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(fmt, ...) printk(fmt, ##__VA_ARGS__)
 103 #else
 104 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 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 = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1135                             ipv6_addr_scope(&addr), addr_flags,
1136                             tmp_valid_lft, tmp_prefered_lft);
1137         if (IS_ERR(ift)) {
1138                 in6_ifa_put(ifp);
1139                 in6_dev_put(idev);
1140                 pr_info("%s: retry temporary address regeneration\n", __func__);
1141                 tmpaddr = &addr;
1142                 write_lock(&idev->lock);
1143                 goto retry;
1144         }
1145
1146         spin_lock_bh(&ift->lock);
1147         ift->ifpub = ifp;
1148         ift->cstamp = now;
1149         ift->tstamp = tmp_tstamp;
1150         spin_unlock_bh(&ift->lock);
1151
1152         addrconf_dad_start(ift);
1153         in6_ifa_put(ift);
1154         in6_dev_put(idev);
1155 out:
1156         return ret;
1157 }
1158 #endif
1159
1160 /*
1161  *      Choose an appropriate source address (RFC3484)
1162  */
1163 enum {
1164         IPV6_SADDR_RULE_INIT = 0,
1165         IPV6_SADDR_RULE_LOCAL,
1166         IPV6_SADDR_RULE_SCOPE,
1167         IPV6_SADDR_RULE_PREFERRED,
1168 #ifdef CONFIG_IPV6_MIP6
1169         IPV6_SADDR_RULE_HOA,
1170 #endif
1171         IPV6_SADDR_RULE_OIF,
1172         IPV6_SADDR_RULE_LABEL,
1173 #ifdef CONFIG_IPV6_PRIVACY
1174         IPV6_SADDR_RULE_PRIVACY,
1175 #endif
1176         IPV6_SADDR_RULE_ORCHID,
1177         IPV6_SADDR_RULE_PREFIX,
1178         IPV6_SADDR_RULE_MAX
1179 };
1180
1181 struct ipv6_saddr_score {
1182         int                     rule;
1183         int                     addr_type;
1184         struct inet6_ifaddr     *ifa;
1185         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1186         int                     scopedist;
1187         int                     matchlen;
1188 };
1189
1190 struct ipv6_saddr_dst {
1191         const struct in6_addr *addr;
1192         int ifindex;
1193         int scope;
1194         int label;
1195         unsigned int prefs;
1196 };
1197
1198 static inline int ipv6_saddr_preferred(int type)
1199 {
1200         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1201                 return 1;
1202         return 0;
1203 }
1204
1205 static int ipv6_get_saddr_eval(struct net *net,
1206                                struct ipv6_saddr_score *score,
1207                                struct ipv6_saddr_dst *dst,
1208                                int i)
1209 {
1210         int ret;
1211
1212         if (i <= score->rule) {
1213                 switch (i) {
1214                 case IPV6_SADDR_RULE_SCOPE:
1215                         ret = score->scopedist;
1216                         break;
1217                 case IPV6_SADDR_RULE_PREFIX:
1218                         ret = score->matchlen;
1219                         break;
1220                 default:
1221                         ret = !!test_bit(i, score->scorebits);
1222                 }
1223                 goto out;
1224         }
1225
1226         switch (i) {
1227         case IPV6_SADDR_RULE_INIT:
1228                 /* Rule 0: remember if hiscore is not ready yet */
1229                 ret = !!score->ifa;
1230                 break;
1231         case IPV6_SADDR_RULE_LOCAL:
1232                 /* Rule 1: Prefer same address */
1233                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1234                 break;
1235         case IPV6_SADDR_RULE_SCOPE:
1236                 /* Rule 2: Prefer appropriate scope
1237                  *
1238                  *      ret
1239                  *       ^
1240                  *    -1 |  d 15
1241                  *    ---+--+-+---> scope
1242                  *       |
1243                  *       |             d is scope of the destination.
1244                  *  B-d  |  \
1245                  *       |   \      <- smaller scope is better if
1246                  *  B-15 |    \        if scope is enough for destinaion.
1247                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1248                  * d-C-1 | /
1249                  *       |/         <- greater is better
1250                  *   -C  /             if scope is not enough for destination.
1251                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1252                  *
1253                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1254                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1255                  * Assume B = 0 and we get C > 29.
1256                  */
1257                 ret = __ipv6_addr_src_scope(score->addr_type);
1258                 if (ret >= dst->scope)
1259                         ret = -ret;
1260                 else
1261                         ret -= 128;     /* 30 is enough */
1262                 score->scopedist = ret;
1263                 break;
1264         case IPV6_SADDR_RULE_PREFERRED:
1265                 /* Rule 3: Avoid deprecated and optimistic addresses */
1266                 ret = ipv6_saddr_preferred(score->addr_type) ||
1267                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1268                 break;
1269 #ifdef CONFIG_IPV6_MIP6
1270         case IPV6_SADDR_RULE_HOA:
1271             {
1272                 /* Rule 4: Prefer home address */
1273                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1274                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1275                 break;
1276             }
1277 #endif
1278         case IPV6_SADDR_RULE_OIF:
1279                 /* Rule 5: Prefer outgoing interface */
1280                 ret = (!dst->ifindex ||
1281                        dst->ifindex == score->ifa->idev->dev->ifindex);
1282                 break;
1283         case IPV6_SADDR_RULE_LABEL:
1284                 /* Rule 6: Prefer matching label */
1285                 ret = ipv6_addr_label(net,
1286                                       &score->ifa->addr, score->addr_type,
1287                                       score->ifa->idev->dev->ifindex) == dst->label;
1288                 break;
1289 #ifdef CONFIG_IPV6_PRIVACY
1290         case IPV6_SADDR_RULE_PRIVACY:
1291             {
1292                 /* Rule 7: Prefer public address
1293                  * Note: prefer temporary address if use_tempaddr >= 2
1294                  */
1295                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1296                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1297                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1298                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1299                 break;
1300             }
1301 #endif
1302         case IPV6_SADDR_RULE_ORCHID:
1303                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1304                  *          non-ORCHID vs non-ORCHID
1305                  */
1306                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1307                         ipv6_addr_orchid(dst->addr));
1308                 break;
1309         case IPV6_SADDR_RULE_PREFIX:
1310                 /* Rule 8: Use longest matching prefix */
1311                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1312                 if (ret > score->ifa->prefix_len)
1313                         ret = score->ifa->prefix_len;
1314                 score->matchlen = ret;
1315                 break;
1316         default:
1317                 ret = 0;
1318         }
1319
1320         if (ret)
1321                 __set_bit(i, score->scorebits);
1322         score->rule = i;
1323 out:
1324         return ret;
1325 }
1326
1327 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1328                        const struct in6_addr *daddr, unsigned int prefs,
1329                        struct in6_addr *saddr)
1330 {
1331         struct ipv6_saddr_score scores[2],
1332                                 *score = &scores[0], *hiscore = &scores[1];
1333         struct ipv6_saddr_dst dst;
1334         struct net_device *dev;
1335         int dst_type;
1336
1337         dst_type = __ipv6_addr_type(daddr);
1338         dst.addr = daddr;
1339         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1340         dst.scope = __ipv6_addr_src_scope(dst_type);
1341         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1342         dst.prefs = prefs;
1343
1344         hiscore->rule = -1;
1345         hiscore->ifa = NULL;
1346
1347         rcu_read_lock();
1348
1349         for_each_netdev_rcu(net, dev) {
1350                 struct inet6_dev *idev;
1351
1352                 /* Candidate Source Address (section 4)
1353                  *  - multicast and link-local destination address,
1354                  *    the set of candidate source address MUST only
1355                  *    include addresses assigned to interfaces
1356                  *    belonging to the same link as the outgoing
1357                  *    interface.
1358                  * (- For site-local destination addresses, the
1359                  *    set of candidate source addresses MUST only
1360                  *    include addresses assigned to interfaces
1361                  *    belonging to the same site as the outgoing
1362                  *    interface.)
1363                  */
1364                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1365                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1366                     dst.ifindex && dev->ifindex != dst.ifindex)
1367                         continue;
1368
1369                 idev = __in6_dev_get(dev);
1370                 if (!idev)
1371                         continue;
1372
1373                 read_lock_bh(&idev->lock);
1374                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1375                         int i;
1376
1377                         /*
1378                          * - Tentative Address (RFC2462 section 5.4)
1379                          *  - A tentative address is not considered
1380                          *    "assigned to an interface" in the traditional
1381                          *    sense, unless it is also flagged as optimistic.
1382                          * - Candidate Source Address (section 4)
1383                          *  - In any case, anycast addresses, multicast
1384                          *    addresses, and the unspecified address MUST
1385                          *    NOT be included in a candidate set.
1386                          */
1387                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1388                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1389                                 continue;
1390
1391                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1392
1393                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1394                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1395                                 LIMIT_NETDEBUG(KERN_DEBUG
1396                                                "ADDRCONF: unspecified / multicast address "
1397                                                "assigned as unicast address on %s",
1398                                                dev->name);
1399                                 continue;
1400                         }
1401
1402                         score->rule = -1;
1403                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1404
1405                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1406                                 int minihiscore, miniscore;
1407
1408                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1409                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1410
1411                                 if (minihiscore > miniscore) {
1412                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1413                                             score->scopedist > 0) {
1414                                                 /*
1415                                                  * special case:
1416                                                  * each remaining entry
1417                                                  * has too small (not enough)
1418                                                  * scope, because ifa entries
1419                                                  * are sorted by their scope
1420                                                  * values.
1421                                                  */
1422                                                 goto try_nextdev;
1423                                         }
1424                                         break;
1425                                 } else if (minihiscore < miniscore) {
1426                                         if (hiscore->ifa)
1427                                                 in6_ifa_put(hiscore->ifa);
1428
1429                                         in6_ifa_hold(score->ifa);
1430
1431                                         swap(hiscore, score);
1432
1433                                         /* restore our iterator */
1434                                         score->ifa = hiscore->ifa;
1435
1436                                         break;
1437                                 }
1438                         }
1439                 }
1440 try_nextdev:
1441                 read_unlock_bh(&idev->lock);
1442         }
1443         rcu_read_unlock();
1444
1445         if (!hiscore->ifa)
1446                 return -EADDRNOTAVAIL;
1447
1448         *saddr = hiscore->ifa->addr;
1449         in6_ifa_put(hiscore->ifa);
1450         return 0;
1451 }
1452 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1453
1454 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1455                       unsigned char banned_flags)
1456 {
1457         struct inet6_ifaddr *ifp;
1458         int err = -EADDRNOTAVAIL;
1459
1460         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1461                 if (ifp->scope == IFA_LINK &&
1462                     !(ifp->flags & banned_flags)) {
1463                         *addr = ifp->addr;
1464                         err = 0;
1465                         break;
1466                 }
1467         }
1468         return err;
1469 }
1470
1471 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1472                     unsigned char banned_flags)
1473 {
1474         struct inet6_dev *idev;
1475         int err = -EADDRNOTAVAIL;
1476
1477         rcu_read_lock();
1478         idev = __in6_dev_get(dev);
1479         if (idev) {
1480                 read_lock_bh(&idev->lock);
1481                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1482                 read_unlock_bh(&idev->lock);
1483         }
1484         rcu_read_unlock();
1485         return err;
1486 }
1487
1488 static int ipv6_count_addresses(struct inet6_dev *idev)
1489 {
1490         int cnt = 0;
1491         struct inet6_ifaddr *ifp;
1492
1493         read_lock_bh(&idev->lock);
1494         list_for_each_entry(ifp, &idev->addr_list, if_list)
1495                 cnt++;
1496         read_unlock_bh(&idev->lock);
1497         return cnt;
1498 }
1499
1500 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1501                   const struct net_device *dev, int strict)
1502 {
1503         struct inet6_ifaddr *ifp;
1504         unsigned int hash = inet6_addr_hash(addr);
1505
1506         rcu_read_lock_bh();
1507         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1508                 if (!net_eq(dev_net(ifp->idev->dev), net))
1509                         continue;
1510                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1511                     !(ifp->flags&IFA_F_TENTATIVE) &&
1512                     (dev == NULL || ifp->idev->dev == dev ||
1513                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1514                         rcu_read_unlock_bh();
1515                         return 1;
1516                 }
1517         }
1518
1519         rcu_read_unlock_bh();
1520         return 0;
1521 }
1522 EXPORT_SYMBOL(ipv6_chk_addr);
1523
1524 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1525                                struct net_device *dev)
1526 {
1527         unsigned int hash = inet6_addr_hash(addr);
1528         struct inet6_ifaddr *ifp;
1529
1530         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1531                 if (!net_eq(dev_net(ifp->idev->dev), net))
1532                         continue;
1533                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1534                         if (dev == NULL || ifp->idev->dev == dev)
1535                                 return true;
1536                 }
1537         }
1538         return false;
1539 }
1540
1541 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1542 {
1543         struct inet6_dev *idev;
1544         struct inet6_ifaddr *ifa;
1545         int     onlink;
1546
1547         onlink = 0;
1548         rcu_read_lock();
1549         idev = __in6_dev_get(dev);
1550         if (idev) {
1551                 read_lock_bh(&idev->lock);
1552                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1553                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1554                                                    ifa->prefix_len);
1555                         if (onlink)
1556                                 break;
1557                 }
1558                 read_unlock_bh(&idev->lock);
1559         }
1560         rcu_read_unlock();
1561         return onlink;
1562 }
1563 EXPORT_SYMBOL(ipv6_chk_prefix);
1564
1565 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1566                                      struct net_device *dev, int strict)
1567 {
1568         struct inet6_ifaddr *ifp, *result = NULL;
1569         unsigned int hash = inet6_addr_hash(addr);
1570
1571         rcu_read_lock_bh();
1572         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1573                 if (!net_eq(dev_net(ifp->idev->dev), net))
1574                         continue;
1575                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1576                         if (dev == NULL || ifp->idev->dev == dev ||
1577                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1578                                 result = ifp;
1579                                 in6_ifa_hold(ifp);
1580                                 break;
1581                         }
1582                 }
1583         }
1584         rcu_read_unlock_bh();
1585
1586         return result;
1587 }
1588
1589 /* Gets referenced address, destroys ifaddr */
1590
1591 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1592 {
1593         if (ifp->flags&IFA_F_PERMANENT) {
1594                 spin_lock_bh(&ifp->lock);
1595                 addrconf_del_dad_timer(ifp);
1596                 ifp->flags |= IFA_F_TENTATIVE;
1597                 if (dad_failed)
1598                         ifp->flags |= IFA_F_DADFAILED;
1599                 spin_unlock_bh(&ifp->lock);
1600                 if (dad_failed)
1601                         ipv6_ifa_notify(0, ifp);
1602                 in6_ifa_put(ifp);
1603 #ifdef CONFIG_IPV6_PRIVACY
1604         } else if (ifp->flags&IFA_F_TEMPORARY) {
1605                 struct inet6_ifaddr *ifpub;
1606                 spin_lock_bh(&ifp->lock);
1607                 ifpub = ifp->ifpub;
1608                 if (ifpub) {
1609                         in6_ifa_hold(ifpub);
1610                         spin_unlock_bh(&ifp->lock);
1611                         ipv6_create_tempaddr(ifpub, ifp);
1612                         in6_ifa_put(ifpub);
1613                 } else {
1614                         spin_unlock_bh(&ifp->lock);
1615                 }
1616                 ipv6_del_addr(ifp);
1617 #endif
1618         } else
1619                 ipv6_del_addr(ifp);
1620 }
1621
1622 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1623 {
1624         int err = -ENOENT;
1625
1626         spin_lock(&ifp->state_lock);
1627         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1628                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1629                 err = 0;
1630         }
1631         spin_unlock(&ifp->state_lock);
1632
1633         return err;
1634 }
1635
1636 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1637 {
1638         struct inet6_dev *idev = ifp->idev;
1639
1640         if (addrconf_dad_end(ifp)) {
1641                 in6_ifa_put(ifp);
1642                 return;
1643         }
1644
1645         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1646                              ifp->idev->dev->name, &ifp->addr);
1647
1648         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1649                 struct in6_addr addr;
1650
1651                 addr.s6_addr32[0] = htonl(0xfe800000);
1652                 addr.s6_addr32[1] = 0;
1653
1654                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1655                     ipv6_addr_equal(&ifp->addr, &addr)) {
1656                         /* DAD failed for link-local based on MAC address */
1657                         idev->cnf.disable_ipv6 = 1;
1658
1659                         pr_info("%s: IPv6 being disabled!\n",
1660                                 ifp->idev->dev->name);
1661                 }
1662         }
1663
1664         addrconf_dad_stop(ifp, 1);
1665 }
1666
1667 /* Join to solicited addr multicast group. */
1668
1669 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1670 {
1671         struct in6_addr maddr;
1672
1673         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1674                 return;
1675
1676         addrconf_addr_solict_mult(addr, &maddr);
1677         ipv6_dev_mc_inc(dev, &maddr);
1678 }
1679
1680 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1681 {
1682         struct in6_addr maddr;
1683
1684         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1685                 return;
1686
1687         addrconf_addr_solict_mult(addr, &maddr);
1688         __ipv6_dev_mc_dec(idev, &maddr);
1689 }
1690
1691 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1692 {
1693         struct in6_addr addr;
1694         if (ifp->prefix_len == 127) /* RFC 6164 */
1695                 return;
1696         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1697         if (ipv6_addr_any(&addr))
1698                 return;
1699         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1700 }
1701
1702 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1703 {
1704         struct in6_addr addr;
1705         if (ifp->prefix_len == 127) /* RFC 6164 */
1706                 return;
1707         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1708         if (ipv6_addr_any(&addr))
1709                 return;
1710         __ipv6_dev_ac_dec(ifp->idev, &addr);
1711 }
1712
1713 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1714 {
1715         if (dev->addr_len != ETH_ALEN)
1716                 return -1;
1717         memcpy(eui, dev->dev_addr, 3);
1718         memcpy(eui + 5, dev->dev_addr + 3, 3);
1719
1720         /*
1721          * The zSeries OSA network cards can be shared among various
1722          * OS instances, but the OSA cards have only one MAC address.
