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/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86
  87 #ifdef CONFIG_IPV6_PRIVACY
  88 #include <linux/random.h>
  89 #endif
  90
  91 #include <linux/uaccess.h>
  92 #include <asm/unaligned.h>
  93
  94 #include <linux/proc_fs.h>
  95 #include <linux/seq_file.h>
  96 #include <linux/export.h>
  97
  98 /* Set to 3 to get tracing... */
  99 #define ACONF_DEBUG 2
 100
 101 #if ACONF_DEBUG >= 3
 102 #define ADBG(x) printk x
 103 #else
 104 #define ADBG(x)
 105 #endif
 106
 107 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 108
 109 static inline u32 cstamp_delta(unsigned long cstamp)
 110 {
 111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112 }
 113
 114 #ifdef CONFIG_SYSCTL
 115 static void addrconf_sysctl_register(struct inet6_dev *idev);
 116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117 #else
 118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 119 {
 120 }
 121
 122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 123 {
 124 }
 125 #endif
 126
 127 #ifdef CONFIG_IPV6_PRIVACY
 128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130 static void ipv6_regen_rndid(unsigned long data);
 131 #endif
 132
 133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134 static int ipv6_count_addresses(struct inet6_dev *idev);
 135
 136 /*
 137  *      Configured unicast address hash table
 138  */
 139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 140 static DEFINE_SPINLOCK(addrconf_hash_lock);
 141
 142 static void addrconf_verify(unsigned long);
 143
 144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 145 static DEFINE_SPINLOCK(addrconf_verify_lock);
 146
 147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 149
 150 static void addrconf_type_change(struct net_device *dev,
 151                                  unsigned long event);
 152 static int addrconf_ifdown(struct net_device *dev, int how);
 153
 154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 155                                                   int plen,
 156                                                   const struct net_device *dev,
 157                                                   u32 flags, u32 noflags);
 158
 159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 160 static void addrconf_dad_timer(unsigned long data);
 161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 162 static void addrconf_dad_run(struct inet6_dev *idev);
 163 static void addrconf_rs_timer(unsigned long data);
 164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 166
 167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 168                                 struct prefix_info *pinfo);
 169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 170                                struct net_device *dev);
 171
 172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 173         .forwarding             = 0,
 174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 175         .mtu6                   = IPV6_MIN_MTU,
 176         .accept_ra              = 1,
 177         .accept_redirects       = 1,
 178         .autoconf               = 1,
 179         .force_mld_version      = 0,
 180         .mldv1_unsolicited_report_interval = 10 * HZ,
 181         .mldv2_unsolicited_report_interval = HZ,
 182         .dad_transmits          = 1,
 183         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 184         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 185         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 186 #ifdef CONFIG_IPV6_PRIVACY
 187         .use_tempaddr           = 0,
 188         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 189         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 190         .regen_max_retry        = REGEN_MAX_RETRY,
 191         .max_desync_factor      = MAX_DESYNC_FACTOR,
 192 #endif
 193         .max_addresses          = IPV6_MAX_ADDRESSES,
 194         .accept_ra_defrtr       = 1,
 195         .accept_ra_pinfo        = 1,
 196 #ifdef CONFIG_IPV6_ROUTER_PREF
 197         .accept_ra_rtr_pref     = 1,
 198         .rtr_probe_interval     = 60 * HZ,
 199 #ifdef CONFIG_IPV6_ROUTE_INFO
 200         .accept_ra_rt_info_max_plen = 0,
 201 #endif
 202 #endif
 203         .proxy_ndp              = 0,
 204         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 205         .disable_ipv6           = 0,
 206         .accept_dad             = 1,
 207 };
 208
 209 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 210         .forwarding             = 0,
 211         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 212         .mtu6                   = IPV6_MIN_MTU,
 213         .accept_ra              = 1,
 214         .accept_redirects       = 1,
 215         .autoconf               = 1,
 216         .force_mld_version      = 0,
 217         .mldv1_unsolicited_report_interval = 10 * HZ,
 218         .mldv2_unsolicited_report_interval = HZ,
 219         .dad_transmits          = 1,
 220         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 221         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 222         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 223 #ifdef CONFIG_IPV6_PRIVACY
 224         .use_tempaddr           = 0,
 225         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 226         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 227         .regen_max_retry        = REGEN_MAX_RETRY,
 228         .max_desync_factor      = MAX_DESYNC_FACTOR,
 229 #endif
 230         .max_addresses          = IPV6_MAX_ADDRESSES,
 231         .accept_ra_defrtr       = 1,
 232         .accept_ra_pinfo        = 1,
 233 #ifdef CONFIG_IPV6_ROUTER_PREF
 234         .accept_ra_rtr_pref     = 1,
 235         .rtr_probe_interval     = 60 * HZ,
 236 #ifdef CONFIG_IPV6_ROUTE_INFO
 237         .accept_ra_rt_info_max_plen = 0,
 238 #endif
 239 #endif
 240         .proxy_ndp              = 0,
 241         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 242         .disable_ipv6           = 0,
 243         .accept_dad             = 1,
 244 };
 245
 246 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 247 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 248 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 249 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 250 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 251 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 252 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 253 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 254
 255 /* Check if a valid qdisc is available */
 256 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 257 {
 258         return !qdisc_tx_is_noop(dev);
 259 }
 260
 261 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 262 {
 263         if (del_timer(&idev->rs_timer))
 264                 __in6_dev_put(idev);
 265 }
 266
 267 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 268 {
 269         if (del_timer(&ifp->dad_timer))
 270                 __in6_ifa_put(ifp);
 271 }
 272
 273 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 274                                   unsigned long when)
 275 {
 276         if (!timer_pending(&idev->rs_timer))
 277                 in6_dev_hold(idev);
 278         mod_timer(&idev->rs_timer, jiffies + when);
 279 }
 280
 281 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 282                                    unsigned long when)
 283 {
 284         if (!timer_pending(&ifp->dad_timer))
 285                 in6_ifa_hold(ifp);
 286         mod_timer(&ifp->dad_timer, jiffies + when);
 287 }
 288
 289 static int snmp6_alloc_dev(struct inet6_dev *idev)
 290 {
 291         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 292                           sizeof(struct ipstats_mib),
 293                           __alignof__(struct ipstats_mib)) < 0)
 294                 goto err_ip;
 295         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 296                                         GFP_KERNEL);
 297         if (!idev->stats.icmpv6dev)
 298                 goto err_icmp;
 299         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 300                                            GFP_KERNEL);
 301         if (!idev->stats.icmpv6msgdev)
 302                 goto err_icmpmsg;
 303
 304         return 0;
 305
 306 err_icmpmsg:
 307         kfree(idev->stats.icmpv6dev);
 308 err_icmp:
 309         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 310 err_ip:
 311         return -ENOMEM;
 312 }
 313
 314 static void snmp6_free_dev(struct inet6_dev *idev)
 315 {
 316         kfree(idev->stats.icmpv6msgdev);
 317         kfree(idev->stats.icmpv6dev);
 318         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 319 }
 320
 321 /* Nobody refers to this device, we may destroy it. */
 322
 323 void in6_dev_finish_destroy(struct inet6_dev *idev)
 324 {
 325         struct net_device *dev = idev->dev;
 326
 327         WARN_ON(!list_empty(&idev->addr_list));
 328         WARN_ON(idev->mc_list != NULL);
 329         WARN_ON(timer_pending(&idev->rs_timer));
 330
 331 #ifdef NET_REFCNT_DEBUG
 332         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 333 #endif
 334         dev_put(dev);
 335         if (!idev->dead) {
 336                 pr_warn("Freeing alive inet6 device %p\n", idev);
 337                 return;
 338         }
 339         snmp6_free_dev(idev);
 340         kfree_rcu(idev, rcu);
 341 }
 342 EXPORT_SYMBOL(in6_dev_finish_destroy);
 343
 344 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 345 {
 346         struct inet6_dev *ndev;
 347
 348         ASSERT_RTNL();
 349
 350         if (dev->mtu < IPV6_MIN_MTU)
 351                 return NULL;
 352
 353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 354
 355         if (ndev == NULL)
 356                 return NULL;
 357
 358         rwlock_init(&ndev->lock);
 359         ndev->dev = dev;
 360         INIT_LIST_HEAD(&ndev->addr_list);
 361         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 362                     (unsigned long)ndev);
 363         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 364         ndev->cnf.mtu6 = dev->mtu;
 365         ndev->cnf.sysctl = NULL;
 366         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 367         if (ndev->nd_parms == NULL) {
 368                 kfree(ndev);
 369                 return NULL;
 370         }
 371         if (ndev->cnf.forwarding)
 372                 dev_disable_lro(dev);
 373         /* We refer to the device */
 374         dev_hold(dev);
 375
 376         if (snmp6_alloc_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 dev_put(dev);
 382                 kfree(ndev);
 383                 return NULL;
 384         }
 385
 386         if (snmp6_register_dev(ndev) < 0) {
 387                 ADBG((KERN_WARNING
 388                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 389                         __func__, dev->name));
 390                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 391                 ndev->dead = 1;
 392                 in6_dev_finish_destroy(ndev);
 393                 return NULL;
 394         }
 395
 396         /* One reference from device.  We must do this before
 397          * we invoke __ipv6_regen_rndid().
 398          */
 399         in6_dev_hold(ndev);
 400
 401         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 402                 ndev->cnf.accept_dad = -1;
 403
 404 #if IS_ENABLED(CONFIG_IPV6_SIT)
 405         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 406                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 407                 ndev->cnf.rtr_solicits = 0;
 408         }
 409 #endif
 410
 411 #ifdef CONFIG_IPV6_PRIVACY
 412         INIT_LIST_HEAD(&ndev->tempaddr_list);
 413         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 414         if ((dev->flags&IFF_LOOPBACK) ||
 415             dev->type == ARPHRD_TUNNEL ||
 416             dev->type == ARPHRD_TUNNEL6 ||
 417             dev->type == ARPHRD_SIT ||
 418             dev->type == ARPHRD_NONE) {
 419                 ndev->cnf.use_tempaddr = -1;
 420         } else {
 421                 in6_dev_hold(ndev);
 422                 ipv6_regen_rndid((unsigned long) ndev);
 423         }
 424 #endif
 425         ndev->token = in6addr_any;
 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 interface-local all-node multicast group */
 437         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 438
 439         /* Join all-node multicast group */
 440         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 441
 442         /* Join all-router multicast group if forwarding is set */
 443         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 444                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 445
 446         return ndev;
 447 }
 448
 449 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 450 {
 451         struct inet6_dev *idev;
 452
 453         ASSERT_RTNL();
 454
 455         idev = __in6_dev_get(dev);
 456         if (!idev) {
 457                 idev = ipv6_add_dev(dev);
 458                 if (!idev)
 459                         return NULL;
 460         }
 461
 462         if (dev->flags&IFF_UP)
 463                 ipv6_mc_up(idev);
 464         return idev;
 465 }
 466
 467 static int inet6_netconf_msgsize_devconf(int type)
 468 {
 469         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 470                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 471
 472         /* type -1 is used for ALL */
 473         if (type == -1 || type == NETCONFA_FORWARDING)
 474                 size += nla_total_size(4);
 475 #ifdef CONFIG_IPV6_MROUTE
 476         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 477                 size += nla_total_size(4);
 478 #endif
 479
 480         return size;
 481 }
 482
 483 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 484                                       struct ipv6_devconf *devconf, u32 portid,
 485                                       u32 seq, int event, unsigned int flags,
 486                                       int type)
 487 {
 488         struct nlmsghdr  *nlh;
 489         struct netconfmsg *ncm;
 490
 491         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 492                         flags);
 493         if (nlh == NULL)
 494                 return -EMSGSIZE;
 495
 496         ncm = nlmsg_data(nlh);
 497         ncm->ncm_family = AF_INET6;
 498
 499         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 500                 goto nla_put_failure;
 501
 502         /* type -1 is used for ALL */
 503         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 504             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 505                 goto nla_put_failure;
 506 #ifdef CONFIG_IPV6_MROUTE
 507         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 508             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 509                         devconf->mc_forwarding) < 0)
 510                 goto nla_put_failure;
 511 #endif
 512         return nlmsg_end(skb, nlh);
 513
 514 nla_put_failure:
 515         nlmsg_cancel(skb, nlh);
 516         return -EMSGSIZE;
 517 }
 518
 519 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 520                                   struct ipv6_devconf *devconf)
 521 {
 522         struct sk_buff *skb;
 523         int err = -ENOBUFS;
 524
 525         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 526         if (skb == NULL)
 527                 goto errout;
 528
 529         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 530                                          RTM_NEWNETCONF, 0, type);
 531         if (err < 0) {
 532                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 533                 WARN_ON(err == -EMSGSIZE);
 534                 kfree_skb(skb);
 535                 goto errout;
 536         }
 537         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 538         return;
 539 errout:
 540         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 541 }
 542
 543 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 544         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 545         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 546 };
 547
 548 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 549                                      struct nlmsghdr *nlh)
 550 {
 551         struct net *net = sock_net(in_skb->sk);
 552         struct nlattr *tb[NETCONFA_MAX+1];
 553         struct netconfmsg *ncm;
 554         struct sk_buff *skb;
 555         struct ipv6_devconf *devconf;
 556         struct inet6_dev *in6_dev;
 557         struct net_device *dev;
 558         int ifindex;
 559         int err;
 560
 561         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 562                           devconf_ipv6_policy);
 563         if (err < 0)
 564                 goto errout;
 565
 566         err = EINVAL;
 567         if (!tb[NETCONFA_IFINDEX])
 568                 goto errout;
 569
 570         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 571         switch (ifindex) {
 572         case NETCONFA_IFINDEX_ALL:
 573                 devconf = net->ipv6.devconf_all;
 574                 break;
 575         case NETCONFA_IFINDEX_DEFAULT:
 576                 devconf = net->ipv6.devconf_dflt;
 577                 break;
 578         default:
 579                 dev = __dev_get_by_index(net, ifindex);
 580                 if (dev == NULL)
 581                         goto errout;
 582                 in6_dev = __in6_dev_get(dev);
 583                 if (in6_dev == NULL)
 584                         goto errout;
 585                 devconf = &in6_dev->cnf;
 586                 break;
 587         }
 588
 589         err = -ENOBUFS;
 590         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 591         if (skb == NULL)
 592                 goto errout;
 593
 594         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 595                                          NETLINK_CB(in_skb).portid,
 596                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 597                                          -1);
 598         if (err < 0) {
 599                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 600                 WARN_ON(err == -EMSGSIZE);
 601                 kfree_skb(skb);
 602                 goto errout;
 603         }
 604         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 605 errout:
 606         return err;
 607 }
 608
 609 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 610                                       struct netlink_callback *cb)
 611 {
 612         struct net *net = sock_net(skb->sk);
 613         int h, s_h;
 614         int idx, s_idx;
 615         struct net_device *dev;
 616         struct inet6_dev *idev;
 617         struct hlist_head *head;
 618
 619         s_h = cb->args[0];
 620         s_idx = idx = cb->args[1];
 621
 622         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 623                 idx = 0;
 624                 head = &net->dev_index_head[h];
 625                 rcu_read_lock();
 626                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 627                           net->dev_base_seq;
 628                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 629                         if (idx < s_idx)
 630                                 goto cont;
 631                         idev = __in6_dev_get(dev);
 632                         if (!idev)
 633                                 goto cont;
 634
 635                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 636                                                        &idev->cnf,
 637                                                        NETLINK_CB(cb->skb).portid,
 638                                                        cb->nlh->nlmsg_seq,
 639                                                        RTM_NEWNETCONF,
 640                                                        NLM_F_MULTI,
 641                                                        -1) <= 0) {
 642                                 rcu_read_unlock();
 643                                 goto done;
 644                         }
 645                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 646 cont:
 647                         idx++;
 648                 }
 649                 rcu_read_unlock();
 650         }
 651         if (h == NETDEV_HASHENTRIES) {
 652                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 653                                                net->ipv6.devconf_all,
 654                                                NETLINK_CB(cb->skb).portid,
 655                                                cb->nlh->nlmsg_seq,
 656                                                RTM_NEWNETCONF, NLM_F_MULTI,
 657                                                -1) <= 0)
 658                         goto done;
 659                 else
 660                         h++;
 661         }
 662         if (h == NETDEV_HASHENTRIES + 1) {
 663                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 664                                                net->ipv6.devconf_dflt,
 665                                                NETLINK_CB(cb->skb).portid,
 666                                                cb->nlh->nlmsg_seq,
 667                                                RTM_NEWNETCONF, NLM_F_MULTI,
 668                                                -1) <= 0)
 669                         goto done;
 670                 else
 671                         h++;
 672         }
 673 done:
 674         cb->args[0] = h;
 675         cb->args[1] = idx;
 676
 677         return skb->len;
 678 }
 679
 680 #ifdef CONFIG_SYSCTL
 681 static void dev_forward_change(struct inet6_dev *idev)
 682 {
 683         struct net_device *dev;
 684         struct inet6_ifaddr *ifa;
 685
 686         if (!idev)
 687                 return;
 688         dev = idev->dev;
 689         if (idev->cnf.forwarding)
 690                 dev_disable_lro(dev);
 691         if (dev->flags & IFF_MULTICAST) {
 692                 if (idev->cnf.forwarding) {
 693                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 694                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 695                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 696                 } else {
 697                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 698                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 699                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 700                 }
 701         }
 702
 703         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 704                 if (ifa->flags&IFA_F_TENTATIVE)
 705                         continue;
 706                 if (idev->cnf.forwarding)
 707                         addrconf_join_anycast(ifa);
 708                 else
 709                         addrconf_leave_anycast(ifa);
 710         }
 711         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 712                                      dev->ifindex, &idev->cnf);
 713 }
 714
 715
 716 static void addrconf_forward_change(struct net *net, __s32 newf)
 717 {
 718         struct net_device *dev;
 719         struct inet6_dev *idev;
 720
 721         for_each_netdev(net, dev) {
 722                 idev = __in6_dev_get(dev);
 723                 if (idev) {
 724                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 725                         idev->cnf.forwarding = newf;
 726                         if (changed)
 727                                 dev_forward_change(idev);
 728                 }
 729         }
 730 }
 731
 732 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 733 {
 734         struct net *net;
 735         int old;
 736
 737         if (!rtnl_trylock())
 738                 return restart_syscall();
 739
 740         net = (struct net *)table->extra2;
 741         old = *p;
 742         *p = newf;
 743
 744         if (p == &net->ipv6.devconf_dflt->forwarding) {
 745                 if ((!newf) ^ (!old))
 746                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 747                                                      NETCONFA_IFINDEX_DEFAULT,
 748                                                      net->ipv6.devconf_dflt);
 749                 rtnl_unlock();
 750                 return 0;
 751         }
 752
 753         if (p == &net->ipv6.devconf_all->forwarding) {
 754                 net->ipv6.devconf_dflt->forwarding = newf;
 755                 addrconf_forward_change(net, newf);
 756                 if ((!newf) ^ (!old))
 757                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 758                                                      NETCONFA_IFINDEX_ALL,
 759                                                      net->ipv6.devconf_all);
 760         } else if ((!newf) ^ (!old))
 761                 dev_forward_change((struct inet6_dev *)table->extra1);
 762         rtnl_unlock();
 763
 764         if (newf)
 765                 rt6_purge_dflt_routers(net);
 766         return 1;
 767 }
 768 #endif
 769
 770 /* Nobody refers to this ifaddr, destroy it */
 771 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 772 {
 773         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 774
 775 #ifdef NET_REFCNT_DEBUG
 776         pr_debug("%s\n", __func__);
 777 #endif
 778
 779         in6_dev_put(ifp->idev);
 780
 781         if (del_timer(&ifp->dad_timer))
 782                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 783
 784         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 785                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 786                 return;
 787         }
 788         ip6_rt_put(ifp->rt);
 789
 790         kfree_rcu(ifp, rcu);
 791 }
 792
 793 static void
 794 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 795 {
 796         struct list_head *p;
 797         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 798
 799         /*
 800          * Each device address list is sorted in order of scope -
 801          * global before linklocal.
