4379                 return 0;
4380         }
4381
4382         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4383                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4384                 addrconf_disable_change(net, newf);
4385         } else if ((!newf) ^ (!old))
4386                 dev_disable_change((struct inet6_dev *)table->extra1);
4387
4388         rtnl_unlock();
4389         return 0;
4390 }
4391
4392 static
4393 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4394                             void __user *buffer, size_t *lenp, loff_t *ppos)
4395 {
4396         int *valp = ctl->data;
4397         int val = *valp;
4398         loff_t pos = *ppos;
4399         ctl_table lctl;
4400         int ret;
4401
4402         /*
4403          * ctl->data points to idev->cnf.disable_ipv6, we should
4404          * not modify it until we get the rtnl lock.
4405          */
4406         lctl = *ctl;
4407         lctl.data = &val;
4408
4409         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4410
4411         if (write)
4412                 ret = addrconf_disable_ipv6(ctl, valp, val);
4413         if (ret)
4414                 *ppos = pos;
4415         return ret;
4416 }
4417
4418 static struct addrconf_sysctl_table
4419 {
4420         struct ctl_table_header *sysctl_header;
4421         ctl_table addrconf_vars[DEVCONF_MAX+1];
4422 } addrconf_sysctl __read_mostly = {
4423         .sysctl_header = NULL,
4424         .addrconf_vars = {
4425                 {
4426                         .procname       = "forwarding",
4427                         .data           = &ipv6_devconf.forwarding,
4428                         .maxlen         = sizeof(int),
4429                         .mode           = 0644,
4430                         .proc_handler   = addrconf_sysctl_forward,
4431                 },
4432                 {
4433                         .procname       = "hop_limit",
4434                         .data           = &ipv6_devconf.hop_limit,
4435                         .maxlen         = sizeof(int),
4436                         .mode           = 0644,
4437                         .proc_handler   = proc_dointvec,
4438                 },
4439                 {
4440                         .procname       = "mtu",
4441                         .data           = &ipv6_devconf.mtu6,
4442                         .maxlen         = sizeof(int),
4443                         .mode           = 0644,
4444                         .proc_handler   = proc_dointvec,
4445                 },
4446                 {
4447                         .procname       = "accept_ra",
4448                         .data           = &ipv6_devconf.accept_ra,
4449                         .maxlen         = sizeof(int),
4450                         .mode           = 0644,
4451                         .proc_handler   = proc_dointvec,
4452                 },
4453                 {
4454                         .procname       = "accept_redirects",
4455                         .data           = &ipv6_devconf.accept_redirects,
4456                         .maxlen         = sizeof(int),
4457                         .mode           = 0644,
4458                         .proc_handler   = proc_dointvec,
4459                 },
4460                 {
4461                         .procname       = "autoconf",
4462                         .data           = &ipv6_devconf.autoconf,
4463                         .maxlen         = sizeof(int),
4464                         .mode           = 0644,
4465                         .proc_handler   = proc_dointvec,
4466                 },
4467                 {
4468                         .procname       = "dad_transmits",
4469                         .data           = &ipv6_devconf.dad_transmits,
4470                         .maxlen         = sizeof(int),
4471                         .mode           = 0644,
4472                         .proc_handler   = proc_dointvec,
4473                 },
4474                 {
4475                         .procname       = "router_solicitations",
4476                         .data           = &ipv6_devconf.rtr_solicits,
4477                         .maxlen         = sizeof(int),
4478                         .mode           = 0644,
4479                         .proc_handler   = proc_dointvec,
4480                 },
4481                 {
4482                         .procname       = "router_solicitation_interval",
4483                         .data           = &ipv6_devconf.rtr_solicit_interval,
4484                         .maxlen         = sizeof(int),
4485                         .mode           = 0644,
4486                         .proc_handler   = proc_dointvec_jiffies,
4487                 },
4488                 {
4489                         .procname       = "router_solicitation_delay",
4490                         .data           = &ipv6_devconf.rtr_solicit_delay,
4491                         .maxlen         = sizeof(int),
4492                         .mode           = 0644,
4493                         .proc_handler   = proc_dointvec_jiffies,
4494                 },
4495                 {
4496                         .procname       = "force_mld_version",
4497                         .data           = &ipv6_devconf.force_mld_version,
4498                         .maxlen         = sizeof(int),