1723          * This leads to duplicate address conflicts in conjunction
1724          * with IPv6 if more than one instance uses the same card.
1725          *
1726          * The driver for these cards can deliver a unique 16-bit
1727          * identifier for each instance sharing the same card.  It is
1728          * placed instead of 0xFFFE in the interface identifier.  The
1729          * "u" bit of the interface identifier is not inverted in this
1730          * case.  Hence the resulting interface identifier has local
1731          * scope according to RFC2373.
1732          */
1733         if (dev->dev_id) {
1734                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1735                 eui[4] = dev->dev_id & 0xFF;
1736         } else {
1737                 eui[3] = 0xFF;
1738                 eui[4] = 0xFE;
1739                 eui[0] ^= 2;
1740         }
1741         return 0;
1742 }
1743
1744 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1745 {
1746         if (dev->addr_len != IEEE802154_ADDR_LEN)
1747                 return -1;
1748         memcpy(eui, dev->dev_addr, 8);
1749         eui[0] ^= 2;
1750         return 0;
1751 }
1752
1753 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1754 {
1755         union fwnet_hwaddr *ha;
1756
1757         if (dev->addr_len != FWNET_ALEN)
1758                 return -1;
1759
1760         ha = (union fwnet_hwaddr *)dev->dev_addr;
1761
1762         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1763         eui[0] ^= 2;
1764         return 0;
1765 }
1766
1767 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1768 {
1769         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1770         if (dev->addr_len != ARCNET_ALEN)
1771                 return -1;
1772         memset(eui, 0, 7);
1773         eui[7] = *(u8 *)dev->dev_addr;
1774         return 0;
1775 }
1776
1777 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1778 {
1779         if (dev->addr_len != INFINIBAND_ALEN)
1780                 return -1;
1781         memcpy(eui, dev->dev_addr + 12, 8);
1782         eui[0] |= 2;
1783         return 0;
1784 }
1785
1786 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1787 {
1788         if (addr == 0)
1789                 return -1;
1790         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1791                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1792                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1793                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1794                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1795                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1796         eui[1] = 0;
1797         eui[2] = 0x5E;
1798         eui[3] = 0xFE;
1799         memcpy(eui + 4, &addr, 4);
1800         return 0;
1801 }
1802
1803 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1804 {
1805         if (dev->priv_flags & IFF_ISATAP)
1806                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1807         return -1;
1808 }
1809
1810 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1811 {
1812         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1813 }
1814
1815 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1816 {
1817         memcpy(eui, dev->perm_addr, 3);
1818         memcpy(eui + 5, dev->perm_addr + 3, 3);
1819         eui[3] = 0xFF;
1820         eui[4] = 0xFE;
1821         eui[0] ^= 2;
1822         return 0;
1823 }
1824
1825 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1826 {
1827         switch (dev->type) {
1828         case ARPHRD_ETHER:
1829         case ARPHRD_FDDI:
1830                 return addrconf_ifid_eui48(eui, dev);
1831         case ARPHRD_ARCNET:
1832                 return addrconf_ifid_arcnet(eui, dev);
1833         case ARPHRD_INFINIBAND:
1834                 return addrconf_ifid_infiniband(eui, dev);
1835         case ARPHRD_SIT:
1836                 return addrconf_ifid_sit(eui, dev);
1837         case ARPHRD_IPGRE:
1838                 return addrconf_ifid_gre(eui, dev);
1839         case ARPHRD_IEEE802154:
1840                 return addrconf_ifid_eui64(eui, dev);
1841         case ARPHRD_IEEE1394:
1842                 return addrconf_ifid_ieee1394(eui, dev);
1843         case ARPHRD_TUNNEL6:
1844                 return addrconf_ifid_ip6tnl(eui, dev);
1845         }
1846         return -1;
1847 }
1848
1849 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1850 {
1851         int err = -1;
1852         struct inet6_ifaddr *ifp;
1853
1854         read_lock_bh(&idev->lock);
1855         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1856                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1857                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1858                         err = 0;
1859                         break;
1860                 }
1861         }
1862         read_unlock_bh(&idev->lock);
1863         return err;
1864 }
1865
1866 #ifdef CONFIG_IPV6_PRIVACY
1867 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1868 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1869 {
1870 regen:
1871         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1872         idev->rndid[0] &= ~0x02;
1873
1874         /*
1875          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1876          * check if generated address is not inappropriate
1877          *
1878          *  - Reserved subnet anycast (RFC 2526)
1879          *      11111101 11....11 1xxxxxxx
1880          *  - ISATAP (RFC4214) 6.1
1881          *      00-00-5E-FE-xx-xx-xx-xx
1882          *  - value 0
1883          *  - XXX: already assigned to an address on the device
1884          */
1885         if (idev->rndid[0] == 0xfd &&
1886             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1887             (idev->rndid[7]&0x80))
1888                 goto regen;
1889         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1890                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1891                         goto regen;
1892                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1893                         goto regen;
1894         }
1895 }
1896
1897 static void ipv6_regen_rndid(unsigned long data)
1898 {
1899         struct inet6_dev *idev = (struct inet6_dev *) data;
1900         unsigned long expires;
1901
1902         rcu_read_lock_bh();
1903         write_lock_bh(&idev->lock);
1904
1905         if (idev->dead)
1906                 goto out;
1907
1908         __ipv6_regen_rndid(idev);
1909
1910         expires = jiffies +
1911                 idev->cnf.temp_prefered_lft * HZ -
1912                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1913                 idev->cnf.max_desync_factor * HZ;
1914         if (time_before(expires, jiffies)) {
1915                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1916                         __func__, idev->dev->name);
1917                 goto out;
1918         }
1919
1920         if (!mod_timer(&idev->regen_timer, expires))
1921                 in6_dev_hold(idev);
1922
1923 out:
1924         write_unlock_bh(&idev->lock);
1925         rcu_read_unlock_bh();
1926         in6_dev_put(idev);
1927 }
1928
1929 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1930 {
1931         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1932                 __ipv6_regen_rndid(idev);
1933 }
1934 #endif
1935
1936 /*
1937  *      Add prefix route.
1938  */
1939
1940 static void
1941 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1942                       unsigned long expires, u32 flags)
1943 {
1944         struct fib6_config cfg = {
1945                 .fc_table = RT6_TABLE_PREFIX,
1946                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1947                 .fc_ifindex = dev->ifindex,
1948                 .fc_expires = expires,
1949                 .fc_dst_len = plen,
1950                 .fc_flags = RTF_UP | flags,
1951                 .fc_nlinfo.nl_net = dev_net(dev),
1952                 .fc_protocol = RTPROT_KERNEL,
1953         };
1954
1955         cfg.fc_dst = *pfx;
1956
1957         /* Prevent useless cloning on PtP SIT.
1958            This thing is done here expecting that the whole
1959            class of non-broadcast devices need not cloning.
1960          */
1961 #if IS_ENABLED(CONFIG_IPV6_SIT)
1962         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1963                 cfg.fc_flags |= RTF_NONEXTHOP;
1964 #endif
1965
1966         ip6_route_add(&cfg);
1967 }
1968
1969
1970 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1971                                                   int plen,
1972                                                   const struct net_device *dev,
1973                                                   u32 flags, u32 noflags)
1974 {
1975         struct fib6_node *fn;
1976         struct rt6_info *rt = NULL;
1977         struct fib6_table *table;
1978
1979         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1980         if (table == NULL)
1981                 return NULL;
1982
1983         read_lock_bh(&table->tb6_lock);
1984         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1985         if (!fn)
1986                 goto out;
1987         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1988                 if (rt->dst.dev->ifindex != dev->ifindex)
1989                         continue;
1990                 if ((rt->rt6i_flags & flags) != flags)
1991                         continue;
1992                 if ((rt->rt6i_flags & noflags) != 0)
1993                         continue;
1994                 dst_hold(&rt->dst);
1995                 break;
1996         }
1997 out:
1998         read_unlock_bh(&table->tb6_lock);
1999         return rt;
2000 }
2001
2002
2003 /* Create "default" multicast route to the interface */
2004
2005 static void addrconf_add_mroute(struct net_device *dev)
2006 {
2007         struct fib6_config cfg = {
2008                 .fc_table = RT6_TABLE_LOCAL,
2009                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2010                 .fc_ifindex = dev->ifindex,
2011                 .fc_dst_len = 8,
2012                 .fc_flags = RTF_UP,
2013                 .fc_nlinfo.nl_net = dev_net(dev),
2014         };
2015
2016         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2017
2018         ip6_route_add(&cfg);
2019 }
2020
2021 #if IS_ENABLED(CONFIG_IPV6_SIT)
2022 static void sit_route_add(struct net_device *dev)
2023 {
2024         struct fib6_config cfg = {
2025                 .fc_table = RT6_TABLE_MAIN,
2026                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2027                 .fc_ifindex = dev->ifindex,
2028                 .fc_dst_len = 96,
2029                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2030                 .fc_nlinfo.nl_net = dev_net(dev),
2031         };
2032
2033         /* prefix length - 96 bits "::d.d.d.d" */
2034         ip6_route_add(&cfg);
2035 }
2036 #endif
2037
2038 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2039 {
2040         struct inet6_dev *idev;
2041
2042         ASSERT_RTNL();
2043
2044         idev = ipv6_find_idev(dev);
2045         if (!idev)
2046                 return ERR_PTR(-ENOBUFS);
2047
2048         if (idev->cnf.disable_ipv6)
2049                 return ERR_PTR(-EACCES);
2050
2051         /* Add default multicast route */
2052         if (!(dev->flags & IFF_LOOPBACK))
2053                 addrconf_add_mroute(dev);
2054
2055         return idev;
2056 }
2057
2058 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2059 {
2060         struct prefix_info *pinfo;
2061         __u32 valid_lft;
2062         __u32 prefered_lft;
2063         int addr_type;
2064         struct inet6_dev *in6_dev;
2065         struct net *net = dev_net(dev);
2066
2067         pinfo = (struct prefix_info *) opt;
2068
2069         if (len < sizeof(struct prefix_info)) {
2070                 ADBG("addrconf: prefix option too short\n");
2071                 return;
2072         }
2073
2074         /*
2075          *      Validation checks ([ADDRCONF], page 19)
2076          */
2077
2078         addr_type = ipv6_addr_type(&pinfo->prefix);
2079
2080         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2081                 return;
2082
2083         valid_lft = ntohl(pinfo->valid);
2084         prefered_lft = ntohl(pinfo->prefered);
2085
2086         if (prefered_lft > valid_lft) {
2087                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2088                 return;
2089         }
2090
2091         in6_dev = in6_dev_get(dev);
2092
2093         if (in6_dev == NULL) {
2094                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2095                                     dev->name);
2096                 return;
2097         }
2098
2099         /*
2100          *      Two things going on here:
2101          *      1) Add routes for on-link prefixes
2102          *      2) Configure prefixes with the auto flag set
2103          */
2104
2105         if (pinfo->onlink) {
2106                 struct rt6_info *rt;
2107                 unsigned long rt_expires;
2108
2109                 /* Avoid arithmetic overflow. Really, we could
2110                  * save rt_expires in seconds, likely valid_lft,
2111                  * but it would require division in fib gc, that it
2112                  * not good.
2113                  */
2114                 if (HZ > USER_HZ)
2115                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2116                 else
2117                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2118
2119                 if (addrconf_finite_timeout(rt_expires))
2120                         rt_expires *= HZ;
2121
2122                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2123                                                pinfo->prefix_len,
2124                                                dev,
2125                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2126                                                RTF_GATEWAY | RTF_DEFAULT);
2127
2128                 if (rt) {
2129                         /* Autoconf prefix route */
2130                         if (valid_lft == 0) {
2131                                 ip6_del_rt(rt);
2132                                 rt = NULL;
2133                         } else if (addrconf_finite_timeout(rt_expires)) {
2134                                 /* not infinity */
2135                                 rt6_set_expires(rt, jiffies + rt_expires);
2136                         } else {
2137                                 rt6_clean_expires(rt);
2138                         }
2139                 } else if (valid_lft) {
2140                         clock_t expires = 0;
2141                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2142                         if (addrconf_finite_timeout(rt_expires)) {
2143                                 /* not infinity */
2144                                 flags |= RTF_EXPIRES;
2145                                 expires = jiffies_to_clock_t(rt_expires);
2146                         }
2147                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2148                                               dev, expires, flags);
2149                 }
2150                 ip6_rt_put(rt);
2151         }
2152
2153         /* Try to figure out our local address for this prefix */
2154
2155         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2156                 struct inet6_ifaddr *ifp;
2157                 struct in6_addr addr;
2158                 int create = 0, update_lft = 0;
2159                 bool tokenized = false;
2160
2161                 if (pinfo->prefix_len == 64) {
2162                         memcpy(&addr, &pinfo->prefix, 8);
2163
2164                         if (!ipv6_addr_any(&in6_dev->token)) {
2165                                 read_lock_bh(&in6_dev->lock);
2166                                 memcpy(addr.s6_addr + 8,
2167                                        in6_dev->token.s6_addr + 8, 8);
2168                                 read_unlock_bh(&in6_dev->lock);
2169                                 tokenized = true;
2170                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2171                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2172                                 in6_dev_put(in6_dev);
2173                                 return;
2174                         }
2175                         goto ok;
2176                 }
2177                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2178                                     pinfo->prefix_len);
2179                 in6_dev_put(in6_dev);
2180                 return;
2181
2182 ok:
2183
2184                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2185
2186                 if (ifp == NULL && valid_lft) {
2187                         int max_addresses = in6_dev->cnf.max_addresses;
2188                         u32 addr_flags = 0;
2189
2190 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2191                         if (in6_dev->cnf.optimistic_dad &&
2192                             !net->ipv6.devconf_all->forwarding && sllao)
2193                                 addr_flags = IFA_F_OPTIMISTIC;
2194 #endif
2195
2196                         /* Do not allow to create too much of autoconfigured
2197                          * addresses; this would be too easy way to crash kernel.
