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

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86
  87 #ifdef CONFIG_IPV6_PRIVACY
  88 #include <linux/random.h>
  89 #endif
  90
  91 #include <linux/uaccess.h>
  92 #include <asm/unaligned.h>
  93
  94 #include <linux/proc_fs.h>
  95 #include <linux/seq_file.h>
  96 #include <linux/export.h>
  97
  98 /* Set to 3 to get tracing... */
  99 #define ACONF_DEBUG 2
 100
 101 #if ACONF_DEBUG >= 3
 102 #define ADBG(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         .suppress_frag_ndisc    = 1,
 208 };
 209
 210 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 211         .forwarding             = 0,
 212         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 213         .mtu6                   = IPV6_MIN_MTU,
 214         .accept_ra              = 1,
 215         .accept_redirects       = 1,
 216         .autoconf               = 1,
 217         .force_mld_version      = 0,
 218         .mldv1_unsolicited_report_interval = 10 * HZ,
 219         .mldv2_unsolicited_report_interval = HZ,
 220         .dad_transmits          = 1,
 221         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 222         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 223         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 224 #ifdef CONFIG_IPV6_PRIVACY
 225         .use_tempaddr           = 0,
 226         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 227         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 228         .regen_max_retry        = REGEN_MAX_RETRY,
 229         .max_desync_factor      = MAX_DESYNC_FACTOR,
 230 #endif
 231         .max_addresses          = IPV6_MAX_ADDRESSES,
 232         .accept_ra_defrtr       = 1,
 233         .accept_ra_pinfo        = 1,
 234 #ifdef CONFIG_IPV6_ROUTER_PREF
 235         .accept_ra_rtr_pref     = 1,
 236         .rtr_probe_interval     = 60 * HZ,
 237 #ifdef CONFIG_IPV6_ROUTE_INFO
 238         .accept_ra_rt_info_max_plen = 0,
 239 #endif
 240 #endif
 241         .proxy_ndp              = 0,
 242         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 243         .disable_ipv6           = 0,
 244         .accept_dad             = 1,
 245         .suppress_frag_ndisc    = 1,
 246 };
 247
 248 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 249 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 250 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 251 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 252 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 253 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 254 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 255 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 256
 257 /* Check if a valid qdisc is available */
 258 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 259 {
 260         return !qdisc_tx_is_noop(dev);
 261 }
 262
 263 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 264 {
 265         if (del_timer(&idev->rs_timer))
 266                 __in6_dev_put(idev);
 267 }
 268
 269 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 270 {
 271         if (del_timer(&ifp->dad_timer))
 272                 __in6_ifa_put(ifp);
 273 }
 274
 275 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 276                                   unsigned long when)
 277 {
 278         if (!timer_pending(&idev->rs_timer))
 279                 in6_dev_hold(idev);
 280         mod_timer(&idev->rs_timer, jiffies + when);
 281 }
 282
 283 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 284                                    unsigned long when)
 285 {
 286         if (!timer_pending(&ifp->dad_timer))
 287                 in6_ifa_hold(ifp);
 288         mod_timer(&ifp->dad_timer, jiffies + when);
 289 }
 290
 291 static int snmp6_alloc_dev(struct inet6_dev *idev)
 292 {
 293         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 294                           sizeof(struct ipstats_mib),
 295                           __alignof__(struct ipstats_mib)) < 0)
 296                 goto err_ip;
 297         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 298                                         GFP_KERNEL);
 299         if (!idev->stats.icmpv6dev)
 300                 goto err_icmp;
 301         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 302                                            GFP_KERNEL);
 303         if (!idev->stats.icmpv6msgdev)
 304                 goto err_icmpmsg;
 305
 306         return 0;
 307
 308 err_icmpmsg:
 309         kfree(idev->stats.icmpv6dev);
 310 err_icmp:
 311         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 312 err_ip:
 313         return -ENOMEM;
 314 }
 315
 316 static void snmp6_free_dev(struct inet6_dev *idev)
 317 {
 318         kfree(idev->stats.icmpv6msgdev);
 319         kfree(idev->stats.icmpv6dev);
 320         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 321 }
 322
 323 /* Nobody refers to this device, we may destroy it. */
 324
 325 void in6_dev_finish_destroy(struct inet6_dev *idev)
 326 {
 327         struct net_device *dev = idev->dev;
 328
 329         WARN_ON(!list_empty(&idev->addr_list));
 330         WARN_ON(idev->mc_list != NULL);
 331         WARN_ON(timer_pending(&idev->rs_timer));
 332
 333 #ifdef NET_REFCNT_DEBUG
 334         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 335 #endif
 336         dev_put(dev);
 337         if (!idev->dead) {
 338                 pr_warn("Freeing alive inet6 device %p\n", idev);
 339                 return;
 340         }
 341         snmp6_free_dev(idev);
 342         kfree_rcu(idev, rcu);
 343 }
 344 EXPORT_SYMBOL(in6_dev_finish_destroy);
 345
 346 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 347 {
 348         struct inet6_dev *ndev;
 349
 350         ASSERT_RTNL();
 351
 352         if (dev->mtu < IPV6_MIN_MTU)
 353                 return NULL;
 354
 355         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 356
 357         if (ndev == NULL)
 358                 return NULL;
 359
 360         rwlock_init(&ndev->lock);
 361         ndev->dev = dev;
 362         INIT_LIST_HEAD(&ndev->addr_list);
 363         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 364                     (unsigned long)ndev);
 365         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 366         ndev->cnf.mtu6 = dev->mtu;
 367         ndev->cnf.sysctl = NULL;
 368         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 369         if (ndev->nd_parms == NULL) {
 370                 kfree(ndev);
 371                 return NULL;
 372         }
 373         if (ndev->cnf.forwarding)
 374                 dev_disable_lro(dev);
 375         /* We refer to the device */
 376         dev_hold(dev);
 377
 378         if (snmp6_alloc_dev(ndev) < 0) {
 379                 ADBG(KERN_WARNING
 380                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 381                         __func__, dev->name);
 382                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 383                 dev_put(dev);
 384                 kfree(ndev);
 385                 return NULL;
 386         }
 387
 388         if (snmp6_register_dev(ndev) < 0) {
 389                 ADBG(KERN_WARNING
 390                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 391                         __func__, dev->name);
 392                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 393                 ndev->dead = 1;
 394                 in6_dev_finish_destroy(ndev);
 395                 return NULL;
 396         }
 397
 398         /* One reference from device.  We must do this before
 399          * we invoke __ipv6_regen_rndid().
 400          */
 401         in6_dev_hold(ndev);
 402
 403         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 404                 ndev->cnf.accept_dad = -1;
 405
 406 #if IS_ENABLED(CONFIG_IPV6_SIT)
 407         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 408                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 409                 ndev->cnf.rtr_solicits = 0;
 410         }
 411 #endif
 412
 413 #ifdef CONFIG_IPV6_PRIVACY
 414         INIT_LIST_HEAD(&ndev->tempaddr_list);
 415         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 416         if ((dev->flags&IFF_LOOPBACK) ||
 417             dev->type == ARPHRD_TUNNEL ||
 418             dev->type == ARPHRD_TUNNEL6 ||
 419             dev->type == ARPHRD_SIT ||
 420             dev->type == ARPHRD_NONE) {
 421                 ndev->cnf.use_tempaddr = -1;
 422         } else {
 423                 in6_dev_hold(ndev);
 424                 ipv6_regen_rndid((unsigned long) ndev);
 425         }
 426 #endif
 427         ndev->token = in6addr_any;
 428
 429         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 430                 ndev->if_flags |= IF_READY;
 431
 432         ipv6_mc_init_dev(ndev);
 433         ndev->tstamp = jiffies;
 434         addrconf_sysctl_register(ndev);
 435         /* protected by rtnl_lock */
 436         rcu_assign_pointer(dev->ip6_ptr, ndev);
 437
 438         /* Join interface-local all-node multicast group */
 439         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 440
 441         /* Join all-node multicast group */
 442         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 443
 444         /* Join all-router multicast group if forwarding is set */
 445         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 446                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 447
 448         return ndev;
 449 }
 450
 451 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 452 {
 453         struct inet6_dev *idev;
 454
 455         ASSERT_RTNL();
 456
 457         idev = __in6_dev_get(dev);
 458         if (!idev) {
 459                 idev = ipv6_add_dev(dev);
 460                 if (!idev)
 461                         return NULL;
 462         }
 463
 464         if (dev->flags&IFF_UP)
 465                 ipv6_mc_up(idev);
 466         return idev;
 467 }
 468
 469 static int inet6_netconf_msgsize_devconf(int type)
 470 {
 471         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 472                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 473
 474         /* type -1 is used for ALL */
 475         if (type == -1 || type == NETCONFA_FORWARDING)
 476                 size += nla_total_size(4);
 477 #ifdef CONFIG_IPV6_MROUTE
 478         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 479                 size += nla_total_size(4);
 480 #endif
 481
 482         return size;
 483 }
 484
 485 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 486                                       struct ipv6_devconf *devconf, u32 portid,
 487                                       u32 seq, int event, unsigned int flags,
 488                                       int type)
 489 {
 490         struct nlmsghdr  *nlh;
 491         struct netconfmsg *ncm;
 492
 493         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 494                         flags);
 495         if (nlh == NULL)
 496                 return -EMSGSIZE;
 497
 498         ncm = nlmsg_data(nlh);
 499         ncm->ncm_family = AF_INET6;
 500
 501         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 502                 goto nla_put_failure;
 503
 504         /* type -1 is used for ALL */
 505         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 506             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 507                 goto nla_put_failure;