4499                         .mode           = 0644,
4500                         .proc_handler   = proc_dointvec,
4501                 },
4502 #ifdef CONFIG_IPV6_PRIVACY
4503                 {
4504                         .procname       = "use_tempaddr",
4505                         .data           = &ipv6_devconf.use_tempaddr,
4506                         .maxlen         = sizeof(int),
4507                         .mode           = 0644,
4508                         .proc_handler   = proc_dointvec,
4509                 },
4510                 {
4511                         .procname       = "temp_valid_lft",
4512                         .data           = &ipv6_devconf.temp_valid_lft,
4513                         .maxlen         = sizeof(int),
4514                         .mode           = 0644,
4515                         .proc_handler   = proc_dointvec,
4516                 },
4517                 {
4518                         .procname       = "temp_prefered_lft",
4519                         .data           = &ipv6_devconf.temp_prefered_lft,
4520                         .maxlen         = sizeof(int),
4521                         .mode           = 0644,
4522                         .proc_handler   = proc_dointvec,
4523                 },
4524                 {
4525                         .procname       = "regen_max_retry",
4526                         .data           = &ipv6_devconf.regen_max_retry,
4527                         .maxlen         = sizeof(int),
4528                         .mode           = 0644,
4529                         .proc_handler   = proc_dointvec,
4530                 },
4531                 {
4532                         .procname       = "max_desync_factor",
4533                         .data           = &ipv6_devconf.max_desync_factor,
4534                         .maxlen         = sizeof(int),
4535                         .mode           = 0644,
4536                         .proc_handler   = proc_dointvec,
4537                 },
4538 #endif
4539                 {
4540                         .procname       = "max_addresses",
4541                         .data           = &ipv6_devconf.max_addresses,
4542                         .maxlen         = sizeof(int),
4543                         .mode           = 0644,
4544                         .proc_handler   = proc_dointvec,
4545                 },
4546                 {
4547                         .procname       = "accept_ra_defrtr",
4548                         .data           = &ipv6_devconf.accept_ra_defrtr,
4549                         .maxlen         = sizeof(int),
4550                         .mode           = 0644,
4551                         .proc_handler   = proc_dointvec,
4552                 },
4553                 {
4554                         .procname       = "accept_ra_pinfo",
4555                         .data           = &ipv6_devconf.accept_ra_pinfo,
4556                         .maxlen         = sizeof(int),
4557                         .mode           = 0644,
4558                         .proc_handler   = proc_dointvec,
4559                 },
4560 #ifdef CONFIG_IPV6_ROUTER_PREF
4561                 {
4562                         .procname       = "accept_ra_rtr_pref",
4563                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4564                         .maxlen         = sizeof(int),
4565                         .mode           = 0644,
4566                         .proc_handler   = proc_dointvec,
4567                 },
4568                 {
4569                         .procname       = "router_probe_interval",
4570                         .data           = &ipv6_devconf.rtr_probe_interval,
4571                         .maxlen         = sizeof(int),
4572                         .mode           = 0644,
4573                         .proc_handler   = proc_dointvec_jiffies,
4574                 },
4575 #ifdef CONFIG_IPV6_ROUTE_INFO
4576                 {
4577                         .procname       = "accept_ra_rt_info_max_plen",
4578                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4579                         .maxlen         = sizeof(int),
4580                         .mode           = 0644,
4581                         .proc_handler   = proc_dointvec,
4582                 },
4583 #endif
4584 #endif
4585                 {
4586                         .procname       = "proxy_ndp",
4587                         .data           = &ipv6_devconf.proxy_ndp,
4588                         .maxlen         = sizeof(int),
4589                         .mode           = 0644,
4590                         .proc_handler   = proc_dointvec,
4591                 },
4592                 {
4593                         .procname       = "accept_source_route",
4594                         .data           = &ipv6_devconf.accept_source_route,
4595                         .maxlen         = sizeof(int),
4596                         .mode           = 0644,
4597                         .proc_handler   = proc_dointvec,
4598                 },
4599 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4600                 {
4601                         .procname       = "optimistic_dad",
4602                         .data           = &ipv6_devconf.optimistic_dad,
4603                         .maxlen         = sizeof(int),
4604                         .mode           = 0644,