2198                          */
2199                         if (!max_addresses ||
2200                             ipv6_count_addresses(in6_dev) < max_addresses)
2201                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2202                                                     pinfo->prefix_len,
2203                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2204                                                     addr_flags, valid_lft,
2205                                                     prefered_lft);
2206
2207                         if (IS_ERR_OR_NULL(ifp)) {
2208                                 in6_dev_put(in6_dev);
2209                                 return;
2210                         }
2211
2212                         update_lft = 0;
2213                         create = 1;
2214                         ifp->cstamp = jiffies;
2215                         ifp->tokenized = tokenized;
2216                         addrconf_dad_start(ifp);
2217                 }
2218
2219                 if (ifp) {
2220                         int flags;
2221                         unsigned long now;
2222 #ifdef CONFIG_IPV6_PRIVACY
2223                         struct inet6_ifaddr *ift;
2224 #endif
2225                         u32 stored_lft;
2226
2227                         /* update lifetime (RFC2462 5.5.3 e) */
2228                         spin_lock(&ifp->lock);
2229                         now = jiffies;
2230                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2231                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2232                         else
2233                                 stored_lft = 0;
2234                         if (!update_lft && !create && stored_lft) {
2235                                 if (valid_lft > MIN_VALID_LIFETIME ||
2236                                     valid_lft > stored_lft)
2237                                         update_lft = 1;
2238                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2239                                         /* valid_lft <= stored_lft is always true */
2240                                         /*
2241                                          * RFC 4862 Section 5.5.3e:
2242                                          * "Note that the preferred lifetime of
2243                                          *  the corresponding address is always
2244                                          *  reset to the Preferred Lifetime in
2245                                          *  the received Prefix Information
2246                                          *  option, regardless of whether the
2247                                          *  valid lifetime is also reset or
2248                                          *  ignored."
2249                                          *
2250                                          *  So if the preferred lifetime in
2251                                          *  this advertisement is different
2252                                          *  than what we have stored, but the
2253                                          *  valid lifetime is invalid, just
2254                                          *  reset prefered_lft.
2255                                          *
2256                                          *  We must set the valid lifetime
2257                                          *  to the stored lifetime since we'll
2258                                          *  be updating the timestamp below,
2259                                          *  else we'll set it back to the
2260                                          *  minimum.
2261                                          */
2262                                         if (prefered_lft != ifp->prefered_lft) {
2263                                                 valid_lft = stored_lft;
2264                                                 update_lft = 1;
2265                                         }
2266                                 } else {
2267                                         valid_lft = MIN_VALID_LIFETIME;
2268                                         if (valid_lft < prefered_lft)
2269                                                 prefered_lft = valid_lft;
2270                                         update_lft = 1;
2271                                 }
2272                         }
2273
2274                         if (update_lft) {
2275                                 ifp->valid_lft = valid_lft;
2276                                 ifp->prefered_lft = prefered_lft;
2277                                 ifp->tstamp = now;
2278                                 flags = ifp->flags;
2279                                 ifp->flags &= ~IFA_F_DEPRECATED;
2280                                 spin_unlock(&ifp->lock);
2281
2282                                 if (!(flags&IFA_F_TENTATIVE))
2283                                         ipv6_ifa_notify(0, ifp);
2284                         } else
2285                                 spin_unlock(&ifp->lock);
2286
2287 #ifdef CONFIG_IPV6_PRIVACY
2288                         read_lock_bh(&in6_dev->lock);
2289                         /* update all temporary addresses in the list */
2290                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2291                                             tmp_list) {
2292                                 int age, max_valid, max_prefered;
2293
2294                                 if (ifp != ift->ifpub)
2295                                         continue;
2296
2297                                 /*
2298                                  * RFC 4941 section 3.3:
2299                                  * If a received option will extend the lifetime
2300                                  * of a public address, the lifetimes of
2301                                  * temporary addresses should be extended,
2302                                  * subject to the overall constraint that no
2303                                  * temporary addresses should ever remain
2304                                  * "valid" or "preferred" for a time longer than
2305                                  * (TEMP_VALID_LIFETIME) or
2306                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2307                                  * respectively.
2308                                  */
2309                                 age = (now - ift->cstamp) / HZ;
2310                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2311                                 if (max_valid < 0)
2312                                         max_valid = 0;
2313
2314                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2315                                                in6_dev->cnf.max_desync_factor -
2316                                                age;
2317                                 if (max_prefered < 0)
2318                                         max_prefered = 0;
2319
2320                                 if (valid_lft > max_valid)
2321                                         valid_lft = max_valid;
2322
2323                                 if (prefered_lft > max_prefered)
2324                                         prefered_lft = max_prefered;
2325
2326                                 spin_lock(&ift->lock);
2327                                 flags = ift->flags;
2328                                 ift->valid_lft = valid_lft;
2329                                 ift->prefered_lft = prefered_lft;
2330                                 ift->tstamp = now;
2331                                 if (prefered_lft > 0)
2332                                         ift->flags &= ~IFA_F_DEPRECATED;
2333
2334                                 spin_unlock(&ift->lock);
2335                                 if (!(flags&IFA_F_TENTATIVE))
2336                                         ipv6_ifa_notify(0, ift);
2337                         }
2338
2339                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2340                                 /*
2341                                  * When a new public address is created as
2342                                  * described in [ADDRCONF], also create a new
2343                                  * temporary address. Also create a temporary
2344                                  * address if it's enabled but no temporary
2345                                  * address currently exists.
2346                                  */
2347                                 read_unlock_bh(&in6_dev->lock);
2348                                 ipv6_create_tempaddr(ifp, NULL);
2349                         } else {
2350                                 read_unlock_bh(&in6_dev->lock);
2351                         }
2352 #endif
2353                         in6_ifa_put(ifp);
2354                         addrconf_verify(0);
2355                 }
2356         }
2357         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2358         in6_dev_put(in6_dev);
2359 }
2360
2361 /*
2362  *      Set destination address.
2363  *      Special case for SIT interfaces where we create a new "virtual"
2364  *      device.
2365  */
2366 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2367 {
2368         struct in6_ifreq ireq;
2369         struct net_device *dev;
2370         int err = -EINVAL;
2371
2372         rtnl_lock();
2373
2374         err = -EFAULT;
2375         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2376                 goto err_exit;
2377
2378         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2379
2380         err = -ENODEV;
2381         if (dev == NULL)
2382                 goto err_exit;
2383
2384 #if IS_ENABLED(CONFIG_IPV6_SIT)
2385         if (dev->type == ARPHRD_SIT) {
2386                 const struct net_device_ops *ops = dev->netdev_ops;
2387                 struct ifreq ifr;
2388                 struct ip_tunnel_parm p;
2389
2390                 err = -EADDRNOTAVAIL;
2391                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2392                         goto err_exit;
2393
2394                 memset(&p, 0, sizeof(p));
2395                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2396                 p.iph.saddr = 0;
2397                 p.iph.version = 4;
2398                 p.iph.ihl = 5;
2399                 p.iph.protocol = IPPROTO_IPV6;
2400                 p.iph.ttl = 64;
2401                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2402
2403                 if (ops->ndo_do_ioctl) {
2404                         mm_segment_t oldfs = get_fs();
2405
2406                         set_fs(KERNEL_DS);
2407                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2408                         set_fs(oldfs);
2409                 } else
2410                         err = -EOPNOTSUPP;
2411
2412                 if (err == 0) {
2413                         err = -ENOBUFS;
2414                         dev = __dev_get_by_name(net, p.name);
2415                         if (!dev)
2416                                 goto err_exit;
2417                         err = dev_open(dev);
2418                 }
2419         }
2420 #endif
2421
2422 err_exit:
2423         rtnl_unlock();
2424         return err;
2425 }
2426
2427 /*
2428  *      Manual configuration of address on an interface
2429  */
2430 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2431                           const struct in6_addr *peer_pfx,
2432                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2433                           __u32 valid_lft)
2434 {
2435         struct inet6_ifaddr *ifp;
2436         struct inet6_dev *idev;
2437         struct net_device *dev;
2438         int scope;
2439         u32 flags;
2440         clock_t expires;
2441         unsigned long timeout;
2442
2443         ASSERT_RTNL();
2444
2445         if (plen > 128)
2446                 return -EINVAL;
2447
2448         /* check the lifetime */
2449         if (!valid_lft || prefered_lft > valid_lft)
2450                 return -EINVAL;
2451
2452         dev = __dev_get_by_index(net, ifindex);
2453         if (!dev)
2454                 return -ENODEV;
2455
2456         idev = addrconf_add_dev(dev);
2457         if (IS_ERR(idev))
2458                 return PTR_ERR(idev);
2459
2460         scope = ipv6_addr_scope(pfx);
2461
2462         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2463         if (addrconf_finite_timeout(timeout)) {
2464                 expires = jiffies_to_clock_t(timeout * HZ);
2465                 valid_lft = timeout;
2466                 flags = RTF_EXPIRES;
2467         } else {
2468                 expires = 0;
2469                 flags = 0;
2470                 ifa_flags |= IFA_F_PERMANENT;
2471         }
2472
2473         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2474         if (addrconf_finite_timeout(timeout)) {
2475                 if (timeout == 0)
2476                         ifa_flags |= IFA_F_DEPRECATED;
2477                 prefered_lft = timeout;
2478         }
2479
2480         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2481                             valid_lft, prefered_lft);
2482
2483         if (!IS_ERR(ifp)) {
2484                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2485                                       expires, flags);
2486                 /*
2487                  * Note that section 3.1 of RFC 4429 indicates
2488                  * that the Optimistic flag should not be set for
2489                  * manually configured addresses
2490                  */
2491                 addrconf_dad_start(ifp);
2492                 in6_ifa_put(ifp);
2493                 addrconf_verify(0);
2494                 return 0;
2495         }
2496
2497         return PTR_ERR(ifp);
2498 }
2499
2500 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2501                           unsigned int plen)
2502 {
2503         struct inet6_ifaddr *ifp;
2504         struct inet6_dev *idev;
2505         struct net_device *dev;
2506
2507         if (plen > 128)
2508                 return -EINVAL;
2509
2510         dev = __dev_get_by_index(net, ifindex);
2511         if (!dev)
2512                 return -ENODEV;
2513
2514         if ((idev = __in6_dev_get(dev)) == NULL)
2515                 return -ENXIO;
2516
2517         read_lock_bh(&idev->lock);
2518         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2519                 if (ifp->prefix_len == plen &&
2520                     ipv6_addr_equal(pfx, &ifp->addr)) {
2521                         in6_ifa_hold(ifp);
2522                         read_unlock_bh(&idev->lock);
2523
2524                         ipv6_del_addr(ifp);
2525                         return 0;
2526                 }
2527         }
2528         read_unlock_bh(&idev->lock);
2529         return -EADDRNOTAVAIL;
2530 }
2531
2532
2533 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2534 {
2535         struct in6_ifreq ireq;
2536         int err;
2537
2538         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2539                 return -EPERM;
2540
2541         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2542                 return -EFAULT;
2543
2544         rtnl_lock();
2545         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2546                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2547                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2548         rtnl_unlock();
2549         return err;
2550 }
2551
2552 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2553 {
2554         struct in6_ifreq ireq;
2555         int err;
2556
2557         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2558                 return -EPERM;
2559
2560         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2561                 return -EFAULT;
2562
2563         rtnl_lock();
2564         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2565                              ireq.ifr6_prefixlen);
2566         rtnl_unlock();
2567         return err;
2568 }
2569
2570 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2571                      int plen, int scope)
2572 {
2573         struct inet6_ifaddr *ifp;
2574
2575         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2576                             scope, IFA_F_PERMANENT, 0, 0);
2577         if (!IS_ERR(ifp)) {
2578                 spin_lock_bh(&ifp->lock);
2579                 ifp->flags &= ~IFA_F_TENTATIVE;
2580                 spin_unlock_bh(&ifp->lock);
2581                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2582                 in6_ifa_put(ifp);
2583         }
2584 }
2585
2586 #if IS_ENABLED(CONFIG_IPV6_SIT)
2587 static void sit_add_v4_addrs(struct inet6_dev *idev)
2588 {
2589         struct in6_addr addr;
2590         struct net_device *dev;
2591         struct net *net = dev_net(idev->dev);
2592         int scope;
2593
2594         ASSERT_RTNL();
2595
2596         memset(&addr, 0, sizeof(struct in6_addr));
2597         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2598
2599         if (idev->dev->flags&IFF_POINTOPOINT) {
2600                 addr.s6_addr32[0] = htonl(0xfe800000);
2601                 scope = IFA_LINK;
2602         } else {
2603                 scope = IPV6_ADDR_COMPATv4;
2604         }
2605
2606         if (addr.s6_addr32[3]) {
2607                 add_addr(idev, &addr, 128, scope);
2608                 return;
2609         }
2610
2611         for_each_netdev(net, dev) {
2612                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2613                 if (in_dev && (dev->flags & IFF_UP)) {
2614                         struct in_ifaddr *ifa;
2615
2616                         int flag = scope;
2617
2618                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2619                                 int plen;
2620
2621                                 addr.s6_addr32[3] = ifa->ifa_local;
2622
2623                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2624                                         continue;
2625                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2626                                         if (idev->dev->flags&IFF_POINTOPOINT)
2627                                                 continue;
2628                                         flag |= IFA_HOST;
2629                                 }
2630                                 if (idev->dev->flags&IFF_POINTOPOINT)
2631                                         plen = 64;
2632                                 else
2633                                         plen = 96;
2634
2635                                 add_addr(idev, &addr, plen, flag);
2636                         }
2637                 }
2638         }
2639 }
2640 #endif
2641
2642 static void init_loopback(struct net_device *dev)
2643 {
2644         struct inet6_dev  *idev;
2645         struct net_device *sp_dev;
2646         struct inet6_ifaddr *sp_ifa;
2647         struct rt6_info *sp_rt;
2648
2649         /* ::1 */
2650
2651         ASSERT_RTNL();
2652
2653         if ((idev = ipv6_find_idev(dev)) == NULL) {
2654                 pr_debug("%s: add_dev failed\n", __func__);