 508 #ifdef CONFIG_IPV6_MROUTE
 509         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 510             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 511                         devconf->mc_forwarding) < 0)
 512                 goto nla_put_failure;
 513 #endif
 514         return nlmsg_end(skb, nlh);
 515
 516 nla_put_failure:
 517         nlmsg_cancel(skb, nlh);
 518         return -EMSGSIZE;
 519 }
 520
 521 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 522                                   struct ipv6_devconf *devconf)
 523 {
 524         struct sk_buff *skb;
 525         int err = -ENOBUFS;
 526
 527         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 528         if (skb == NULL)
 529                 goto errout;
 530
 531         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 532                                          RTM_NEWNETCONF, 0, type);
 533         if (err < 0) {
 534                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 535                 WARN_ON(err == -EMSGSIZE);
 536                 kfree_skb(skb);
 537                 goto errout;
 538         }
 539         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 540         return;
 541 errout:
 542         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 543 }
 544
 545 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 546         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 547         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 548 };
 549
 550 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 551                                      struct nlmsghdr *nlh)
 552 {
 553         struct net *net = sock_net(in_skb->sk);
 554         struct nlattr *tb[NETCONFA_MAX+1];
 555         struct netconfmsg *ncm;
 556         struct sk_buff *skb;
 557         struct ipv6_devconf *devconf;
 558         struct inet6_dev *in6_dev;
 559         struct net_device *dev;
 560         int ifindex;
 561         int err;
 562
 563         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 564                           devconf_ipv6_policy);
 565         if (err < 0)
 566                 goto errout;
 567
 568         err = EINVAL;
 569         if (!tb[NETCONFA_IFINDEX])
 570                 goto errout;
 571
 572         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 573         switch (ifindex) {
 574         case NETCONFA_IFINDEX_ALL:
 575                 devconf = net->ipv6.devconf_all;
 576                 break;
 577         case NETCONFA_IFINDEX_DEFAULT:
 578                 devconf = net->ipv6.devconf_dflt;
 579                 break;
 580         default:
 581                 dev = __dev_get_by_index(net, ifindex);
 582                 if (dev == NULL)
 583                         goto errout;
 584                 in6_dev = __in6_dev_get(dev);
 585                 if (in6_dev == NULL)
 586                         goto errout;
 587                 devconf = &in6_dev->cnf;
 588                 break;
 589         }
 590
 591         err = -ENOBUFS;
 592         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 593         if (skb == NULL)
 594                 goto errout;
 595
 596         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 597                                          NETLINK_CB(in_skb).portid,
 598                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 599                                          -1);
 600         if (err < 0) {
 601                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 602                 WARN_ON(err == -EMSGSIZE);
 603                 kfree_skb(skb);
 604                 goto errout;
 605         }
 606         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 607 errout:
 608         return err;
 609 }
 610
 611 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 612                                       struct netlink_callback *cb)
 613 {
 614         struct net *net = sock_net(skb->sk);
 615         int h, s_h;
 616         int idx, s_idx;
 617         struct net_device *dev;
 618         struct inet6_dev *idev;
 619         struct hlist_head *head;
 620
 621         s_h = cb->args[0];
 622         s_idx = idx = cb->args[1];
 623
 624         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 625                 idx = 0;
 626                 head = &net->dev_index_head[h];
 627                 rcu_read_lock();
 628                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 629                           net->dev_base_seq;
 630                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 631                         if (idx < s_idx)
 632                                 goto cont;
 633                         idev = __in6_dev_get(dev);
 634                         if (!idev)
 635                                 goto cont;
 636
 637                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 638                                                        &idev->cnf,
 639                                                        NETLINK_CB(cb->skb).portid,
 640                                                        cb->nlh->nlmsg_seq,
 641                                                        RTM_NEWNETCONF,
 642                                                        NLM_F_MULTI,
 643                                                        -1) <= 0) {
 644                                 rcu_read_unlock();
 645                                 goto done;
 646                         }
 647                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 648 cont:
 649                         idx++;
 650                 }
 651                 rcu_read_unlock();
 652         }
 653         if (h == NETDEV_HASHENTRIES) {
 654                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 655                                                net->ipv6.devconf_all,
 656                                                NETLINK_CB(cb->skb).portid,
 657                                                cb->nlh->nlmsg_seq,
 658                                                RTM_NEWNETCONF, NLM_F_MULTI,
 659                                                -1) <= 0)
 660                         goto done;
 661                 else
 662                         h++;
 663         }
 664         if (h == NETDEV_HASHENTRIES + 1) {
 665                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 666                                                net->ipv6.devconf_dflt,
 667                                                NETLINK_CB(cb->skb).portid,
 668                                                cb->nlh->nlmsg_seq,
 669                                                RTM_NEWNETCONF, NLM_F_MULTI,
 670                                                -1) <= 0)
 671                         goto done;
 672                 else
 673                         h++;
 674         }
 675 done:
 676         cb->args[0] = h;
 677         cb->args[1] = idx;
 678
 679         return skb->len;
 680 }
 681
 682 #ifdef CONFIG_SYSCTL
 683 static void dev_forward_change(struct inet6_dev *idev)
 684 {
 685         struct net_device *dev;
 686         struct inet6_ifaddr *ifa;
 687
 688         if (!idev)
 689                 return;
 690         dev = idev->dev;
 691         if (idev->cnf.forwarding)
 692                 dev_disable_lro(dev);
 693         if (dev->flags & IFF_MULTICAST) {
 694                 if (idev->cnf.forwarding) {
 695                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 696                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 697                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 698                 } else {
 699                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 700                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 701                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 702                 }
 703         }
 704
 705         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 706                 if (ifa->flags&IFA_F_TENTATIVE)
 707                         continue;
 708                 if (idev->cnf.forwarding)
 709                         addrconf_join_anycast(ifa);
 710                 else
 711                         addrconf_leave_anycast(ifa);
 712         }
 713         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 714                                      dev->ifindex, &idev->cnf);
 715 }
 716
 717
 718 static void addrconf_forward_change(struct net *net, __s32 newf)
 719 {
 720         struct net_device *dev;
 721         struct inet6_dev *idev;
 722
 723         for_each_netdev(net, dev) {
 724                 idev = __in6_dev_get(dev);
 725                 if (idev) {
 726                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 727                         idev->cnf.forwarding = newf;
 728                         if (changed)
 729                                 dev_forward_change(idev);
 730                 }
 731         }
 732 }
 733
 734 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 735 {
 736         struct net *net;
 737         int old;
 738
 739         if (!rtnl_trylock())
 740                 return restart_syscall();
 741
 742         net = (struct net *)table->extra2;
 743         old = *p;
 744         *p = newf;
 745
 746         if (p == &net->ipv6.devconf_dflt->forwarding) {
 747                 if ((!newf) ^ (!old))
 748                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 749                                                      NETCONFA_IFINDEX_DEFAULT,
 750                                                      net->ipv6.devconf_dflt);
 751                 rtnl_unlock();
 752                 return 0;
 753         }
 754
 755         if (p == &net->ipv6.devconf_all->forwarding) {
 756                 net->ipv6.devconf_dflt->forwarding = newf;
 757                 addrconf_forward_change(net, newf);
 758                 if ((!newf) ^ (!old))
 759                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 760                                                      NETCONFA_IFINDEX_ALL,
 761                                                      net->ipv6.devconf_all);
 762         } else if ((!newf) ^ (!old))
 763                 dev_forward_change((struct inet6_dev *)table->extra1);
 764         rtnl_unlock();
 765
 766         if (newf)
 767                 rt6_purge_dflt_routers(net);
 768         return 1;
 769 }
 770 #endif
 771
 772 /* Nobody refers to this ifaddr, destroy it */
 773 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 774 {
 775         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 776
 777 #ifdef NET_REFCNT_DEBUG
 778         pr_debug("%s\n", __func__);
 779 #endif
 780
 781         in6_dev_put(ifp->idev);
 782
 783         if (del_timer(&ifp->dad_timer))
 784                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 785
 786         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 787                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 788                 return;
 789         }
 790         ip6_rt_put(ifp->rt);
 791
 792         kfree_rcu(ifp, rcu);
 793 }
 794
 795 static void
 796 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 797 {
 798         struct list_head *p;
 799         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 800
 801         /*
 802          * Each device address list is sorted in order of scope -
 803          * global before linklocal.