4605                         .proc_handler   = proc_dointvec,
4606
4607                 },
4608 #endif
4609 #ifdef CONFIG_IPV6_MROUTE
4610                 {
4611                         .procname       = "mc_forwarding",
4612                         .data           = &ipv6_devconf.mc_forwarding,
4613                         .maxlen         = sizeof(int),
4614                         .mode           = 0444,
4615                         .proc_handler   = proc_dointvec,
4616                 },
4617 #endif
4618                 {
4619                         .procname       = "disable_ipv6",
4620                         .data           = &ipv6_devconf.disable_ipv6,
4621                         .maxlen         = sizeof(int),
4622                         .mode           = 0644,
4623                         .proc_handler   = addrconf_sysctl_disable,
4624                 },
4625                 {
4626                         .procname       = "accept_dad",
4627                         .data           = &ipv6_devconf.accept_dad,
4628                         .maxlen         = sizeof(int),
4629                         .mode           = 0644,
4630                         .proc_handler   = proc_dointvec,
4631                 },
4632                 {
4633                         .procname       = "force_tllao",
4634                         .data           = &ipv6_devconf.force_tllao,
4635                         .maxlen         = sizeof(int),
4636                         .mode           = 0644,
4637                         .proc_handler   = proc_dointvec
4638                 },
4639                 {
4640                         /* sentinel */
4641                 }
4642         },
4643 };
4644
4645 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4646                 struct inet6_dev *idev, struct ipv6_devconf *p)
4647 {
4648         int i;
4649         struct addrconf_sysctl_table *t;
4650         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4651
4652         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4653         if (t == NULL)
4654                 goto out;
4655
4656         for (i = 0; t->addrconf_vars[i].data; i++) {
4657                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4658                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4659                 t->addrconf_vars[i].extra2 = net;
4660         }
4661
4662         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4663
4664         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4665         if (t->sysctl_header == NULL)
4666                 goto free;
4667
4668         p->sysctl = t;
4669         return 0;
4670
4671 free:
4672         kfree(t);
4673 out:
4674         return -ENOBUFS;
4675 }
4676
4677 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4678 {
4679         struct addrconf_sysctl_table *t;
4680
4681         if (p->sysctl == NULL)
4682                 return;
4683
4684         t = p->sysctl;
4685         p->sysctl = NULL;
4686         unregister_net_sysctl_table(t->sysctl_header);
4687         kfree(t);
4688 }
4689
4690 static void addrconf_sysctl_register(struct inet6_dev *idev)
4691 {
4692         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4693                               &ndisc_ifinfo_sysctl_change);
4694         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4695                                         idev, &idev->cnf);
4696 }
4697
4698 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4699 {
4700         __addrconf_sysctl_unregister(&idev->cnf);
4701         neigh_sysctl_unregister(idev->nd_parms);
4702 }
4703
4704
4705 #endif
4706
4707 static int __net_init addrconf_init_net(struct net *net)
4708 {
4709         int err;
4710         struct ipv6_devconf *all, *dflt;
4711
4712         err = -ENOMEM;
4713         all = &ipv6_devconf;
4714         dflt = &ipv6_devconf_dflt;
4715
4716         if (!net_eq(net, &init_net)) {
4717                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4718                 if (all == NULL)
4719                         goto err_alloc_all;
4720
4721                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4722                 if (dflt == NULL)
4723                         goto err_alloc_dflt;
4724         } else {
4725                 /* these will be inherited by all namespaces */
4726                 dflt->autoconf = ipv6_defaults.autoconf;
4727                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4728         }
4729
4730         net->ipv6.devconf_all = all;
4731         net->ipv6.devconf_dflt = dflt;
4732
4733 #ifdef CONFIG_SYSCTL
4734         err = __addrconf_sysctl_register(net, "all", NULL, all);
4735         if (err < 0)
4736                 goto err_reg_all;
4737
4738         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4739         if (err < 0)
4740                 goto err_reg_dflt;
4741 #endif
4742         return 0;
4743
4744 #ifdef CONFIG_SYSCTL
4745 err_reg_dflt:
4746         __addrconf_sysctl_unregister(all);
4747 err_reg_all:
4748         kfree(dflt);
4749 #endif
4750 err_alloc_dflt:
4751         kfree(all);
4752 err_alloc_all:
4753         return err;
4754 }
4755
4756 static void __net_exit addrconf_exit_net(struct net *net)
4757 {
4758 #ifdef CONFIG_SYSCTL