2655                 return;
2656         }
2657
2658         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2659
2660         /* Add routes to other interface's IPv6 addresses */
2661         for_each_netdev(dev_net(dev), sp_dev) {
2662                 if (!strcmp(sp_dev->name, dev->name))
2663                         continue;
2664
2665                 idev = __in6_dev_get(sp_dev);
2666                 if (!idev)
2667                         continue;
2668
2669                 read_lock_bh(&idev->lock);
2670                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2671
2672                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2673                                 continue;
2674
2675                         if (sp_ifa->rt)
2676                                 continue;
2677
2678                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2679
2680                         /* Failure cases are ignored */
2681                         if (!IS_ERR(sp_rt)) {
2682                                 sp_ifa->rt = sp_rt;
2683                                 ip6_ins_rt(sp_rt);
2684                         }
2685                 }
2686                 read_unlock_bh(&idev->lock);
2687         }
2688 }
2689
2690 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2691 {
2692         struct inet6_ifaddr *ifp;
2693         u32 addr_flags = IFA_F_PERMANENT;
2694
2695 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2696         if (idev->cnf.optimistic_dad &&
2697             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2698                 addr_flags |= IFA_F_OPTIMISTIC;
2699 #endif
2700
2701
2702         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2703         if (!IS_ERR(ifp)) {
2704                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2705                 addrconf_dad_start(ifp);
2706                 in6_ifa_put(ifp);
2707         }
2708 }
2709
2710 static void addrconf_dev_config(struct net_device *dev)
2711 {
2712         struct in6_addr addr;
2713         struct inet6_dev *idev;
2714
2715         ASSERT_RTNL();
2716
2717         if ((dev->type != ARPHRD_ETHER) &&
2718             (dev->type != ARPHRD_FDDI) &&
2719             (dev->type != ARPHRD_ARCNET) &&
2720             (dev->type != ARPHRD_INFINIBAND) &&
2721             (dev->type != ARPHRD_IEEE802154) &&
2722             (dev->type != ARPHRD_IEEE1394) &&
2723             (dev->type != ARPHRD_TUNNEL6)) {
2724                 /* Alas, we support only Ethernet autoconfiguration. */
2725                 return;
2726         }
2727
2728         idev = addrconf_add_dev(dev);
2729         if (IS_ERR(idev))
2730                 return;
2731
2732         memset(&addr, 0, sizeof(struct in6_addr));
2733         addr.s6_addr32[0] = htonl(0xFE800000);
2734
2735         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2736                 addrconf_add_linklocal(idev, &addr);
2737 }
2738
2739 #if IS_ENABLED(CONFIG_IPV6_SIT)
2740 static void addrconf_sit_config(struct net_device *dev)
2741 {
2742         struct inet6_dev *idev;
2743
2744         ASSERT_RTNL();
2745
2746         /*
2747          * Configure the tunnel with one of our IPv4
2748          * addresses... we should configure all of
2749          * our v4 addrs in the tunnel
2750          */
2751
2752         if ((idev = ipv6_find_idev(dev)) == NULL) {
2753                 pr_debug("%s: add_dev failed\n", __func__);
2754                 return;
2755         }
2756
2757         if (dev->priv_flags & IFF_ISATAP) {
2758                 struct in6_addr addr;
2759
2760                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2761                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2762                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2763                         addrconf_add_linklocal(idev, &addr);
2764                 return;
2765         }
2766
2767         sit_add_v4_addrs(idev);
2768
2769         if (dev->flags&IFF_POINTOPOINT)
2770                 addrconf_add_mroute(dev);
2771         else
2772                 sit_route_add(dev);
2773 }
2774 #endif
2775
2776 #if IS_ENABLED(CONFIG_NET_IPGRE)
2777 static void addrconf_gre_config(struct net_device *dev)
2778 {
2779         struct inet6_dev *idev;
2780         struct in6_addr addr;
2781
2782         ASSERT_RTNL();
2783
2784         if ((idev = ipv6_find_idev(dev)) == NULL) {
2785                 pr_debug("%s: add_dev failed\n", __func__);
2786                 return;
2787         }
2788
2789         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2790         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2791
2792         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2793                 addrconf_add_linklocal(idev, &addr);
2794 }
2795 #endif
2796
2797 static inline int
2798 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2799 {
2800         struct in6_addr lladdr;
2801
2802         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2803                 addrconf_add_linklocal(idev, &lladdr);
2804                 return 0;
2805         }
2806         return -1;
2807 }
2808
2809 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2810                            void *ptr)
2811 {
2812         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2813         struct inet6_dev *idev = __in6_dev_get(dev);
2814         int run_pending = 0;
2815         int err;
2816
2817         switch (event) {
2818         case NETDEV_REGISTER:
2819                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2820                         idev = ipv6_add_dev(dev);
2821                         if (!idev)
2822                                 return notifier_from_errno(-ENOMEM);
2823                 }
2824                 break;
2825
2826         case NETDEV_UP:
2827         case NETDEV_CHANGE:
2828                 if (dev->flags & IFF_SLAVE)
2829                         break;
2830
2831                 if (event == NETDEV_UP) {
2832                         if (!addrconf_qdisc_ok(dev)) {
2833                                 /* device is not ready yet. */
2834                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2835                                         dev->name);
2836                                 break;
2837                         }
2838
2839                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2840                                 idev = ipv6_add_dev(dev);
2841
2842                         if (idev) {
2843                                 idev->if_flags |= IF_READY;
2844                                 run_pending = 1;
2845                         }
2846                 } else {
2847                         if (!addrconf_qdisc_ok(dev)) {
2848                                 /* device is still not ready. */
2849                                 break;
2850                         }
2851
2852                         if (idev) {
2853                                 if (idev->if_flags & IF_READY)
2854                                         /* device is already configured. */
2855                                         break;
2856                                 idev->if_flags |= IF_READY;
2857                         }
2858
2859                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2860                                 dev->name);
2861
2862                         run_pending = 1;
2863                 }
2864
2865                 switch (dev->type) {
2866 #if IS_ENABLED(CONFIG_IPV6_SIT)
2867                 case ARPHRD_SIT:
2868                         addrconf_sit_config(dev);
2869                         break;
2870 #endif
2871 #if IS_ENABLED(CONFIG_NET_IPGRE)
2872                 case ARPHRD_IPGRE:
2873                         addrconf_gre_config(dev);
2874                         break;
2875 #endif
2876                 case ARPHRD_LOOPBACK:
2877                         init_loopback(dev);
2878                         break;
2879
2880                 default:
2881                         addrconf_dev_config(dev);
2882                         break;
2883                 }
2884
2885                 if (idev) {
2886                         if (run_pending)
2887                                 addrconf_dad_run(idev);
2888
2889                         /*
2890                          * If the MTU changed during the interface down,
2891                          * when the interface up, the changed MTU must be
2892                          * reflected in the idev as well as routers.
2893                          */
2894                         if (idev->cnf.mtu6 != dev->mtu &&
2895                             dev->mtu >= IPV6_MIN_MTU) {
2896                                 rt6_mtu_change(dev, dev->mtu);
2897                                 idev->cnf.mtu6 = dev->mtu;
2898                         }
2899                         idev->tstamp = jiffies;
2900                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2901
2902                         /*
2903                          * If the changed mtu during down is lower than
2904                          * IPV6_MIN_MTU stop IPv6 on this interface.
2905                          */
2906                         if (dev->mtu < IPV6_MIN_MTU)
2907                                 addrconf_ifdown(dev, 1);
2908                 }
2909                 break;
2910
2911         case NETDEV_CHANGEMTU:
2912                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2913                         rt6_mtu_change(dev, dev->mtu);
2914                         idev->cnf.mtu6 = dev->mtu;
2915                         break;
2916                 }
2917
2918                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2919                         idev = ipv6_add_dev(dev);
2920                         if (idev)
2921                                 break;
2922                 }
2923
2924                 /*
2925                  * MTU falled under IPV6_MIN_MTU.
2926                  * Stop IPv6 on this interface.
2927                  */
2928
2929         case NETDEV_DOWN:
2930         case NETDEV_UNREGISTER:
2931                 /*
2932                  *      Remove all addresses from this interface.
2933                  */
2934                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2935                 break;
2936
2937         case NETDEV_CHANGENAME:
2938                 if (idev) {
2939                         snmp6_unregister_dev(idev);
2940                         addrconf_sysctl_unregister(idev);
2941                         addrconf_sysctl_register(idev);
2942                         err = snmp6_register_dev(idev);
2943                         if (err)
2944                                 return notifier_from_errno(err);
2945                 }
2946                 break;
2947
2948         case NETDEV_PRE_TYPE_CHANGE:
2949         case NETDEV_POST_TYPE_CHANGE:
2950                 addrconf_type_change(dev, event);
2951                 break;
2952         }
2953
2954         return NOTIFY_OK;
2955 }
2956
2957 /*
2958  *      addrconf module should be notified of a device going up
2959  */
2960 static struct notifier_block ipv6_dev_notf = {
2961         .notifier_call = addrconf_notify,
2962 };
2963
2964 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2965 {
2966         struct inet6_dev *idev;
2967         ASSERT_RTNL();
2968
2969         idev = __in6_dev_get(dev);
2970
2971         if (event == NETDEV_POST_TYPE_CHANGE)
2972                 ipv6_mc_remap(idev);
2973         else if (event == NETDEV_PRE_TYPE_CHANGE)
2974                 ipv6_mc_unmap(idev);
2975 }
2976
2977 static int addrconf_ifdown(struct net_device *dev, int how)
2978 {
2979         struct net *net = dev_net(dev);
2980         struct inet6_dev *idev;
2981         struct inet6_ifaddr *ifa;
2982         int state, i;
2983
2984         ASSERT_RTNL();
2985
2986         rt6_ifdown(net, dev);
2987         neigh_ifdown(&nd_tbl, dev);
2988
2989         idev = __in6_dev_get(dev);
2990         if (idev == NULL)
2991                 return -ENODEV;
2992
2993         /*
2994          * Step 1: remove reference to ipv6 device from parent device.
2995          *         Do not dev_put!
2996          */
2997         if (how) {
2998                 idev->dead = 1;
2999
3000                 /* protected by rtnl_lock */
3001                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3002
3003                 /* Step 1.5: remove snmp6 entry */
3004                 snmp6_unregister_dev(idev);
3005
3006         }
3007
3008         /* Step 2: clear hash table */
3009         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3010                 struct hlist_head *h = &inet6_addr_lst[i];
3011
3012                 spin_lock_bh(&addrconf_hash_lock);
3013         restart:
3014                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3015                         if (ifa->idev == idev) {
3016                                 hlist_del_init_rcu(&ifa->addr_lst);
3017                                 addrconf_del_dad_timer(ifa);
3018                                 goto restart;
3019                         }
3020                 }
3021                 spin_unlock_bh(&addrconf_hash_lock);
3022         }
3023
3024         write_lock_bh(&idev->lock);
3025
3026         addrconf_del_rs_timer(idev);
3027
3028         /* Step 2: clear flags for stateless addrconf */
3029         if (!how)
3030                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3031
3032 #ifdef CONFIG_IPV6_PRIVACY
3033         if (how && del_timer(&idev->regen_timer))
3034                 in6_dev_put(idev);
3035
3036         /* Step 3: clear tempaddr list */
3037         while (!list_empty(&idev->tempaddr_list)) {
3038                 ifa = list_first_entry(&idev->tempaddr_list,
3039                                        struct inet6_ifaddr, tmp_list);
3040                 list_del(&ifa->tmp_list);
3041                 write_unlock_bh(&idev->lock);
3042                 spin_lock_bh(&ifa->lock);
3043
3044                 if (ifa->ifpub) {
3045                         in6_ifa_put(ifa->ifpub);
3046                         ifa->ifpub = NULL;
3047                 }
3048                 spin_unlock_bh(&ifa->lock);
3049                 in6_ifa_put(ifa);
3050                 write_lock_bh(&idev->lock);
3051         }
3052 #endif
3053
3054         while (!list_empty(&idev->addr_list)) {
3055                 ifa = list_first_entry(&idev->addr_list,
3056                                        struct inet6_ifaddr, if_list);
3057                 addrconf_del_dad_timer(ifa);
3058
3059                 list_del(&ifa->if_list);
3060
3061                 write_unlock_bh(&idev->lock);
3062
3063                 spin_lock_bh(&ifa->state_lock);
3064                 state = ifa->state;
3065                 ifa->state = INET6_IFADDR_STATE_DEAD;
3066                 spin_unlock_bh(&ifa->state_lock);
3067
3068                 if (state != INET6_IFADDR_STATE_DEAD) {
3069                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3070                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3071                 }
3072                 in6_ifa_put(ifa);
3073
3074                 write_lock_bh(&idev->lock);
3075         }
3076
3077         write_unlock_bh(&idev->lock);
3078
3079         /* Step 5: Discard multicast list */
3080         if (how)
3081                 ipv6_mc_destroy_dev(idev);
3082         else
3083                 ipv6_mc_down(idev);
3084
3085         idev->tstamp = jiffies;
3086
3087         /* Last: Shot the device (if unregistered) */
3088         if (how) {
3089                 addrconf_sysctl_unregister(idev);
3090                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3091                 neigh_ifdown(&nd_tbl, dev);
3092                 in6_dev_put(idev);
3093         }
3094         return 0;
3095 }
3096
3097 static void addrconf_rs_timer(unsigned long data)
3098 {
3099         struct inet6_dev *idev = (struct inet6_dev *)data;
3100         struct in6_addr lladdr;
3101
3102         write_lock(&idev->lock);
3103         if (idev->dead || !(idev->if_flags & IF_READY))
3104                 goto out;
3105
3106         if (!ipv6_accept_ra(idev))
3107                 goto out;
3108
3109         /* Announcement received after solicitation was sent */
3110         if (idev->if_flags & IF_RA_RCVD)
3111                 goto out;
3112
3113         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3114                 if (!__ipv6_get_lladdr(idev, &lladdr, IFA_F_TENTATIVE))
3115                         ndisc_send_rs(idev->dev, &lladdr,
3116                                       &in6addr_linklocal_allrouters);
3117                 else
3118                         goto out;
3119
3120                 /* The wait after the last probe can be shorter */
3121                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3122                                              idev->cnf.rtr_solicits) ?
3123                                       idev->cnf.rtr_solicit_delay :
3124                                       idev->cnf.rtr_solicit_interval);
3125         } else {
3126                 /*
3127                  * Note: we do not support deprecated "all on-link"
3128                  * assumption any longer.
3129                  */
3130                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3131         }
3132
3133 out:
3134         write_unlock(&idev->lock);
3135         in6_dev_put(idev);
3136 }
3137
3138 /*
3139  *      Duplicate Address Detection
3140  */
3141 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3142 {
3143         unsigned long rand_num;
3144         struct inet6_dev *idev = ifp->idev;
3145
3146         if (ifp->flags & IFA_F_OPTIMISTIC)
3147                 rand_num = 0;
3148         else
3149                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3150
3151         ifp->dad_probes = idev->cnf.dad_transmits;
3152         addrconf_mod_dad_timer(ifp, rand_num);
3153 }
3154
3155 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3156 {
3157         struct inet6_dev *idev = ifp->idev;
3158         struct net_device *dev = idev->dev;
3159
3160         addrconf_join_solict(dev, &ifp->addr);
3161
3162         net_srandom(ifp->addr.s6_addr32[3]);
3163
3164         read_lock_bh(&idev->lock);
3165         spin_lock(&ifp->lock);
3166         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3167                 goto out;
3168
3169         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3170             idev->cnf.accept_dad < 1 ||
3171             !(ifp->flags&IFA_F_TENTATIVE) ||
3172             ifp->flags & IFA_F_NODAD) {
3173                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3174                 spin_unlock(&ifp->lock);
3175                 read_unlock_bh(&idev->lock);
3176
3177                 addrconf_dad_completed(ifp);
3178                 return;
3179         }
3180
3181         if (!(idev->if_flags & IF_READY)) {
3182                 spin_unlock(&ifp->lock);
3183                 read_unlock_bh(&idev->lock);
3184                 /*
3185                  * If the device is not ready:
3186                  * - keep it tentative if it is a permanent address.