 804          */
 805         list_for_each(p, &idev->addr_list) {
 806                 struct inet6_ifaddr *ifa
 807                         = list_entry(p, struct inet6_ifaddr, if_list);
 808                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 809                         break;
 810         }
 811
 812         list_add_tail(&ifp->if_list, p);
 813 }
 814
 815 static u32 inet6_addr_hash(const struct in6_addr *addr)
 816 {
 817         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 818 }
 819
 820 /* On success it returns ifp with increased reference count */
 821
 822 static struct inet6_ifaddr *
 823 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 824               const struct in6_addr *peer_addr, int pfxlen,
 825               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 826 {
 827         struct inet6_ifaddr *ifa = NULL;
 828         struct rt6_info *rt;
 829         unsigned int hash;
 830         int err = 0;
 831         int addr_type = ipv6_addr_type(addr);
 832
 833         if (addr_type == IPV6_ADDR_ANY ||
 834             addr_type & IPV6_ADDR_MULTICAST ||
 835             (!(idev->dev->flags & IFF_LOOPBACK) &&
 836              addr_type & IPV6_ADDR_LOOPBACK))
 837                 return ERR_PTR(-EADDRNOTAVAIL);
 838
 839         rcu_read_lock_bh();
 840         if (idev->dead) {
 841                 err = -ENODEV;                  /*XXX*/
 842                 goto out2;
 843         }
 844
 845         if (idev->cnf.disable_ipv6) {
 846                 err = -EACCES;
 847                 goto out2;
 848         }
 849
 850         spin_lock(&addrconf_hash_lock);
 851
 852         /* Ignore adding duplicate addresses on an interface */
 853         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 854                 ADBG("ipv6_add_addr: already assigned\n");
 855                 err = -EEXIST;
 856                 goto out;
 857         }
 858
 859         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 860
 861         if (ifa == NULL) {
 862                 ADBG("ipv6_add_addr: malloc failed\n");
 863                 err = -ENOBUFS;
 864                 goto out;
 865         }
 866
 867         rt = addrconf_dst_alloc(idev, addr, false);
 868         if (IS_ERR(rt)) {
 869                 err = PTR_ERR(rt);
 870                 goto out;
 871         }
 872
 873         ifa->addr = *addr;
 874         if (peer_addr)
 875                 ifa->peer_addr = *peer_addr;
 876
 877         spin_lock_init(&ifa->lock);
 878         spin_lock_init(&ifa->state_lock);
 879         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 880                     (unsigned long)ifa);
 881         INIT_HLIST_NODE(&ifa->addr_lst);
 882         ifa->scope = scope;
 883         ifa->prefix_len = pfxlen;
 884         ifa->flags = flags | IFA_F_TENTATIVE;
 885         ifa->valid_lft = valid_lft;
 886         ifa->prefered_lft = prefered_lft;
 887         ifa->cstamp = ifa->tstamp = jiffies;
 888         ifa->tokenized = false;
 889
 890         ifa->rt = rt;
 891
 892         ifa->idev = idev;
 893         in6_dev_hold(idev);
 894         /* For caller */
 895         in6_ifa_hold(ifa);
 896
 897         /* Add to big hash table */
 898         hash = inet6_addr_hash(addr);
 899
 900         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 901         spin_unlock(&addrconf_hash_lock);
 902
 903         write_lock(&idev->lock);
 904         /* Add to inet6_dev unicast addr list. */
 905         ipv6_link_dev_addr(idev, ifa);
 906
 907 #ifdef CONFIG_IPV6_PRIVACY
 908         if (ifa->flags&IFA_F_TEMPORARY) {
 909                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 910                 in6_ifa_hold(ifa);
 911         }
 912 #endif
 913
 914         in6_ifa_hold(ifa);
 915         write_unlock(&idev->lock);
 916 out2:
 917         rcu_read_unlock_bh();
 918
 919         if (likely(err == 0))
 920                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 921         else {
 922                 kfree(ifa);
 923                 ifa = ERR_PTR(err);
 924         }
 925
 926         return ifa;
 927 out:
 928         spin_unlock(&addrconf_hash_lock);
 929         goto out2;
 930 }
 931
 932 /* This function wants to get referenced ifp and releases it before return */
 933
 934 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 935 {
 936         struct inet6_ifaddr *ifa, *ifn;
 937         struct inet6_dev *idev = ifp->idev;
 938         int state;
 939         int deleted = 0, onlink = 0;
 940         unsigned long expires = jiffies;
 941
 942         spin_lock_bh(&ifp->state_lock);
 943         state = ifp->state;
 944         ifp->state = INET6_IFADDR_STATE_DEAD;
 945         spin_unlock_bh(&ifp->state_lock);
 946
 947         if (state == INET6_IFADDR_STATE_DEAD)
 948                 goto out;
 949
 950         spin_lock_bh(&addrconf_hash_lock);
 951         hlist_del_init_rcu(&ifp->addr_lst);
 952         spin_unlock_bh(&addrconf_hash_lock);
 953
 954         write_lock_bh(&idev->lock);
 955 #ifdef CONFIG_IPV6_PRIVACY
 956         if (ifp->flags&IFA_F_TEMPORARY) {
 957                 list_del(&ifp->tmp_list);
 958                 if (ifp->ifpub) {
 959                         in6_ifa_put(ifp->ifpub);
 960                         ifp->ifpub = NULL;
 961                 }
 962                 __in6_ifa_put(ifp);
 963         }
 964 #endif
 965
 966         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 967                 if (ifa == ifp) {
 968                         list_del_init(&ifp->if_list);
 969                         __in6_ifa_put(ifp);
 970
 971                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 972                                 break;
 973                         deleted = 1;
 974                         continue;
 975                 } else if (ifp->flags & IFA_F_PERMANENT) {
 976                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 977                                               ifp->prefix_len)) {
 978                                 if (ifa->flags & IFA_F_PERMANENT) {
 979                                         onlink = 1;
 980                                         if (deleted)
 981                                                 break;
 982                                 } else {
 983                                         unsigned long lifetime;
 984
 985                                         if (!onlink)
 986                                                 onlink = -1;
 987
 988                                         spin_lock(&ifa->lock);
 989
 990                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 991                                         /*
 992                                          * Note: Because this address is
 993                                          * not permanent, lifetime <
 994                                          * LONG_MAX / HZ here.
 995                                          */
 996                                         if (time_before(expires,
 997                                                         ifa->tstamp + lifetime * HZ))
 998                                                 expires = ifa->tstamp + lifetime * HZ;
 999                                         spin_unlock(&ifa->lock);
1000                                 }
1001                         }
1002                 }
1003         }
1004         write_unlock_bh(&idev->lock);
1005
1006         addrconf_del_dad_timer(ifp);
1007
1008         ipv6_ifa_notify(RTM_DELADDR, ifp);
1009
1010         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1011
1012         /*
1013          * Purge or update corresponding prefix
1014          *
1015          * 1) we don't purge prefix here if address was not permanent.
1016          *    prefix is managed by its own lifetime.
1017          * 2) if there're no addresses, delete prefix.
1018          * 3) if there're still other permanent address(es),
1019          *    corresponding prefix is still permanent.
1020          * 4) otherwise, update prefix lifetime to the
1021          *    longest valid lifetime among the corresponding
1022          *    addresses on the device.
1023          *    Note: subsequent RA will update lifetime.
1024          *
1025          * --yoshfuji
1026          */
1027         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1028                 struct in6_addr prefix;
1029                 struct rt6_info *rt;
1030
1031                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1032
1033                 rt = addrconf_get_prefix_route(&prefix,
1034                                                ifp->prefix_len,
1035                                                ifp->idev->dev,
1036                                                0, RTF_GATEWAY | RTF_DEFAULT);
1037
1038                 if (rt) {
1039                         if (onlink == 0) {
1040                                 ip6_del_rt(rt);
1041                                 rt = NULL;
1042                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1043                                 rt6_set_expires(rt, expires);
1044                         }
1045                 }
1046                 ip6_rt_put(rt);
1047         }
1048
1049         /* clean up prefsrc entries */
1050         rt6_remove_prefsrc(ifp);
1051 out:
1052         in6_ifa_put(ifp);
1053 }
1054
1055 #ifdef CONFIG_IPV6_PRIVACY
1056 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1057 {
1058         struct inet6_dev *idev = ifp->idev;
1059         struct in6_addr addr, *tmpaddr;
1060         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1061         unsigned long regen_advance;
1062         int tmp_plen;
1063         int ret = 0;
1064         int max_addresses;
1065         u32 addr_flags;
1066         unsigned long now = jiffies;
1067
1068         write_lock(&idev->lock);
1069         if (ift) {
1070                 spin_lock_bh(&ift->lock);
1071                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1072                 spin_unlock_bh(&ift->lock);
1073                 tmpaddr = &addr;
1074         } else {
1075                 tmpaddr = NULL;
1076         }
1077 retry:
1078         in6_dev_hold(idev);
1079         if (idev->cnf.use_tempaddr <= 0) {
1080                 write_unlock(&idev->lock);
1081                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1082                 in6_dev_put(idev);
1083                 ret = -1;
1084                 goto out;
1085         }
1086         spin_lock_bh(&ifp->lock);
1087         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1088                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1089                 spin_unlock_bh(&ifp->lock);
1090                 write_unlock(&idev->lock);
1091                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1092                         __func__);
1093                 in6_dev_put(idev);
1094                 ret = -1;
1095                 goto out;
1096         }
1097         in6_ifa_hold(ifp);
1098         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1099         __ipv6_try_regen_rndid(idev, tmpaddr);
1100         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1101         age = (now - ifp->tstamp) / HZ;
1102         tmp_valid_lft = min_t(__u32,
1103                               ifp->valid_lft,
1104                               idev->cnf.temp_valid_lft + age);
1105         tmp_prefered_lft = min_t(__u32,
1106                                  ifp->prefered_lft,
1107                                  idev->cnf.temp_prefered_lft + age -
1108                                  idev->cnf.max_desync_factor);
1109         tmp_plen = ifp->prefix_len;
1110         max_addresses = idev->cnf.max_addresses;
1111         tmp_tstamp = ifp->tstamp;
1112         spin_unlock_bh(&ifp->lock);
1113
1114         regen_advance = idev->cnf.regen_max_retry *
1115                         idev->cnf.dad_transmits *
1116                         idev->nd_parms->retrans_time / HZ;
1117         write_unlock(&idev->lock);
1118
1119         /* A temporary address is created only if this calculated Preferred
1120          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1121          * an implementation must not create a temporary address with a zero
1122          * Preferred Lifetime.
1123          */
1124         if (tmp_prefered_lft <= regen_advance) {
1125                 in6_ifa_put(ifp);
1126                 in6_dev_put(idev);
1127                 ret = -1;
1128                 goto out;
1129         }
1130
1131         addr_flags = IFA_F_TEMPORARY;
1132         /* set in addrconf_prefix_rcv() */
1133         if (ifp->flags & IFA_F_OPTIMISTIC)
1134                 addr_flags |= IFA_F_OPTIMISTIC;
1135
1136         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1137                             ipv6_addr_scope(&addr), addr_flags,
1138                             tmp_valid_lft, tmp_prefered_lft);
1139         if (IS_ERR(ift)) {
1140                 in6_ifa_put(ifp);
1141                 in6_dev_put(idev);
1142                 pr_info("%s: retry temporary address regeneration\n", __func__);
1143                 tmpaddr = &addr;
1144                 write_lock(&idev->lock);
1145                 goto retry;
1146         }
1147
1148         spin_lock_bh(&ift->lock);
1149         ift->ifpub = ifp;
1150         ift->cstamp = now;
1151         ift->tstamp = tmp_tstamp;
1152         spin_unlock_bh(&ift->lock);
1153
1154         addrconf_dad_start(ift);
1155         in6_ifa_put(ift);
1156         in6_dev_put(idev);
1157 out:
1158         return ret;
1159 }
1160 #endif
1161
1162 /*
1163  *      Choose an appropriate source address (RFC3484)
1164  */
1165 enum {
1166         IPV6_SADDR_RULE_INIT = 0,
1167         IPV6_SADDR_RULE_LOCAL,
1168         IPV6_SADDR_RULE_SCOPE,
1169         IPV6_SADDR_RULE_PREFERRED,
1170 #ifdef CONFIG_IPV6_MIP6
1171         IPV6_SADDR_RULE_HOA,
1172 #endif
1173         IPV6_SADDR_RULE_OIF,
1174         IPV6_SADDR_RULE_LABEL,
1175 #ifdef CONFIG_IPV6_PRIVACY
1176         IPV6_SADDR_RULE_PRIVACY,
1177 #endif
1178         IPV6_SADDR_RULE_ORCHID,
1179         IPV6_SADDR_RULE_PREFIX,
1180         IPV6_SADDR_RULE_MAX
1181 };
1182
1183 struct ipv6_saddr_score {
1184         int                     rule;
1185         int                     addr_type;
1186         struct inet6_ifaddr     *ifa;
1187         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1188         int                     scopedist;
1189         int                     matchlen;
1190 };
1191
1192 struct ipv6_saddr_dst {
1193         const struct in6_addr *addr;
1194         int ifindex;
1195         int scope;
1196         int label;
1197         unsigned int prefs;
1198 };
1199
1200 static inline int ipv6_saddr_preferred(int type)
1201 {
1202         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1203                 return 1;
1204         return 0;
1205 }
1206
1207 static int ipv6_get_saddr_eval(struct net *net,
1208                                struct ipv6_saddr_score *score,
1209                                struct ipv6_saddr_dst *dst,
1210                                int i)
1211 {
1212         int ret;
1213
1214         if (i <= score->rule) {
1215                 switch (i) {
1216                 case IPV6_SADDR_RULE_SCOPE:
1217                         ret = score->scopedist;
1218                         break;
1219                 case IPV6_SADDR_RULE_PREFIX:
1220                         ret = score->matchlen;
1221                         break;
1222                 default:
1223                         ret = !!test_bit(i, score->scorebits);
1224                 }
1225                 goto out;
1226         }
1227
1228         switch (i) {
1229         case IPV6_SADDR_RULE_INIT:
1230                 /* Rule 0: remember if hiscore is not ready yet */
1231                 ret = !!score->ifa;
1232                 break;
1233         case IPV6_SADDR_RULE_LOCAL:
1234                 /* Rule 1: Prefer same address */
1235                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1236                 break;
1237         case IPV6_SADDR_RULE_SCOPE:
1238                 /* Rule 2: Prefer appropriate scope
1239                  *
1240                  *      ret
1241                  *       ^
1242                  *    -1 |  d 15
1243                  *    ---+--+-+---> scope
1244                  *       |
1245                  *       |             d is scope of the destination.