 802          */
 803         list_for_each(p, &idev->addr_list) {
 804                 struct inet6_ifaddr *ifa
 805                         = list_entry(p, struct inet6_ifaddr, if_list);
 806                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 807                         break;
 808         }
 809
 810         list_add_tail(&ifp->if_list, p);
 811 }
 812
 813 static u32 inet6_addr_hash(const struct in6_addr *addr)
 814 {
 815         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 816 }
 817
 818 /* On success it returns ifp with increased reference count */
 819
 820 static struct inet6_ifaddr *
 821 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 822               const struct in6_addr *peer_addr, int pfxlen,
 823               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 824 {
 825         struct inet6_ifaddr *ifa = NULL;
 826         struct rt6_info *rt;
 827         unsigned int hash;
 828         int err = 0;
 829         int addr_type = ipv6_addr_type(addr);
 830
 831         if (addr_type == IPV6_ADDR_ANY ||
 832             addr_type & IPV6_ADDR_MULTICAST ||
 833             (!(idev->dev->flags & IFF_LOOPBACK) &&
 834              addr_type & IPV6_ADDR_LOOPBACK))
 835                 return ERR_PTR(-EADDRNOTAVAIL);
 836
 837         rcu_read_lock_bh();
 838         if (idev->dead) {
 839                 err = -ENODEV;                  /*XXX*/
 840                 goto out2;
 841         }
 842
 843         if (idev->cnf.disable_ipv6) {
 844                 err = -EACCES;
 845                 goto out2;
 846         }
 847
 848         spin_lock(&addrconf_hash_lock);
 849
 850         /* Ignore adding duplicate addresses on an interface */
 851         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 852                 ADBG(("ipv6_add_addr: already assigned\n"));
 853                 err = -EEXIST;
 854                 goto out;
 855         }
 856
 857         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 858
 859         if (ifa == NULL) {
 860                 ADBG(("ipv6_add_addr: malloc failed\n"));
 861                 err = -ENOBUFS;
 862                 goto out;
 863         }
 864
 865         rt = addrconf_dst_alloc(idev, addr, false);
 866         if (IS_ERR(rt)) {
 867                 err = PTR_ERR(rt);
 868                 goto out;
 869         }
 870
 871         ifa->addr = *addr;
 872         if (peer_addr)
 873                 ifa->peer_addr = *peer_addr;
 874
 875         spin_lock_init(&ifa->lock);
 876         spin_lock_init(&ifa->state_lock);
 877         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 878                     (unsigned long)ifa);
 879         INIT_HLIST_NODE(&ifa->addr_lst);
 880         ifa->scope = scope;
 881         ifa->prefix_len = pfxlen;
 882         ifa->flags = flags | IFA_F_TENTATIVE;
 883         ifa->valid_lft = valid_lft;
 884         ifa->prefered_lft = prefered_lft;
 885         ifa->cstamp = ifa->tstamp = jiffies;
 886         ifa->tokenized = false;
 887
 888         ifa->rt = rt;
 889
 890         ifa->idev = idev;
 891         in6_dev_hold(idev);
 892         /* For caller */
 893         in6_ifa_hold(ifa);
 894
 895         /* Add to big hash table */
 896         hash = inet6_addr_hash(addr);
 897
 898         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 899         spin_unlock(&addrconf_hash_lock);
 900
 901         write_lock(&idev->lock);
 902         /* Add to inet6_dev unicast addr list. */
 903         ipv6_link_dev_addr(idev, ifa);
 904
 905 #ifdef CONFIG_IPV6_PRIVACY
 906         if (ifa->flags&IFA_F_TEMPORARY) {
 907                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 908                 in6_ifa_hold(ifa);
 909         }
 910 #endif
 911
 912         in6_ifa_hold(ifa);
 913         write_unlock(&idev->lock);
 914 out2:
 915         rcu_read_unlock_bh();
 916
 917         if (likely(err == 0))
 918                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 919         else {
 920                 kfree(ifa);
 921                 ifa = ERR_PTR(err);
 922         }
 923
 924         return ifa;
 925 out:
 926         spin_unlock(&addrconf_hash_lock);
 927         goto out2;
 928 }
 929
 930 /* This function wants to get referenced ifp and releases it before return */
 931
 932 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 933 {
 934         struct inet6_ifaddr *ifa, *ifn;
 935         struct inet6_dev *idev = ifp->idev;
 936         int state;
 937         int deleted = 0, onlink = 0;
 938         unsigned long expires = jiffies;
 939
 940         spin_lock_bh(&ifp->state_lock);
 941         state = ifp->state;
 942         ifp->state = INET6_IFADDR_STATE_DEAD;
 943         spin_unlock_bh(&ifp->state_lock);
 944
 945         if (state == INET6_IFADDR_STATE_DEAD)
 946                 goto out;
 947
 948         spin_lock_bh(&addrconf_hash_lock);
 949         hlist_del_init_rcu(&ifp->addr_lst);
 950         spin_unlock_bh(&addrconf_hash_lock);
 951
 952         write_lock_bh(&idev->lock);
 953 #ifdef CONFIG_IPV6_PRIVACY
 954         if (ifp->flags&IFA_F_TEMPORARY) {
 955                 list_del(&ifp->tmp_list);
 956                 if (ifp->ifpub) {
 957                         in6_ifa_put(ifp->ifpub);
 958                         ifp->ifpub = NULL;
 959                 }
 960                 __in6_ifa_put(ifp);
 961         }
 962 #endif
 963
 964         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 965                 if (ifa == ifp) {
 966                         list_del_init(&ifp->if_list);
 967                         __in6_ifa_put(ifp);
 968
 969                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 970                                 break;
 971                         deleted = 1;
 972                         continue;
 973                 } else if (ifp->flags & IFA_F_PERMANENT) {
 974                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 975                                               ifp->prefix_len)) {
 976                                 if (ifa->flags & IFA_F_PERMANENT) {
 977                                         onlink = 1;
 978                                         if (deleted)
 979                                                 break;
 980                                 } else {
 981                                         unsigned long lifetime;
 982
 983                                         if (!onlink)
 984                                                 onlink = -1;
 985
 986                                         spin_lock(&ifa->lock);
 987
 988                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 989                                         /*
 990                                          * Note: Because this address is
 991                                          * not permanent, lifetime <
 992                                          * LONG_MAX / HZ here.
 993                                          */
 994                                         if (time_before(expires,
 995                                                         ifa->tstamp + lifetime * HZ))
 996                                                 expires = ifa->tstamp + lifetime * HZ;
 997                                         spin_unlock(&ifa->lock);
 998                                 }
 999                         }
1000                 }
1001         }
1002         write_unlock_bh(&idev->lock);
1003
1004         addrconf_del_dad_timer(ifp);
1005
1006         ipv6_ifa_notify(RTM_DELADDR, ifp);
1007
1008         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1009
1010         /*
1011          * Purge or update corresponding prefix
1012          *
1013          * 1) we don't purge prefix here if address was not permanent.
1014          *    prefix is managed by its own lifetime.
1015          * 2) if there're no addresses, delete prefix.
1016          * 3) if there're still other permanent address(es),
1017          *    corresponding prefix is still permanent.
1018          * 4) otherwise, update prefix lifetime to the
1019          *    longest valid lifetime among the corresponding
1020          *    addresses on the device.
1021          *    Note: subsequent RA will update lifetime.
1022          *
1023          * --yoshfuji
1024          */
1025         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1026                 struct in6_addr prefix;
1027                 struct rt6_info *rt;
1028
1029                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1030
1031                 rt = addrconf_get_prefix_route(&prefix,
1032                                                ifp->prefix_len,
1033                                                ifp->idev->dev,
1034                                                0, RTF_GATEWAY | RTF_DEFAULT);
1035
1036                 if (rt) {
1037                         if (onlink == 0) {
1038                                 ip6_del_rt(rt);
1039                                 rt = NULL;
1040                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1041                                 rt6_set_expires(rt, expires);
1042                         }
1043                 }
1044                 ip6_rt_put(rt);
1045         }
1046
1047         /* clean up prefsrc entries */
1048         rt6_remove_prefsrc(ifp);
1049 out:
1050         in6_ifa_put(ifp);
1051 }
1052
1053 #ifdef CONFIG_IPV6_PRIVACY
1054 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1055 {
1056         struct inet6_dev *idev = ifp->idev;
1057         struct in6_addr addr, *tmpaddr;
1058         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1059         unsigned long regen_advance;
1060         int tmp_plen;
1061         int ret = 0;
1062         int max_addresses;
1063         u32 addr_flags;
1064         unsigned long now = jiffies;
1065
1066         write_lock(&idev->lock);
1067         if (ift) {
1068                 spin_lock_bh(&ift->lock);
1069                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1070                 spin_unlock_bh(&ift->lock);
1071                 tmpaddr = &addr;
1072         } else {
1073                 tmpaddr = NULL;
1074         }
1075 retry:
1076         in6_dev_hold(idev);
1077         if (idev->cnf.use_tempaddr <= 0) {
1078                 write_unlock(&idev->lock);
1079                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1080                 in6_dev_put(idev);
1081                 ret = -1;
1082                 goto out;
1083         }
1084         spin_lock_bh(&ifp->lock);
1085         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1086                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1087                 spin_unlock_bh(&ifp->lock);
1088                 write_unlock(&idev->lock);
1089                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1090                         __func__);
1091                 in6_dev_put(idev);
1092                 ret = -1;
1093                 goto out;
1094         }
1095         in6_ifa_hold(ifp);
1096         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1097         __ipv6_try_regen_rndid(idev, tmpaddr);
1098         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1099         age = (now - ifp->tstamp) / HZ;
1100         tmp_valid_lft = min_t(__u32,
1101                               ifp->valid_lft,
1102                               idev->cnf.temp_valid_lft + age);
1103         tmp_prefered_lft = min_t(__u32,
1104                                  ifp->prefered_lft,
1105                                  idev->cnf.temp_prefered_lft + age -
1106                                  idev->cnf.max_desync_factor);
1107         tmp_plen = ifp->prefix_len;
1108         max_addresses = idev->cnf.max_addresses;
1109         tmp_tstamp = ifp->tstamp;
1110         spin_unlock_bh(&ifp->lock);
1111
1112         regen_advance = idev->cnf.regen_max_retry *
1113                         idev->cnf.dad_transmits *
1114                         idev->nd_parms->retrans_time / HZ;
1115         write_unlock(&idev->lock);
1116
1117         /* A temporary address is created only if this calculated Preferred
1118          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1119          * an implementation must not create a temporary address with a zero
1120          * Preferred Lifetime.
1121          */
1122         if (tmp_prefered_lft <= regen_advance) {
1123                 in6_ifa_put(ifp);
1124                 in6_dev_put(idev);
1125                 ret = -1;
1126                 goto out;
1127         }
1128
1129         addr_flags = IFA_F_TEMPORARY;
1130         /* set in addrconf_prefix_rcv() */
1131         if (ifp->flags & IFA_F_OPTIMISTIC)
1132                 addr_flags |= IFA_F_OPTIMISTIC;
1133
1134         ift = !max_addresses ||
1135               ipv6_count_addresses(idev) < max_addresses ?
1136                 ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1137                               ipv6_addr_scope(&addr), addr_flags,
1138                               tmp_valid_lft, tmp_prefered_lft) : NULL;
1139         if (IS_ERR_OR_NULL(ift)) {
1140                 in6_ifa_put(ifp);
1141                 in6_dev_put(idev);
1142                 pr_info("%s: retry temporary address regeneration\n", __func__);
1143                 tmpaddr = &addr;
1144                 write_lock(&idev->lock);
1145                 goto retry;
1146         }
1147
1148         spin_lock_bh(&ift->lock);
1149         ift->ifpub = ifp;
1150         ift->cstamp = now;
1151         ift->tstamp = tmp_tstamp;
1152         spin_unlock_bh(&ift->lock);
1153
1154         addrconf_dad_start(ift);
1155         in6_ifa_put(ift);
1156         in6_dev_put(idev);
1157 out:
1158         return ret;
1159 }
1160 #endif
1161
1162 /*
1163  *      Choose an appropriate source address (RFC3484)
1164  */
1165 enum {
1166         IPV6_SADDR_RULE_INIT = 0,
1167         IPV6_SADDR_RULE_LOCAL,
1168         IPV6_SADDR_RULE_SCOPE,
1169         IPV6_SADDR_RULE_PREFERRED,
1170 #ifdef CONFIG_IPV6_MIP6
1171         IPV6_SADDR_RULE_HOA,
1172 #endif
1173         IPV6_SADDR_RULE_OIF,
1174         IPV6_SADDR_RULE_LABEL,
1175 #ifdef CONFIG_IPV6_PRIVACY
1176         IPV6_SADDR_RULE_PRIVACY,
1177 #endif
1178         IPV6_SADDR_RULE_ORCHID,
1179         IPV6_SADDR_RULE_PREFIX,
1180         IPV6_SADDR_RULE_MAX
1181 };
1182
1183 struct ipv6_saddr_score {
1184         int                     rule;
1185         int                     addr_type;
1186         struct inet6_ifaddr     *ifa;
1187         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1188         int                     scopedist;
1189         int                     matchlen;
1190 };
1191
1192 struct ipv6_saddr_dst {
1193         const struct in6_addr *addr;
1194         int ifindex;
1195         int scope;
1196         int label;
1197         unsigned int prefs;
1198 };
1199
1200 static inline int ipv6_saddr_preferred(int type)
1201 {
1202         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1203                 return 1;
1204         return 0;
1205 }
1206
1207 static int ipv6_get_saddr_eval(struct net *net,
1208                                struct ipv6_saddr_score *score,
1209                                struct ipv6_saddr_dst *dst,
1210                                int i)
1211 {
1212         int ret;
1213
1214         if (i <= score->rule) {
1215                 switch (i) {
1216                 case IPV6_SADDR_RULE_SCOPE:
1217                         ret = score->scopedist;
1218                         break;
1219                 case IPV6_SADDR_RULE_PREFIX:
1220                         ret = score->matchlen;
1221                         break;
1222                 default:
1223                         ret = !!test_bit(i, score->scorebits);
1224                 }
1225                 goto out;
1226         }
1227
1228         switch (i) {
1229         case IPV6_SADDR_RULE_INIT:
1230                 /* Rule 0: remember if hiscore is not ready yet */
1231                 ret = !!score->ifa;
1232                 break;
1233         case IPV6_SADDR_RULE_LOCAL:
1234                 /* Rule 1: Prefer same address */
1235                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1236                 break;
1237         case IPV6_SADDR_RULE_SCOPE:
1238                 /* Rule 2: Prefer appropriate scope
1239                  *
1240                  *      ret
1241                  *       ^
1242                  *    -1 |  d 15
1243                  *    ---+--+-+---> scope
1244                  *       |
1245                  *       |             d is scope of the destination.
1246                  *  B-d  |  \
1247                  *       |   \      <- smaller scope is better if
1248                  *  B-15 |    \        if scope is enough for destinaion.
1249                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1250                  * d-C-1 | /
1251                  *       |/         <- greater is better
1252                  *   -C  /             if scope is not enough for destination.
1253                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1254                  *
1255                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1256                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1257                  * Assume B = 0 and we get C > 29.
1258                  */
1259                 ret = __ipv6_addr_src_scope(score->addr_type);
1260                 if (ret >= dst->scope)
1261                         ret = -ret;
1262                 else
1263                         ret -= 128;     /* 30 is enough */
1264                 score->scopedist = ret;
1265                 break;
1266         case IPV6_SADDR_RULE_PREFERRED:
1267                 /* Rule 3: Avoid deprecated and optimistic addresses */
1268                 ret = ipv6_saddr_preferred(score->addr_type) ||
1269                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1270                 break;
1271 #ifdef CONFIG_IPV6_MIP6
1272         case IPV6_SADDR_RULE_HOA:
1273             {
1274                 /* Rule 4: Prefer home address */
1275                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1276                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1277                 break;
1278             }
1279 #endif
1280         case IPV6_SADDR_RULE_OIF:
1281                 /* Rule 5: Prefer outgoing interface */
1282                 ret = (!dst->ifindex ||
1283                        dst->ifindex == score->ifa->idev->dev->ifindex);
1284                 break;
1285         case IPV6_SADDR_RULE_LABEL:
1286                 /* Rule 6: Prefer matching label */
1287                 ret = ipv6_addr_label(net,
1288                                       &score->ifa->addr, score->addr_type,
1289                                       score->ifa->idev->dev->ifindex) == dst->label;
1290                 break;
1291 #ifdef CONFIG_IPV6_PRIVACY
1292         case IPV6_SADDR_RULE_PRIVACY:
1293             {
1294                 /* Rule 7: Prefer public address
1295                  * Note: prefer temporary address if use_tempaddr >= 2
1296                  */
1297                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1298                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1299                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1300                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1301                 break;
1302             }
1303 #endif
1304         case IPV6_SADDR_RULE_ORCHID:
1305                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1306                  *          non-ORCHID vs non-ORCHID
1307                  */
1308                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1309                         ipv6_addr_orchid(dst->addr));
1310                 break;
1311         case IPV6_SADDR_RULE_PREFIX:
1312                 /* Rule 8: Use longest matching prefix */
1313                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1314                 if (ret > score->ifa->prefix_len)
1315                         ret = score->ifa->prefix_len;
1316                 score->matchlen = ret;
1317                 break;
1318         default:
1319                 ret = 0;
1320         }
1321
1322         if (ret)
1323                 __set_bit(i, score->scorebits);
1324         score->rule = i;
1325 out:
1326         return ret;
1327 }
1328
1329 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1330                        const struct in6_addr *daddr, unsigned int prefs,
1331                        struct in6_addr *saddr)
1332 {
1333         struct ipv6_saddr_score scores[2],
1334                                 *score = &scores[0], *hiscore = &scores[1];
1335         struct ipv6_saddr_dst dst;
1336         struct net_device *dev;
1337         int dst_type;
1338
1339         dst_type = __ipv6_addr_type(daddr);
1340         dst.addr = daddr;
1341         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1342         dst.scope = __ipv6_addr_src_scope(dst_type);
1343         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1344         dst.prefs = prefs;
1345
1346         hiscore->rule = -1;
1347         hiscore->ifa = NULL;
1348
1349         rcu_read_lock();
1350
1351         for_each_netdev_rcu(net, dev) {
1352                 struct inet6_dev *idev;
1353
1354                 /* Candidate Source Address (section 4)
1355                  *  - multicast and link-local destination address,
1356                  *    the set of candidate source address MUST only
1357                  *    include addresses assigned to interfaces
1358                  *    belonging to the same link as the outgoing
1359                  *    interface.