3187                  * - otherwise, kill it.
3188                  */
3189                 in6_ifa_hold(ifp);
3190                 addrconf_dad_stop(ifp, 0);
3191                 return;
3192         }
3193
3194         /*
3195          * Optimistic nodes can start receiving
3196          * Frames right away
3197          */
3198         if (ifp->flags & IFA_F_OPTIMISTIC)
3199                 ip6_ins_rt(ifp->rt);
3200
3201         addrconf_dad_kick(ifp);
3202 out:
3203         spin_unlock(&ifp->lock);
3204         read_unlock_bh(&idev->lock);
3205 }
3206
3207 static void addrconf_dad_timer(unsigned long data)
3208 {
3209         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3210         struct inet6_dev *idev = ifp->idev;
3211         struct in6_addr mcaddr;
3212
3213         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3214                 goto out;
3215
3216         write_lock(&idev->lock);
3217         if (idev->dead || !(idev->if_flags & IF_READY)) {
3218                 write_unlock(&idev->lock);
3219                 goto out;
3220         }
3221
3222         spin_lock(&ifp->lock);
3223         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3224                 spin_unlock(&ifp->lock);
3225                 write_unlock(&idev->lock);
3226                 goto out;
3227         }
3228
3229         if (ifp->dad_probes == 0) {
3230                 /*
3231                  * DAD was successful
3232                  */
3233
3234                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3235                 spin_unlock(&ifp->lock);
3236                 write_unlock(&idev->lock);
3237
3238                 addrconf_dad_completed(ifp);
3239
3240                 goto out;
3241         }
3242
3243         ifp->dad_probes--;
3244         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3245         spin_unlock(&ifp->lock);
3246         write_unlock(&idev->lock);
3247
3248         /* send a neighbour solicitation for our addr */
3249         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3250         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3251 out:
3252         in6_ifa_put(ifp);
3253 }
3254
3255 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3256 {
3257         struct net_device *dev = ifp->idev->dev;
3258         struct in6_addr lladdr;
3259         bool send_rs, send_mld;
3260
3261         addrconf_del_dad_timer(ifp);
3262
3263         /*
3264          *      Configure the address for reception. Now it is valid.
3265          */
3266
3267         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3268
3269         /* If added prefix is link local and we are prepared to process
3270            router advertisements, start sending router solicitations.
3271          */
3272
3273         read_lock_bh(&ifp->idev->lock);
3274         spin_lock(&ifp->lock);
3275         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3276                    ifp->idev->valid_ll_addr_cnt == 1;
3277         send_rs = send_mld &&
3278                   ipv6_accept_ra(ifp->idev) &&
3279                   ifp->idev->cnf.rtr_solicits > 0 &&
3280                   (dev->flags&IFF_LOOPBACK) == 0;
3281         spin_unlock(&ifp->lock);
3282         read_unlock_bh(&ifp->idev->lock);
3283
3284         /* While dad is in progress mld report's source address is in6_addrany.
3285          * Resend with proper ll now.
3286          */
3287         if (send_mld)
3288                 ipv6_mc_dad_complete(ifp->idev);
3289
3290         if (send_rs) {
3291                 /*
3292                  *      If a host as already performed a random delay
3293                  *      [...] as part of DAD [...] there is no need
3294                  *      to delay again before sending the first RS
3295                  */
3296                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3297                         return;
3298                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3299
3300                 write_lock_bh(&ifp->idev->lock);
3301                 spin_lock(&ifp->lock);
3302                 ifp->idev->rs_probes = 1;
3303                 ifp->idev->if_flags |= IF_RS_SENT;
3304                 addrconf_mod_rs_timer(ifp->idev,
3305                                       ifp->idev->cnf.rtr_solicit_interval);
3306                 spin_unlock(&ifp->lock);
3307                 write_unlock_bh(&ifp->idev->lock);
3308         }
3309 }
3310
3311 static void addrconf_dad_run(struct inet6_dev *idev)
3312 {
3313         struct inet6_ifaddr *ifp;
3314
3315         read_lock_bh(&idev->lock);
3316         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3317                 spin_lock(&ifp->lock);
3318                 if (ifp->flags & IFA_F_TENTATIVE &&
3319                     ifp->state == INET6_IFADDR_STATE_DAD)
3320                         addrconf_dad_kick(ifp);
3321                 spin_unlock(&ifp->lock);
3322         }
3323         read_unlock_bh(&idev->lock);
3324 }
3325
3326 #ifdef CONFIG_PROC_FS
3327 struct if6_iter_state {
3328         struct seq_net_private p;
3329         int bucket;
3330         int offset;
3331 };
3332
3333 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3334 {
3335         struct inet6_ifaddr *ifa = NULL;
3336         struct if6_iter_state *state = seq->private;
3337         struct net *net = seq_file_net(seq);
3338         int p = 0;
3339
3340         /* initial bucket if pos is 0 */
3341         if (pos == 0) {
3342                 state->bucket = 0;
3343                 state->offset = 0;
3344         }
3345
3346         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3347                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3348                                          addr_lst) {
3349                         if (!net_eq(dev_net(ifa->idev->dev), net))
3350                                 continue;
3351                         /* sync with offset */
3352                         if (p < state->offset) {
3353                                 p++;
3354                                 continue;
3355                         }
3356                         state->offset++;
3357                         return ifa;
3358                 }
3359
3360                 /* prepare for next bucket */
3361                 state->offset = 0;
3362                 p = 0;
3363         }
3364         return NULL;
3365 }
3366
3367 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3368                                          struct inet6_ifaddr *ifa)
3369 {
3370         struct if6_iter_state *state = seq->private;
3371         struct net *net = seq_file_net(seq);
3372
3373         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3374                 if (!net_eq(dev_net(ifa->idev->dev), net))
3375                         continue;
3376                 state->offset++;
3377                 return ifa;
3378         }
3379
3380         while (++state->bucket < IN6_ADDR_HSIZE) {
3381                 state->offset = 0;
3382                 hlist_for_each_entry_rcu_bh(ifa,
3383                                      &inet6_addr_lst[state->bucket], addr_lst) {
3384                         if (!net_eq(dev_net(ifa->idev->dev), net))
3385                                 continue;
3386                         state->offset++;
3387                         return ifa;
3388                 }
3389         }
3390
3391         return NULL;
3392 }
3393
3394 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3395         __acquires(rcu_bh)
3396 {
3397         rcu_read_lock_bh();
3398         return if6_get_first(seq, *pos);
3399 }
3400
3401 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3402 {
3403         struct inet6_ifaddr *ifa;
3404
3405         ifa = if6_get_next(seq, v);
3406         ++*pos;
3407         return ifa;
3408 }
3409
3410 static void if6_seq_stop(struct seq_file *seq, void *v)
3411         __releases(rcu_bh)
3412 {
3413         rcu_read_unlock_bh();
3414 }
3415
3416 static int if6_seq_show(struct seq_file *seq, void *v)
3417 {
3418         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3419         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3420                    &ifp->addr,
3421                    ifp->idev->dev->ifindex,
3422                    ifp->prefix_len,
3423                    ifp->scope,
3424                    ifp->flags,
3425                    ifp->idev->dev->name);
3426         return 0;
3427 }
3428
3429 static const struct seq_operations if6_seq_ops = {
3430         .start  = if6_seq_start,
3431         .next   = if6_seq_next,
3432         .show   = if6_seq_show,
3433         .stop   = if6_seq_stop,
3434 };
3435
3436 static int if6_seq_open(struct inode *inode, struct file *file)
3437 {
3438         return seq_open_net(inode, file, &if6_seq_ops,
3439                             sizeof(struct if6_iter_state));
3440 }
3441
3442 static const struct file_operations if6_fops = {
3443         .owner          = THIS_MODULE,
3444         .open           = if6_seq_open,
3445         .read           = seq_read,
3446         .llseek         = seq_lseek,
3447         .release        = seq_release_net,
3448 };
3449
3450 static int __net_init if6_proc_net_init(struct net *net)
3451 {
3452         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3453                 return -ENOMEM;
3454         return 0;
3455 }
3456
3457 static void __net_exit if6_proc_net_exit(struct net *net)
3458 {
3459         remove_proc_entry("if_inet6", net->proc_net);
3460 }
3461
3462 static struct pernet_operations if6_proc_net_ops = {
3463        .init = if6_proc_net_init,
3464        .exit = if6_proc_net_exit,
3465 };
3466
3467 int __init if6_proc_init(void)
3468 {
3469         return register_pernet_subsys(&if6_proc_net_ops);
3470 }
3471
3472 void if6_proc_exit(void)
3473 {
3474         unregister_pernet_subsys(&if6_proc_net_ops);
3475 }
3476 #endif  /* CONFIG_PROC_FS */
3477
3478 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3479 /* Check if address is a home address configured on any interface. */
3480 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3481 {
3482         int ret = 0;
3483         struct inet6_ifaddr *ifp = NULL;
3484         unsigned int hash = inet6_addr_hash(addr);
3485
3486         rcu_read_lock_bh();
3487         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3488                 if (!net_eq(dev_net(ifp->idev->dev), net))
3489                         continue;
3490                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3491                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3492                         ret = 1;
3493                         break;
3494                 }
3495         }
3496         rcu_read_unlock_bh();
3497         return ret;
3498 }
3499 #endif
3500
3501 /*
3502  *      Periodic address status verification
3503  */
3504
3505 static void addrconf_verify(unsigned long foo)
3506 {
3507         unsigned long now, next, next_sec, next_sched;
3508         struct inet6_ifaddr *ifp;
3509         int i;
3510
3511         rcu_read_lock_bh();
3512         spin_lock(&addrconf_verify_lock);
3513         now = jiffies;
3514         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3515
3516         del_timer(&addr_chk_timer);
3517
3518         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3519 restart:
3520                 hlist_for_each_entry_rcu_bh(ifp,
3521                                          &inet6_addr_lst[i], addr_lst) {
3522                         unsigned long age;
3523
3524                         if (ifp->flags & IFA_F_PERMANENT)
3525                                 continue;
3526
3527                         spin_lock(&ifp->lock);
3528                         /* We try to batch several events at once. */
3529                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3530
3531                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3532                             age >= ifp->valid_lft) {
3533                                 spin_unlock(&ifp->lock);
3534                                 in6_ifa_hold(ifp);
3535                                 ipv6_del_addr(ifp);
3536                                 goto restart;
3537                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3538                                 spin_unlock(&ifp->lock);
3539                                 continue;
3540                         } else if (age >= ifp->prefered_lft) {
3541                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3542                                 int deprecate = 0;
3543
3544                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3545                                         deprecate = 1;
3546                                         ifp->flags |= IFA_F_DEPRECATED;
3547                                 }
3548
3549                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3550                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3551
3552                                 spin_unlock(&ifp->lock);
3553
3554                                 if (deprecate) {
3555                                         in6_ifa_hold(ifp);
3556
3557                                         ipv6_ifa_notify(0, ifp);
3558                                         in6_ifa_put(ifp);
3559                                         goto restart;
3560                                 }
3561 #ifdef CONFIG_IPV6_PRIVACY
3562                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3563                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3564                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3565                                         ifp->idev->cnf.dad_transmits *
3566                                         ifp->idev->nd_parms->retrans_time / HZ;
3567
3568                                 if (age >= ifp->prefered_lft - regen_advance) {
3569                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3570                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3571                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3572                                         if (!ifp->regen_count && ifpub) {
3573                                                 ifp->regen_count++;
3574                                                 in6_ifa_hold(ifp);
3575                                                 in6_ifa_hold(ifpub);
3576                                                 spin_unlock(&ifp->lock);
3577
3578                                                 spin_lock(&ifpub->lock);
3579                                                 ifpub->regen_count = 0;
3580                                                 spin_unlock(&ifpub->lock);
3581                                                 ipv6_create_tempaddr(ifpub, ifp);
3582                                                 in6_ifa_put(ifpub);
3583                                                 in6_ifa_put(ifp);
3584                                                 goto restart;
3585                                         }
3586                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3587                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3588                                 spin_unlock(&ifp->lock);
3589 #endif
3590                         } else {
3591                                 /* ifp->prefered_lft <= ifp->valid_lft */
3592                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3593                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3594                                 spin_unlock(&ifp->lock);
3595                         }
3596                 }
3597         }
3598
3599         next_sec = round_jiffies_up(next);
3600         next_sched = next;
3601
3602         /* If rounded timeout is accurate enough, accept it. */
3603         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3604                 next_sched = next_sec;
3605
3606         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3607         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3608                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3609
3610         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3611               now, next, next_sec, next_sched);
3612
3613         addr_chk_timer.expires = next_sched;
3614         add_timer(&addr_chk_timer);
3615         spin_unlock(&addrconf_verify_lock);
3616         rcu_read_unlock_bh();
3617 }
3618
3619 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3620                                      struct in6_addr **peer_pfx)
3621 {
3622         struct in6_addr *pfx = NULL;
3623
3624         *peer_pfx = NULL;
3625
3626         if (addr)
3627                 pfx = nla_data(addr);
3628
3629         if (local) {
3630                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3631                         *peer_pfx = pfx;
3632                 pfx = nla_data(local);
3633         }
3634
3635         return pfx;
3636 }
3637
3638 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3639         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3640         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3641         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3642 };
3643
3644 static int
3645 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3646 {
3647         struct net *net = sock_net(skb->sk);
3648         struct ifaddrmsg *ifm;
3649         struct nlattr *tb[IFA_MAX+1];
3650         struct in6_addr *pfx, *peer_pfx;
3651         int err;
3652
3653         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3654         if (err < 0)
3655                 return err;
3656
3657         ifm = nlmsg_data(nlh);
3658         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3659         if (pfx == NULL)
3660                 return -EINVAL;
3661
3662         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3663 }
3664
3665 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3666                              u32 prefered_lft, u32 valid_lft)
3667 {
3668         u32 flags;
3669         clock_t expires;
3670         unsigned long timeout;
3671
3672         if (!valid_lft || (prefered_lft > valid_lft))
3673                 return -EINVAL;
3674
3675         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3676         if (addrconf_finite_timeout(timeout)) {
3677                 expires = jiffies_to_clock_t(timeout * HZ);
3678                 valid_lft = timeout;
3679                 flags = RTF_EXPIRES;
3680         } else {
3681                 expires = 0;
3682                 flags = 0;
3683                 ifa_flags |= IFA_F_PERMANENT;
3684         }
3685
3686         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3687         if (addrconf_finite_timeout(timeout)) {
3688                 if (timeout == 0)
3689                         ifa_flags |= IFA_F_DEPRECATED;
3690                 prefered_lft = timeout;
3691         }
3692
3693         spin_lock_bh(&ifp->lock);
3694         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3695         ifp->tstamp = jiffies;
3696         ifp->valid_lft = valid_lft;
3697         ifp->prefered_lft = prefered_lft;
3698
3699         spin_unlock_bh(&ifp->lock);
3700         if (!(ifp->flags&IFA_F_TENTATIVE))
3701                 ipv6_ifa_notify(0, ifp);
3702
3703         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3704                               expires, flags);
3705         addrconf_verify(0);
3706
3707         return 0;
3708 }
3709
3710 static int
3711 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3712 {
3713         struct net *net = sock_net(skb->sk);
3714         struct ifaddrmsg *ifm;
3715         struct nlattr *tb[IFA_MAX+1];
3716         struct in6_addr *pfx, *peer_pfx;
3717         struct inet6_ifaddr *ifa;
3718         struct net_device *dev;
3719         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3720         u8 ifa_flags;
3721         int err;
3722
3723         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3724         if (err < 0)
3725                 return err;
3726
3727         ifm = nlmsg_data(nlh);
3728         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3729         if (pfx == NULL)
3730                 return -EINVAL;
3731
3732         if (tb[IFA_CACHEINFO]) {
3733                 struct ifa_cacheinfo *ci;
3734
3735                 ci = nla_data(tb[IFA_CACHEINFO]);
3736                 valid_lft = ci->ifa_valid;
3737                 preferred_lft = ci->ifa_prefered;
3738         } else {
3739                 preferred_lft = INFINITY_LIFE_TIME;
3740                 valid_lft = INFINITY_LIFE_TIME;
3741         }
3742
3743         dev =  __dev_get_by_index(net, ifm->ifa_index);
3744         if (dev == NULL)
3745                 return -ENODEV;
3746
3747         /* We ignore other flags so far. */
3748         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3749
3750         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3751         if (ifa == NULL) {
3752                 /*
3753                  * It would be best to check for !NLM_F_CREATE here but
3754                  * userspace alreay relies on not having to provide this.