1246                  *  B-d  |  \
1247                  *       |   \      <- smaller scope is better if
1248                  *  B-15 |    \        if scope is enough for destinaion.
1249                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1250                  * d-C-1 | /
1251                  *       |/         <- greater is better
1252                  *   -C  /             if scope is not enough for destination.
1253                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1254                  *
1255                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1256                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1257                  * Assume B = 0 and we get C > 29.
1258                  */
1259                 ret = __ipv6_addr_src_scope(score->addr_type);
1260                 if (ret >= dst->scope)
1261                         ret = -ret;
1262                 else
1263                         ret -= 128;     /* 30 is enough */
1264                 score->scopedist = ret;
1265                 break;
1266         case IPV6_SADDR_RULE_PREFERRED:
1267                 /* Rule 3: Avoid deprecated and optimistic addresses */
1268                 ret = ipv6_saddr_preferred(score->addr_type) ||
1269                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1270                 break;
1271 #ifdef CONFIG_IPV6_MIP6
1272         case IPV6_SADDR_RULE_HOA:
1273             {
1274                 /* Rule 4: Prefer home address */
1275                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1276                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1277                 break;
1278             }
1279 #endif
1280         case IPV6_SADDR_RULE_OIF:
1281                 /* Rule 5: Prefer outgoing interface */
1282                 ret = (!dst->ifindex ||
1283                        dst->ifindex == score->ifa->idev->dev->ifindex);
1284                 break;
1285         case IPV6_SADDR_RULE_LABEL:
1286                 /* Rule 6: Prefer matching label */
1287                 ret = ipv6_addr_label(net,
1288                                       &score->ifa->addr, score->addr_type,
1289                                       score->ifa->idev->dev->ifindex) == dst->label;
1290                 break;
1291 #ifdef CONFIG_IPV6_PRIVACY
1292         case IPV6_SADDR_RULE_PRIVACY:
1293             {
1294                 /* Rule 7: Prefer public address
1295                  * Note: prefer temporary address if use_tempaddr >= 2
1296                  */
1297                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1298                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1299                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1300                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1301                 break;
1302             }
1303 #endif
1304         case IPV6_SADDR_RULE_ORCHID:
1305                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1306                  *          non-ORCHID vs non-ORCHID
1307                  */
1308                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1309                         ipv6_addr_orchid(dst->addr));
1310                 break;
1311         case IPV6_SADDR_RULE_PREFIX:
1312                 /* Rule 8: Use longest matching prefix */
1313                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1314                 if (ret > score->ifa->prefix_len)
1315                         ret = score->ifa->prefix_len;
1316                 score->matchlen = ret;
1317                 break;
1318         default:
1319                 ret = 0;
1320         }
1321
1322         if (ret)
1323                 __set_bit(i, score->scorebits);
1324         score->rule = i;
1325 out:
1326         return ret;
1327 }
1328
1329 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1330                        const struct in6_addr *daddr, unsigned int prefs,
1331                        struct in6_addr *saddr)
1332 {
1333         struct ipv6_saddr_score scores[2],
1334                                 *score = &scores[0], *hiscore = &scores[1];
1335         struct ipv6_saddr_dst dst;
1336         struct net_device *dev;
1337         int dst_type;
1338
1339         dst_type = __ipv6_addr_type(daddr);
1340         dst.addr = daddr;
1341         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1342         dst.scope = __ipv6_addr_src_scope(dst_type);
1343         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1344         dst.prefs = prefs;
1345
1346         hiscore->rule = -1;
1347         hiscore->ifa = NULL;
1348
1349         rcu_read_lock();
1350
1351         for_each_netdev_rcu(net, dev) {
1352                 struct inet6_dev *idev;
1353
1354                 /* Candidate Source Address (section 4)
1355                  *  - multicast and link-local destination address,
1356                  *    the set of candidate source address MUST only
1357                  *    include addresses assigned to interfaces
1358                  *    belonging to the same link as the outgoing
1359                  *    interface.
1360                  * (- For site-local destination addresses, the
1361                  *    set of candidate source addresses MUST only
1362                  *    include addresses assigned to interfaces
1363                  *    belonging to the same site as the outgoing
1364                  *    interface.)
1365                  */
1366                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1367                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1368                     dst.ifindex && dev->ifindex != dst.ifindex)
1369                         continue;
1370
1371                 idev = __in6_dev_get(dev);
1372                 if (!idev)
1373                         continue;
1374
1375                 read_lock_bh(&idev->lock);
1376                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1377                         int i;
1378
1379                         /*
1380                          * - Tentative Address (RFC2462 section 5.4)
1381                          *  - A tentative address is not considered
1382                          *    "assigned to an interface" in the traditional
1383                          *    sense, unless it is also flagged as optimistic.
1384                          * - Candidate Source Address (section 4)
1385                          *  - In any case, anycast addresses, multicast
1386                          *    addresses, and the unspecified address MUST
1387                          *    NOT be included in a candidate set.
1388                          */
1389                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1390                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1391                                 continue;
1392
1393                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1394
1395                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1396                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1397                                 LIMIT_NETDEBUG(KERN_DEBUG
1398                                                "ADDRCONF: unspecified / multicast address "
1399                                                "assigned as unicast address on %s",
1400                                                dev->name);
1401                                 continue;
1402                         }
1403
1404                         score->rule = -1;
1405                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1406
1407                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1408                                 int minihiscore, miniscore;
1409
1410                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1411                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1412
1413                                 if (minihiscore > miniscore) {
1414                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1415                                             score->scopedist > 0) {
1416                                                 /*
1417                                                  * special case:
1418                                                  * each remaining entry
1419                                                  * has too small (not enough)
1420                                                  * scope, because ifa entries
1421                                                  * are sorted by their scope
1422                                                  * values.