1360                  * (- For site-local destination addresses, the
1361                  *    set of candidate source addresses MUST only
1362                  *    include addresses assigned to interfaces
1363                  *    belonging to the same site as the outgoing
1364                  *    interface.)
1365                  */
1366                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1367                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1368                     dst.ifindex && dev->ifindex != dst.ifindex)
1369                         continue;
1370
1371                 idev = __in6_dev_get(dev);
1372                 if (!idev)
1373                         continue;
1374
1375                 read_lock_bh(&idev->lock);
1376                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1377                         int i;
1378
1379                         /*
1380                          * - Tentative Address (RFC2462 section 5.4)
1381                          *  - A tentative address is not considered
1382                          *    "assigned to an interface" in the traditional
1383                          *    sense, unless it is also flagged as optimistic.
1384                          * - Candidate Source Address (section 4)
1385                          *  - In any case, anycast addresses, multicast
1386                          *    addresses, and the unspecified address MUST
1387                          *    NOT be included in a candidate set.
1388                          */
1389                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1390                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1391                                 continue;
1392
1393                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1394
1395                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1396                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1397                                 LIMIT_NETDEBUG(KERN_DEBUG
1398                                                "ADDRCONF: unspecified / multicast address "
1399                                                "assigned as unicast address on %s",
1400                                                dev->name);
1401                                 continue;
1402                         }
1403
1404                         score->rule = -1;
1405                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1406
1407                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1408                                 int minihiscore, miniscore;
1409
1410                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1411                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1412
1413                                 if (minihiscore > miniscore) {
1414                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1415                                             score->scopedist > 0) {
1416                                                 /*
1417                                                  * special case:
1418                                                  * each remaining entry
1419                                                  * has too small (not enough)
1420                                                  * scope, because ifa entries
1421                                                  * are sorted by their scope
1422                                                  * values.
1423                                                  */
1424                                                 goto try_nextdev;
1425                                         }
1426                                         break;
1427                                 } else if (minihiscore < miniscore) {
1428                                         if (hiscore->ifa)
1429                                                 in6_ifa_put(hiscore->ifa);
1430
1431                                         in6_ifa_hold(score->ifa);
1432
1433                                         swap(hiscore, score);
1434
1435                                         /* restore our iterator */
1436                                         score->ifa = hiscore->ifa;
1437
1438                                         break;
1439                                 }
1440                         }
1441                 }
1442 try_nextdev:
1443                 read_unlock_bh(&idev->lock);
1444         }
1445         rcu_read_unlock();
1446
1447         if (!hiscore->ifa)
1448                 return -EADDRNOTAVAIL;
1449
1450         *saddr = hiscore->ifa->addr;
1451         in6_ifa_put(hiscore->ifa);
1452         return 0;
1453 }
1454 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1455
1456 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1457                       unsigned char banned_flags)
1458 {
1459         struct inet6_ifaddr *ifp;
1460         int err = -EADDRNOTAVAIL;
1461
1462         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1463                 if (ifp->scope == IFA_LINK &&
1464                     !(ifp->flags & banned_flags)) {
1465                         *addr = ifp->addr;
1466                         err = 0;
1467                         break;
1468                 }
1469         }
1470         return err;
1471 }
1472
1473 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1474                     unsigned char banned_flags)
1475 {
1476         struct inet6_dev *idev;
1477         int err = -EADDRNOTAVAIL;
1478
1479         rcu_read_lock();
1480         idev = __in6_dev_get(dev);
1481         if (idev) {
1482                 read_lock_bh(&idev->lock);
1483                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1484                 read_unlock_bh(&idev->lock);
1485         }
1486         rcu_read_unlock();
1487         return err;
1488 }
1489
1490 static int ipv6_count_addresses(struct inet6_dev *idev)
1491 {
1492         int cnt = 0;
1493         struct inet6_ifaddr *ifp;
1494
1495         read_lock_bh(&idev->lock);
1496         list_for_each_entry(ifp, &idev->addr_list, if_list)
1497                 cnt++;
1498         read_unlock_bh(&idev->lock);
1499         return cnt;
1500 }
1501
1502 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1503                   const struct net_device *dev, int strict)
1504 {
1505         struct inet6_ifaddr *ifp;
1506         unsigned int hash = inet6_addr_hash(addr);
1507
1508         rcu_read_lock_bh();
1509         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1510                 if (!net_eq(dev_net(ifp->idev->dev), net))
1511                         continue;
1512                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1513                     !(ifp->flags&IFA_F_TENTATIVE) &&
1514                     (dev == NULL || ifp->idev->dev == dev ||
1515                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1516                         rcu_read_unlock_bh();
1517                         return 1;
1518                 }
1519         }
1520
1521         rcu_read_unlock_bh();
1522         return 0;
1523 }
1524 EXPORT_SYMBOL(ipv6_chk_addr);
1525
1526 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1527                                struct net_device *dev)
1528 {
1529         unsigned int hash = inet6_addr_hash(addr);
1530         struct inet6_ifaddr *ifp;
1531
1532         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1533                 if (!net_eq(dev_net(ifp->idev->dev), net))
1534                         continue;
1535                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1536                         if (dev == NULL || ifp->idev->dev == dev)
1537                                 return true;
1538                 }
1539         }
1540         return false;
1541 }
1542
1543 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1544 {
1545         struct inet6_dev *idev;
1546         struct inet6_ifaddr *ifa;
1547         int     onlink;
1548
1549         onlink = 0;
1550         rcu_read_lock();
1551         idev = __in6_dev_get(dev);
1552         if (idev) {
1553                 read_lock_bh(&idev->lock);
1554                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1555                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1556                                                    ifa->prefix_len);
1557                         if (onlink)
1558                                 break;
1559                 }
1560                 read_unlock_bh(&idev->lock);
1561         }
1562         rcu_read_unlock();
1563         return onlink;
1564 }
1565 EXPORT_SYMBOL(ipv6_chk_prefix);
1566
1567 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1568                                      struct net_device *dev, int strict)
1569 {
1570         struct inet6_ifaddr *ifp, *result = NULL;
1571         unsigned int hash = inet6_addr_hash(addr);
1572
1573         rcu_read_lock_bh();
1574         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1575                 if (!net_eq(dev_net(ifp->idev->dev), net))
1576                         continue;
1577                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1578                         if (dev == NULL || ifp->idev->dev == dev ||
1579                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1580                                 result = ifp;
1581                                 in6_ifa_hold(ifp);
1582                                 break;
1583                         }
1584                 }
1585         }
1586         rcu_read_unlock_bh();
1587
1588         return result;
1589 }
1590
1591 /* Gets referenced address, destroys ifaddr */
1592
1593 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1594 {
1595         if (ifp->flags&IFA_F_PERMANENT) {
1596                 spin_lock_bh(&ifp->lock);
1597                 addrconf_del_dad_timer(ifp);
1598                 ifp->flags |= IFA_F_TENTATIVE;
1599                 if (dad_failed)
1600                         ifp->flags |= IFA_F_DADFAILED;
1601                 spin_unlock_bh(&ifp->lock);
1602                 if (dad_failed)
1603                         ipv6_ifa_notify(0, ifp);
1604                 in6_ifa_put(ifp);
1605 #ifdef CONFIG_IPV6_PRIVACY
1606         } else if (ifp->flags&IFA_F_TEMPORARY) {
1607                 struct inet6_ifaddr *ifpub;
1608                 spin_lock_bh(&ifp->lock);
1609                 ifpub = ifp->ifpub;
1610                 if (ifpub) {
1611                         in6_ifa_hold(ifpub);
1612                         spin_unlock_bh(&ifp->lock);
1613                         ipv6_create_tempaddr(ifpub, ifp);
1614                         in6_ifa_put(ifpub);
1615                 } else {
1616                         spin_unlock_bh(&ifp->lock);
1617                 }
1618                 ipv6_del_addr(ifp);
1619 #endif
1620         } else
1621                 ipv6_del_addr(ifp);
1622 }
1623
1624 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1625 {
1626         int err = -ENOENT;
1627
1628         spin_lock(&ifp->state_lock);
1629         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1630                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1631                 err = 0;
1632         }
1633         spin_unlock(&ifp->state_lock);
1634
1635         return err;
1636 }
1637
1638 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1639 {
1640         struct inet6_dev *idev = ifp->idev;
1641
1642         if (addrconf_dad_end(ifp)) {
1643                 in6_ifa_put(ifp);
1644                 return;
1645         }
1646
1647         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1648                              ifp->idev->dev->name, &ifp->addr);
1649
1650         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1651                 struct in6_addr addr;
1652
1653                 addr.s6_addr32[0] = htonl(0xfe800000);
1654                 addr.s6_addr32[1] = 0;
1655
1656                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1657                     ipv6_addr_equal(&ifp->addr, &addr)) {
1658                         /* DAD failed for link-local based on MAC address */
1659                         idev->cnf.disable_ipv6 = 1;
1660
1661                         pr_info("%s: IPv6 being disabled!\n",
1662                                 ifp->idev->dev->name);
1663                 }
1664         }
1665
1666         addrconf_dad_stop(ifp, 1);
1667 }
1668
1669 /* Join to solicited addr multicast group. */
1670
1671 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1672 {
1673         struct in6_addr maddr;
1674
1675         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1676                 return;
1677
1678         addrconf_addr_solict_mult(addr, &maddr);
1679         ipv6_dev_mc_inc(dev, &maddr);
1680 }
1681
1682 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1683 {
1684         struct in6_addr maddr;
1685
1686         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1687                 return;
1688
1689         addrconf_addr_solict_mult(addr, &maddr);
1690         __ipv6_dev_mc_dec(idev, &maddr);
1691 }
1692
1693 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1694 {
1695         struct in6_addr addr;
1696         if (ifp->prefix_len == 127) /* RFC 6164 */
1697                 return;
1698         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1699         if (ipv6_addr_any(&addr))
1700                 return;
1701         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1702 }
1703
1704 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1705 {
1706         struct in6_addr addr;
1707         if (ifp->prefix_len == 127) /* RFC 6164 */
1708                 return;
1709         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1710         if (ipv6_addr_any(&addr))
1711                 return;
1712         __ipv6_dev_ac_dec(ifp->idev, &addr);
1713 }
1714
1715 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1716 {
1717         if (dev->addr_len != ETH_ALEN)
1718                 return -1;
1719         memcpy(eui, dev->dev_addr, 3);
1720         memcpy(eui + 5, dev->dev_addr + 3, 3);
1721
1722         /*
1723          * The zSeries OSA network cards can be shared among various
1724          * OS instances, but the OSA cards have only one MAC address.
1725          * This leads to duplicate address conflicts in conjunction
1726          * with IPv6 if more than one instance uses the same card.
1727          *
1728          * The driver for these cards can deliver a unique 16-bit
1729          * identifier for each instance sharing the same card.  It is
1730          * placed instead of 0xFFFE in the interface identifier.  The
1731          * "u" bit of the interface identifier is not inverted in this
1732          * case.  Hence the resulting interface identifier has local
1733          * scope according to RFC2373.
1734          */
1735         if (dev->dev_id) {
1736                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1737                 eui[4] = dev->dev_id & 0xFF;
1738         } else {
1739                 eui[3] = 0xFF;
1740                 eui[4] = 0xFE;
1741                 eui[0] ^= 2;
1742         }
1743         return 0;
1744 }
1745
1746 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1747 {
1748         if (dev->addr_len != IEEE802154_ADDR_LEN)
1749                 return -1;
1750         memcpy(eui, dev->dev_addr, 8);
1751         eui[0] ^= 2;
1752         return 0;
1753 }
1754
1755 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1756 {
1757         union fwnet_hwaddr *ha;
1758
1759         if (dev->addr_len != FWNET_ALEN)
1760                 return -1;
1761
1762         ha = (union fwnet_hwaddr *)dev->dev_addr;
1763
1764         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1765         eui[0] ^= 2;
1766         return 0;
1767 }
1768
1769 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1770 {
1771         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1772         if (dev->addr_len != ARCNET_ALEN)
1773                 return -1;
1774         memset(eui, 0, 7);
1775         eui[7] = *(u8 *)dev->dev_addr;
1776         return 0;
1777 }
1778
1779 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1780 {
1781         if (dev->addr_len != INFINIBAND_ALEN)
1782                 return -1;
1783         memcpy(eui, dev->dev_addr + 12, 8);
1784         eui[0] |= 2;
1785         return 0;
1786 }
1787
1788 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1789 {
1790         if (addr == 0)
1791                 return -1;
1792         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1793                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1794                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1795                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1796                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1797                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1798         eui[1] = 0;
1799         eui[2] = 0x5E;
1800         eui[3] = 0xFE;
1801         memcpy(eui + 4, &addr, 4);
1802         return 0;
1803 }
1804
1805 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1806 {
1807         if (dev->priv_flags & IFF_ISATAP)
1808                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1809         return -1;
1810 }
1811
1812 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1813 {
1814         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1815 }
1816
1817 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1818 {
1819         switch (dev->type) {
1820         case ARPHRD_ETHER:
1821         case ARPHRD_FDDI:
1822                 return addrconf_ifid_eui48(eui, dev);
1823         case ARPHRD_ARCNET:
1824                 return addrconf_ifid_arcnet(eui, dev);
1825         case ARPHRD_INFINIBAND:
1826                 return addrconf_ifid_infiniband(eui, dev);
1827         case ARPHRD_SIT:
1828                 return addrconf_ifid_sit(eui, dev);
1829         case ARPHRD_IPGRE:
1830                 return addrconf_ifid_gre(eui, dev);
1831         case ARPHRD_IEEE802154:
1832                 return addrconf_ifid_eui64(eui, dev);
1833         case ARPHRD_IEEE1394:
1834                 return addrconf_ifid_ieee1394(eui, dev);
1835         }
1836         return -1;
1837 }
1838
1839 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1840 {
1841         int err = -1;
1842         struct inet6_ifaddr *ifp;
1843
1844         read_lock_bh(&idev->lock);
1845         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1846                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1847                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1848                         err = 0;
1849                         break;
1850                 }
1851         }
1852         read_unlock_bh(&idev->lock);
1853         return err;
1854 }
1855
1856 #ifdef CONFIG_IPV6_PRIVACY
1857 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1858 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1859 {
1860 regen:
1861         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1862         idev->rndid[0] &= ~0x02;
1863
1864         /*
1865          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1866          * check if generated address is not inappropriate
1867          *
1868          *  - Reserved subnet anycast (RFC 2526)
1869          *      11111101 11....11 1xxxxxxx
1870          *  - ISATAP (RFC4214) 6.1
1871          *      00-00-5E-FE-xx-xx-xx-xx
1872          *  - value 0
1873          *  - XXX: already assigned to an address on the device
1874          */
1875         if (idev->rndid[0] == 0xfd &&
1876             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1877             (idev->rndid[7]&0x80))
1878                 goto regen;
1879         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1880                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1881                         goto regen;
1882                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1883                         goto regen;
1884         }
1885 }
1886
1887 static void ipv6_regen_rndid(unsigned long data)
1888 {
1889         struct inet6_dev *idev = (struct inet6_dev *) data;
1890         unsigned long expires;
1891
1892         rcu_read_lock_bh();
1893         write_lock_bh(&idev->lock);
1894
1895         if (idev->dead)
1896                 goto out;
1897
1898         __ipv6_regen_rndid(idev);
1899
1900         expires = jiffies +
1901                 idev->cnf.temp_prefered_lft * HZ -
1902                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1903                 idev->cnf.max_desync_factor * HZ;
1904         if (time_before(expires, jiffies)) {
1905                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1906                         __func__, idev->dev->name);
1907                 goto out;
1908         }
1909
1910         if (!mod_timer(&idev->regen_timer, expires))
1911                 in6_dev_hold(idev);
1912
1913 out:
1914         write_unlock_bh(&idev->lock);
1915         rcu_read_unlock_bh();
1916         in6_dev_put(idev);
1917 }
1918
1919 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1920 {
1921         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1922                 __ipv6_regen_rndid(idev);
1923 }
1924 #endif
1925
1926 /*
1927  *      Add prefix route.
1928  */
1929
1930 static void
1931 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1932                       unsigned long expires, u32 flags)
1933 {
1934         struct fib6_config cfg = {
1935                 .fc_table = RT6_TABLE_PREFIX,
1936                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1937                 .fc_ifindex = dev->ifindex,
1938                 .fc_expires = expires,
1939                 .fc_dst_len = plen,
1940                 .fc_flags = RTF_UP | flags,
1941                 .fc_nlinfo.nl_net = dev_net(dev),
1942                 .fc_protocol = RTPROT_KERNEL,
1943         };
1944
1945         cfg.fc_dst = *pfx;
1946
1947         /* Prevent useless cloning on PtP SIT.
1948            This thing is done here expecting that the whole
1949            class of non-broadcast devices need not cloning.
1950          */
1951 #if IS_ENABLED(CONFIG_IPV6_SIT)
1952         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1953                 cfg.fc_flags |= RTF_NONEXTHOP;
1954 #endif
1955
1956         ip6_route_add(&cfg);
1957 }
1958
1959
1960 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1961                                                   int plen,
1962                                                   const struct net_device *dev,
1963                                                   u32 flags, u32 noflags)
1964 {
1965         struct fib6_node *fn;
1966         struct rt6_info *rt = NULL;
1967         struct fib6_table *table;
1968
1969         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1970         if (table == NULL)
1971                 return NULL;
1972
1973         read_lock_bh(&table->tb6_lock);
1974         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1975         if (!fn)
1976                 goto out;
1977         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1978                 if (rt->dst.dev->ifindex != dev->ifindex)
1979                         continue;
1980                 if ((rt->rt6i_flags & flags) != flags)
1981                         continue;
1982                 if ((rt->rt6i_flags & noflags) != 0)
1983                         continue;
1984                 dst_hold(&rt->dst);
1985                 break;
1986         }
1987 out:
1988         read_unlock_bh(&table->tb6_lock);
1989         return rt;
1990 }
1991
1992
1993 /* Create "default" multicast route to the interface */
1994
1995 static void addrconf_add_mroute(struct net_device *dev)
1996 {
1997         struct fib6_config cfg = {
1998                 .fc_table = RT6_TABLE_LOCAL,
1999                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2000                 .fc_ifindex = dev->ifindex,
2001                 .fc_dst_len = 8,
2002                 .fc_flags = RTF_UP,
2003                 .fc_nlinfo.nl_net = dev_net(dev),
2004         };
2005
2006         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2007
2008         ip6_route_add(&cfg);
2009 }
2010
2011 #if IS_ENABLED(CONFIG_IPV6_SIT)
2012 static void sit_route_add(struct net_device *dev)
2013 {
2014         struct fib6_config cfg = {
2015                 .fc_table = RT6_TABLE_MAIN,
2016                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2017                 .fc_ifindex = dev->ifindex,
2018                 .fc_dst_len = 96,
2019                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2020                 .fc_nlinfo.nl_net = dev_net(dev),
2021         };
2022
2023         /* prefix length - 96 bits "::d.d.d.d" */
2024         ip6_route_add(&cfg);
2025 }
2026 #endif
2027
2028 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2029 {
2030         struct inet6_dev *idev;
2031
2032         ASSERT_RTNL();
2033
2034         idev = ipv6_find_idev(dev);
2035         if (!idev)
2036                 return ERR_PTR(-ENOBUFS);
2037
2038         if (idev->cnf.disable_ipv6)
2039                 return ERR_PTR(-EACCES);
2040
2041         /* Add default multicast route */
2042         if (!(dev->flags & IFF_LOOPBACK))
2043                 addrconf_add_mroute(dev);
2044
2045         return idev;
2046 }
2047
2048 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2049 {
2050         struct prefix_info *pinfo;
2051         __u32 valid_lft;
2052         __u32 prefered_lft;
2053         int addr_type;
2054         struct inet6_dev *in6_dev;
2055         struct net *net = dev_net(dev);
2056
2057         pinfo = (struct prefix_info *) opt;
2058
2059         if (len < sizeof(struct prefix_info)) {
2060                 ADBG(("addrconf: prefix option too short\n"));
2061                 return;
2062         }
2063
2064         /*
2065          *      Validation checks ([ADDRCONF], page 19)
2066          */
2067
2068         addr_type = ipv6_addr_type(&pinfo->prefix);
2069
2070         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2071                 return;
2072
2073         valid_lft = ntohl(pinfo->valid);
2074         prefered_lft = ntohl(pinfo->prefered);
2075
2076         if (prefered_lft > valid_lft) {
2077                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2078                 return;
2079         }
2080
2081         in6_dev = in6_dev_get(dev);
2082
2083         if (in6_dev == NULL) {
2084                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2085                                     dev->name);
2086                 return;
2087         }
2088
2089         /*
2090          *      Two things going on here:
2091          *      1) Add routes for on-link prefixes
2092          *      2) Configure prefixes with the auto flag set
2093          */
2094
2095         if (pinfo->onlink) {
2096                 struct rt6_info *rt;
2097                 unsigned long rt_expires;
2098
2099                 /* Avoid arithmetic overflow. Really, we could
2100                  * save rt_expires in seconds, likely valid_lft,
2101                  * but it would require division in fib gc, that it
2102                  * not good.