3755                  */
3756                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3757                                       ifm->ifa_prefixlen, ifa_flags,
3758                                       preferred_lft, valid_lft);
3759         }
3760
3761         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3762             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3763                 err = -EEXIST;
3764         else
3765                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3766
3767         in6_ifa_put(ifa);
3768
3769         return err;
3770 }
3771
3772 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3773                           u8 scope, int ifindex)
3774 {
3775         struct ifaddrmsg *ifm;
3776
3777         ifm = nlmsg_data(nlh);
3778         ifm->ifa_family = AF_INET6;
3779         ifm->ifa_prefixlen = prefixlen;
3780         ifm->ifa_flags = flags;
3781         ifm->ifa_scope = scope;
3782         ifm->ifa_index = ifindex;
3783 }
3784
3785 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3786                          unsigned long tstamp, u32 preferred, u32 valid)
3787 {
3788         struct ifa_cacheinfo ci;
3789
3790         ci.cstamp = cstamp_delta(cstamp);
3791         ci.tstamp = cstamp_delta(tstamp);
3792         ci.ifa_prefered = preferred;
3793         ci.ifa_valid = valid;
3794
3795         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3796 }
3797
3798 static inline int rt_scope(int ifa_scope)
3799 {
3800         if (ifa_scope & IFA_HOST)
3801                 return RT_SCOPE_HOST;
3802         else if (ifa_scope & IFA_LINK)
3803                 return RT_SCOPE_LINK;
3804         else if (ifa_scope & IFA_SITE)
3805                 return RT_SCOPE_SITE;
3806         else
3807                 return RT_SCOPE_UNIVERSE;
3808 }
3809
3810 static inline int inet6_ifaddr_msgsize(void)
3811 {
3812         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3813                + nla_total_size(16) /* IFA_LOCAL */
3814                + nla_total_size(16) /* IFA_ADDRESS */
3815                + nla_total_size(sizeof(struct ifa_cacheinfo));
3816 }
3817
3818 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3819                              u32 portid, u32 seq, int event, unsigned int flags)
3820 {
3821         struct nlmsghdr  *nlh;
3822         u32 preferred, valid;
3823
3824         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3825         if (nlh == NULL)
3826                 return -EMSGSIZE;
3827
3828         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3829                       ifa->idev->dev->ifindex);
3830
3831         if (!(ifa->flags&IFA_F_PERMANENT)) {
3832                 preferred = ifa->prefered_lft;
3833                 valid = ifa->valid_lft;
3834                 if (preferred != INFINITY_LIFE_TIME) {
3835                         long tval = (jiffies - ifa->tstamp)/HZ;
3836                         if (preferred > tval)
3837                                 preferred -= tval;
3838                         else
3839                                 preferred = 0;
3840                         if (valid != INFINITY_LIFE_TIME) {
3841                                 if (valid > tval)
3842                                         valid -= tval;
3843                                 else
3844                                         valid = 0;
3845                         }
3846                 }
3847         } else {
3848                 preferred = INFINITY_LIFE_TIME;
3849                 valid = INFINITY_LIFE_TIME;
3850         }
3851
3852         if (!ipv6_addr_any(&ifa->peer_addr)) {
3853                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3854                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3855                         goto error;
3856         } else
3857                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3858                         goto error;
3859
3860         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3861                 goto error;
3862
3863         return nlmsg_end(skb, nlh);
3864
3865 error:
3866         nlmsg_cancel(skb, nlh);
3867         return -EMSGSIZE;
3868 }
3869
3870 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3871                                 u32 portid, u32 seq, int event, u16 flags)
3872 {
3873         struct nlmsghdr  *nlh;
3874         u8 scope = RT_SCOPE_UNIVERSE;
3875         int ifindex = ifmca->idev->dev->ifindex;
3876
3877         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3878                 scope = RT_SCOPE_SITE;
3879
3880         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3881         if (nlh == NULL)
3882                 return -EMSGSIZE;
3883
3884         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3885         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3886             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3887                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3888                 nlmsg_cancel(skb, nlh);
3889                 return -EMSGSIZE;
3890         }
3891
3892         return nlmsg_end(skb, nlh);
3893 }
3894
3895 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3896                                 u32 portid, u32 seq, int event, unsigned int flags)
3897 {
3898         struct nlmsghdr  *nlh;
3899         u8 scope = RT_SCOPE_UNIVERSE;
3900         int ifindex = ifaca->aca_idev->dev->ifindex;
3901
3902         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3903                 scope = RT_SCOPE_SITE;
3904
3905         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3906         if (nlh == NULL)
3907                 return -EMSGSIZE;
3908
3909         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3910         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3911             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3912                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3913                 nlmsg_cancel(skb, nlh);
3914                 return -EMSGSIZE;
3915         }
3916
3917         return nlmsg_end(skb, nlh);
3918 }
3919
3920 enum addr_type_t {
3921         UNICAST_ADDR,
3922         MULTICAST_ADDR,
3923         ANYCAST_ADDR,
3924 };
3925
3926 /* called with rcu_read_lock() */
3927 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3928                           struct netlink_callback *cb, enum addr_type_t type,
3929                           int s_ip_idx, int *p_ip_idx)
3930 {
3931         struct ifmcaddr6 *ifmca;
3932         struct ifacaddr6 *ifaca;
3933         int err = 1;
3934         int ip_idx = *p_ip_idx;
3935
3936         read_lock_bh(&idev->lock);
3937         switch (type) {
3938         case UNICAST_ADDR: {
3939                 struct inet6_ifaddr *ifa;
3940
3941                 /* unicast address incl. temp addr */
3942                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3943                         if (++ip_idx < s_ip_idx)
3944                                 continue;
3945                         err = inet6_fill_ifaddr(skb, ifa,
3946                                                 NETLINK_CB(cb->skb).portid,
3947                                                 cb->nlh->nlmsg_seq,
3948                                                 RTM_NEWADDR,
3949                                                 NLM_F_MULTI);
3950                         if (err <= 0)
3951                                 break;
3952                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3953                 }
3954                 break;
3955         }
3956         case MULTICAST_ADDR:
3957                 /* multicast address */
3958                 for (ifmca = idev->mc_list; ifmca;
3959                      ifmca = ifmca->next, ip_idx++) {
3960                         if (ip_idx < s_ip_idx)
3961                                 continue;
3962                         err = inet6_fill_ifmcaddr(skb, ifmca,
3963                                                   NETLINK_CB(cb->skb).portid,
3964                                                   cb->nlh->nlmsg_seq,
3965                                                   RTM_GETMULTICAST,
3966                                                   NLM_F_MULTI);
3967                         if (err <= 0)
3968                                 break;
3969                 }
3970                 break;
3971         case ANYCAST_ADDR:
3972                 /* anycast address */
3973                 for (ifaca = idev->ac_list; ifaca;
3974                      ifaca = ifaca->aca_next, ip_idx++) {
3975                         if (ip_idx < s_ip_idx)
3976                                 continue;
3977                         err = inet6_fill_ifacaddr(skb, ifaca,
3978                                                   NETLINK_CB(cb->skb).portid,
3979                                                   cb->nlh->nlmsg_seq,
3980                                                   RTM_GETANYCAST,
3981                                                   NLM_F_MULTI);
3982                         if (err <= 0)
3983                                 break;
3984                 }
3985                 break;
3986         default:
3987                 break;
3988         }
3989         read_unlock_bh(&idev->lock);
3990         *p_ip_idx = ip_idx;
3991         return err;
3992 }
3993
3994 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3995                            enum addr_type_t type)
3996 {
3997         struct net *net = sock_net(skb->sk);
3998         int h, s_h;
3999         int idx, ip_idx;
4000         int s_idx, s_ip_idx;
4001         struct net_device *dev;
4002         struct inet6_dev *idev;
4003         struct hlist_head *head;
4004
4005         s_h = cb->args[0];
4006         s_idx = idx = cb->args[1];
4007         s_ip_idx = ip_idx = cb->args[2];
4008
4009         rcu_read_lock();
4010         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4011         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4012                 idx = 0;
4013                 head = &net->dev_index_head[h];
4014                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4015                         if (idx < s_idx)
4016                                 goto cont;
4017                         if (h > s_h || idx > s_idx)
4018                                 s_ip_idx = 0;
4019                         ip_idx = 0;
4020                         idev = __in6_dev_get(dev);
4021                         if (!idev)
4022                                 goto cont;
4023
4024                         if (in6_dump_addrs(idev, skb, cb, type,
4025                                            s_ip_idx, &ip_idx) <= 0)
4026                                 goto done;
4027 cont:
4028                         idx++;
4029                 }
4030         }
4031 done:
4032         rcu_read_unlock();
4033         cb->args[0] = h;
4034         cb->args[1] = idx;
4035         cb->args[2] = ip_idx;
4036
4037         return skb->len;
4038 }
4039
4040 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4041 {
4042         enum addr_type_t type = UNICAST_ADDR;
4043
4044         return inet6_dump_addr(skb, cb, type);
4045 }
4046
4047 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4048 {
4049         enum addr_type_t type = MULTICAST_ADDR;
4050
4051         return inet6_dump_addr(skb, cb, type);
4052 }
4053
4054
4055 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4056 {
4057         enum addr_type_t type = ANYCAST_ADDR;
4058
4059         return inet6_dump_addr(skb, cb, type);
4060 }
4061
4062 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4063 {
4064         struct net *net = sock_net(in_skb->sk);
4065         struct ifaddrmsg *ifm;
4066         struct nlattr *tb[IFA_MAX+1];
4067         struct in6_addr *addr = NULL, *peer;
4068         struct net_device *dev = NULL;
4069         struct inet6_ifaddr *ifa;
4070         struct sk_buff *skb;
4071         int err;
4072
4073         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4074         if (err < 0)
4075                 goto errout;
4076
4077         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4078         if (addr == NULL) {
4079                 err = -EINVAL;
4080                 goto errout;
4081         }
4082
4083         ifm = nlmsg_data(nlh);
4084         if (ifm->ifa_index)
4085                 dev = __dev_get_by_index(net, ifm->ifa_index);
4086
4087         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4088         if (!ifa) {
4089                 err = -EADDRNOTAVAIL;
4090                 goto errout;
4091         }
4092
4093         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4094         if (!skb) {
4095                 err = -ENOBUFS;
4096                 goto errout_ifa;
4097         }
4098
4099         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4100                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4101         if (err < 0) {
4102                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4103                 WARN_ON(err == -EMSGSIZE);
4104                 kfree_skb(skb);
4105                 goto errout_ifa;
4106         }
4107         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4108 errout_ifa:
4109         in6_ifa_put(ifa);
4110 errout:
4111         return err;
4112 }
4113
4114 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4115 {
4116         struct sk_buff *skb;
4117         struct net *net = dev_net(ifa->idev->dev);
4118         int err = -ENOBUFS;
4119
4120         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4121         if (skb == NULL)
4122                 goto errout;
4123
4124         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4125         if (err < 0) {
4126                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4127                 WARN_ON(err == -EMSGSIZE);
4128                 kfree_skb(skb);
4129                 goto errout;
4130         }
4131         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4132         return;
4133 errout:
4134         if (err < 0)
4135                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4136 }
4137
4138 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4139                                 __s32 *array, int bytes)
4140 {
4141         BUG_ON(bytes < (DEVCONF_MAX * 4));
4142
4143         memset(array, 0, bytes);
4144         array[DEVCONF_FORWARDING] = cnf->forwarding;
4145         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4146         array[DEVCONF_MTU6] = cnf->mtu6;
4147         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4148         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4149         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4150         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4151         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4152         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4153                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4154         array[DEVCONF_RTR_SOLICIT_DELAY] =
4155                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4156         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4157         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4158                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4159         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4160                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4161 #ifdef CONFIG_IPV6_PRIVACY
4162         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4163         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4164         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4165         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4166         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4167 #endif
4168         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4169         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4170         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4171 #ifdef CONFIG_IPV6_ROUTER_PREF
4172         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4173         array[DEVCONF_RTR_PROBE_INTERVAL] =
4174                 jiffies_to_msecs(cnf->rtr_probe_interval);
4175 #ifdef CONFIG_IPV6_ROUTE_INFO
4176         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4177 #endif
4178 #endif
4179         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4180         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4181 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4182         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4183 #endif
4184 #ifdef CONFIG_IPV6_MROUTE
4185         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4186 #endif
4187         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4188         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4189         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4190         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4191 }
4192
4193 static inline size_t inet6_ifla6_size(void)
4194 {
4195         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4196              + nla_total_size(sizeof(struct ifla_cacheinfo))
4197              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4198              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4199              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4200              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4201 }
4202
4203 static inline size_t inet6_if_nlmsg_size(void)
4204 {
4205         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4206                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4207                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4208                + nla_total_size(4) /* IFLA_MTU */
4209                + nla_total_size(4) /* IFLA_LINK */
4210                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4211 }
4212
4213 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4214                                       int items, int bytes)
4215 {
4216         int i;
4217         int pad = bytes - sizeof(u64) * items;
4218         BUG_ON(pad < 0);
4219
4220         /* Use put_unaligned() because stats may not be aligned for u64. */
4221         put_unaligned(items, &stats[0]);
4222         for (i = 1; i < items; i++)
4223                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4224
4225         memset(&stats[items], 0, pad);
4226 }
4227
4228 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4229                                       int items, int bytes, size_t syncpoff)
4230 {
4231         int i;
4232         int pad = bytes - sizeof(u64) * items;
4233         BUG_ON(pad < 0);
4234
4235         /* Use put_unaligned() because stats may not be aligned for u64. */
4236         put_unaligned(items, &stats[0]);
4237         for (i = 1; i < items; i++)
4238                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4239
4240         memset(&stats[items], 0, pad);
4241 }
4242
4243 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4244                              int bytes)
4245 {
4246         switch (attrtype) {
4247         case IFLA_INET6_STATS:
4248                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4249                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4250                 break;
4251         case IFLA_INET6_ICMP6STATS:
4252                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4253                 break;
4254         }
4255 }
4256
4257 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4258 {
4259         struct nlattr *nla;
4260         struct ifla_cacheinfo ci;
4261
4262         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4263                 goto nla_put_failure;
4264         ci.max_reasm_len = IPV6_MAXPLEN;
4265         ci.tstamp = cstamp_delta(idev->tstamp);
4266         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4267         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4268         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4269                 goto nla_put_failure;
4270         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4271         if (nla == NULL)
4272                 goto nla_put_failure;
4273         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4274
4275         /* XXX - MC not implemented */
4276
4277         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4278         if (nla == NULL)
4279                 goto nla_put_failure;
4280         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4281
4282         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4283         if (nla == NULL)
4284                 goto nla_put_failure;
4285         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4286
4287         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4288         if (nla == NULL)
4289                 goto nla_put_failure;
4290         read_lock_bh(&idev->lock);
4291         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4292         read_unlock_bh(&idev->lock);
4293
4294         return 0;
4295
4296 nla_put_failure:
4297         return -EMSGSIZE;
4298 }
4299
4300 static size_t inet6_get_link_af_size(const struct net_device *dev)
4301 {
4302         if (!__in6_dev_get(dev))
4303                 return 0;
4304
4305         return inet6_ifla6_size();
4306 }
4307
4308 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4309 {
4310         struct inet6_dev *idev = __in6_dev_get(dev);
4311
4312         if (!idev)
4313                 return -ENODATA;
4314
4315         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4316                 return -EMSGSIZE;
4317
4318         return 0;
4319 }
4320
4321 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4322 {
4323         struct inet6_ifaddr *ifp;
4324         struct net_device *dev = idev->dev;
4325         bool update_rs = false;
4326         struct in6_addr ll_addr;
4327
4328         if (token == NULL)
4329                 return -EINVAL;
4330         if (ipv6_addr_any(token))
4331                 return -EINVAL;
4332         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4333                 return -EINVAL;
4334         if (!ipv6_accept_ra(idev))
4335                 return -EINVAL;
4336         if (idev->cnf.rtr_solicits <= 0)
4337                 return -EINVAL;
4338
4339         write_lock_bh(&idev->lock);
4340
4341         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4342         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4343
4344         write_unlock_bh(&idev->lock);
4345
4346         if (!idev->dead && (idev->if_flags & IF_READY) &&
4347             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4348                              IFA_F_OPTIMISTIC)) {
4349
4350                 /* If we're not ready, then normal ifup will take care
4351                  * of this. Otherwise, we need to request our rs here.