1423                                                  */
1424                                                 goto try_nextdev;
1425                                         }
1426                                         break;
1427                                 } else if (minihiscore < miniscore) {
1428                                         if (hiscore->ifa)
1429                                                 in6_ifa_put(hiscore->ifa);
1430
1431                                         in6_ifa_hold(score->ifa);
1432
1433                                         swap(hiscore, score);
1434
1435                                         /* restore our iterator */
1436                                         score->ifa = hiscore->ifa;
1437
1438                                         break;
1439                                 }
1440                         }
1441                 }
1442 try_nextdev:
1443                 read_unlock_bh(&idev->lock);
1444         }
1445         rcu_read_unlock();
1446
1447         if (!hiscore->ifa)
1448                 return -EADDRNOTAVAIL;
1449
1450         *saddr = hiscore->ifa->addr;
1451         in6_ifa_put(hiscore->ifa);
1452         return 0;
1453 }
1454 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1455
1456 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1457                       unsigned char banned_flags)
1458 {
1459         struct inet6_ifaddr *ifp;
1460         int err = -EADDRNOTAVAIL;
1461
1462         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1463                 if (ifp->scope == IFA_LINK &&
1464                     !(ifp->flags & banned_flags)) {
1465                         *addr = ifp->addr;
1466                         err = 0;
1467                         break;
1468                 }
1469         }
1470         return err;
1471 }
1472
1473 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1474                     unsigned char banned_flags)
1475 {
1476         struct inet6_dev *idev;
1477         int err = -EADDRNOTAVAIL;
1478
1479         rcu_read_lock();
1480         idev = __in6_dev_get(dev);
1481         if (idev) {
1482                 read_lock_bh(&idev->lock);
1483                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1484                 read_unlock_bh(&idev->lock);
1485         }
1486         rcu_read_unlock();
1487         return err;
1488 }
1489
1490 static int ipv6_count_addresses(struct inet6_dev *idev)
1491 {
1492         int cnt = 0;
1493         struct inet6_ifaddr *ifp;
1494
1495         read_lock_bh(&idev->lock);
1496         list_for_each_entry(ifp, &idev->addr_list, if_list)
1497                 cnt++;
1498         read_unlock_bh(&idev->lock);
1499         return cnt;
1500 }
1501
1502 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1503                   const struct net_device *dev, int strict)
1504 {
1505         struct inet6_ifaddr *ifp;
1506         unsigned int hash = inet6_addr_hash(addr);
1507
1508         rcu_read_lock_bh();
1509         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1510                 if (!net_eq(dev_net(ifp->idev->dev), net))
1511                         continue;
1512                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1513                     !(ifp->flags&IFA_F_TENTATIVE) &&
1514                     (dev == NULL || ifp->idev->dev == dev ||
1515                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1516                         rcu_read_unlock_bh();
1517                         return 1;
1518                 }
1519         }
1520
1521         rcu_read_unlock_bh();
1522         return 0;
1523 }
1524 EXPORT_SYMBOL(ipv6_chk_addr);
1525
1526 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1527                                struct net_device *dev)
1528 {
1529         unsigned int hash = inet6_addr_hash(addr);
1530         struct inet6_ifaddr *ifp;
1531
1532         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1533                 if (!net_eq(dev_net(ifp->idev->dev), net))
1534                         continue;
1535                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1536                         if (dev == NULL || ifp->idev->dev == dev)
1537                                 return true;
1538                 }
1539         }
1540         return false;
1541 }
1542
1543 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1544 {
1545         struct inet6_dev *idev;
1546         struct inet6_ifaddr *ifa;
1547         int     onlink;
1548
1549         onlink = 0;
1550         rcu_read_lock();
1551         idev = __in6_dev_get(dev);
1552         if (idev) {
1553                 read_lock_bh(&idev->lock);
1554                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1555                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1556                                                    ifa->prefix_len);
1557                         if (onlink)
1558                                 break;
1559                 }
1560                 read_unlock_bh(&idev->lock);
1561         }
1562         rcu_read_unlock();
1563         return onlink;
1564 }
1565 EXPORT_SYMBOL(ipv6_chk_prefix);
1566
1567 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1568                                      struct net_device *dev, int strict)
1569 {
1570         struct inet6_ifaddr *ifp, *result = NULL;
1571         unsigned int hash = inet6_addr_hash(addr);
1572
1573         rcu_read_lock_bh();
1574         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1575                 if (!net_eq(dev_net(ifp->idev->dev), net))
1576                         continue;
1577                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1578                         if (dev == NULL || ifp->idev->dev == dev ||
1579                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1580                                 result = ifp;
1581                                 in6_ifa_hold(ifp);
1582                                 break;
1583                         }
1584                 }
1585         }
1586         rcu_read_unlock_bh();
1587
1588         return result;
1589 }
1590
1591 /* Gets referenced address, destroys ifaddr */
1592
1593 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1594 {
1595         if (ifp->flags&IFA_F_PERMANENT) {
1596                 spin_lock_bh(&ifp->lock);
1597                 addrconf_del_dad_timer(ifp);
1598                 ifp->flags |= IFA_F_TENTATIVE;
1599                 if (dad_failed)
1600                         ifp->flags |= IFA_F_DADFAILED;
1601                 spin_unlock_bh(&ifp->lock);
1602                 if (dad_failed)
1603                         ipv6_ifa_notify(0, ifp);
1604                 in6_ifa_put(ifp);
1605 #ifdef CONFIG_IPV6_PRIVACY
1606         } else if (ifp->flags&IFA_F_TEMPORARY) {
1607                 struct inet6_ifaddr *ifpub;
1608                 spin_lock_bh(&ifp->lock);
1609                 ifpub = ifp->ifpub;
1610                 if (ifpub) {
1611                         in6_ifa_hold(ifpub);
1612                         spin_unlock_bh(&ifp->lock);
1613                         ipv6_create_tempaddr(ifpub, ifp);
1614                         in6_ifa_put(ifpub);
1615                 } else {
1616                         spin_unlock_bh(&ifp->lock);
1617                 }
1618                 ipv6_del_addr(ifp);
1619 #endif
1620         } else
1621                 ipv6_del_addr(ifp);
1622 }
1623
1624 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1625 {
1626         int err = -ENOENT;
1627
1628         spin_lock(&ifp->state_lock);
1629         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1630                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1631                 err = 0;
1632         }
1633         spin_unlock(&ifp->state_lock);
1634
1635         return err;
1636 }
1637
1638 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1639 {
1640         struct inet6_dev *idev = ifp->idev;
1641
1642         if (addrconf_dad_end(ifp)) {
1643                 in6_ifa_put(ifp);
1644                 return;
1645         }
1646
1647         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1648                              ifp->idev->dev->name, &ifp->addr);
1649
1650         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1651                 struct in6_addr addr;
1652
1653                 addr.s6_addr32[0] = htonl(0xfe800000);
1654                 addr.s6_addr32[1] = 0;
1655
1656                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1657                     ipv6_addr_equal(&ifp->addr, &addr)) {
1658                         /* DAD failed for link-local based on MAC address */
1659                         idev->cnf.disable_ipv6 = 1;
1660
1661                         pr_info("%s: IPv6 being disabled!\n",
1662                                 ifp->idev->dev->name);
1663                 }
1664         }
1665
1666         addrconf_dad_stop(ifp, 1);
1667 }
1668
1669 /* Join to solicited addr multicast group. */
1670
1671 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1672 {
1673         struct in6_addr maddr;
1674
1675         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1676                 return;
1677
1678         addrconf_addr_solict_mult(addr, &maddr);
1679         ipv6_dev_mc_inc(dev, &maddr);
1680 }
1681
1682 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1683 {
1684         struct in6_addr maddr;
1685
1686         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1687                 return;
1688
1689         addrconf_addr_solict_mult(addr, &maddr);
1690         __ipv6_dev_mc_dec(idev, &maddr);
1691 }
1692
1693 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1694 {
1695         struct in6_addr addr;
1696         if (ifp->prefix_len == 127) /* RFC 6164 */
1697                 return;
1698         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1699         if (ipv6_addr_any(&addr))
1700                 return;
1701         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1702 }
1703
1704 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1705 {
1706         struct in6_addr addr;
1707         if (ifp->prefix_len == 127) /* RFC 6164 */
1708                 return;
1709         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1710         if (ipv6_addr_any(&addr))
1711                 return;
1712         __ipv6_dev_ac_dec(ifp->idev, &addr);
1713 }
1714
1715 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1716 {
1717         if (dev->addr_len != ETH_ALEN)
1718                 return -1;
1719         memcpy(eui, dev->dev_addr, 3);
1720         memcpy(eui + 5, dev->dev_addr + 3, 3);
1721
1722         /*
1723          * The zSeries OSA network cards can be shared among various
1724          * OS instances, but the OSA cards have only one MAC address.
1725          * This leads to duplicate address conflicts in conjunction
1726          * with IPv6 if more than one instance uses the same card.
1727          *
1728          * The driver for these cards can deliver a unique 16-bit
1729          * identifier for each instance sharing the same card.  It is
1730          * placed instead of 0xFFFE in the interface identifier.  The
1731          * "u" bit of the interface identifier is not inverted in this
1732          * case.  Hence the resulting interface identifier has local
1733          * scope according to RFC2373.
1734          */
1735         if (dev->dev_id) {
1736                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1737                 eui[4] = dev->dev_id & 0xFF;
1738         } else {
1739                 eui[3] = 0xFF;
1740                 eui[4] = 0xFE;
1741                 eui[0] ^= 2;
1742         }
1743         return 0;
1744 }
1745
1746 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1747 {
1748         if (dev->addr_len != IEEE802154_ADDR_LEN)
1749                 return -1;
1750         memcpy(eui, dev->dev_addr, 8);
1751         eui[0] ^= 2;
1752         return 0;
1753 }
1754
1755 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1756 {
1757         union fwnet_hwaddr *ha;
1758
1759         if (dev->addr_len != FWNET_ALEN)
1760                 return -1;
1761
1762         ha = (union fwnet_hwaddr *)dev->dev_addr;
1763
1764         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1765         eui[0] ^= 2;
1766         return 0;
1767 }
1768
1769 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1770 {
1771         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1772         if (dev->addr_len != ARCNET_ALEN)
1773                 return -1;
1774         memset(eui, 0, 7);
1775         eui[7] = *(u8 *)dev->dev_addr;
1776         return 0;
1777 }
1778
1779 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1780 {
1781         if (dev->addr_len != INFINIBAND_ALEN)
1782                 return -1;
1783         memcpy(eui, dev->dev_addr + 12, 8);
1784         eui[0] |= 2;
1785         return 0;
1786 }
1787
1788 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1789 {
1790         if (addr == 0)
1791                 return -1;
1792         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1793                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1794                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1795                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1796                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1797                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1798         eui[1] = 0;
1799         eui[2] = 0x5E;
1800         eui[3] = 0xFE;
1801         memcpy(eui + 4, &addr, 4);
1802         return 0;
1803 }
1804
1805 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1806 {
1807         if (dev->priv_flags & IFF_ISATAP)
1808                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1809         return -1;
1810 }
1811
1812 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1813 {
1814         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1815 }
1816
1817 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1818 {
1819         memcpy(eui, dev->perm_addr, 3);
1820         memcpy(eui + 5, dev->perm_addr + 3, 3);
1821         eui[3] = 0xFF;
1822         eui[4] = 0xFE;
1823         eui[0] ^= 2;
1824         return 0;
1825 }
1826
1827 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1828 {
1829         switch (dev->type) {
1830         case ARPHRD_ETHER:
1831         case ARPHRD_FDDI:
1832                 return addrconf_ifid_eui48(eui, dev);
1833         case ARPHRD_ARCNET:
1834                 return addrconf_ifid_arcnet(eui, dev);
1835         case ARPHRD_INFINIBAND:
1836                 return addrconf_ifid_infiniband(eui, dev);
1837         case ARPHRD_SIT:
1838                 return addrconf_ifid_sit(eui, dev);
1839         case ARPHRD_IPGRE:
1840                 return addrconf_ifid_gre(eui, dev);
1841         case ARPHRD_IEEE802154:
1842                 return addrconf_ifid_eui64(eui, dev);
1843         case ARPHRD_IEEE1394:
1844                 return addrconf_ifid_ieee1394(eui, dev);
1845         case ARPHRD_TUNNEL6:
1846                 return addrconf_ifid_ip6tnl(eui, dev);
1847         }
1848         return -1;
1849 }
1850
1851 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1852 {
1853         int err = -1;
1854         struct inet6_ifaddr *ifp;
1855
1856         read_lock_bh(&idev->lock);
1857         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1858                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1859                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1860                         err = 0;
1861                         break;
1862                 }
1863         }
1864         read_unlock_bh(&idev->lock);
1865         return err;
1866 }
1867
1868 #ifdef CONFIG_IPV6_PRIVACY
1869 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1870 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1871 {
1872 regen:
1873         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1874         idev->rndid[0] &= ~0x02;
1875
1876         /*
1877          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1878          * check if generated address is not inappropriate
1879          *
1880          *  - Reserved subnet anycast (RFC 2526)
1881          *      11111101 11....11 1xxxxxxx
1882          *  - ISATAP (RFC4214) 6.1
1883          *      00-00-5E-FE-xx-xx-xx-xx
1884          *  - value 0
1885          *  - XXX: already assigned to an address on the device
1886          */
1887         if (idev->rndid[0] == 0xfd &&
1888             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1889             (idev->rndid[7]&0x80))
1890                 goto regen;
1891         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1892                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1893                         goto regen;
1894                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1895                         goto regen;
1896         }
1897 }
1898
1899 static void ipv6_regen_rndid(unsigned long data)
1900 {
1901         struct inet6_dev *idev = (struct inet6_dev *) data;
1902         unsigned long expires;
1903
1904         rcu_read_lock_bh();
1905         write_lock_bh(&idev->lock);
1906
1907         if (idev->dead)
1908                 goto out;
1909
1910         __ipv6_regen_rndid(idev);
1911
1912         expires = jiffies +
1913                 idev->cnf.temp_prefered_lft * HZ -
1914                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1915                 idev->cnf.max_desync_factor * HZ;
1916         if (time_before(expires, jiffies)) {
1917                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1918                         __func__, idev->dev->name);
1919                 goto out;
1920         }
1921
1922         if (!mod_timer(&idev->regen_timer, expires))
1923                 in6_dev_hold(idev);
1924
1925 out:
1926         write_unlock_bh(&idev->lock);
1927         rcu_read_unlock_bh();
1928         in6_dev_put(idev);
1929 }
1930
1931 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1932 {
1933         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1934                 __ipv6_regen_rndid(idev);
1935 }
1936 #endif
1937
1938 /*
1939  *      Add prefix route.