2103                  */
2104                 if (HZ > USER_HZ)
2105                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2106                 else
2107                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2108
2109                 if (addrconf_finite_timeout(rt_expires))
2110                         rt_expires *= HZ;
2111
2112                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2113                                                pinfo->prefix_len,
2114                                                dev,
2115                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2116                                                RTF_GATEWAY | RTF_DEFAULT);
2117
2118                 if (rt) {
2119                         /* Autoconf prefix route */
2120                         if (valid_lft == 0) {
2121                                 ip6_del_rt(rt);
2122                                 rt = NULL;
2123                         } else if (addrconf_finite_timeout(rt_expires)) {
2124                                 /* not infinity */
2125                                 rt6_set_expires(rt, jiffies + rt_expires);
2126                         } else {
2127                                 rt6_clean_expires(rt);
2128                         }
2129                 } else if (valid_lft) {
2130                         clock_t expires = 0;
2131                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2132                         if (addrconf_finite_timeout(rt_expires)) {
2133                                 /* not infinity */
2134                                 flags |= RTF_EXPIRES;
2135                                 expires = jiffies_to_clock_t(rt_expires);
2136                         }
2137                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2138                                               dev, expires, flags);
2139                 }
2140                 ip6_rt_put(rt);
2141         }
2142
2143         /* Try to figure out our local address for this prefix */
2144
2145         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2146                 struct inet6_ifaddr *ifp;
2147                 struct in6_addr addr;
2148                 int create = 0, update_lft = 0;
2149                 bool tokenized = false;
2150
2151                 if (pinfo->prefix_len == 64) {
2152                         memcpy(&addr, &pinfo->prefix, 8);
2153
2154                         if (!ipv6_addr_any(&in6_dev->token)) {
2155                                 read_lock_bh(&in6_dev->lock);
2156                                 memcpy(addr.s6_addr + 8,
2157                                        in6_dev->token.s6_addr + 8, 8);
2158                                 read_unlock_bh(&in6_dev->lock);
2159                                 tokenized = true;
2160                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2161                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2162                                 in6_dev_put(in6_dev);
2163                                 return;
2164                         }
2165                         goto ok;
2166                 }
2167                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2168                                     pinfo->prefix_len);
2169                 in6_dev_put(in6_dev);
2170                 return;
2171
2172 ok:
2173
2174                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2175
2176                 if (ifp == NULL && valid_lft) {
2177                         int max_addresses = in6_dev->cnf.max_addresses;
2178                         u32 addr_flags = 0;
2179
2180 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2181                         if (in6_dev->cnf.optimistic_dad &&
2182                             !net->ipv6.devconf_all->forwarding && sllao)
2183                                 addr_flags = IFA_F_OPTIMISTIC;
2184 #endif
2185
2186                         /* Do not allow to create too much of autoconfigured
2187                          * addresses; this would be too easy way to crash kernel.
2188                          */
2189                         if (!max_addresses ||
2190                             ipv6_count_addresses(in6_dev) < max_addresses)
2191                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2192                                                     pinfo->prefix_len,
2193                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2194                                                     addr_flags, valid_lft,
2195                                                     prefered_lft);
2196
2197                         if (IS_ERR_OR_NULL(ifp)) {
2198                                 in6_dev_put(in6_dev);
2199                                 return;
2200                         }
2201
2202                         update_lft = 0;
2203                         create = 1;
2204                         ifp->cstamp = jiffies;
2205                         ifp->tokenized = tokenized;
2206                         addrconf_dad_start(ifp);
2207                 }
2208
2209                 if (ifp) {
2210                         int flags;
2211                         unsigned long now;
2212 #ifdef CONFIG_IPV6_PRIVACY
2213                         struct inet6_ifaddr *ift;
2214 #endif
2215                         u32 stored_lft;
2216
2217                         /* update lifetime (RFC2462 5.5.3 e) */
2218                         spin_lock(&ifp->lock);
2219                         now = jiffies;
2220                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2221                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2222                         else
2223                                 stored_lft = 0;
2224                         if (!update_lft && !create && stored_lft) {
2225                                 if (valid_lft > MIN_VALID_LIFETIME ||
2226                                     valid_lft > stored_lft)
2227                                         update_lft = 1;
2228                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2229                                         /* valid_lft <= stored_lft is always true */
2230                                         /*
2231                                          * RFC 4862 Section 5.5.3e:
2232                                          * "Note that the preferred lifetime of
2233                                          *  the corresponding address is always
2234                                          *  reset to the Preferred Lifetime in
2235                                          *  the received Prefix Information
2236                                          *  option, regardless of whether the
2237                                          *  valid lifetime is also reset or
2238                                          *  ignored."
2239                                          *
2240                                          *  So if the preferred lifetime in
2241                                          *  this advertisement is different
2242                                          *  than what we have stored, but the
2243                                          *  valid lifetime is invalid, just
2244                                          *  reset prefered_lft.
2245                                          *
2246                                          *  We must set the valid lifetime
2247                                          *  to the stored lifetime since we'll
2248                                          *  be updating the timestamp below,
2249                                          *  else we'll set it back to the
2250                                          *  minimum.
2251                                          */
2252                                         if (prefered_lft != ifp->prefered_lft) {
2253                                                 valid_lft = stored_lft;
2254                                                 update_lft = 1;
2255                                         }
2256                                 } else {
2257                                         valid_lft = MIN_VALID_LIFETIME;
2258                                         if (valid_lft < prefered_lft)
2259                                                 prefered_lft = valid_lft;
2260                                         update_lft = 1;
2261                                 }
2262                         }
2263
2264                         if (update_lft) {
2265                                 ifp->valid_lft = valid_lft;
2266                                 ifp->prefered_lft = prefered_lft;
2267                                 ifp->tstamp = now;
2268                                 flags = ifp->flags;
2269                                 ifp->flags &= ~IFA_F_DEPRECATED;
2270                                 spin_unlock(&ifp->lock);
2271
2272                                 if (!(flags&IFA_F_TENTATIVE))
2273                                         ipv6_ifa_notify(0, ifp);
2274                         } else
2275                                 spin_unlock(&ifp->lock);
2276
2277 #ifdef CONFIG_IPV6_PRIVACY
2278                         read_lock_bh(&in6_dev->lock);
2279                         /* update all temporary addresses in the list */
2280                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2281                                             tmp_list) {
2282                                 int age, max_valid, max_prefered;
2283
2284                                 if (ifp != ift->ifpub)
2285                                         continue;
2286
2287                                 /*
2288                                  * RFC 4941 section 3.3:
2289                                  * If a received option will extend the lifetime
2290                                  * of a public address, the lifetimes of
2291                                  * temporary addresses should be extended,
2292                                  * subject to the overall constraint that no
2293                                  * temporary addresses should ever remain
2294                                  * "valid" or "preferred" for a time longer than
2295                                  * (TEMP_VALID_LIFETIME) or
2296                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2297                                  * respectively.
2298                                  */
2299                                 age = (now - ift->cstamp) / HZ;
2300                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2301                                 if (max_valid < 0)
2302                                         max_valid = 0;
2303
2304                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2305                                                in6_dev->cnf.max_desync_factor -
2306                                                age;
2307                                 if (max_prefered < 0)
2308                                         max_prefered = 0;
2309
2310                                 if (valid_lft > max_valid)
2311                                         valid_lft = max_valid;
2312
2313                                 if (prefered_lft > max_prefered)
2314                                         prefered_lft = max_prefered;
2315
2316                                 spin_lock(&ift->lock);
2317                                 flags = ift->flags;
2318                                 ift->valid_lft = valid_lft;
2319                                 ift->prefered_lft = prefered_lft;
2320                                 ift->tstamp = now;
2321                                 if (prefered_lft > 0)
2322                                         ift->flags &= ~IFA_F_DEPRECATED;
2323
2324                                 spin_unlock(&ift->lock);
2325                                 if (!(flags&IFA_F_TENTATIVE))
2326                                         ipv6_ifa_notify(0, ift);
2327                         }
2328
2329                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2330                                 /*
2331                                  * When a new public address is created as
2332                                  * described in [ADDRCONF], also create a new
2333                                  * temporary address. Also create a temporary
2334                                  * address if it's enabled but no temporary
2335                                  * address currently exists.
2336                                  */
2337                                 read_unlock_bh(&in6_dev->lock);
2338                                 ipv6_create_tempaddr(ifp, NULL);
2339                         } else {
2340                                 read_unlock_bh(&in6_dev->lock);
2341                         }
2342 #endif
2343                         in6_ifa_put(ifp);
2344                         addrconf_verify(0);
2345                 }
2346         }
2347         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2348         in6_dev_put(in6_dev);
2349 }
2350
2351 /*
2352  *      Set destination address.
2353  *      Special case for SIT interfaces where we create a new "virtual"
2354  *      device.
2355  */
2356 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2357 {
2358         struct in6_ifreq ireq;
2359         struct net_device *dev;
2360         int err = -EINVAL;
2361
2362         rtnl_lock();
2363
2364         err = -EFAULT;
2365         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2366                 goto err_exit;
2367
2368         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2369
2370         err = -ENODEV;
2371         if (dev == NULL)
2372                 goto err_exit;
2373
2374 #if IS_ENABLED(CONFIG_IPV6_SIT)
2375         if (dev->type == ARPHRD_SIT) {
2376                 const struct net_device_ops *ops = dev->netdev_ops;
2377                 struct ifreq ifr;
2378                 struct ip_tunnel_parm p;
2379
2380                 err = -EADDRNOTAVAIL;
2381                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2382                         goto err_exit;
2383
2384                 memset(&p, 0, sizeof(p));
2385                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2386                 p.iph.saddr = 0;
2387                 p.iph.version = 4;
2388                 p.iph.ihl = 5;
2389                 p.iph.protocol = IPPROTO_IPV6;
2390                 p.iph.ttl = 64;
2391                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2392
2393                 if (ops->ndo_do_ioctl) {
2394                         mm_segment_t oldfs = get_fs();
2395
2396                         set_fs(KERNEL_DS);
2397                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2398                         set_fs(oldfs);
2399                 } else
2400                         err = -EOPNOTSUPP;
2401
2402                 if (err == 0) {
2403                         err = -ENOBUFS;
2404                         dev = __dev_get_by_name(net, p.name);
2405                         if (!dev)
2406                                 goto err_exit;
2407                         err = dev_open(dev);
2408                 }
2409         }
2410 #endif
2411
2412 err_exit:
2413         rtnl_unlock();
2414         return err;
2415 }
2416
2417 /*
2418  *      Manual configuration of address on an interface
2419  */
2420 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2421                           const struct in6_addr *peer_pfx,
2422                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2423                           __u32 valid_lft)
2424 {
2425         struct inet6_ifaddr *ifp;
2426         struct inet6_dev *idev;
2427         struct net_device *dev;
2428         int scope;
2429         u32 flags;
2430         clock_t expires;
2431         unsigned long timeout;
2432
2433         ASSERT_RTNL();
2434
2435         if (plen > 128)
2436                 return -EINVAL;
2437
2438         /* check the lifetime */
2439         if (!valid_lft || prefered_lft > valid_lft)
2440                 return -EINVAL;
2441
2442         dev = __dev_get_by_index(net, ifindex);
2443         if (!dev)
2444                 return -ENODEV;
2445
2446         idev = addrconf_add_dev(dev);
2447         if (IS_ERR(idev))
2448                 return PTR_ERR(idev);
2449
2450         scope = ipv6_addr_scope(pfx);
2451
2452         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2453         if (addrconf_finite_timeout(timeout)) {
2454                 expires = jiffies_to_clock_t(timeout * HZ);
2455                 valid_lft = timeout;
2456                 flags = RTF_EXPIRES;
2457         } else {
2458                 expires = 0;
2459                 flags = 0;
2460                 ifa_flags |= IFA_F_PERMANENT;
2461         }
2462
2463         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2464         if (addrconf_finite_timeout(timeout)) {
2465                 if (timeout == 0)
2466                         ifa_flags |= IFA_F_DEPRECATED;
2467                 prefered_lft = timeout;
2468         }
2469
2470         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2471                             valid_lft, prefered_lft);
2472
2473         if (!IS_ERR(ifp)) {
2474                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2475                                       expires, flags);
2476                 /*
2477                  * Note that section 3.1 of RFC 4429 indicates
2478                  * that the Optimistic flag should not be set for
2479                  * manually configured addresses
2480                  */
2481                 addrconf_dad_start(ifp);
2482                 in6_ifa_put(ifp);
2483                 addrconf_verify(0);
2484                 return 0;
2485         }
2486
2487         return PTR_ERR(ifp);
2488 }
2489
2490 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2491                           unsigned int plen)
2492 {
2493         struct inet6_ifaddr *ifp;
2494         struct inet6_dev *idev;
2495         struct net_device *dev;
2496
2497         if (plen > 128)
2498                 return -EINVAL;
2499
2500         dev = __dev_get_by_index(net, ifindex);
2501         if (!dev)
2502                 return -ENODEV;
2503
2504         if ((idev = __in6_dev_get(dev)) == NULL)
2505                 return -ENXIO;
2506
2507         read_lock_bh(&idev->lock);
2508         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2509                 if (ifp->prefix_len == plen &&
2510                     ipv6_addr_equal(pfx, &ifp->addr)) {
2511                         in6_ifa_hold(ifp);
2512                         read_unlock_bh(&idev->lock);
2513
2514                         ipv6_del_addr(ifp);
2515                         return 0;
2516                 }
2517         }
2518         read_unlock_bh(&idev->lock);
2519         return -EADDRNOTAVAIL;
2520 }
2521
2522
2523 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2524 {
2525         struct in6_ifreq ireq;
2526         int err;
2527
2528         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2529                 return -EPERM;
2530
2531         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2532                 return -EFAULT;
2533
2534         rtnl_lock();
2535         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2536                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2537                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2538         rtnl_unlock();
2539         return err;
2540 }
2541
2542 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2543 {
2544         struct in6_ifreq ireq;
2545         int err;
2546
2547         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2548                 return -EPERM;
2549
2550         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2551                 return -EFAULT;
2552
2553         rtnl_lock();
2554         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2555                              ireq.ifr6_prefixlen);
2556         rtnl_unlock();
2557         return err;
2558 }
2559
2560 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2561                      int plen, int scope)
2562 {
2563         struct inet6_ifaddr *ifp;
2564
2565         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2566                             scope, IFA_F_PERMANENT, 0, 0);
2567         if (!IS_ERR(ifp)) {
2568                 spin_lock_bh(&ifp->lock);
2569                 ifp->flags &= ~IFA_F_TENTATIVE;
2570                 spin_unlock_bh(&ifp->lock);
2571                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2572                 in6_ifa_put(ifp);
2573         }
2574 }
2575
2576 #if IS_ENABLED(CONFIG_IPV6_SIT)
2577 static void sit_add_v4_addrs(struct inet6_dev *idev)
2578 {
2579         struct in6_addr addr;
2580         struct net_device *dev;
2581         struct net *net = dev_net(idev->dev);
2582         int scope;
2583
2584         ASSERT_RTNL();
2585
2586         memset(&addr, 0, sizeof(struct in6_addr));
2587         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2588
2589         if (idev->dev->flags&IFF_POINTOPOINT) {
2590                 addr.s6_addr32[0] = htonl(0xfe800000);
2591                 scope = IFA_LINK;
2592         } else {
2593                 scope = IPV6_ADDR_COMPATv4;
2594         }
2595
2596         if (addr.s6_addr32[3]) {
2597                 add_addr(idev, &addr, 128, scope);
2598                 return;
2599         }
2600
2601         for_each_netdev(net, dev) {
2602                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2603                 if (in_dev && (dev->flags & IFF_UP)) {
2604                         struct in_ifaddr *ifa;
2605
2606                         int flag = scope;
2607
2608                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2609                                 int plen;
2610
2611                                 addr.s6_addr32[3] = ifa->ifa_local;
2612
2613                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2614                                         continue;
2615                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2616                                         if (idev->dev->flags&IFF_POINTOPOINT)
2617                                                 continue;
2618                                         flag |= IFA_HOST;
2619                                 }
2620                                 if (idev->dev->flags&IFF_POINTOPOINT)
2621                                         plen = 64;
2622                                 else
2623                                         plen = 96;
2624
2625                                 add_addr(idev, &addr, plen, flag);
2626                         }
2627                 }
2628         }
2629 }
2630 #endif
2631
2632 static void init_loopback(struct net_device *dev)
2633 {
2634         struct inet6_dev  *idev;
2635         struct net_device *sp_dev;
2636         struct inet6_ifaddr *sp_ifa;
2637         struct rt6_info *sp_rt;
2638
2639         /* ::1 */
2640
2641         ASSERT_RTNL();
2642
2643         if ((idev = ipv6_find_idev(dev)) == NULL) {
2644                 pr_debug("%s: add_dev failed\n", __func__);
2645                 return;
2646         }
2647
2648         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2649
2650         /* Add routes to other interface's IPv6 addresses */
2651         for_each_netdev(dev_net(dev), sp_dev) {
2652                 if (!strcmp(sp_dev->name, dev->name))
2653                         continue;