4352                  */
4353                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4354                 update_rs = true;
4355         }
4356
4357         write_lock_bh(&idev->lock);
4358
4359         if (update_rs) {
4360                 idev->if_flags |= IF_RS_SENT;
4361                 idev->rs_probes = 1;
4362                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4363         }
4364
4365         /* Well, that's kinda nasty ... */
4366         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4367                 spin_lock(&ifp->lock);
4368                 if (ifp->tokenized) {
4369                         ifp->valid_lft = 0;
4370                         ifp->prefered_lft = 0;
4371                 }
4372                 spin_unlock(&ifp->lock);
4373         }
4374
4375         write_unlock_bh(&idev->lock);
4376         addrconf_verify(0);
4377         return 0;
4378 }
4379
4380 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4381 {
4382         int err = -EINVAL;
4383         struct inet6_dev *idev = __in6_dev_get(dev);
4384         struct nlattr *tb[IFLA_INET6_MAX + 1];
4385
4386         if (!idev)
4387                 return -EAFNOSUPPORT;
4388
4389         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4390                 BUG();
4391
4392         if (tb[IFLA_INET6_TOKEN])
4393                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4394
4395         return err;
4396 }
4397
4398 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4399                              u32 portid, u32 seq, int event, unsigned int flags)
4400 {
4401         struct net_device *dev = idev->dev;
4402         struct ifinfomsg *hdr;
4403         struct nlmsghdr *nlh;
4404         void *protoinfo;
4405
4406         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4407         if (nlh == NULL)
4408                 return -EMSGSIZE;
4409
4410         hdr = nlmsg_data(nlh);
4411         hdr->ifi_family = AF_INET6;
4412         hdr->__ifi_pad = 0;
4413         hdr->ifi_type = dev->type;
4414         hdr->ifi_index = dev->ifindex;
4415         hdr->ifi_flags = dev_get_flags(dev);
4416         hdr->ifi_change = 0;
4417
4418         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4419             (dev->addr_len &&
4420              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4421             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4422             (dev->ifindex != dev->iflink &&
4423              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4424                 goto nla_put_failure;
4425         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4426         if (protoinfo == NULL)
4427                 goto nla_put_failure;
4428
4429         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4430                 goto nla_put_failure;
4431
4432         nla_nest_end(skb, protoinfo);
4433         return nlmsg_end(skb, nlh);
4434
4435 nla_put_failure:
4436         nlmsg_cancel(skb, nlh);
4437         return -EMSGSIZE;
4438 }
4439
4440 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4441 {
4442         struct net *net = sock_net(skb->sk);
4443         int h, s_h;
4444         int idx = 0, s_idx;
4445         struct net_device *dev;
4446         struct inet6_dev *idev;
4447         struct hlist_head *head;
4448
4449         s_h = cb->args[0];
4450         s_idx = cb->args[1];
4451
4452         rcu_read_lock();
4453         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4454                 idx = 0;
4455                 head = &net->dev_index_head[h];
4456                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4457                         if (idx < s_idx)
4458                                 goto cont;
4459                         idev = __in6_dev_get(dev);
4460                         if (!idev)
4461                                 goto cont;
4462                         if (inet6_fill_ifinfo(skb, idev,
4463                                               NETLINK_CB(cb->skb).portid,
4464                                               cb->nlh->nlmsg_seq,
4465                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4466                                 goto out;
4467 cont:
4468                         idx++;
4469                 }
4470         }
4471 out:
4472         rcu_read_unlock();
4473         cb->args[1] = idx;
4474         cb->args[0] = h;
4475
4476         return skb->len;
4477 }
4478
4479 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4480 {
4481         struct sk_buff *skb;
4482         struct net *net = dev_net(idev->dev);
4483         int err = -ENOBUFS;
4484
4485         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4486         if (skb == NULL)
4487                 goto errout;
4488
4489         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4490         if (err < 0) {
4491                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4492                 WARN_ON(err == -EMSGSIZE);
4493                 kfree_skb(skb);
4494                 goto errout;
4495         }
4496         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4497         return;
4498 errout:
4499         if (err < 0)
4500                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4501 }
4502
4503 static inline size_t inet6_prefix_nlmsg_size(void)
4504 {
4505         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4506                + nla_total_size(sizeof(struct in6_addr))
4507                + nla_total_size(sizeof(struct prefix_cacheinfo));
4508 }
4509
4510 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4511                              struct prefix_info *pinfo, u32 portid, u32 seq,
4512                              int event, unsigned int flags)
4513 {
4514         struct prefixmsg *pmsg;
4515         struct nlmsghdr *nlh;
4516         struct prefix_cacheinfo ci;
4517
4518         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4519         if (nlh == NULL)
4520                 return -EMSGSIZE;
4521
4522         pmsg = nlmsg_data(nlh);
4523         pmsg->prefix_family = AF_INET6;
4524         pmsg->prefix_pad1 = 0;
4525         pmsg->prefix_pad2 = 0;
4526         pmsg->prefix_ifindex = idev->dev->ifindex;
4527         pmsg->prefix_len = pinfo->prefix_len;
4528         pmsg->prefix_type = pinfo->type;
4529         pmsg->prefix_pad3 = 0;
4530         pmsg->prefix_flags = 0;
4531         if (pinfo->onlink)
4532                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4533         if (pinfo->autoconf)
4534                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4535
4536         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4537                 goto nla_put_failure;
4538         ci.preferred_time = ntohl(pinfo->prefered);
4539         ci.valid_time = ntohl(pinfo->valid);
4540         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4541                 goto nla_put_failure;
4542         return nlmsg_end(skb, nlh);
4543
4544 nla_put_failure:
4545         nlmsg_cancel(skb, nlh);
4546         return -EMSGSIZE;
4547 }
4548
4549 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4550                          struct prefix_info *pinfo)
4551 {
4552         struct sk_buff *skb;
4553         struct net *net = dev_net(idev->dev);
4554         int err = -ENOBUFS;
4555
4556         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4557         if (skb == NULL)
4558                 goto errout;
4559
4560         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4561         if (err < 0) {
4562                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4563                 WARN_ON(err == -EMSGSIZE);
4564                 kfree_skb(skb);
4565                 goto errout;
4566         }
4567         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4568         return;
4569 errout:
4570         if (err < 0)
4571                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4572 }
4573
4574 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4575 {
4576         write_lock_bh(&ifp->idev->lock);
4577         spin_lock(&ifp->lock);
4578         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4579                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4580             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4581                 ifp->idev->valid_ll_addr_cnt += count;
4582         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4583         spin_unlock(&ifp->lock);
4584         write_unlock_bh(&ifp->idev->lock);
4585 }
4586
4587 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4588 {
4589         struct net *net = dev_net(ifp->idev->dev);
4590
4591         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4592
4593         switch (event) {
4594         case RTM_NEWADDR:
4595                 update_valid_ll_addr_cnt(ifp, 1);
4596
4597                 /*
4598                  * If the address was optimistic
4599                  * we inserted the route at the start of
4600                  * our DAD process, so we don't need
4601                  * to do it again
4602                  */
4603                 if (!(ifp->rt->rt6i_node))
4604                         ip6_ins_rt(ifp->rt);
4605                 if (ifp->idev->cnf.forwarding)
4606                         addrconf_join_anycast(ifp);
4607                 if (!ipv6_addr_any(&ifp->peer_addr))
4608                         addrconf_prefix_route(&ifp->peer_addr, 128,
4609                                               ifp->idev->dev, 0, 0);
4610                 break;
4611         case RTM_DELADDR:
4612                 update_valid_ll_addr_cnt(ifp, -1);
4613
4614                 if (ifp->idev->cnf.forwarding)
4615                         addrconf_leave_anycast(ifp);
4616                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4617                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4618                         struct rt6_info *rt;
4619                         struct net_device *dev = ifp->idev->dev;
4620
4621                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4622                                         dev->ifindex, 1);
4623                         if (rt) {
4624                                 dst_hold(&rt->dst);
4625                                 if (ip6_del_rt(rt))
4626                                         dst_free(&rt->dst);
4627                         }
4628                 }
4629                 dst_hold(&ifp->rt->dst);
4630
4631                 if (ip6_del_rt(ifp->rt))
4632                         dst_free(&ifp->rt->dst);
4633                 break;
4634         }
4635         atomic_inc(&net->ipv6.dev_addr_genid);
4636         rt_genid_bump_ipv6(net);
4637 }
4638
4639 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4640 {
4641         rcu_read_lock_bh();
4642         if (likely(ifp->idev->dead == 0))
4643                 __ipv6_ifa_notify(event, ifp);
4644         rcu_read_unlock_bh();
4645 }
4646
4647 #ifdef CONFIG_SYSCTL
4648
4649 static
4650 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4651                            void __user *buffer, size_t *lenp, loff_t *ppos)
4652 {
4653         int *valp = ctl->data;
4654         int val = *valp;
4655         loff_t pos = *ppos;
4656         struct ctl_table lctl;
4657         int ret;
4658
4659         /*
4660          * ctl->data points to idev->cnf.forwarding, we should
4661          * not modify it until we get the rtnl lock.
4662          */
4663         lctl = *ctl;
4664         lctl.data = &val;
4665
4666         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4667
4668         if (write)
4669                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4670         if (ret)
4671                 *ppos = pos;
4672         return ret;
4673 }
4674
4675 static void dev_disable_change(struct inet6_dev *idev)
4676 {
4677         struct netdev_notifier_info info;
4678
4679         if (!idev || !idev->dev)
4680                 return;
4681
4682         netdev_notifier_info_init(&info, idev->dev);
4683         if (idev->cnf.disable_ipv6)
4684                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4685         else
4686                 addrconf_notify(NULL, NETDEV_UP, &info);
4687 }
4688
4689 static void addrconf_disable_change(struct net *net, __s32 newf)
4690 {
4691         struct net_device *dev;
4692         struct inet6_dev *idev;
4693
4694         rcu_read_lock();
4695         for_each_netdev_rcu(net, dev) {
4696                 idev = __in6_dev_get(dev);
4697                 if (idev) {
4698                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4699                         idev->cnf.disable_ipv6 = newf;
4700                         if (changed)
4701                                 dev_disable_change(idev);
4702                 }
4703         }
4704         rcu_read_unlock();
4705 }
4706
4707 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4708 {
4709         struct net *net;
4710         int old;
4711
4712         if (!rtnl_trylock())
4713                 return restart_syscall();
4714
4715         net = (struct net *)table->extra2;
4716         old = *p;
4717         *p = newf;
4718
4719         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4720                 rtnl_unlock();
4721                 return 0;
4722         }
4723
4724         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4725                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4726                 addrconf_disable_change(net, newf);
4727         } else if ((!newf) ^ (!old))
4728                 dev_disable_change((struct inet6_dev *)table->extra1);
4729
4730         rtnl_unlock();
4731         return 0;
4732 }
4733
4734 static
4735 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4736                             void __user *buffer, size_t *lenp, loff_t *ppos)
4737 {
4738         int *valp = ctl->data;
4739         int val = *valp;
4740         loff_t pos = *ppos;
4741         struct ctl_table lctl;
4742         int ret;
4743
4744         /*
4745          * ctl->data points to idev->cnf.disable_ipv6, we should
4746          * not modify it until we get the rtnl lock.