1940  */
1941
1942 static void
1943 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1944                       unsigned long expires, u32 flags)
1945 {
1946         struct fib6_config cfg = {
1947                 .fc_table = RT6_TABLE_PREFIX,
1948                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1949                 .fc_ifindex = dev->ifindex,
1950                 .fc_expires = expires,
1951                 .fc_dst_len = plen,
1952                 .fc_flags = RTF_UP | flags,
1953                 .fc_nlinfo.nl_net = dev_net(dev),
1954                 .fc_protocol = RTPROT_KERNEL,
1955         };
1956
1957         cfg.fc_dst = *pfx;
1958
1959         /* Prevent useless cloning on PtP SIT.
1960            This thing is done here expecting that the whole
1961            class of non-broadcast devices need not cloning.
1962          */
1963 #if IS_ENABLED(CONFIG_IPV6_SIT)
1964         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1965                 cfg.fc_flags |= RTF_NONEXTHOP;
1966 #endif
1967
1968         ip6_route_add(&cfg);
1969 }
1970
1971
1972 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1973                                                   int plen,
1974                                                   const struct net_device *dev,
1975                                                   u32 flags, u32 noflags)
1976 {
1977         struct fib6_node *fn;
1978         struct rt6_info *rt = NULL;
1979         struct fib6_table *table;
1980
1981         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1982         if (table == NULL)
1983                 return NULL;
1984
1985         read_lock_bh(&table->tb6_lock);
1986         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1987         if (!fn)
1988                 goto out;
1989         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1990                 if (rt->dst.dev->ifindex != dev->ifindex)
1991                         continue;
1992                 if ((rt->rt6i_flags & flags) != flags)
1993                         continue;
1994                 if ((rt->rt6i_flags & noflags) != 0)
1995                         continue;
1996                 dst_hold(&rt->dst);
1997                 break;
1998         }
1999 out:
2000         read_unlock_bh(&table->tb6_lock);
2001         return rt;
2002 }
2003
2004
2005 /* Create "default" multicast route to the interface */
2006
2007 static void addrconf_add_mroute(struct net_device *dev)
2008 {
2009         struct fib6_config cfg = {
2010                 .fc_table = RT6_TABLE_LOCAL,
2011                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2012                 .fc_ifindex = dev->ifindex,
2013                 .fc_dst_len = 8,
2014                 .fc_flags = RTF_UP,
2015                 .fc_nlinfo.nl_net = dev_net(dev),
2016         };
2017
2018         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2019
2020         ip6_route_add(&cfg);
2021 }
2022
2023 #if IS_ENABLED(CONFIG_IPV6_SIT)
2024 static void sit_route_add(struct net_device *dev)
2025 {
2026         struct fib6_config cfg = {
2027                 .fc_table = RT6_TABLE_MAIN,
2028                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2029                 .fc_ifindex = dev->ifindex,
2030                 .fc_dst_len = 96,
2031                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2032                 .fc_nlinfo.nl_net = dev_net(dev),
2033         };
2034
2035         /* prefix length - 96 bits "::d.d.d.d" */
2036         ip6_route_add(&cfg);
2037 }
2038 #endif
2039
2040 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2041 {
2042         struct inet6_dev *idev;
2043
2044         ASSERT_RTNL();
2045
2046         idev = ipv6_find_idev(dev);
2047         if (!idev)
2048                 return ERR_PTR(-ENOBUFS);
2049
2050         if (idev->cnf.disable_ipv6)
2051                 return ERR_PTR(-EACCES);
2052
2053         /* Add default multicast route */
2054         if (!(dev->flags & IFF_LOOPBACK))
2055                 addrconf_add_mroute(dev);
2056
2057         return idev;
2058 }
2059
2060 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2061 {
2062         struct prefix_info *pinfo;
2063         __u32 valid_lft;
2064         __u32 prefered_lft;
2065         int addr_type;
2066         struct inet6_dev *in6_dev;
2067         struct net *net = dev_net(dev);
2068
2069         pinfo = (struct prefix_info *) opt;
2070
2071         if (len < sizeof(struct prefix_info)) {
2072                 ADBG("addrconf: prefix option too short\n");
2073                 return;
2074         }
2075
2076         /*
2077          *      Validation checks ([ADDRCONF], page 19)
2078          */
2079
2080         addr_type = ipv6_addr_type(&pinfo->prefix);
2081
2082         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2083                 return;
2084
2085         valid_lft = ntohl(pinfo->valid);
2086         prefered_lft = ntohl(pinfo->prefered);
2087
2088         if (prefered_lft > valid_lft) {
2089                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2090                 return;
2091         }
2092
2093         in6_dev = in6_dev_get(dev);
2094
2095         if (in6_dev == NULL) {
2096                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2097                                     dev->name);
2098                 return;
2099         }
2100
2101         /*
2102          *      Two things going on here:
2103          *      1) Add routes for on-link prefixes
2104          *      2) Configure prefixes with the auto flag set
2105          */
2106
2107         if (pinfo->onlink) {
2108                 struct rt6_info *rt;
2109                 unsigned long rt_expires;
2110
2111                 /* Avoid arithmetic overflow. Really, we could
2112                  * save rt_expires in seconds, likely valid_lft,
2113                  * but it would require division in fib gc, that it
2114                  * not good.
2115                  */
2116                 if (HZ > USER_HZ)
2117                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2118                 else
2119                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2120
2121                 if (addrconf_finite_timeout(rt_expires))
2122                         rt_expires *= HZ;
2123
2124                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2125                                                pinfo->prefix_len,
2126                                                dev,
2127                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2128                                                RTF_GATEWAY | RTF_DEFAULT);
2129
2130                 if (rt) {
2131                         /* Autoconf prefix route */
2132                         if (valid_lft == 0) {
2133                                 ip6_del_rt(rt);
2134                                 rt = NULL;
2135                         } else if (addrconf_finite_timeout(rt_expires)) {
2136                                 /* not infinity */
2137                                 rt6_set_expires(rt, jiffies + rt_expires);
2138                         } else {
2139                                 rt6_clean_expires(rt);
2140                         }
2141                 } else if (valid_lft) {
2142                         clock_t expires = 0;
2143                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2144                         if (addrconf_finite_timeout(rt_expires)) {
2145                                 /* not infinity */
2146                                 flags |= RTF_EXPIRES;
2147                                 expires = jiffies_to_clock_t(rt_expires);
2148                         }
2149                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2150                                               dev, expires, flags);
2151                 }
2152                 ip6_rt_put(rt);
2153         }
2154
2155         /* Try to figure out our local address for this prefix */
2156
2157         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2158                 struct inet6_ifaddr *ifp;
2159                 struct in6_addr addr;
2160                 int create = 0, update_lft = 0;
2161                 bool tokenized = false;
2162
2163                 if (pinfo->prefix_len == 64) {
2164                         memcpy(&addr, &pinfo->prefix, 8);
2165
2166                         if (!ipv6_addr_any(&in6_dev->token)) {
2167                                 read_lock_bh(&in6_dev->lock);
2168                                 memcpy(addr.s6_addr + 8,
2169                                        in6_dev->token.s6_addr + 8, 8);
2170                                 read_unlock_bh(&in6_dev->lock);
2171                                 tokenized = true;
2172                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2173                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2174                                 in6_dev_put(in6_dev);
2175                                 return;
2176                         }
2177                         goto ok;
2178                 }
2179                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2180                                     pinfo->prefix_len);
2181                 in6_dev_put(in6_dev);
2182                 return;
2183
2184 ok:
2185
2186                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2187
2188                 if (ifp == NULL && valid_lft) {
2189                         int max_addresses = in6_dev->cnf.max_addresses;
2190                         u32 addr_flags = 0;
2191
2192 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2193                         if (in6_dev->cnf.optimistic_dad &&
2194                             !net->ipv6.devconf_all->forwarding && sllao)
2195                                 addr_flags = IFA_F_OPTIMISTIC;
2196 #endif
2197
2198                         /* Do not allow to create too much of autoconfigured
2199                          * addresses; this would be too easy way to crash kernel.