2654
2655                 idev = __in6_dev_get(sp_dev);
2656                 if (!idev)
2657                         continue;
2658
2659                 read_lock_bh(&idev->lock);
2660                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2661
2662                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2663                                 continue;
2664
2665                         if (sp_ifa->rt)
2666                                 continue;
2667
2668                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2669
2670                         /* Failure cases are ignored */
2671                         if (!IS_ERR(sp_rt)) {
2672                                 sp_ifa->rt = sp_rt;
2673                                 ip6_ins_rt(sp_rt);
2674                         }
2675                 }
2676                 read_unlock_bh(&idev->lock);
2677         }
2678 }
2679
2680 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2681 {
2682         struct inet6_ifaddr *ifp;
2683         u32 addr_flags = IFA_F_PERMANENT;
2684
2685 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2686         if (idev->cnf.optimistic_dad &&
2687             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2688                 addr_flags |= IFA_F_OPTIMISTIC;
2689 #endif
2690
2691
2692         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2693         if (!IS_ERR(ifp)) {
2694                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2695                 addrconf_dad_start(ifp);
2696                 in6_ifa_put(ifp);
2697         }
2698 }
2699
2700 static void addrconf_dev_config(struct net_device *dev)
2701 {
2702         struct in6_addr addr;
2703         struct inet6_dev *idev;
2704
2705         ASSERT_RTNL();
2706
2707         if ((dev->type != ARPHRD_ETHER) &&
2708             (dev->type != ARPHRD_FDDI) &&
2709             (dev->type != ARPHRD_ARCNET) &&
2710             (dev->type != ARPHRD_INFINIBAND) &&
2711             (dev->type != ARPHRD_IEEE802154) &&
2712             (dev->type != ARPHRD_IEEE1394)) {
2713                 /* Alas, we support only Ethernet autoconfiguration. */
2714                 return;
2715         }
2716
2717         idev = addrconf_add_dev(dev);
2718         if (IS_ERR(idev))
2719                 return;
2720
2721         memset(&addr, 0, sizeof(struct in6_addr));
2722         addr.s6_addr32[0] = htonl(0xFE800000);
2723
2724         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2725                 addrconf_add_linklocal(idev, &addr);
2726 }
2727
2728 #if IS_ENABLED(CONFIG_IPV6_SIT)
2729 static void addrconf_sit_config(struct net_device *dev)
2730 {
2731         struct inet6_dev *idev;
2732
2733         ASSERT_RTNL();
2734
2735         /*
2736          * Configure the tunnel with one of our IPv4
2737          * addresses... we should configure all of
2738          * our v4 addrs in the tunnel
2739          */
2740
2741         if ((idev = ipv6_find_idev(dev)) == NULL) {
2742                 pr_debug("%s: add_dev failed\n", __func__);
2743                 return;
2744         }
2745
2746         if (dev->priv_flags & IFF_ISATAP) {
2747                 struct in6_addr addr;
2748
2749                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2750                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2751                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2752                         addrconf_add_linklocal(idev, &addr);
2753                 return;
2754         }
2755
2756         sit_add_v4_addrs(idev);
2757
2758         if (dev->flags&IFF_POINTOPOINT)
2759                 addrconf_add_mroute(dev);
2760         else
2761                 sit_route_add(dev);
2762 }
2763 #endif
2764
2765 #if IS_ENABLED(CONFIG_NET_IPGRE)
2766 static void addrconf_gre_config(struct net_device *dev)
2767 {
2768         struct inet6_dev *idev;
2769         struct in6_addr addr;
2770
2771         ASSERT_RTNL();
2772
2773         if ((idev = ipv6_find_idev(dev)) == NULL) {
2774                 pr_debug("%s: add_dev failed\n", __func__);
2775                 return;
2776         }
2777
2778         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2779         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2780
2781         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2782                 addrconf_add_linklocal(idev, &addr);
2783 }
2784 #endif
2785
2786 static inline int
2787 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2788 {
2789         struct in6_addr lladdr;
2790
2791         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2792                 addrconf_add_linklocal(idev, &lladdr);
2793                 return 0;
2794         }
2795         return -1;
2796 }
2797
2798 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2799 {
2800         struct net_device *link_dev;
2801         struct net *net = dev_net(idev->dev);
2802
2803         /* first try to inherit the link-local address from the link device */
2804         if (idev->dev->iflink &&
2805             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2806                 if (!ipv6_inherit_linklocal(idev, link_dev))
2807                         return;
2808         }
2809         /* then try to inherit it from any device */
2810         for_each_netdev(net, link_dev) {
2811                 if (!ipv6_inherit_linklocal(idev, link_dev))
2812                         return;
2813         }
2814         pr_debug("init ip6-ip6: add_linklocal failed\n");
2815 }
2816
2817 /*
2818  * Autoconfigure tunnel with a link-local address so routing protocols,
2819  * DHCPv6, MLD etc. can be run over the virtual link
2820  */
2821
2822 static void addrconf_ip6_tnl_config(struct net_device *dev)
2823 {
2824         struct inet6_dev *idev;
2825
2826         ASSERT_RTNL();
2827
2828         idev = addrconf_add_dev(dev);
2829         if (IS_ERR(idev)) {
2830                 pr_debug("init ip6-ip6: add_dev failed\n");
2831                 return;
2832         }
2833         ip6_tnl_add_linklocal(idev);
2834 }
2835
2836 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2837                            void *ptr)
2838 {
2839         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2840         struct inet6_dev *idev = __in6_dev_get(dev);
2841         int run_pending = 0;
2842         int err;
2843
2844         switch (event) {
2845         case NETDEV_REGISTER:
2846                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2847                         idev = ipv6_add_dev(dev);
2848                         if (!idev)
2849                                 return notifier_from_errno(-ENOMEM);
2850                 }
2851                 break;
2852
2853         case NETDEV_UP:
2854         case NETDEV_CHANGE:
2855                 if (dev->flags & IFF_SLAVE)
2856                         break;
2857
2858                 if (event == NETDEV_UP) {
2859                         if (!addrconf_qdisc_ok(dev)) {
2860                                 /* device is not ready yet. */
2861                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2862                                         dev->name);
2863                                 break;
2864                         }
2865
2866                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2867                                 idev = ipv6_add_dev(dev);
2868
2869                         if (idev) {
2870                                 idev->if_flags |= IF_READY;
2871                                 run_pending = 1;
2872                         }
2873                 } else {
2874                         if (!addrconf_qdisc_ok(dev)) {
2875                                 /* device is still not ready. */
2876                                 break;
2877                         }
2878
2879                         if (idev) {
2880                                 if (idev->if_flags & IF_READY)
2881                                         /* device is already configured. */
2882                                         break;
2883                                 idev->if_flags |= IF_READY;
2884                         }
2885
2886                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2887                                 dev->name);
2888
2889                         run_pending = 1;
2890                 }
2891
2892                 switch (dev->type) {
2893 #if IS_ENABLED(CONFIG_IPV6_SIT)
2894                 case ARPHRD_SIT:
2895                         addrconf_sit_config(dev);
2896                         break;
2897 #endif
2898 #if IS_ENABLED(CONFIG_NET_IPGRE)
2899                 case ARPHRD_IPGRE:
2900                         addrconf_gre_config(dev);
2901                         break;
2902 #endif
2903                 case ARPHRD_TUNNEL6:
2904                         addrconf_ip6_tnl_config(dev);
2905                         break;
2906                 case ARPHRD_LOOPBACK:
2907                         init_loopback(dev);
2908                         break;
2909
2910                 default:
2911                         addrconf_dev_config(dev);
2912                         break;
2913                 }
2914
2915                 if (idev) {
2916                         if (run_pending)
2917                                 addrconf_dad_run(idev);
2918
2919                         /*
2920                          * If the MTU changed during the interface down,
2921                          * when the interface up, the changed MTU must be
2922                          * reflected in the idev as well as routers.
2923                          */
2924                         if (idev->cnf.mtu6 != dev->mtu &&
2925                             dev->mtu >= IPV6_MIN_MTU) {
2926                                 rt6_mtu_change(dev, dev->mtu);
2927                                 idev->cnf.mtu6 = dev->mtu;
2928                         }
2929                         idev->tstamp = jiffies;
2930                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2931
2932                         /*
2933                          * If the changed mtu during down is lower than
2934                          * IPV6_MIN_MTU stop IPv6 on this interface.
2935                          */
2936                         if (dev->mtu < IPV6_MIN_MTU)
2937                                 addrconf_ifdown(dev, 1);
2938                 }
2939                 break;
2940
2941         case NETDEV_CHANGEMTU:
2942                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2943                         rt6_mtu_change(dev, dev->mtu);
2944                         idev->cnf.mtu6 = dev->mtu;
2945                         break;
2946                 }
2947
2948                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2949                         idev = ipv6_add_dev(dev);
2950                         if (idev)
2951                                 break;
2952                 }
2953
2954                 /*
2955                  * MTU falled under IPV6_MIN_MTU.
2956                  * Stop IPv6 on this interface.
2957                  */
2958
2959         case NETDEV_DOWN:
2960         case NETDEV_UNREGISTER:
2961                 /*
2962                  *      Remove all addresses from this interface.
2963                  */
2964                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2965                 break;
2966
2967         case NETDEV_CHANGENAME:
2968                 if (idev) {
2969                         snmp6_unregister_dev(idev);
2970                         addrconf_sysctl_unregister(idev);
2971                         addrconf_sysctl_register(idev);
2972                         err = snmp6_register_dev(idev);
2973                         if (err)
2974                                 return notifier_from_errno(err);
2975                 }
2976                 break;
2977
2978         case NETDEV_PRE_TYPE_CHANGE:
2979         case NETDEV_POST_TYPE_CHANGE:
2980                 addrconf_type_change(dev, event);
2981                 break;
2982         }
2983
2984         return NOTIFY_OK;
2985 }
2986
2987 /*
2988  *      addrconf module should be notified of a device going up
2989  */
2990 static struct notifier_block ipv6_dev_notf = {
2991         .notifier_call = addrconf_notify,
2992 };
2993
2994 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2995 {
2996         struct inet6_dev *idev;
2997         ASSERT_RTNL();
2998
2999         idev = __in6_dev_get(dev);
3000
3001         if (event == NETDEV_POST_TYPE_CHANGE)
3002                 ipv6_mc_remap(idev);
3003         else if (event == NETDEV_PRE_TYPE_CHANGE)
3004                 ipv6_mc_unmap(idev);
3005 }
3006
3007 static int addrconf_ifdown(struct net_device *dev, int how)
3008 {
3009         struct net *net = dev_net(dev);
3010         struct inet6_dev *idev;
3011         struct inet6_ifaddr *ifa;
3012         int state, i;
3013
3014         ASSERT_RTNL();
3015
3016         rt6_ifdown(net, dev);
3017         neigh_ifdown(&nd_tbl, dev);
3018
3019         idev = __in6_dev_get(dev);
3020         if (idev == NULL)
3021                 return -ENODEV;
3022
3023         /*
3024          * Step 1: remove reference to ipv6 device from parent device.
3025          *         Do not dev_put!
3026          */
3027         if (how) {
3028                 idev->dead = 1;
3029
3030                 /* protected by rtnl_lock */
3031                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3032
3033                 /* Step 1.5: remove snmp6 entry */
3034                 snmp6_unregister_dev(idev);
3035
3036         }
3037
3038         /* Step 2: clear hash table */
3039         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3040                 struct hlist_head *h = &inet6_addr_lst[i];
3041
3042                 spin_lock_bh(&addrconf_hash_lock);
3043         restart:
3044                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3045                         if (ifa->idev == idev) {
3046                                 hlist_del_init_rcu(&ifa->addr_lst);
3047                                 addrconf_del_dad_timer(ifa);
3048                                 goto restart;
3049                         }
3050                 }
3051                 spin_unlock_bh(&addrconf_hash_lock);
3052         }
3053
3054         write_lock_bh(&idev->lock);
3055
3056         addrconf_del_rs_timer(idev);
3057
3058         /* Step 2: clear flags for stateless addrconf */
3059         if (!how)
3060                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3061
3062 #ifdef CONFIG_IPV6_PRIVACY
3063         if (how && del_timer(&idev->regen_timer))
3064                 in6_dev_put(idev);
3065
3066         /* Step 3: clear tempaddr list */
3067         while (!list_empty(&idev->tempaddr_list)) {
3068                 ifa = list_first_entry(&idev->tempaddr_list,
3069                                        struct inet6_ifaddr, tmp_list);
3070                 list_del(&ifa->tmp_list);
3071                 write_unlock_bh(&idev->lock);
3072                 spin_lock_bh(&ifa->lock);
3073
3074                 if (ifa->ifpub) {
3075                         in6_ifa_put(ifa->ifpub);
3076                         ifa->ifpub = NULL;
3077                 }
3078                 spin_unlock_bh(&ifa->lock);
3079                 in6_ifa_put(ifa);
3080                 write_lock_bh(&idev->lock);
3081         }
3082 #endif
3083
3084         while (!list_empty(&idev->addr_list)) {
3085                 ifa = list_first_entry(&idev->addr_list,
3086                                        struct inet6_ifaddr, if_list);
3087                 addrconf_del_dad_timer(ifa);
3088
3089                 list_del(&ifa->if_list);
3090
3091                 write_unlock_bh(&idev->lock);
3092
3093                 spin_lock_bh(&ifa->state_lock);
3094                 state = ifa->state;
3095                 ifa->state = INET6_IFADDR_STATE_DEAD;
3096                 spin_unlock_bh(&ifa->state_lock);
3097
3098                 if (state != INET6_IFADDR_STATE_DEAD) {
3099                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3100                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3101                 }
3102                 in6_ifa_put(ifa);
3103
3104                 write_lock_bh(&idev->lock);
3105         }
3106
3107         write_unlock_bh(&idev->lock);
3108
3109         /* Step 5: Discard multicast list */
3110         if (how)
3111                 ipv6_mc_destroy_dev(idev);
3112         else
3113                 ipv6_mc_down(idev);
3114
3115         idev->tstamp = jiffies;
3116
3117         /* Last: Shot the device (if unregistered) */
3118         if (how) {
3119                 addrconf_sysctl_unregister(idev);
3120                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3121                 neigh_ifdown(&nd_tbl, dev);
3122                 in6_dev_put(idev);
3123         }
3124         return 0;
3125 }
3126
3127 static void addrconf_rs_timer(unsigned long data)
3128 {
3129         struct inet6_dev *idev = (struct inet6_dev *)data;
3130         struct in6_addr lladdr;
3131
3132         write_lock(&idev->lock);
3133         if (idev->dead || !(idev->if_flags & IF_READY))
3134                 goto out;
3135
3136         if (!ipv6_accept_ra(idev))
3137                 goto out;
3138
3139         /* Announcement received after solicitation was sent */
3140         if (idev->if_flags & IF_RA_RCVD)
3141                 goto out;
3142
3143         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3144                 if (!__ipv6_get_lladdr(idev, &lladdr, IFA_F_TENTATIVE))
3145                         ndisc_send_rs(idev->dev, &lladdr,
3146                                       &in6addr_linklocal_allrouters);
3147                 else
3148                         goto out;
3149
3150                 /* The wait after the last probe can be shorter */
3151                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3152                                              idev->cnf.rtr_solicits) ?
3153                                       idev->cnf.rtr_solicit_delay :
3154                                       idev->cnf.rtr_solicit_interval);
3155         } else {
3156                 /*
3157                  * Note: we do not support deprecated "all on-link"
3158                  * assumption any longer.
3159                  */
3160                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3161         }
3162
3163 out:
3164         write_unlock(&idev->lock);
3165         in6_dev_put(idev);
3166 }
3167
3168 /*
3169  *      Duplicate Address Detection
3170  */
3171 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3172 {
3173         unsigned long rand_num;
3174         struct inet6_dev *idev = ifp->idev;
3175
3176         if (ifp->flags & IFA_F_OPTIMISTIC)
3177                 rand_num = 0;
3178         else
3179                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3180
3181         ifp->dad_probes = idev->cnf.dad_transmits;
3182         addrconf_mod_dad_timer(ifp, rand_num);
3183 }
3184
3185 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3186 {
3187         struct inet6_dev *idev = ifp->idev;
3188         struct net_device *dev = idev->dev;
3189
3190         addrconf_join_solict(dev, &ifp->addr);
3191
3192         net_srandom(ifp->addr.s6_addr32[3]);
3193
3194         read_lock_bh(&idev->lock);
3195         spin_lock(&ifp->lock);
3196         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3197                 goto out;
3198
3199         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3200             idev->cnf.accept_dad < 1 ||
3201             !(ifp->flags&IFA_F_TENTATIVE) ||
3202             ifp->flags & IFA_F_NODAD) {
3203                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3204                 spin_unlock(&ifp->lock);
3205                 read_unlock_bh(&idev->lock);
3206
3207                 addrconf_dad_completed(ifp);
3208                 return;
3209         }
3210
3211         if (!(idev->if_flags & IF_READY)) {
3212                 spin_unlock(&ifp->lock);
3213                 read_unlock_bh(&idev->lock);
3214                 /*
3215                  * If the device is not ready:
3216                  * - keep it tentative if it is a permanent address.
3217                  * - otherwise, kill it.
3218                  */
3219                 in6_ifa_hold(ifp);
3220                 addrconf_dad_stop(ifp, 0);
3221                 return;
3222         }
3223
3224         /*
3225          * Optimistic nodes can start receiving
3226          * Frames right away
3227          */
3228         if (ifp->flags & IFA_F_OPTIMISTIC)
3229                 ip6_ins_rt(ifp->rt);
3230
3231         addrconf_dad_kick(ifp);
3232 out:
3233         spin_unlock(&ifp->lock);
3234         read_unlock_bh(&idev->lock);
3235 }
3236
3237 static void addrconf_dad_timer(unsigned long data)
3238 {
3239         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3240         struct inet6_dev *idev = ifp->idev;
3241         struct in6_addr mcaddr;
3242
3243         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3244                 goto out;
3245
3246         write_lock(&idev->lock);
3247         if (idev->dead || !(idev->if_flags & IF_READY)) {
3248                 write_unlock(&idev->lock);
3249                 goto out;
3250         }
3251
3252         spin_lock(&ifp->lock);
3253         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3254                 spin_unlock(&ifp->lock);
3255                 write_unlock(&idev->lock);
3256                 goto out;
3257         }
3258
3259         if (ifp->dad_probes == 0) {
3260                 /*
3261                  * DAD was successful
3262                  */
3263
3264                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3265                 spin_unlock(&ifp->lock);
3266                 write_unlock(&idev->lock);
3267
3268                 addrconf_dad_completed(ifp);
3269
3270                 goto out;
3271         }
3272
3273         ifp->dad_probes--;
3274         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3275         spin_unlock(&ifp->lock);
3276         write_unlock(&idev->lock);
3277
3278         /* send a neighbour solicitation for our addr */
3279         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3280         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3281 out:
3282         in6_ifa_put(ifp);
3283 }
3284
3285 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3286 {
3287         struct net_device *dev = ifp->idev->dev;
3288         struct in6_addr lladdr;
3289         bool send_rs, send_mld;
3290
3291         addrconf_del_dad_timer(ifp);
3292
3293         /*
3294          *      Configure the address for reception. Now it is valid.
3295          */
3296
3297         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3298
3299         /* If added prefix is link local and we are prepared to process
3300            router advertisements, start sending router solicitations.
3301          */
3302
3303         read_lock_bh(&ifp->idev->lock);
3304         spin_lock(&ifp->lock);
3305         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3306                    ifp->idev->valid_ll_addr_cnt == 1;
3307         send_rs = send_mld &&
3308                   ipv6_accept_ra(ifp->idev) &&
3309                   ifp->idev->cnf.rtr_solicits > 0 &&
3310                   (dev->flags&IFF_LOOPBACK) == 0;
3311         spin_unlock(&ifp->lock);
3312         read_unlock_bh(&ifp->idev->lock);
3313
3314         /* While dad is in progress mld report's source address is in6_addrany.