4747          */
4748         lctl = *ctl;
4749         lctl.data = &val;
4750
4751         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4752
4753         if (write)
4754                 ret = addrconf_disable_ipv6(ctl, valp, val);
4755         if (ret)
4756                 *ppos = pos;
4757         return ret;
4758 }
4759
4760 static struct addrconf_sysctl_table
4761 {
4762         struct ctl_table_header *sysctl_header;
4763         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4764 } addrconf_sysctl __read_mostly = {
4765         .sysctl_header = NULL,
4766         .addrconf_vars = {
4767                 {
4768                         .procname       = "forwarding",
4769                         .data           = &ipv6_devconf.forwarding,
4770                         .maxlen         = sizeof(int),
4771                         .mode           = 0644,
4772                         .proc_handler   = addrconf_sysctl_forward,
4773                 },
4774                 {
4775                         .procname       = "hop_limit",
4776                         .data           = &ipv6_devconf.hop_limit,
4777                         .maxlen         = sizeof(int),
4778                         .mode           = 0644,
4779                         .proc_handler   = proc_dointvec,
4780                 },
4781                 {
4782                         .procname       = "mtu",
4783                         .data           = &ipv6_devconf.mtu6,
4784                         .maxlen         = sizeof(int),
4785                         .mode           = 0644,
4786                         .proc_handler   = proc_dointvec,
4787                 },
4788                 {
4789                         .procname       = "accept_ra",
4790                         .data           = &ipv6_devconf.accept_ra,
4791                         .maxlen         = sizeof(int),
4792                         .mode           = 0644,
4793                         .proc_handler   = proc_dointvec,
4794                 },
4795                 {
4796                         .procname       = "accept_redirects",
4797                         .data           = &ipv6_devconf.accept_redirects,
4798                         .maxlen         = sizeof(int),
4799                         .mode           = 0644,
4800                         .proc_handler   = proc_dointvec,
4801                 },
4802                 {
4803                         .procname       = "autoconf",
4804                         .data           = &ipv6_devconf.autoconf,
4805                         .maxlen         = sizeof(int),
4806                         .mode           = 0644,
4807                         .proc_handler   = proc_dointvec,
4808                 },
4809                 {
4810                         .procname       = "dad_transmits",
4811                         .data           = &ipv6_devconf.dad_transmits,
4812                         .maxlen         = sizeof(int),
4813                         .mode           = 0644,
4814                         .proc_handler   = proc_dointvec,
4815                 },
4816                 {
4817                         .procname       = "router_solicitations",
4818                         .data           = &ipv6_devconf.rtr_solicits,
4819                         .maxlen         = sizeof(int),
4820                         .mode           = 0644,
4821                         .proc_handler   = proc_dointvec,
4822                 },
4823                 {
4824                         .procname       = "router_solicitation_interval",
4825                         .data           = &ipv6_devconf.rtr_solicit_interval,
4826                         .maxlen         = sizeof(int),
4827                         .mode           = 0644,
4828                         .proc_handler   = proc_dointvec_jiffies,
4829                 },
4830                 {
4831                         .procname       = "router_solicitation_delay",
4832                         .data           = &ipv6_devconf.rtr_solicit_delay,
4833                         .maxlen         = sizeof(int),
4834                         .mode           = 0644,
4835                         .proc_handler   = proc_dointvec_jiffies,
4836                 },
4837                 {
4838                         .procname       = "force_mld_version",
4839                         .data           = &ipv6_devconf.force_mld_version,
4840                         .maxlen         = sizeof(int),
4841                         .mode           = 0644,
4842                         .proc_handler   = proc_dointvec,
4843                 },
4844                 {
4845                         .procname       = "mldv1_unsolicited_report_interval",
4846                         .data           =
4847                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4848                         .maxlen         = sizeof(int),
4849                         .mode           = 0644,
4850                         .proc_handler   = proc_dointvec_ms_jiffies,
4851                 },
4852                 {
4853                         .procname       = "mldv2_unsolicited_report_interval",
4854                         .data           =
4855                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4856                         .maxlen         = sizeof(int),
4857                         .mode           = 0644,
4858                         .proc_handler   = proc_dointvec_ms_jiffies,
4859                 },
4860 #ifdef CONFIG_IPV6_PRIVACY
4861                 {
4862                         .procname       = "use_tempaddr",
4863                         .data           = &ipv6_devconf.use_tempaddr,
4864                         .maxlen         = sizeof(int),
4865                         .mode           = 0644,
4866                         .proc_handler   = proc_dointvec,
4867                 },
4868                 {
4869                         .procname       = "temp_valid_lft",
4870                         .data           = &ipv6_devconf.temp_valid_lft,
4871                         .maxlen         = sizeof(int),
4872                         .mode           = 0644,
4873                         .proc_handler   = proc_dointvec,
4874                 },
4875                 {
4876                         .procname       = "temp_prefered_lft",
4877                         .data           = &ipv6_devconf.temp_prefered_lft,
4878                         .maxlen         = sizeof(int),
4879                         .mode           = 0644,
4880                         .proc_handler   = proc_dointvec,
4881                 },
4882                 {
4883                         .procname       = "regen_max_retry",
4884                         .data           = &ipv6_devconf.regen_max_retry,
4885                         .maxlen         = sizeof(int),
4886                         .mode           = 0644,
4887                         .proc_handler   = proc_dointvec,
4888                 },
4889                 {
4890                         .procname       = "max_desync_factor",
4891                         .data           = &ipv6_devconf.max_desync_factor,
4892                         .maxlen         = sizeof(int),
4893                         .mode           = 0644,
4894                         .proc_handler   = proc_dointvec,
4895                 },
4896 #endif
4897                 {
4898                         .procname       = "max_addresses",
4899                         .data           = &ipv6_devconf.max_addresses,
4900                         .maxlen         = sizeof(int),
4901                         .mode           = 0644,
4902                         .proc_handler   = proc_dointvec,
4903                 },
4904                 {
4905                         .procname       = "accept_ra_defrtr",
4906                         .data           = &ipv6_devconf.accept_ra_defrtr,
4907                         .maxlen         = sizeof(int),
4908                         .mode           = 0644,
4909                         .proc_handler   = proc_dointvec,
4910                 },
4911                 {
4912                         .procname       = "accept_ra_pinfo",
4913                         .data           = &ipv6_devconf.accept_ra_pinfo,
4914                         .maxlen         = sizeof(int),
4915                         .mode           = 0644,
4916                         .proc_handler   = proc_dointvec,
4917                 },
4918 #ifdef CONFIG_IPV6_ROUTER_PREF
4919                 {
4920                         .procname       = "accept_ra_rtr_pref",
4921                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4922                         .maxlen         = sizeof(int),
4923                         .mode           = 0644,
4924                         .proc_handler   = proc_dointvec,
4925                 },
4926                 {
4927                         .procname       = "router_probe_interval",
4928                         .data           = &ipv6_devconf.rtr_probe_interval,
4929                         .maxlen         = sizeof(int),
4930                         .mode           = 0644,
4931                         .proc_handler   = proc_dointvec_jiffies,
4932                 },
4933 #ifdef CONFIG_IPV6_ROUTE_INFO
4934                 {
4935                         .procname       = "accept_ra_rt_info_max_plen",
4936                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4937                         .maxlen         = sizeof(int),
4938                         .mode           = 0644,
4939                         .proc_handler   = proc_dointvec,
4940                 },
4941 #endif
4942 #endif
4943                 {
4944                         .procname       = "proxy_ndp",
4945                         .data           = &ipv6_devconf.proxy_ndp,
4946                         .maxlen         = sizeof(int),
4947                         .mode           = 0644,
4948                         .proc_handler   = proc_dointvec,
4949                 },
4950                 {
4951                         .procname       = "accept_source_route",
4952                         .data           = &ipv6_devconf.accept_source_route,
4953                         .maxlen         = sizeof(int),
4954                         .mode           = 0644,
4955                         .proc_handler   = proc_dointvec,
4956                 },
4957 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4958                 {
4959                         .procname       = "optimistic_dad",
4960                         .data           = &ipv6_devconf.optimistic_dad,
4961                         .maxlen         = sizeof(int),
4962                         .mode           = 0644,
4963                         .proc_handler   = proc_dointvec,
4964
4965                 },
4966 #endif
4967 #ifdef CONFIG_IPV6_MROUTE
4968                 {
4969                         .procname       = "mc_forwarding",
4970                         .data           = &ipv6_devconf.mc_forwarding,
4971                         .maxlen         = sizeof(int),
4972                         .mode           = 0444,
4973                         .proc_handler   = proc_dointvec,
4974                 },
4975 #endif
4976                 {
4977                         .procname       = "disable_ipv6",
4978                         .data           = &ipv6_devconf.disable_ipv6,
4979                         .maxlen         = sizeof(int),
4980                         .mode           = 0644,
4981                         .proc_handler   = addrconf_sysctl_disable,
4982                 },
4983                 {
4984                         .procname       = "accept_dad",
4985                         .data           = &ipv6_devconf.accept_dad,
4986                         .maxlen         = sizeof(int),
4987                         .mode           = 0644,
4988                         .proc_handler   = proc_dointvec,
4989                 },
4990                 {
4991                         .procname       = "force_tllao",
4992                         .data           = &ipv6_devconf.force_tllao,
4993                         .maxlen         = sizeof(int),
4994                         .mode           = 0644,
4995                         .proc_handler   = proc_dointvec
4996                 },
4997                 {
4998                         .procname       = "ndisc_notify",
4999                         .data           = &ipv6_devconf.ndisc_notify,
5000                         .maxlen         = sizeof(int),
5001                         .mode           = 0644,
5002                         .proc_handler   = proc_dointvec
5003                 },
5004                 {
5005                         /* sentinel */
5006                 }
5007         },
5008 };
5009
5010 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5011                 struct inet6_dev *idev, struct ipv6_devconf *p)
5012 {
5013         int i;
5014         struct addrconf_sysctl_table *t;
5015         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5016
5017         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5018         if (t == NULL)
5019                 goto out;
5020
5021         for (i = 0; t->addrconf_vars[i].data; i++) {
5022                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5023                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5024                 t->addrconf_vars[i].extra2 = net;
5025         }
5026
5027         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5028
5029         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5030         if (t->sysctl_header == NULL)
5031                 goto free;
5032
5033         p->sysctl = t;
5034         return 0;
5035
5036 free:
5037         kfree(t);
5038 out:
5039         return -ENOBUFS;
5040 }
5041
5042 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5043 {
5044         struct addrconf_sysctl_table *t;
5045
5046         if (p->sysctl == NULL)
5047                 return;
5048
5049         t = p->sysctl;
5050         p->sysctl = NULL;
5051         unregister_net_sysctl_table(t->sysctl_header);
5052         kfree(t);
5053 }
5054
5055 static void addrconf_sysctl_register(struct inet6_dev *idev)
5056 {
5057         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5058                               &ndisc_ifinfo_sysctl_change);
5059         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5060                                         idev, &idev->cnf);
5061 }
5062
5063 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5064 {
5065         __addrconf_sysctl_unregister(&idev->cnf);
5066         neigh_sysctl_unregister(idev->nd_parms);
5067 }
5068
5069
5070 #endif
5071
5072 static int __net_init addrconf_init_net(struct net *net)
5073 {
5074         int err = -ENOMEM;
5075         struct ipv6_devconf *all, *dflt;
5076
5077         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5078         if (all == NULL)
5079                 goto err_alloc_all;
5080
5081         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5082         if (dflt == NULL)
5083                 goto err_alloc_dflt;
5084
5085         /* these will be inherited by all namespaces */
5086         dflt->autoconf = ipv6_defaults.autoconf;
5087         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5088
5089         net->ipv6.devconf_all = all;
5090         net->ipv6.devconf_dflt = dflt;
5091
5092 #ifdef CONFIG_SYSCTL
5093         err = __addrconf_sysctl_register(net, "all", NULL, all);
5094         if (err < 0)
5095                 goto err_reg_all;
5096
5097         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5098         if (err < 0)
5099                 goto err_reg_dflt;
5100 #endif
5101         return 0;
5102
5103 #ifdef CONFIG_SYSCTL
5104 err_reg_dflt:
5105         __addrconf_sysctl_unregister(all);
5106 err_reg_all:
5107         kfree(dflt);
5108 #endif
5109 err_alloc_dflt:
5110         kfree(all);
5111 err_alloc_all:
5112         return err;
5113 }
5114
5115 static void __net_exit addrconf_exit_net(struct net *net)
5116 {
5117 #ifdef CONFIG_SYSCTL
5118         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5119         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5120 #endif
5121         if (!net_eq(net, &init_net)) {
5122                 kfree(net->ipv6.devconf_dflt);
5123                 kfree(net->ipv6.devconf_all);
5124         }
5125 }
5126
5127 static struct pernet_operations addrconf_ops = {
5128         .init = addrconf_init_net,
5129         .exit = addrconf_exit_net,
5130 };
5131
5132 static struct rtnl_af_ops inet6_ops = {
5133         .family           = AF_INET6,
5134         .fill_link_af     = inet6_fill_link_af,
5135         .get_link_af_size = inet6_get_link_af_size,
5136         .set_link_af      = inet6_set_link_af,
5137 };
5138
5139 /*
5140  *      Init / cleanup code
5141  */
5142
5143 int __init addrconf_init(void)
5144 {
5145         int i, err;
5146
5147         err = ipv6_addr_label_init();
5148         if (err < 0) {
5149                 pr_crit("%s: cannot initialize default policy table: %d\n",
5150                         __func__, err);
5151                 goto out;
5152         }
5153
5154         err = register_pernet_subsys(&addrconf_ops);
5155         if (err < 0)
5156                 goto out_addrlabel;
5157
5158         /* The addrconf netdev notifier requires that loopback_dev
5159          * has it's ipv6 private information allocated and setup
5160          * before it can bring up and give link-local addresses
5161          * to other devices which are up.
5162          *
5163          * Unfortunately, loopback_dev is not necessarily the first
5164          * entry in the global dev_base list of net devices.  In fact,
5165          * it is likely to be the very last entry on that list.
5166          * So this causes the notifier registry below to try and
5167          * give link-local addresses to all devices besides loopback_dev
5168          * first, then loopback_dev, which cases all the non-loopback_dev
5169          * devices to fail to get a link-local address.
5170          *
5171          * So, as a temporary fix, allocate the ipv6 structure for
5172          * loopback_dev first by hand.
5173          * Longer term, all of the dependencies ipv6 has upon the loopback
5174          * device and it being up should be removed.
5175          */
5176         rtnl_lock();
5177         if (!ipv6_add_dev(init_net.loopback_dev))
5178                 err = -ENOMEM;
5179         rtnl_unlock();
5180         if (err)
5181                 goto errlo;
5182
5183         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5184                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5185
5186         register_netdevice_notifier(&ipv6_dev_notf);
5187
5188         addrconf_verify(0);
5189
5190         err = rtnl_af_register(&inet6_ops);
5191         if (err < 0)
5192                 goto errout_af;
5193
5194         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5195                               NULL);
5196         if (err < 0)
5197                 goto errout;
5198
5199         /* Only the first call to __rtnl_register can fail */
5200         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5201         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5202         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5203                         inet6_dump_ifaddr, NULL);
5204         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5205                         inet6_dump_ifmcaddr, NULL);
5206         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5207                         inet6_dump_ifacaddr, NULL);
5208         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5209                         inet6_netconf_dump_devconf, NULL);
5210
5211         ipv6_addr_label_rtnl_register();
5212
5213         return 0;
5214 errout:
5215         rtnl_af_unregister(&inet6_ops);
5216 errout_af:
5217         unregister_netdevice_notifier(&ipv6_dev_notf);
5218 errlo:
5219         unregister_pernet_subsys(&addrconf_ops);
5220 out_addrlabel:
5221         ipv6_addr_label_cleanup();
5222 out:
5223         return err;
5224 }
5225
5226 void addrconf_cleanup(void)
5227 {
5228         struct net_device *dev;
5229         int i;
5230
5231         unregister_netdevice_notifier(&ipv6_dev_notf);
5232         unregister_pernet_subsys(&addrconf_ops);
5233         ipv6_addr_label_cleanup();
5234
5235         rtnl_lock();
5236
5237         __rtnl_af_unregister(&inet6_ops);
5238
5239         /* clean dev list */
5240         for_each_netdev(&init_net, dev) {
5241                 if (__in6_dev_get(dev) == NULL)
5242                         continue;
5243                 addrconf_ifdown(dev, 1);
5244         }
5245         addrconf_ifdown(init_net.loopback_dev, 2);
5246
5247         /*
5248          *      Check hash table.
5249          */
5250         spin_lock_bh(&addrconf_hash_lock);
5251         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5252                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5253         spin_unlock_bh(&addrconf_hash_lock);
5254
5255         del_timer(&addr_chk_timer);
5256         rtnl_unlock();
5257 }