2200                          */
2201                         if (!max_addresses ||
2202                             ipv6_count_addresses(in6_dev) < max_addresses)
2203                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2204                                                     pinfo->prefix_len,
2205                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2206                                                     addr_flags, valid_lft,
2207                                                     prefered_lft);
2208
2209                         if (IS_ERR_OR_NULL(ifp)) {
2210                                 in6_dev_put(in6_dev);
2211                                 return;
2212                         }
2213
2214                         update_lft = 0;
2215                         create = 1;
2216                         ifp->cstamp = jiffies;
2217                         ifp->tokenized = tokenized;
2218                         addrconf_dad_start(ifp);
2219                 }
2220
2221                 if (ifp) {
2222                         int flags;
2223                         unsigned long now;
2224 #ifdef CONFIG_IPV6_PRIVACY
2225                         struct inet6_ifaddr *ift;
2226 #endif
2227                         u32 stored_lft;
2228
2229                         /* update lifetime (RFC2462 5.5.3 e) */
2230                         spin_lock(&ifp->lock);
2231                         now = jiffies;
2232                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2233                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2234                         else
2235                                 stored_lft = 0;
2236                         if (!update_lft && !create && stored_lft) {
2237                                 if (valid_lft > MIN_VALID_LIFETIME ||
2238                                     valid_lft > stored_lft)
2239                                         update_lft = 1;
2240                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2241                                         /* valid_lft <= stored_lft is always true */
2242                                         /*
2243                                          * RFC 4862 Section 5.5.3e:
2244                                          * "Note that the preferred lifetime of
2245                                          *  the corresponding address is always
2246                                          *  reset to the Preferred Lifetime in
2247                                          *  the received Prefix Information
2248                                          *  option, regardless of whether the
2249                                          *  valid lifetime is also reset or
2250                                          *  ignored."
2251                                          *
2252                                          *  So if the preferred lifetime in
2253                                          *  this advertisement is different
2254                                          *  than what we have stored, but the
2255                                          *  valid lifetime is invalid, just
2256                                          *  reset prefered_lft.
2257                                          *
2258                                          *  We must set the valid lifetime
2259                                          *  to the stored lifetime since we'll
2260                                          *  be updating the timestamp below,
2261                                          *  else we'll set it back to the
2262                                          *  minimum.
2263                                          */
2264                                         if (prefered_lft != ifp->prefered_lft) {
2265                                                 valid_lft = stored_lft;
2266                                                 update_lft = 1;
2267                                         }
2268                                 } else {
2269                                         valid_lft = MIN_VALID_LIFETIME;
2270                                         if (valid_lft < prefered_lft)
2271                                                 prefered_lft = valid_lft;
2272                                         update_lft = 1;
2273                                 }
2274                         }
2275
2276                         if (update_lft) {
2277                                 ifp->valid_lft = valid_lft;
2278                                 ifp->prefered_lft = prefered_lft;
2279                                 ifp->tstamp = now;
2280                                 flags = ifp->flags;
2281                                 ifp->flags &= ~IFA_F_DEPRECATED;
2282                                 spin_unlock(&ifp->lock);
2283
2284                                 if (!(flags&IFA_F_TENTATIVE))
2285                                         ipv6_ifa_notify(0, ifp);
2286                         } else
2287                                 spin_unlock(&ifp->lock);
2288
2289 #ifdef CONFIG_IPV6_PRIVACY
2290                         read_lock_bh(&in6_dev->lock);
2291                         /* update all temporary addresses in the list */
2292                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2293                                             tmp_list) {
2294                                 int age, max_valid, max_prefered;
2295
2296                                 if (ifp != ift->ifpub)
2297                                         continue;
2298
2299                                 /*
2300                                  * RFC 4941 section 3.3:
2301                                  * If a received option will extend the lifetime
2302                                  * of a public address, the lifetimes of
2303                                  * temporary addresses should be extended,
2304                                  * subject to the overall constraint that no
2305                                  * temporary addresses should ever remain
2306                                  * "valid" or "preferred" for a time longer than
2307                                  * (TEMP_VALID_LIFETIME) or
2308                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2309                                  * respectively.
2310                                  */
2311                                 age = (now - ift->cstamp) / HZ;
2312                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2313                                 if (max_valid < 0)
2314                                         max_valid = 0;
2315
2316                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2317                                                in6_dev->cnf.max_desync_factor -
2318                                                age;
2319                                 if (max_prefered < 0)
2320                                         max_prefered = 0;
2321
2322                                 if (valid_lft > max_valid)
2323                                         valid_lft = max_valid;
2324
2325                                 if (prefered_lft > max_prefered)
2326                                         prefered_lft = max_prefered;
2327
2328                                 spin_lock(&ift->lock);
2329                                 flags = ift->flags;
2330                                 ift->valid_lft = valid_lft;
2331                                 ift->prefered_lft = prefered_lft;
2332                                 ift->tstamp = now;
2333                                 if (prefered_lft > 0)
2334                                         ift->flags &= ~IFA_F_DEPRECATED;
2335
2336                                 spin_unlock(&ift->lock);
2337                                 if (!(flags&IFA_F_TENTATIVE))
2338                                         ipv6_ifa_notify(0, ift);
2339                         }
2340
2341                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2342                                 /*
2343                                  * When a new public address is created as
2344                                  * described in [ADDRCONF], also create a new
2345                                  * temporary address. Also create a temporary
2346                                  * address if it's enabled but no temporary
2347                                  * address currently exists.
2348                                  */
2349                                 read_unlock_bh(&in6_dev->lock);
2350                                 ipv6_create_tempaddr(ifp, NULL);
2351                         } else {
2352                                 read_unlock_bh(&in6_dev->lock);
2353                         }
2354 #endif
2355                         in6_ifa_put(ifp);
2356                         addrconf_verify(0);
2357                 }
2358         }
2359         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2360         in6_dev_put(in6_dev);
2361 }
2362
2363 /*
2364  *      Set destination address.
2365  *      Special case for SIT interfaces where we create a new "virtual"
2366  *      device.
2367  */
2368 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2369 {
2370         struct in6_ifreq ireq;
2371         struct net_device *dev;
2372         int err = -EINVAL;
2373
2374         rtnl_lock();
2375
2376         err = -EFAULT;
2377         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2378                 goto err_exit;
2379
2380         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2381
2382         err = -ENODEV;
2383         if (dev == NULL)
2384                 goto err_exit;
2385
2386 #if IS_ENABLED(CONFIG_IPV6_SIT)
2387         if (dev->type == ARPHRD_SIT) {
2388                 const struct net_device_ops *ops = dev->netdev_ops;
2389                 struct ifreq ifr;
2390                 struct ip_tunnel_parm p;
2391
2392                 err = -EADDRNOTAVAIL;
2393                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2394                         goto err_exit;
2395
2396                 memset(&p, 0, sizeof(p));
2397                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2398                 p.iph.saddr = 0;
2399                 p.iph.version = 4;
2400                 p.iph.ihl = 5;
2401                 p.iph.protocol = IPPROTO_IPV6;
2402                 p.iph.ttl = 64;
2403                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2404
2405                 if (ops->ndo_do_ioctl) {
2406                         mm_segment_t oldfs = get_fs();
2407
2408                         set_fs(KERNEL_DS);
2409                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2410                         set_fs(oldfs);
2411                 } else
2412                         err = -EOPNOTSUPP;
2413
2414                 if (err == 0) {
2415                         err = -ENOBUFS;
2416                         dev = __dev_get_by_name(net, p.name);
2417                         if (!dev)
2418                                 goto err_exit;
2419                         err = dev_open(dev);
2420                 }
2421         }
2422 #endif
2423
2424 err_exit:
2425         rtnl_unlock();
2426         return err;
2427 }
2428
2429 /*
2430  *      Manual configuration of address on an interface
2431  */
2432 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2433                           const struct in6_addr *peer_pfx,
2434                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2435                           __u32 valid_lft)
2436 {
2437         struct inet6_ifaddr *ifp;
2438         struct inet6_dev *idev;
2439         struct net_device *dev;
2440         int scope;
2441         u32 flags;
2442         clock_t expires;
2443         unsigned long timeout;
2444
2445         ASSERT_RTNL();
2446
2447         if (plen > 128)
2448                 return -EINVAL;
2449
2450         /* check the lifetime */
2451         if (!valid_lft || prefered_lft > valid_lft)
2452                 return -EINVAL;
2453
2454         dev = __dev_get_by_index(net, ifindex);
2455         if (!dev)
2456                 return -ENODEV;
2457
2458         idev = addrconf_add_dev(dev);
2459         if (IS_ERR(idev))
2460                 return PTR_ERR(idev);
2461
2462         scope = ipv6_addr_scope(pfx);
2463
2464         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2465         if (addrconf_finite_timeout(timeout)) {
2466                 expires = jiffies_to_clock_t(timeout * HZ);
2467                 valid_lft = timeout;
2468                 flags = RTF_EXPIRES;
2469         } else {
2470                 expires = 0;
2471                 flags = 0;
2472                 ifa_flags |= IFA_F_PERMANENT;
2473         }
2474
2475         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2476         if (addrconf_finite_timeout(timeout)) {
2477                 if (timeout == 0)
2478                         ifa_flags |= IFA_F_DEPRECATED;
2479                 prefered_lft = timeout;
2480         }
2481
2482         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2483                             valid_lft, prefered_lft);
2484
2485         if (!IS_ERR(ifp)) {
2486                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2487                                       expires, flags);
2488                 /*
2489                  * Note that section 3.1 of RFC 4429 indicates
2490                  * that the Optimistic flag should not be set for
2491                  * manually configured addresses