4759         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4760         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4761 #endif
4762         if (!net_eq(net, &init_net)) {
4763                 kfree(net->ipv6.devconf_dflt);
4764                 kfree(net->ipv6.devconf_all);
4765         }
4766 }
4767
4768 static struct pernet_operations addrconf_ops = {
4769         .init = addrconf_init_net,
4770         .exit = addrconf_exit_net,
4771 };
4772
4773 /*
4774  *      Device notifier
4775  */
4776
4777 int register_inet6addr_notifier(struct notifier_block *nb)
4778 {
4779         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4780 }
4781 EXPORT_SYMBOL(register_inet6addr_notifier);
4782
4783 int unregister_inet6addr_notifier(struct notifier_block *nb)
4784 {
4785         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4786 }
4787 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4788
4789 static struct rtnl_af_ops inet6_ops = {
4790         .family           = AF_INET6,
4791         .fill_link_af     = inet6_fill_link_af,
4792         .get_link_af_size = inet6_get_link_af_size,
4793 };
4794
4795 /*
4796  *      Init / cleanup code
4797  */
4798
4799 int __init addrconf_init(void)
4800 {
4801         int i, err;
4802
4803         err = ipv6_addr_label_init();
4804         if (err < 0) {
4805                 pr_crit("%s: cannot initialize default policy table: %d\n",
4806                         __func__, err);
4807                 goto out;
4808         }
4809
4810         err = register_pernet_subsys(&addrconf_ops);
4811         if (err < 0)
4812                 goto out_addrlabel;
4813
4814         /* The addrconf netdev notifier requires that loopback_dev
4815          * has it's ipv6 private information allocated and setup
4816          * before it can bring up and give link-local addresses
4817          * to other devices which are up.
4818          *
4819          * Unfortunately, loopback_dev is not necessarily the first
4820          * entry in the global dev_base list of net devices.  In fact,
4821          * it is likely to be the very last entry on that list.
4822          * So this causes the notifier registry below to try and
4823          * give link-local addresses to all devices besides loopback_dev
4824          * first, then loopback_dev, which cases all the non-loopback_dev
4825          * devices to fail to get a link-local address.
4826          *
4827          * So, as a temporary fix, allocate the ipv6 structure for
4828          * loopback_dev first by hand.
4829          * Longer term, all of the dependencies ipv6 has upon the loopback
4830          * device and it being up should be removed.
4831          */
4832         rtnl_lock();
4833         if (!ipv6_add_dev(init_net.loopback_dev))
4834                 err = -ENOMEM;
4835         rtnl_unlock();
4836         if (err)
4837                 goto errlo;
4838
4839         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4840                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4841
4842         register_netdevice_notifier(&ipv6_dev_notf);
4843
4844         addrconf_verify(0);
4845
4846         err = rtnl_af_register(&inet6_ops);
4847         if (err < 0)
4848                 goto errout_af;
4849
4850         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4851                               NULL);
4852         if (err < 0)
4853                 goto errout;
4854
4855         /* Only the first call to __rtnl_register can fail */
4856         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4857         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4858         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4859                         inet6_dump_ifaddr, NULL);
4860         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4861                         inet6_dump_ifmcaddr, NULL);
4862         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4863                         inet6_dump_ifacaddr, NULL);
4864
4865         ipv6_addr_label_rtnl_register();
4866
4867         return 0;
4868 errout:
4869         rtnl_af_unregister(&inet6_ops);
4870 errout_af:
4871         unregister_netdevice_notifier(&ipv6_dev_notf);
4872 errlo:
4873         unregister_pernet_subsys(&addrconf_ops);
4874 out_addrlabel:
4875         ipv6_addr_label_cleanup();
4876 out:
4877         return err;
4878 }
4879
4880 void addrconf_cleanup(void)
4881 {
4882         struct net_device *dev;
4883         int i;
4884
4885         unregister_netdevice_notifier(&ipv6_dev_notf);
4886         unregister_pernet_subsys(&addrconf_ops);
4887         ipv6_addr_label_cleanup();
4888
4889         rtnl_lock();
4890
4891         __rtnl_af_unregister(&inet6_ops);
4892
4893         /* clean dev list */
4894         for_each_netdev(&init_net, dev) {
4895                 if (__in6_dev_get(dev) == NULL)
4896                         continue;
4897                 addrconf_ifdown(dev, 1);
4898         }
4899         addrconf_ifdown(init_net.loopback_dev, 2);
4900
4901         /*
4902          *      Check hash table.
4903          */
4904         spin_lock_bh(&addrconf_hash_lock);
4905         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4906                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4907         spin_unlock_bh(&addrconf_hash_lock);
4908
4909         del_timer(&addr_chk_timer);
4910         rtnl_unlock();
4911 }