3315          * Resend with proper ll now.
3316          */
3317         if (send_mld)
3318                 ipv6_mc_dad_complete(ifp->idev);
3319
3320         if (send_rs) {
3321                 /*
3322                  *      If a host as already performed a random delay
3323                  *      [...] as part of DAD [...] there is no need
3324                  *      to delay again before sending the first RS
3325                  */
3326                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3327                         return;
3328                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3329
3330                 write_lock_bh(&ifp->idev->lock);
3331                 spin_lock(&ifp->lock);
3332                 ifp->idev->rs_probes = 1;
3333                 ifp->idev->if_flags |= IF_RS_SENT;
3334                 addrconf_mod_rs_timer(ifp->idev,
3335                                       ifp->idev->cnf.rtr_solicit_interval);
3336                 spin_unlock(&ifp->lock);
3337                 write_unlock_bh(&ifp->idev->lock);
3338         }
3339 }
3340
3341 static void addrconf_dad_run(struct inet6_dev *idev)
3342 {
3343         struct inet6_ifaddr *ifp;
3344
3345         read_lock_bh(&idev->lock);
3346         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3347                 spin_lock(&ifp->lock);
3348                 if (ifp->flags & IFA_F_TENTATIVE &&
3349                     ifp->state == INET6_IFADDR_STATE_DAD)
3350                         addrconf_dad_kick(ifp);
3351                 spin_unlock(&ifp->lock);
3352         }
3353         read_unlock_bh(&idev->lock);
3354 }
3355
3356 #ifdef CONFIG_PROC_FS
3357 struct if6_iter_state {
3358         struct seq_net_private p;
3359         int bucket;
3360         int offset;
3361 };
3362
3363 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3364 {
3365         struct inet6_ifaddr *ifa = NULL;
3366         struct if6_iter_state *state = seq->private;
3367         struct net *net = seq_file_net(seq);
3368         int p = 0;
3369
3370         /* initial bucket if pos is 0 */
3371         if (pos == 0) {
3372                 state->bucket = 0;
3373                 state->offset = 0;
3374         }
3375
3376         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3377                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3378                                          addr_lst) {
3379                         if (!net_eq(dev_net(ifa->idev->dev), net))
3380                                 continue;
3381                         /* sync with offset */
3382                         if (p < state->offset) {
3383                                 p++;
3384                                 continue;
3385                         }
3386                         state->offset++;
3387                         return ifa;
3388                 }
3389
3390                 /* prepare for next bucket */
3391                 state->offset = 0;
3392                 p = 0;
3393         }
3394         return NULL;
3395 }
3396
3397 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3398                                          struct inet6_ifaddr *ifa)
3399 {
3400         struct if6_iter_state *state = seq->private;
3401         struct net *net = seq_file_net(seq);
3402
3403         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3404                 if (!net_eq(dev_net(ifa->idev->dev), net))
3405                         continue;
3406                 state->offset++;
3407                 return ifa;
3408         }
3409
3410         while (++state->bucket < IN6_ADDR_HSIZE) {
3411                 state->offset = 0;
3412                 hlist_for_each_entry_rcu_bh(ifa,
3413                                      &inet6_addr_lst[state->bucket], addr_lst) {
3414                         if (!net_eq(dev_net(ifa->idev->dev), net))
3415                                 continue;
3416                         state->offset++;
3417                         return ifa;
3418                 }
3419         }
3420
3421         return NULL;
3422 }
3423
3424 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3425         __acquires(rcu_bh)
3426 {
3427         rcu_read_lock_bh();
3428         return if6_get_first(seq, *pos);
3429 }
3430
3431 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3432 {
3433         struct inet6_ifaddr *ifa;
3434
3435         ifa = if6_get_next(seq, v);
3436         ++*pos;
3437         return ifa;
3438 }
3439
3440 static void if6_seq_stop(struct seq_file *seq, void *v)
3441         __releases(rcu_bh)
3442 {
3443         rcu_read_unlock_bh();
3444 }
3445
3446 static int if6_seq_show(struct seq_file *seq, void *v)
3447 {
3448         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3449         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3450                    &ifp->addr,
3451                    ifp->idev->dev->ifindex,
3452                    ifp->prefix_len,
3453                    ifp->scope,
3454                    ifp->flags,
3455                    ifp->idev->dev->name);
3456         return 0;
3457 }
3458
3459 static const struct seq_operations if6_seq_ops = {
3460         .start  = if6_seq_start,
3461         .next   = if6_seq_next,
3462         .show   = if6_seq_show,
3463         .stop   = if6_seq_stop,
3464 };
3465
3466 static int if6_seq_open(struct inode *inode, struct file *file)
3467 {
3468         return seq_open_net(inode, file, &if6_seq_ops,
3469                             sizeof(struct if6_iter_state));
3470 }
3471
3472 static const struct file_operations if6_fops = {
3473         .owner          = THIS_MODULE,
3474         .open           = if6_seq_open,
3475         .read           = seq_read,
3476         .llseek         = seq_lseek,
3477         .release        = seq_release_net,
3478 };
3479
3480 static int __net_init if6_proc_net_init(struct net *net)
3481 {
3482         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3483                 return -ENOMEM;
3484         return 0;
3485 }
3486
3487 static void __net_exit if6_proc_net_exit(struct net *net)
3488 {
3489         remove_proc_entry("if_inet6", net->proc_net);
3490 }
3491
3492 static struct pernet_operations if6_proc_net_ops = {
3493        .init = if6_proc_net_init,
3494        .exit = if6_proc_net_exit,
3495 };
3496
3497 int __init if6_proc_init(void)
3498 {
3499         return register_pernet_subsys(&if6_proc_net_ops);
3500 }
3501
3502 void if6_proc_exit(void)
3503 {
3504         unregister_pernet_subsys(&if6_proc_net_ops);
3505 }
3506 #endif  /* CONFIG_PROC_FS */
3507
3508 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3509 /* Check if address is a home address configured on any interface. */
3510 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3511 {
3512         int ret = 0;
3513         struct inet6_ifaddr *ifp = NULL;
3514         unsigned int hash = inet6_addr_hash(addr);
3515
3516         rcu_read_lock_bh();
3517         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3518                 if (!net_eq(dev_net(ifp->idev->dev), net))
3519                         continue;
3520                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3521                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3522                         ret = 1;
3523                         break;
3524                 }
3525         }
3526         rcu_read_unlock_bh();
3527         return ret;
3528 }
3529 #endif
3530
3531 /*
3532  *      Periodic address status verification
3533  */
3534
3535 static void addrconf_verify(unsigned long foo)
3536 {
3537         unsigned long now, next, next_sec, next_sched;
3538         struct inet6_ifaddr *ifp;
3539         int i;
3540
3541         rcu_read_lock_bh();
3542         spin_lock(&addrconf_verify_lock);
3543         now = jiffies;
3544         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3545
3546         del_timer(&addr_chk_timer);
3547
3548         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3549 restart:
3550                 hlist_for_each_entry_rcu_bh(ifp,
3551                                          &inet6_addr_lst[i], addr_lst) {
3552                         unsigned long age;
3553
3554                         if (ifp->flags & IFA_F_PERMANENT)
3555                                 continue;
3556
3557                         spin_lock(&ifp->lock);
3558                         /* We try to batch several events at once. */
3559                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3560
3561                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3562                             age >= ifp->valid_lft) {
3563                                 spin_unlock(&ifp->lock);
3564                                 in6_ifa_hold(ifp);
3565                                 ipv6_del_addr(ifp);
3566                                 goto restart;
3567                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3568                                 spin_unlock(&ifp->lock);
3569                                 continue;
3570                         } else if (age >= ifp->prefered_lft) {
3571                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3572                                 int deprecate = 0;
3573
3574                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3575                                         deprecate = 1;
3576                                         ifp->flags |= IFA_F_DEPRECATED;
3577                                 }
3578
3579                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3580                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3581
3582                                 spin_unlock(&ifp->lock);
3583
3584                                 if (deprecate) {
3585                                         in6_ifa_hold(ifp);
3586
3587                                         ipv6_ifa_notify(0, ifp);
3588                                         in6_ifa_put(ifp);
3589                                         goto restart;
3590                                 }
3591 #ifdef CONFIG_IPV6_PRIVACY
3592                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3593                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3594                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3595                                         ifp->idev->cnf.dad_transmits *
3596                                         ifp->idev->nd_parms->retrans_time / HZ;
3597
3598                                 if (age >= ifp->prefered_lft - regen_advance) {
3599                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3600                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3601                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3602                                         if (!ifp->regen_count && ifpub) {
3603                                                 ifp->regen_count++;
3604                                                 in6_ifa_hold(ifp);
3605                                                 in6_ifa_hold(ifpub);
3606                                                 spin_unlock(&ifp->lock);
3607
3608                                                 spin_lock(&ifpub->lock);
3609                                                 ifpub->regen_count = 0;
3610                                                 spin_unlock(&ifpub->lock);
3611                                                 ipv6_create_tempaddr(ifpub, ifp);
3612                                                 in6_ifa_put(ifpub);
3613                                                 in6_ifa_put(ifp);
3614                                                 goto restart;
3615                                         }
3616                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3617                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3618                                 spin_unlock(&ifp->lock);
3619 #endif
3620                         } else {
3621                                 /* ifp->prefered_lft <= ifp->valid_lft */
3622                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3623                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3624                                 spin_unlock(&ifp->lock);
3625                         }
3626                 }
3627         }
3628
3629         next_sec = round_jiffies_up(next);
3630         next_sched = next;
3631
3632         /* If rounded timeout is accurate enough, accept it. */
3633         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3634                 next_sched = next_sec;
3635
3636         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3637         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3638                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3639
3640         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3641               now, next, next_sec, next_sched));
3642
3643         addr_chk_timer.expires = next_sched;
3644         add_timer(&addr_chk_timer);
3645         spin_unlock(&addrconf_verify_lock);
3646         rcu_read_unlock_bh();
3647 }
3648
3649 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3650                                      struct in6_addr **peer_pfx)
3651 {
3652         struct in6_addr *pfx = NULL;
3653
3654         *peer_pfx = NULL;
3655
3656         if (addr)
3657                 pfx = nla_data(addr);
3658
3659         if (local) {
3660                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3661                         *peer_pfx = pfx;
3662                 pfx = nla_data(local);
3663         }
3664
3665         return pfx;
3666 }
3667
3668 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3669         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3670         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3671         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3672 };
3673
3674 static int
3675 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3676 {
3677         struct net *net = sock_net(skb->sk);
3678         struct ifaddrmsg *ifm;
3679         struct nlattr *tb[IFA_MAX+1];
3680         struct in6_addr *pfx, *peer_pfx;
3681         int err;
3682
3683         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3684         if (err < 0)
3685                 return err;
3686
3687         ifm = nlmsg_data(nlh);
3688         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3689         if (pfx == NULL)
3690                 return -EINVAL;
3691
3692         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3693 }
3694
3695 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3696                              u32 prefered_lft, u32 valid_lft)
3697 {
3698         u32 flags;
3699         clock_t expires;
3700         unsigned long timeout;
3701
3702         if (!valid_lft || (prefered_lft > valid_lft))
3703                 return -EINVAL;
3704
3705         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3706         if (addrconf_finite_timeout(timeout)) {
3707                 expires = jiffies_to_clock_t(timeout * HZ);
3708                 valid_lft = timeout;
3709                 flags = RTF_EXPIRES;
3710         } else {
3711                 expires = 0;
3712                 flags = 0;
3713                 ifa_flags |= IFA_F_PERMANENT;
3714         }
3715
3716         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3717         if (addrconf_finite_timeout(timeout)) {
3718                 if (timeout == 0)
3719                         ifa_flags |= IFA_F_DEPRECATED;
3720                 prefered_lft = timeout;
3721         }
3722
3723         spin_lock_bh(&ifp->lock);
3724         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3725         ifp->tstamp = jiffies;
3726         ifp->valid_lft = valid_lft;
3727         ifp->prefered_lft = prefered_lft;
3728
3729         spin_unlock_bh(&ifp->lock);
3730         if (!(ifp->flags&IFA_F_TENTATIVE))
3731                 ipv6_ifa_notify(0, ifp);
3732
3733         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3734                               expires, flags);
3735         addrconf_verify(0);
3736
3737         return 0;
3738 }
3739
3740 static int
3741 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3742 {
3743         struct net *net = sock_net(skb->sk);
3744         struct ifaddrmsg *ifm;
3745         struct nlattr *tb[IFA_MAX+1];
3746         struct in6_addr *pfx, *peer_pfx;
3747         struct inet6_ifaddr *ifa;
3748         struct net_device *dev;
3749         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3750         u8 ifa_flags;
3751         int err;
3752
3753         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3754         if (err < 0)
3755                 return err;
3756
3757         ifm = nlmsg_data(nlh);
3758         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3759         if (pfx == NULL)
3760                 return -EINVAL;
3761
3762         if (tb[IFA_CACHEINFO]) {
3763                 struct ifa_cacheinfo *ci;
3764
3765                 ci = nla_data(tb[IFA_CACHEINFO]);
3766                 valid_lft = ci->ifa_valid;
3767                 preferred_lft = ci->ifa_prefered;
3768         } else {
3769                 preferred_lft = INFINITY_LIFE_TIME;
3770                 valid_lft = INFINITY_LIFE_TIME;
3771         }
3772
3773         dev =  __dev_get_by_index(net, ifm->ifa_index);
3774         if (dev == NULL)
3775                 return -ENODEV;
3776
3777         /* We ignore other flags so far. */
3778         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3779
3780         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3781         if (ifa == NULL) {
3782                 /*
3783                  * It would be best to check for !NLM_F_CREATE here but
3784                  * userspace alreay relies on not having to provide this.
3785                  */
3786                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3787                                       ifm->ifa_prefixlen, ifa_flags,
3788                                       preferred_lft, valid_lft);
3789         }
3790
3791         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3792             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3793                 err = -EEXIST;
3794         else
3795                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3796
3797         in6_ifa_put(ifa);
3798
3799         return err;
3800 }
3801
3802 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3803                           u8 scope, int ifindex)
3804 {
3805         struct ifaddrmsg *ifm;
3806
3807         ifm = nlmsg_data(nlh);
3808         ifm->ifa_family = AF_INET6;
3809         ifm->ifa_prefixlen = prefixlen;
3810         ifm->ifa_flags = flags;
3811         ifm->ifa_scope = scope;
3812         ifm->ifa_index = ifindex;
3813 }
3814
3815 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3816                          unsigned long tstamp, u32 preferred, u32 valid)
3817 {
3818         struct ifa_cacheinfo ci;
3819
3820         ci.cstamp = cstamp_delta(cstamp);
3821         ci.tstamp = cstamp_delta(tstamp);
3822         ci.ifa_prefered = preferred;
3823         ci.ifa_valid = valid;
3824
3825         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3826 }
3827
3828 static inline int rt_scope(int ifa_scope)
3829 {
3830         if (ifa_scope & IFA_HOST)
3831                 return RT_SCOPE_HOST;
3832         else if (ifa_scope & IFA_LINK)
3833                 return RT_SCOPE_LINK;
3834         else if (ifa_scope & IFA_SITE)
3835                 return RT_SCOPE_SITE;
3836         else
3837                 return RT_SCOPE_UNIVERSE;
3838 }
3839
3840 static inline int inet6_ifaddr_msgsize(void)
3841 {
3842         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3843                + nla_total_size(16) /* IFA_LOCAL */
3844                + nla_total_size(16) /* IFA_ADDRESS */
3845                + nla_total_size(sizeof(struct ifa_cacheinfo));
3846 }
3847
3848 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3849                              u32 portid, u32 seq, int event, unsigned int flags)
3850 {
3851         struct nlmsghdr  *nlh;
3852         u32 preferred, valid;
3853
3854         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3855         if (nlh == NULL)
3856                 return -EMSGSIZE;
3857
3858         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3859                       ifa->idev->dev->ifindex);
3860
3861         if (!(ifa->flags&IFA_F_PERMANENT)) {
3862                 preferred = ifa->prefered_lft;
3863                 valid = ifa->valid_lft;
3864                 if (preferred != INFINITY_LIFE_TIME) {
3865                         long tval = (jiffies - ifa->tstamp)/HZ;
3866                         if (preferred > tval)
3867                                 preferred -= tval;
3868                         else
3869                                 preferred = 0;
3870                         if (valid != INFINITY_LIFE_TIME) {
3871                                 if (valid > tval)
3872                                         valid -= tval;
3873                                 else
3874                                         valid = 0;
3875                         }
3876                 }
3877         } else {
3878                 preferred = INFINITY_LIFE_TIME;
3879                 valid = INFINITY_LIFE_TIME;
3880         }
3881
3882         if (!ipv6_addr_any(&ifa->peer_addr)) {
3883                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3884                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3885                         goto error;
3886         } else
3887                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3888                         goto error;
3889
3890         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3891                 goto error;
3892
3893         return nlmsg_end(skb, nlh);
3894
3895 error:
3896         nlmsg_cancel(skb, nlh);
3897         return -EMSGSIZE;
3898 }
3899
3900 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3901                                 u32 portid, u32 seq, int event, u16 flags)
3902 {
3903         struct nlmsghdr  *nlh;
3904         u8 scope = RT_SCOPE_UNIVERSE;
3905         int ifindex = ifmca->idev->dev->ifindex;
3906
3907         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3908                 scope = RT_SCOPE_SITE;
3909
3910         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3911         if (nlh == NULL)
3912                 return -EMSGSIZE;
3913
3914         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3915         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3916             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3917                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3918                 nlmsg_cancel(skb, nlh);
3919                 return -EMSGSIZE;
3920         }
3921
3922         return nlmsg_end(skb, nlh);
3923 }
3924
3925 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3926                                 u32 portid, u32 seq, int event, unsigned int flags)
3927 {
3928         struct nlmsghdr  *nlh;
3929         u8 scope = RT_SCOPE_UNIVERSE;
3930         int ifindex = ifaca->aca_idev->dev->ifindex;
3931
3932         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3933                 scope = RT_SCOPE_SITE;
3934
3935         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3936         if (nlh == NULL)
3937                 return -EMSGSIZE;
3938
3939         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3940         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3941             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3942                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3943                 nlmsg_cancel(skb, nlh);
3944                 return -EMSGSIZE;
3945         }
3946
3947         return nlmsg_end(skb, nlh);
3948 }
3949
3950 enum addr_type_t {
3951         UNICAST_ADDR,
3952         MULTICAST_ADDR,
3953         ANYCAST_ADDR,
3954 };
3955
3956 /* called with rcu_read_lock() */
3957 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3958                           struct netlink_callback *cb, enum addr_type_t type,
3959                           int s_ip_idx, int *p_ip_idx)
3960 {
3961         struct ifmcaddr6 *ifmca;
3962         struct ifacaddr6 *ifaca;
3963         int err = 1;
3964         int ip_idx = *p_ip_idx;
3965
3966         read_lock_bh(&idev->lock);
3967         switch (type) {
3968         case UNICAST_ADDR: {
3969                 struct inet6_ifaddr *ifa;
3970
3971                 /* unicast address incl. temp addr */
3972                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3973                         if (++ip_idx < s_ip_idx)
3974                                 continue;
3975                         err = inet6_fill_ifaddr(skb, ifa,
3976                                                 NETLINK_CB(cb->skb).portid,
3977                                                 cb->nlh->nlmsg_seq,
3978                                                 RTM_NEWADDR,
3979                                                 NLM_F_MULTI);
3980                         if (err <= 0)
3981                                 break;
3982                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3983                 }
3984                 break;
3985         }
3986         case MULTICAST_ADDR:
3987                 /* multicast address */
3988                 for (ifmca = idev->mc_list; ifmca;
3989                      ifmca = ifmca->next, ip_idx++) {
3990                         if (ip_idx < s_ip_idx)
3991                                 continue;
3992                         err = inet6_fill_ifmcaddr(skb, ifmca,
3993                                                   NETLINK_CB(cb->skb).portid,
3994                                                   cb->nlh->nlmsg_seq,
3995                                                   RTM_GETMULTICAST,
3996                                                   NLM_F_MULTI);
3997                         if (err <= 0)
3998                                 break;
3999                 }
4000                 break;
4001         case ANYCAST_ADDR:
4002                 /* anycast address */
4003                 for (ifaca = idev->ac_list; ifaca;
4004                      ifaca = ifaca->aca_next, ip_idx++) {
4005                         if (ip_idx < s_ip_idx)
4006                                 continue;
4007                         err = inet6_fill_ifacaddr(skb, ifaca,
4008                                                   NETLINK_CB(cb->skb).portid,
4009                                                   cb->nlh->nlmsg_seq,
4010                                                   RTM_GETANYCAST,
4011                                                   NLM_F_MULTI);
4012                         if (err <= 0)
4013                                 break;
4014                 }
4015                 break;
4016         default:
4017                 break;
4018         }
4019         read_unlock_bh(&idev->lock);
4020         *p_ip_idx = ip_idx;
4021         return err;
4022 }
4023
4024 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4025                            enum addr_type_t type)
4026 {
4027         struct net *net = sock_net(skb->sk);
4028         int h, s_h;
4029         int idx, ip_idx;
4030         int s_idx, s_ip_idx;
4031         struct net_device *dev;
4032         struct inet6_dev *idev;
4033         struct hlist_head *head;
4034
4035         s_h = cb->args[0];
4036         s_idx = idx = cb->args[1];
4037         s_ip_idx = ip_idx = cb->args[2];
4038
4039         rcu_read_lock();
4040         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4041         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4042                 idx = 0;
4043                 head = &net->dev_index_head[h];
4044                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4045                         if (idx < s_idx)
4046                                 goto cont;
4047                         if (h > s_h || idx > s_idx)
4048                                 s_ip_idx = 0;
4049                         ip_idx = 0;
4050                         idev = __in6_dev_get(dev);
4051                         if (!idev)
4052                                 goto cont;
4053
4054                         if (in6_dump_addrs(idev, skb, cb, type,
4055                                            s_ip_idx, &ip_idx) <= 0)
4056                                 goto done;
4057 cont:
4058                         idx++;
4059                 }
4060         }
4061 done:
4062         rcu_read_unlock();
4063         cb->args[0] = h;
4064         cb->args[1] = idx;
4065         cb->args[2] = ip_idx;
4066
4067         return skb->len;
4068 }
4069
4070 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4071 {
4072         enum addr_type_t type = UNICAST_ADDR;
4073
4074         return inet6_dump_addr(skb, cb, type);
4075 }
4076
4077 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4078 {
4079         enum addr_type_t type = MULTICAST_ADDR;
4080
4081         return inet6_dump_addr(skb, cb, type);
4082 }
4083
4084
4085 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4086 {
4087         enum addr_type_t type = ANYCAST_ADDR;
4088
4089         return inet6_dump_addr(skb, cb, type);
4090 }
4091
4092 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4093 {
4094         struct net *net = sock_net(in_skb->sk);
4095         struct ifaddrmsg *ifm;
4096         struct nlattr *tb[IFA_MAX+1];
4097         struct in6_addr *addr = NULL, *peer;
4098         struct net_device *dev = NULL;
4099         struct inet6_ifaddr *ifa;
4100         struct sk_buff *skb;
4101         int err;
4102
4103         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4104         if (err < 0)
4105                 goto errout;
4106
4107         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4108         if (addr == NULL) {
4109                 err = -EINVAL;
4110                 goto errout;
4111         }
4112
4113         ifm = nlmsg_data(nlh);
4114         if (ifm->ifa_index)
4115                 dev = __dev_get_by_index(net, ifm->ifa_index);
4116
4117         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4118         if (!ifa) {
4119                 err = -EADDRNOTAVAIL;
4120                 goto errout;
4121         }
4122
4123         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4124         if (!skb) {
4125                 err = -ENOBUFS;
4126                 goto errout_ifa;
4127         }
4128
4129         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4130                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4131         if (err < 0) {
4132                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4133                 WARN_ON(err == -EMSGSIZE);
4134                 kfree_skb(skb);
4135                 goto errout_ifa;
4136         }
4137         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4138 errout_ifa:
4139         in6_ifa_put(ifa);
4140 errout:
4141         return err;
4142 }
4143
4144 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4145 {
4146         struct sk_buff *skb;
4147         struct net *net = dev_net(ifa->idev->dev);
4148         int err = -ENOBUFS;
4149
4150         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4151         if (skb == NULL)
4152                 goto errout;
4153
4154         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4155         if (err < 0) {
4156                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4157                 WARN_ON(err == -EMSGSIZE);
4158                 kfree_skb(skb);
4159                 goto errout;
4160         }
4161         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4162         return;
4163 errout:
4164         if (err < 0)
4165                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4166 }
4167
4168 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4169                                 __s32 *array, int bytes)
4170 {
4171         BUG_ON(bytes < (DEVCONF_MAX * 4));
4172
4173         memset(array, 0, bytes);
4174         array[DEVCONF_FORWARDING] = cnf->forwarding;
4175         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4176         array[DEVCONF_MTU6] = cnf->mtu6;
4177         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4178         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4179         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4180         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4181         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4182         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4183                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4184         array[DEVCONF_RTR_SOLICIT_DELAY] =
4185                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4186         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4187         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4188                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4189         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4190                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4191 #ifdef CONFIG_IPV6_PRIVACY
4192         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4193         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4194         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4195         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4196         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4197 #endif
4198         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4199         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4200         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4201 #ifdef CONFIG_IPV6_ROUTER_PREF
4202         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4203         array[DEVCONF_RTR_PROBE_INTERVAL] =
4204                 jiffies_to_msecs(cnf->rtr_probe_interval);
4205 #ifdef CONFIG_IPV6_ROUTE_INFO
4206         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4207 #endif
4208 #endif
4209         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4210         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4211 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4212         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4213 #endif
4214 #ifdef CONFIG_IPV6_MROUTE
4215         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4216 #endif
4217         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4218         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4219         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4220         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4221 }
4222
4223 static inline size_t inet6_ifla6_size(void)
4224 {
4225         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4226              + nla_total_size(sizeof(struct ifla_cacheinfo))
4227              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4228              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4229              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4230              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4231 }
4232
4233 static inline size_t inet6_if_nlmsg_size(void)
4234 {
4235         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4236                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4237                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4238                + nla_total_size(4) /* IFLA_MTU */
4239                + nla_total_size(4) /* IFLA_LINK */
4240                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4241 }
4242
4243 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4244                                       int items, int bytes)
4245 {
4246         int i;
4247         int pad = bytes - sizeof(u64) * items;
4248         BUG_ON(pad < 0);
4249
4250         /* Use put_unaligned() because stats may not be aligned for u64. */
4251         put_unaligned(items, &stats[0]);
4252         for (i = 1; i < items; i++)
4253                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4254
4255         memset(&stats[items], 0, pad);
4256 }
4257
4258 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4259                                       int items, int bytes, size_t syncpoff)
4260 {
4261         int i;
4262         int pad = bytes - sizeof(u64) * items;
4263         BUG_ON(pad < 0);
4264
4265         /* Use put_unaligned() because stats may not be aligned for u64. */
4266         put_unaligned(items, &stats[0]);
4267         for (i = 1; i < items; i++)
4268                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4269
4270         memset(&stats[items], 0, pad);
4271 }
4272
4273 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4274                              int bytes)
4275 {
4276         switch (attrtype) {
4277         case IFLA_INET6_STATS:
4278                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4279                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4280                 break;
4281         case IFLA_INET6_ICMP6STATS:
4282                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4283                 break;
4284         }
4285 }
4286
4287 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4288 {
4289         struct nlattr *nla;
4290         struct ifla_cacheinfo ci;
4291
4292         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4293                 goto nla_put_failure;
4294         ci.max_reasm_len = IPV6_MAXPLEN;
4295         ci.tstamp = cstamp_delta(idev->tstamp);
4296         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4297         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4298         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4299                 goto nla_put_failure;
4300         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4301         if (nla == NULL)
4302                 goto nla_put_failure;
4303         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4304
4305         /* XXX - MC not implemented */
4306
4307         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4308         if (nla == NULL)
4309                 goto nla_put_failure;
4310         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4311
4312         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4313         if (nla == NULL)
4314                 goto nla_put_failure;
4315         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4316
4317         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4318         if (nla == NULL)
4319                 goto nla_put_failure;
4320         read_lock_bh(&idev->lock);
4321         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4322         read_unlock_bh(&idev->lock);
4323
4324         return 0;
4325
4326 nla_put_failure:
4327         return -EMSGSIZE;
4328 }
4329
4330 static size_t inet6_get_link_af_size(const struct net_device *dev)
4331 {
4332         if (!__in6_dev_get(dev))
4333                 return 0;
4334
4335         return inet6_ifla6_size();
4336 }
4337
4338 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4339 {
4340         struct inet6_dev *idev = __in6_dev_get(dev);
4341
4342         if (!idev)
4343                 return -ENODATA;
4344
4345         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4346                 return -EMSGSIZE;
4347
4348         return 0;
4349 }
4350
4351 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4352 {
4353         struct inet6_ifaddr *ifp;
4354         struct net_device *dev = idev->dev;
4355         bool update_rs = false;
4356         struct in6_addr ll_addr;
4357
4358         if (token == NULL)
4359                 return -EINVAL;
4360         if (ipv6_addr_any(token))
4361                 return -EINVAL;
4362         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4363                 return -EINVAL;
4364         if (!ipv6_accept_ra(idev))
4365                 return -EINVAL;
4366         if (idev->cnf.rtr_solicits <= 0)
4367                 return -EINVAL;
4368
4369         write_lock_bh(&idev->lock);
4370
4371         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4372         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4373
4374         write_unlock_bh(&idev->lock);
4375
4376         if (!idev->dead && (idev->if_flags & IF_READY) &&
4377             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4378                              IFA_F_OPTIMISTIC)) {
4379
4380                 /* If we're not ready, then normal ifup will take care
4381                  * of this. Otherwise, we need to request our rs here.
4382                  */
4383                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4384                 update_rs = true;
4385         }
4386
4387         write_lock_bh(&idev->lock);
4388
4389         if (update_rs) {
4390                 idev->if_flags |= IF_RS_SENT;
4391                 idev->rs_probes = 1;
4392                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4393         }
4394
4395         /* Well, that's kinda nasty ... */
4396         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4397                 spin_lock(&ifp->lock);
4398                 if (ifp->tokenized) {
4399                         ifp->valid_lft = 0;
4400                         ifp->prefered_lft = 0;
4401                 }
4402                 spin_unlock(&ifp->lock);
4403         }
4404
4405         write_unlock_bh(&idev->lock);
4406         addrconf_verify(0);
4407         return 0;
4408 }
4409
4410 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4411 {
4412         int err = -EINVAL;
4413         struct inet6_dev *idev = __in6_dev_get(dev);
4414         struct nlattr *tb[IFLA_INET6_MAX + 1];
4415
4416         if (!idev)
4417                 return -EAFNOSUPPORT;
4418
4419         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4420                 BUG();
4421
4422         if (tb[IFLA_INET6_TOKEN])
4423                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4424
4425         return err;
4426 }
4427
4428 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4429                              u32 portid, u32 seq, int event, unsigned int flags)
4430 {
4431         struct net_device *dev = idev->dev;
4432         struct ifinfomsg *hdr;
4433         struct nlmsghdr *nlh;
4434         void *protoinfo;
4435
4436         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4437         if (nlh == NULL)
4438                 return -EMSGSIZE;
4439
4440         hdr = nlmsg_data(nlh);
4441         hdr->ifi_family = AF_INET6;
4442         hdr->__ifi_pad = 0;
4443         hdr->ifi_type = dev->type;
4444         hdr->ifi_index = dev->ifindex;
4445         hdr->ifi_flags = dev_get_flags(dev);
4446         hdr->ifi_change = 0;
4447
4448         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4449             (dev->addr_len &&
4450              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4451             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4452             (dev->ifindex != dev->iflink &&
4453              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4454                 goto nla_put_failure;
4455         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4456         if (protoinfo == NULL)
4457                 goto nla_put_failure;
4458
4459         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4460                 goto nla_put_failure;
4461
4462         nla_nest_end(skb, protoinfo);
4463         return nlmsg_end(skb, nlh);
4464
4465 nla_put_failure:
4466         nlmsg_cancel(skb, nlh);
4467         return -EMSGSIZE;
4468 }
4469
4470 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4471 {
4472         struct net *net = sock_net(skb->sk);
4473         int h, s_h;
4474         int idx = 0, s_idx;
4475         struct net_device *dev;
4476         struct inet6_dev *idev;
4477         struct hlist_head *head;
4478
4479         s_h = cb->args[0];
4480         s_idx = cb->args[1];
4481
4482         rcu_read_lock();
4483         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4484                 idx = 0;
4485                 head = &net->dev_index_head[h];
4486                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4487                         if (idx < s_idx)
4488                                 goto cont;
4489                         idev = __in6_dev_get(dev);
4490                         if (!idev)
4491                                 goto cont;
4492                         if (inet6_fill_ifinfo(skb, idev,
4493                                               NETLINK_CB(cb->skb).portid,
4494                                               cb->nlh->nlmsg_seq,
4495                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4496                                 goto out;
4497 cont:
4498                         idx++;
4499                 }
4500         }
4501 out:
4502         rcu_read_unlock();
4503         cb->args[1] = idx;
4504         cb->args[0] = h;
4505
4506         return skb->len;
4507 }
4508
4509 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4510 {
4511         struct sk_buff *skb;
4512         struct net *net = dev_net(idev->dev);
4513         int err = -ENOBUFS;
4514
4515         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4516         if (skb == NULL)
4517                 goto errout;
4518
4519         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4520         if (err < 0) {
4521                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4522                 WARN_ON(err == -EMSGSIZE);
4523                 kfree_skb(skb);
4524                 goto errout;
4525         }
4526         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4527         return;
4528 errout:
4529         if (err < 0)
4530                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4531 }
4532
4533 static inline size_t inet6_prefix_nlmsg_size(void)
4534 {
4535         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4536                + nla_total_size(sizeof(struct in6_addr))
4537                + nla_total_size(sizeof(struct prefix_cacheinfo));
4538 }
4539
4540 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4541                              struct prefix_info *pinfo, u32 portid, u32 seq,
4542                              int event, unsigned int flags)
4543 {
4544         struct prefixmsg *pmsg;
4545         struct nlmsghdr *nlh;
4546         struct prefix_cacheinfo ci;
4547
4548         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4549         if (nlh == NULL)
4550                 return -EMSGSIZE;
4551
4552         pmsg = nlmsg_data(nlh);
4553         pmsg->prefix_family = AF_INET6;
4554         pmsg->prefix_pad1 = 0;
4555         pmsg->prefix_pad2 = 0;
4556         pmsg->prefix_ifindex = idev->dev->ifindex;
4557         pmsg->prefix_len = pinfo->prefix_len;
4558         pmsg->prefix_type = pinfo->type;
4559         pmsg->prefix_pad3 = 0;
4560         pmsg->prefix_flags = 0;
4561         if (pinfo->onlink)
4562                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4563         if (pinfo->autoconf)
4564                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4565
4566         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4567                 goto nla_put_failure;
4568         ci.preferred_time = ntohl(pinfo->prefered);
4569         ci.valid_time = ntohl(pinfo->valid);
4570         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4571                 goto nla_put_failure;
4572         return nlmsg_end(skb, nlh);
4573
4574 nla_put_failure:
4575         nlmsg_cancel(skb, nlh);
4576         return -EMSGSIZE;
4577 }
4578
4579 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4580                          struct prefix_info *pinfo)
4581 {
4582         struct sk_buff *skb;
4583         struct net *net = dev_net(idev->dev);
4584         int err = -ENOBUFS;
4585
4586         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4587         if (skb == NULL)
4588                 goto errout;
4589
4590         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4591         if (err < 0) {
4592                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4593                 WARN_ON(err == -EMSGSIZE);
4594                 kfree_skb(skb);
4595                 goto errout;
4596         }
4597         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4598         return;
4599 errout:
4600         if (err < 0)
4601                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4602 }
4603
4604 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4605 {
4606         write_lock_bh(&ifp->idev->lock);
4607         spin_lock(&ifp->lock);
4608         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4609                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4610             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4611                 ifp->idev->valid_ll_addr_cnt += count;
4612         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4613         spin_unlock(&ifp->lock);
4614         write_unlock_bh(&ifp->idev->lock);
4615 }
4616
4617 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4618 {
4619         struct net *net = dev_net(ifp->idev->dev);
4620
4621         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4622
4623         switch (event) {
4624         case RTM_NEWADDR:
4625                 update_valid_ll_addr_cnt(ifp, 1);
4626
4627                 /*
4628                  * If the address was optimistic
4629                  * we inserted the route at the start of
4630                  * our DAD process, so we don't need
4631                  * to do it again
4632                  */
4633                 if (!(ifp->rt->rt6i_node))
4634                         ip6_ins_rt(ifp->rt);
4635                 if (ifp->idev->cnf.forwarding)
4636                         addrconf_join_anycast(ifp);
4637                 if (!ipv6_addr_any(&ifp->peer_addr))
4638                         addrconf_prefix_route(&ifp->peer_addr, 128,
4639                                               ifp->idev->dev, 0, 0);
4640                 break;
4641         case RTM_DELADDR:
4642                 update_valid_ll_addr_cnt(ifp, -1);
4643
4644                 if (ifp->idev->cnf.forwarding)
4645                         addrconf_leave_anycast(ifp);
4646                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4647                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4648                         struct rt6_info *rt;
4649                         struct net_device *dev = ifp->idev->dev;
4650
4651                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4652                                         dev->ifindex, 1);
4653                         if (rt) {
4654                                 dst_hold(&rt->dst);
4655                                 if (ip6_del_rt(rt))
4656                                         dst_free(&rt->dst);
4657                         }
4658                 }
4659                 dst_hold(&ifp->rt->dst);
4660
4661                 if (ip6_del_rt(ifp->rt))
4662                         dst_free(&ifp->rt->dst);
4663                 break;
4664         }
4665         atomic_inc(&net->ipv6.dev_addr_genid);
4666         rt_genid_bump_ipv6(net);
4667 }
4668
4669 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4670 {
4671         rcu_read_lock_bh();
4672         if (likely(ifp->idev->dead == 0))
4673                 __ipv6_ifa_notify(event, ifp);
4674         rcu_read_unlock_bh();
4675 }
4676
4677 #ifdef CONFIG_SYSCTL
4678
4679 static
4680 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4681                            void __user *buffer, size_t *lenp, loff_t *ppos)
4682 {
4683         int *valp = ctl->data;
4684         int val = *valp;
4685         loff_t pos = *ppos;
4686         struct ctl_table lctl;
4687         int ret;
4688
4689         /*
4690          * ctl->data points to idev->cnf.forwarding, we should
4691          * not modify it until we get the rtnl lock.
4692          */
4693         lctl = *ctl;
4694         lctl.data = &val;
4695
4696         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4697
4698         if (write)
4699                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4700         if (ret)
4701                 *ppos = pos;
4702         return ret;
4703 }
4704
4705 static void dev_disable_change(struct inet6_dev *idev)
4706 {
4707         struct netdev_notifier_info info;
4708
4709         if (!idev || !idev->dev)
4710                 return;
4711
4712         netdev_notifier_info_init(&info, idev->dev);
4713         if (idev->cnf.disable_ipv6)
4714                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4715         else
4716                 addrconf_notify(NULL, NETDEV_UP, &info);
4717 }
4718
4719 static void addrconf_disable_change(struct net *net, __s32 newf)
4720 {
4721         struct net_device *dev;
4722         struct inet6_dev *idev;
4723
4724         rcu_read_lock();
4725         for_each_netdev_rcu(net, dev) {
4726                 idev = __in6_dev_get(dev);
4727                 if (idev) {
4728                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4729                         idev->cnf.disable_ipv6 = newf;
4730                         if (changed)
4731                                 dev_disable_change(idev);
4732                 }
4733         }
4734         rcu_read_unlock();
4735 }
4736
4737 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4738 {
4739         struct net *net;
4740         int old;
4741
4742         if (!rtnl_trylock())
4743                 return restart_syscall();
4744
4745         net = (struct net *)table->extra2;
4746         old = *p;
4747         *p = newf;
4748
4749         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4750                 rtnl_unlock();
4751                 return 0;
4752         }
4753
4754         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4755                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4756                 addrconf_disable_change(net, newf);
4757         } else if ((!newf) ^ (!old))
4758                 dev_disable_change((struct inet6_dev *)table->extra1);
4759
4760         rtnl_unlock();
4761         return 0;
4762 }
4763
4764 static
4765 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4766                             void __user *buffer, size_t *lenp, loff_t *ppos)
4767 {
4768         int *valp = ctl->data;
4769         int val = *valp;
4770         loff_t pos = *ppos;
4771         struct ctl_table lctl;
4772         int ret;
4773
4774         /*
4775          * ctl->data points to idev->cnf.disable_ipv6, we should
4776          * not modify it until we get the rtnl lock.
4777          */
4778         lctl = *ctl;
4779         lctl.data = &val;
4780
4781         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4782
4783         if (write)
4784                 ret = addrconf_disable_ipv6(ctl, valp, val);
4785         if (ret)
4786                 *ppos = pos;
4787         return ret;
4788 }
4789
4790 static struct addrconf_sysctl_table
4791 {
4792         struct ctl_table_header *sysctl_header;
4793         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4794 } addrconf_sysctl __read_mostly = {
4795         .sysctl_header = NULL,
4796         .addrconf_vars = {
4797                 {
4798                         .procname       = "forwarding",
4799                         .data           = &ipv6_devconf.forwarding,
4800                         .maxlen         = sizeof(int),
4801                         .mode           = 0644,
4802                         .proc_handler   = addrconf_sysctl_forward,
4803                 },
4804                 {
4805                         .procname       = "hop_limit",
4806                         .data           = &ipv6_devconf.hop_limit,
4807                         .maxlen         = sizeof(int),
4808                         .mode           = 0644,
4809                         .proc_handler   = proc_dointvec,
4810                 },
4811                 {
4812                         .procname       = "mtu",
4813                         .data           = &ipv6_devconf.mtu6,
4814                         .maxlen         = sizeof(int),
4815                         .mode           = 0644,
4816                         .proc_handler   = proc_dointvec,
4817                 },
4818                 {
4819                         .procname       = "accept_ra",
4820                         .data           = &ipv6_devconf.accept_ra,
4821                         .maxlen         = sizeof(int),
4822                         .mode           = 0644,
4823                         .proc_handler   = proc_dointvec,
4824                 },
4825                 {
4826                         .procname       = "accept_redirects",
4827                         .data           = &ipv6_devconf.accept_redirects,
4828                         .maxlen         = sizeof(int),
4829                         .mode           = 0644,
4830                         .proc_handler   = proc_dointvec,
4831                 },
4832                 {
4833                         .procname       = "autoconf",
4834                         .data           = &ipv6_devconf.autoconf,
4835                         .maxlen         = sizeof(int),
4836                         .mode           = 0644,
4837                         .proc_handler   = proc_dointvec,
4838                 },
4839                 {
4840                         .procname       = "dad_transmits",
4841                         .data           = &ipv6_devconf.dad_transmits,
4842                         .maxlen         = sizeof(int),
4843                         .mode           = 0644,
4844                         .proc_handler   = proc_dointvec,
4845                 },
4846                 {
4847                         .procname       = "router_solicitations",
4848                         .data           = &ipv6_devconf.rtr_solicits,
4849                         .maxlen         = sizeof(int),
4850                         .mode           = 0644,
4851                         .proc_handler   = proc_dointvec,
4852                 },
4853                 {
4854                         .procname       = "router_solicitation_interval",
4855                         .data           = &ipv6_devconf.rtr_solicit_interval,
4856                         .maxlen         = sizeof(int),
4857                         .mode           = 0644,
4858                         .proc_handler   = proc_dointvec_jiffies,
4859                 },
4860                 {
4861                         .procname       = "router_solicitation_delay",
4862                         .data           = &ipv6_devconf.rtr_solicit_delay,
4863                         .maxlen         = sizeof(int),
4864                         .mode           = 0644,
4865                         .proc_handler   = proc_dointvec_jiffies,
4866                 },
4867                 {
4868                         .procname       = "force_mld_version",
4869                         .data           = &ipv6_devconf.force_mld_version,
4870                         .maxlen         = sizeof(int),
4871                         .mode           = 0644,
4872                         .proc_handler   = proc_dointvec,
4873                 },
4874                 {
4875                         .procname       = "mldv1_unsolicited_report_interval",
4876                         .data           =
4877                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4878                         .maxlen         = sizeof(int),
4879                         .mode           = 0644,
4880                         .proc_handler   = proc_dointvec_ms_jiffies,
4881                 },
4882                 {
4883                         .procname       = "mldv2_unsolicited_report_interval",
4884                         .data           =
4885                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4886                         .maxlen         = sizeof(int),
4887                         .mode           = 0644,
4888                         .proc_handler   = proc_dointvec_ms_jiffies,
4889                 },
4890 #ifdef CONFIG_IPV6_PRIVACY
4891                 {
4892                         .procname       = "use_tempaddr",
4893                         .data           = &ipv6_devconf.use_tempaddr,
4894                         .maxlen         = sizeof(int),
4895                         .mode           = 0644,
4896                         .proc_handler   = proc_dointvec,
4897                 },
4898                 {
4899                         .procname       = "temp_valid_lft",
4900                         .data           = &ipv6_devconf.temp_valid_lft,
4901                         .maxlen         = sizeof(int),
4902                         .mode           = 0644,
4903                         .proc_handler   = proc_dointvec,
4904                 },
4905                 {
4906                         .procname       = "temp_prefered_lft",
4907                         .data           = &ipv6_devconf.temp_prefered_lft,
4908                         .maxlen         = sizeof(int),
4909                         .mode           = 0644,
4910                         .proc_handler   = proc_dointvec,
4911                 },
4912                 {
4913                         .procname       = "regen_max_retry",
4914                         .data           = &ipv6_devconf.regen_max_retry,
4915                         .maxlen         = sizeof(int),
4916                         .mode           = 0644,
4917                         .proc_handler   = proc_dointvec,
4918                 },
4919                 {
4920                         .procname       = "max_desync_factor",
4921                         .data           = &ipv6_devconf.max_desync_factor,
4922                         .maxlen         = sizeof(int),
4923                         .mode           = 0644,
4924                         .proc_handler   = proc_dointvec,
4925                 },
4926 #endif
4927                 {
4928                         .procname       = "max_addresses",
4929                         .data           = &ipv6_devconf.max_addresses,
4930                         .maxlen         = sizeof(int),
4931                         .mode           = 0644,
4932                         .proc_handler   = proc_dointvec,
4933                 },
4934                 {
4935                         .procname       = "accept_ra_defrtr",
4936                         .data           = &ipv6_devconf.accept_ra_defrtr,
4937                         .maxlen         = sizeof(int),
4938                         .mode           = 0644,
4939                         .proc_handler   = proc_dointvec,
4940                 },
4941                 {
4942                         .procname       = "accept_ra_pinfo",
4943                         .data           = &ipv6_devconf.accept_ra_pinfo,
4944                         .maxlen         = sizeof(int),
4945                         .mode           = 0644,
4946                         .proc_handler   = proc_dointvec,
4947                 },
4948 #ifdef CONFIG_IPV6_ROUTER_PREF
4949                 {
4950                         .procname       = "accept_ra_rtr_pref",
4951                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4952                         .maxlen         = sizeof(int),
4953                         .mode           = 0644,
4954                         .proc_handler   = proc_dointvec,
4955                 },
4956                 {
4957                         .procname       = "router_probe_interval",
4958                         .data           = &ipv6_devconf.rtr_probe_interval,
4959                         .maxlen         = sizeof(int),
4960                         .mode           = 0644,
4961                         .proc_handler   = proc_dointvec_jiffies,
4962                 },
4963 #ifdef CONFIG_IPV6_ROUTE_INFO
4964                 {
4965                         .procname       = "accept_ra_rt_info_max_plen",
4966                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4967                         .maxlen         = sizeof(int),
4968                         .mode           = 0644,
4969                         .proc_handler   = proc_dointvec,
4970                 },
4971 #endif
4972 #endif
4973                 {
4974                         .procname       = "proxy_ndp",
4975                         .data           = &ipv6_devconf.proxy_ndp,
4976                         .maxlen         = sizeof(int),
4977                         .mode           = 0644,
4978                         .proc_handler   = proc_dointvec,
4979                 },
4980                 {
4981                         .procname       = "accept_source_route",
4982                         .data           = &ipv6_devconf.accept_source_route,
4983                         .maxlen         = sizeof(int),
4984                         .mode           = 0644,
4985                         .proc_handler   = proc_dointvec,
4986                 },
4987 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4988                 {
4989                         .procname       = "optimistic_dad",
4990                         .data           = &ipv6_devconf.optimistic_dad,
4991                         .maxlen         = sizeof(int),
4992                         .mode           = 0644,
4993                         .proc_handler   = proc_dointvec,
4994
4995                 },
4996 #endif
4997 #ifdef CONFIG_IPV6_MROUTE
4998                 {
4999                         .procname       = "mc_forwarding",
5000                         .data           = &ipv6_devconf.mc_forwarding,
5001                         .maxlen         = sizeof(int),
5002                         .mode           = 0444,
5003                         .proc_handler   = proc_dointvec,
5004                 },
5005 #endif
5006                 {
5007                         .procname       = "disable_ipv6",
5008                         .data           = &ipv6_devconf.disable_ipv6,
5009                         .maxlen         = sizeof(int),
5010                         .mode           = 0644,
5011                         .proc_handler   = addrconf_sysctl_disable,
5012                 },
5013                 {
5014                         .procname       = "accept_dad",
5015                         .data           = &ipv6_devconf.accept_dad,
5016                         .maxlen         = sizeof(int),
5017                         .mode           = 0644,
5018                         .proc_handler   = proc_dointvec,
5019                 },
5020                 {
5021                         .procname       = "force_tllao",
5022                         .data           = &ipv6_devconf.force_tllao,
5023                         .maxlen         = sizeof(int),
5024                         .mode           = 0644,
5025                         .proc_handler   = proc_dointvec
5026                 },
5027                 {
5028                         .procname       = "ndisc_notify",
5029                         .data           = &ipv6_devconf.ndisc_notify,
5030                         .maxlen         = sizeof(int),
5031                         .mode           = 0644,
5032                         .proc_handler   = proc_dointvec
5033                 },
5034                 {
5035                         /* sentinel */
5036                 }
5037         },
5038 };
5039
5040 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5041                 struct inet6_dev *idev, struct ipv6_devconf *p)
5042 {
5043         int i;
5044         struct addrconf_sysctl_table *t;
5045         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5046
5047         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5048         if (t == NULL)
5049                 goto out;
5050
5051         for (i = 0; t->addrconf_vars[i].data; i++) {
5052                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5053                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5054                 t->addrconf_vars[i].extra2 = net;
5055         }
5056
5057         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5058
5059         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5060         if (t->sysctl_header == NULL)
5061                 goto free;
5062
5063         p->sysctl = t;
5064         return 0;
5065
5066 free:
5067         kfree(t);
5068 out:
5069         return -ENOBUFS;
5070 }
5071
5072 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5073 {
5074         struct addrconf_sysctl_table *t;
5075
5076         if (p->sysctl == NULL)
5077                 return;
5078
5079         t = p->sysctl;
5080         p->sysctl = NULL;
5081         unregister_net_sysctl_table(t->sysctl_header);
5082         kfree(t);
5083 }
5084
5085 static void addrconf_sysctl_register(struct inet6_dev *idev)
5086 {
5087         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5088                               &ndisc_ifinfo_sysctl_change);
5089         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5090                                         idev, &idev->cnf);
5091 }
5092
5093 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5094 {
5095         __addrconf_sysctl_unregister(&idev->cnf);
5096         neigh_sysctl_unregister(idev->nd_parms);
5097 }
5098
5099
5100 #endif
5101
5102 static int __net_init addrconf_init_net(struct net *net)
5103 {
5104         int err = -ENOMEM;
5105         struct ipv6_devconf *all, *dflt;
5106
5107         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5108         if (all == NULL)
5109                 goto err_alloc_all;
5110
5111         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5112         if (dflt == NULL)
5113                 goto err_alloc_dflt;
5114
5115         /* these will be inherited by all namespaces */
5116         dflt->autoconf = ipv6_defaults.autoconf;
5117         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5118
5119         net->ipv6.devconf_all = all;
5120         net->ipv6.devconf_dflt = dflt;
5121
5122 #ifdef CONFIG_SYSCTL
5123         err = __addrconf_sysctl_register(net, "all", NULL, all);
5124         if (err < 0)
5125                 goto err_reg_all;
5126
5127         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5128         if (err < 0)
5129                 goto err_reg_dflt;
5130 #endif
5131         return 0;
5132
5133 #ifdef CONFIG_SYSCTL
5134 err_reg_dflt:
5135         __addrconf_sysctl_unregister(all);
5136 err_reg_all:
5137         kfree(dflt);
5138 #endif
5139 err_alloc_dflt:
5140         kfree(all);
5141 err_alloc_all:
5142         return err;
5143 }
5144
5145 static void __net_exit addrconf_exit_net(struct net *net)
5146 {
5147 #ifdef CONFIG_SYSCTL
5148         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5149         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5150 #endif
5151         if (!net_eq(net, &init_net)) {
5152                 kfree(net->ipv6.devconf_dflt);
5153                 kfree(net->ipv6.devconf_all);
5154         }
5155 }
5156
5157 static struct pernet_operations addrconf_ops = {
5158         .init = addrconf_init_net,
5159         .exit = addrconf_exit_net,
5160 };
5161
5162 static struct rtnl_af_ops inet6_ops = {
5163         .family           = AF_INET6,
5164         .fill_link_af     = inet6_fill_link_af,
5165         .get_link_af_size = inet6_get_link_af_size,
5166         .set_link_af      = inet6_set_link_af,
5167 };
5168
5169 /*
5170  *      Init / cleanup code
5171  */
5172
5173 int __init addrconf_init(void)
5174 {
5175         int i, err;
5176
5177         err = ipv6_addr_label_init();
5178         if (err < 0) {
5179                 pr_crit("%s: cannot initialize default policy table: %d\n",
5180                         __func__, err);
5181                 goto out;
5182         }
5183
5184         err = register_pernet_subsys(&addrconf_ops);
5185         if (err < 0)
5186                 goto out_addrlabel;
5187
5188         /* The addrconf netdev notifier requires that loopback_dev
5189          * has it's ipv6 private information allocated and setup
5190          * before it can bring up and give link-local addresses
5191          * to other devices which are up.
5192          *
5193          * Unfortunately, loopback_dev is not necessarily the first
5194          * entry in the global dev_base list of net devices.  In fact,
5195          * it is likely to be the very last entry on that list.
5196          * So this causes the notifier registry below to try and
5197          * give link-local addresses to all devices besides loopback_dev
5198          * first, then loopback_dev, which cases all the non-loopback_dev
5199          * devices to fail to get a link-local address.
5200          *
5201          * So, as a temporary fix, allocate the ipv6 structure for
5202          * loopback_dev first by hand.
5203          * Longer term, all of the dependencies ipv6 has upon the loopback
5204          * device and it being up should be removed.
5205          */
5206         rtnl_lock();
5207         if (!ipv6_add_dev(init_net.loopback_dev))
5208                 err = -ENOMEM;
5209         rtnl_unlock();
5210         if (err)
5211                 goto errlo;
5212
5213         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5214                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5215
5216         register_netdevice_notifier(&ipv6_dev_notf);
5217
5218         addrconf_verify(0);
5219
5220         err = rtnl_af_register(&inet6_ops);
5221         if (err < 0)
5222                 goto errout_af;
5223
5224         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5225                               NULL);
5226         if (err < 0)
5227                 goto errout;
5228
5229         /* Only the first call to __rtnl_register can fail */
5230         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5231         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5232         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5233                         inet6_dump_ifaddr, NULL);
5234         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5235                         inet6_dump_ifmcaddr, NULL);
5236         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5237                         inet6_dump_ifacaddr, NULL);
5238         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5239                         inet6_netconf_dump_devconf, NULL);
5240
5241         ipv6_addr_label_rtnl_register();
5242
5243         return 0;
5244 errout:
5245         rtnl_af_unregister(&inet6_ops);
5246 errout_af:
5247         unregister_netdevice_notifier(&ipv6_dev_notf);
5248 errlo:
5249         unregister_pernet_subsys(&addrconf_ops);
5250 out_addrlabel:
5251         ipv6_addr_label_cleanup();
5252 out:
5253         return err;
5254 }
5255
5256 void addrconf_cleanup(void)
5257 {
5258         struct net_device *dev;
5259         int i;
5260
5261         unregister_netdevice_notifier(&ipv6_dev_notf);
5262         unregister_pernet_subsys(&addrconf_ops);
5263         ipv6_addr_label_cleanup();
5264
5265         rtnl_lock();
5266
5267         __rtnl_af_unregister(&inet6_ops);
5268
5269         /* clean dev list */
5270         for_each_netdev(&init_net, dev) {
5271                 if (__in6_dev_get(dev) == NULL)
5272                         continue;
5273                 addrconf_ifdown(dev, 1);
5274         }
5275         addrconf_ifdown(init_net.loopback_dev, 2);
5276
5277         /*
5278          *      Check hash table.
5279          */
5280         spin_lock_bh(&addrconf_hash_lock);
5281         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5282                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5283         spin_unlock_bh(&addrconf_hash_lock);
5284
5285         del_timer(&addr_chk_timer);
5286         rtnl_unlock();
5287 }