2492                  */
2493                 addrconf_dad_start(ifp);
2494                 in6_ifa_put(ifp);
2495                 addrconf_verify(0);
2496                 return 0;
2497         }
2498
2499         return PTR_ERR(ifp);
2500 }
2501
2502 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2503                           unsigned int plen)
2504 {
2505         struct inet6_ifaddr *ifp;
2506         struct inet6_dev *idev;
2507         struct net_device *dev;
2508
2509         if (plen > 128)
2510                 return -EINVAL;
2511
2512         dev = __dev_get_by_index(net, ifindex);
2513         if (!dev)
2514                 return -ENODEV;
2515
2516         if ((idev = __in6_dev_get(dev)) == NULL)
2517                 return -ENXIO;
2518
2519         read_lock_bh(&idev->lock);
2520         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2521                 if (ifp->prefix_len == plen &&
2522                     ipv6_addr_equal(pfx, &ifp->addr)) {
2523                         in6_ifa_hold(ifp);
2524                         read_unlock_bh(&idev->lock);
2525
2526                         ipv6_del_addr(ifp);
2527                         return 0;
2528                 }
2529         }
2530         read_unlock_bh(&idev->lock);
2531         return -EADDRNOTAVAIL;
2532 }
2533
2534
2535 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2536 {
2537         struct in6_ifreq ireq;
2538         int err;
2539
2540         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2541                 return -EPERM;
2542
2543         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2544                 return -EFAULT;
2545
2546         rtnl_lock();
2547         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2548                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2549                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2550         rtnl_unlock();
2551         return err;
2552 }
2553
2554 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2555 {
2556         struct in6_ifreq ireq;
2557         int err;
2558
2559         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2560                 return -EPERM;
2561
2562         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2563                 return -EFAULT;
2564
2565         rtnl_lock();
2566         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2567                              ireq.ifr6_prefixlen);
2568         rtnl_unlock();
2569         return err;
2570 }
2571
2572 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2573                      int plen, int scope)
2574 {
2575         struct inet6_ifaddr *ifp;
2576
2577         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2578                             scope, IFA_F_PERMANENT, 0, 0);
2579         if (!IS_ERR(ifp)) {
2580                 spin_lock_bh(&ifp->lock);
2581                 ifp->flags &= ~IFA_F_TENTATIVE;
2582                 spin_unlock_bh(&ifp->lock);
2583                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2584                 in6_ifa_put(ifp);
2585         }
2586 }
2587
2588 #if IS_ENABLED(CONFIG_IPV6_SIT)
2589 static void sit_add_v4_addrs(struct inet6_dev *idev)
2590 {
2591         struct in6_addr addr;
2592         struct net_device *dev;
2593         struct net *net = dev_net(idev->dev);
2594         int scope;
2595
2596         ASSERT_RTNL();
2597
2598         memset(&addr, 0, sizeof(struct in6_addr));
2599         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2600
2601         if (idev->dev->flags&IFF_POINTOPOINT) {
2602                 addr.s6_addr32[0] = htonl(0xfe800000);
2603                 scope = IFA_LINK;
2604         } else {
2605                 scope = IPV6_ADDR_COMPATv4;
2606         }
2607
2608         if (addr.s6_addr32[3]) {
2609                 add_addr(idev, &addr, 128, scope);
2610                 return;
2611         }
2612
2613         for_each_netdev(net, dev) {
2614                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2615                 if (in_dev && (dev->flags & IFF_UP)) {
2616                         struct in_ifaddr *ifa;
2617
2618                         int flag = scope;
2619
2620                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2621                                 int plen;
2622
2623                                 addr.s6_addr32[3] = ifa->ifa_local;
2624
2625                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2626                                         continue;
2627                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2628                                         if (idev->dev->flags&IFF_POINTOPOINT)
2629                                                 continue;
2630                                         flag |= IFA_HOST;
2631                                 }
2632                                 if (idev->dev->flags&IFF_POINTOPOINT)
2633                                         plen = 64;
2634                                 else
2635                                         plen = 96;
2636
2637                                 add_addr(idev, &addr, plen, flag);
2638                         }
2639                 }
2640         }
2641 }
2642 #endif
2643
2644 static void init_loopback(struct net_device *dev)
2645 {
2646         struct inet6_dev  *idev;
2647         struct net_device *sp_dev;
2648         struct inet6_ifaddr *sp_ifa;
2649         struct rt6_info *sp_rt;
2650
2651         /* ::1 */
2652
2653         ASSERT_RTNL();
2654
2655         if ((idev = ipv6_find_idev(dev)) == NULL) {
2656                 pr_debug("%s: add_dev failed\n", __func__);
2657                 return;
2658         }
2659
2660         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2661
2662         /* Add routes to other interface's IPv6 addresses */
2663         for_each_netdev(dev_net(dev), sp_dev) {
2664                 if (!strcmp(sp_dev->name, dev->name))
2665                         continue;
2666
2667                 idev = __in6_dev_get(sp_dev);
2668                 if (!idev)
2669                         continue;
2670
2671                 read_lock_bh(&idev->lock);
2672                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2673
2674                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2675                                 continue;
2676
2677                         if (sp_ifa->rt)
2678                                 continue;
2679
2680                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2681
2682                         /* Failure cases are ignored */
2683                         if (!IS_ERR(sp_rt)) {
2684                                 sp_ifa->rt = sp_rt;
2685                                 ip6_ins_rt(sp_rt);
2686                         }
2687                 }
2688                 read_unlock_bh(&idev->lock);
2689         }
2690 }
2691
2692 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2693 {
2694         struct inet6_ifaddr *ifp;
2695         u32 addr_flags = IFA_F_PERMANENT;
2696
2697 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2698         if (idev->cnf.optimistic_dad &&
2699             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2700                 addr_flags |= IFA_F_OPTIMISTIC;
2701 #endif
2702
2703
2704         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2705         if (!IS_ERR(ifp)) {
2706                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2707                 addrconf_dad_start(ifp);
2708                 in6_ifa_put(ifp);
2709         }
2710 }
2711
2712 static void addrconf_dev_config(struct net_device *dev)
2713 {
2714         struct in6_addr addr;
2715         struct inet6_dev *idev;
2716
2717         ASSERT_RTNL();
2718
2719         if ((dev->type != ARPHRD_ETHER) &&
2720             (dev->type != ARPHRD_FDDI) &&
2721             (dev->type != ARPHRD_ARCNET) &&
2722             (dev->type != ARPHRD_INFINIBAND) &&
2723             (dev->type != ARPHRD_IEEE802154) &&
2724             (dev->type != ARPHRD_IEEE1394) &&
2725             (dev->type != ARPHRD_TUNNEL6)) {
2726                 /* Alas, we support only Ethernet autoconfiguration. */
2727                 return;
2728         }
2729
2730         idev = addrconf_add_dev(dev);
2731         if (IS_ERR(idev))
2732                 return;
2733
2734         memset(&addr, 0, sizeof(struct in6_addr));
2735         addr.s6_addr32[0] = htonl(0xFE800000);
2736
2737         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2738                 addrconf_add_linklocal(idev, &addr);
2739 }
2740
2741 #if IS_ENABLED(CONFIG_IPV6_SIT)
2742 static void addrconf_sit_config(struct net_device *dev)
2743 {
2744         struct inet6_dev *idev;
2745
2746         ASSERT_RTNL();
2747
2748         /*
2749          * Configure the tunnel with one of our IPv4
2750          * addresses... we should configure all of
2751          * our v4 addrs in the tunnel
2752          */
2753
2754         if ((idev = ipv6_find_idev(dev)) == NULL) {
2755                 pr_debug("%s: add_dev failed\n", __func__);
2756                 return;
2757         }
2758
2759         if (dev->priv_flags & IFF_ISATAP) {
2760                 struct in6_addr addr;
2761
2762                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2763                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2764                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2765                         addrconf_add_linklocal(idev, &addr);
2766                 return;
2767         }
2768
2769         sit_add_v4_addrs(idev);
2770
2771         if (dev->flags&IFF_POINTOPOINT)
2772                 addrconf_add_mroute(dev);
2773         else
2774                 sit_route_add(dev);
2775 }
2776 #endif
2777
2778 #if IS_ENABLED(CONFIG_NET_IPGRE)
2779 static void addrconf_gre_config(struct net_device *dev)
2780 {
2781         struct inet6_dev *idev;
2782         struct in6_addr addr;
2783
2784         ASSERT_RTNL();
2785
2786         if ((idev = ipv6_find_idev(dev)) == NULL) {
2787                 pr_debug("%s: add_dev failed\n", __func__);
2788                 return;
2789         }
2790
2791         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2792         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2793
2794         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2795                 addrconf_add_linklocal(idev, &addr);
2796 }
2797 #endif
2798
2799 static inline int
2800 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2801 {
2802         struct in6_addr lladdr;
2803
2804         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2805                 addrconf_add_linklocal(idev, &lladdr);
2806                 return 0;
2807         }
2808         return -1;
2809 }
2810
2811 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2812                            void *ptr)
2813 {
2814         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2815         struct inet6_dev *idev = __in6_dev_get(dev);
2816         int run_pending = 0;
2817         int err;
2818
2819         switch (event) {
2820         case NETDEV_REGISTER:
2821                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2822                         idev = ipv6_add_dev(dev);
2823                         if (!idev)
2824                                 return notifier_from_errno(-ENOMEM);
2825                 }
2826                 break;
2827
2828         case NETDEV_UP:
2829         case NETDEV_CHANGE:
2830                 if (dev->flags & IFF_SLAVE)
2831                         break;
2832
2833                 if (event == NETDEV_UP) {
2834                         if (!addrconf_qdisc_ok(dev)) {
2835                                 /* device is not ready yet. */
2836                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2837                                         dev->name);
2838                                 break;
2839                         }
2840
2841                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2842                                 idev = ipv6_add_dev(dev);
2843
2844                         if (idev) {
2845                                 idev->if_flags |= IF_READY;
2846                                 run_pending = 1;
2847                         }
2848                 } else {
2849                         if (!addrconf_qdisc_ok(dev)) {
2850                                 /* device is still not ready. */
2851                                 break;
2852                         }
2853
2854                         if (idev) {
2855                                 if (idev->if_flags & IF_READY)
2856                                         /* device is already configured. */
2857                                         break;
2858                                 idev->if_flags |= IF_READY;
2859                         }
2860
2861                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2862                                 dev->name);
2863
2864                         run_pending = 1;
2865                 }
2866
2867                 switch (dev->type) {
2868 #if IS_ENABLED(CONFIG_IPV6_SIT)
2869                 case ARPHRD_SIT:
2870                         addrconf_sit_config(dev);
2871                         break;
2872 #endif
2873 #if IS_ENABLED(CONFIG_NET_IPGRE)
2874                 case ARPHRD_IPGRE:
2875                         addrconf_gre_config(dev);
2876                         break;
2877 #endif
2878                 case ARPHRD_LOOPBACK:
2879                         init_loopback(dev);
2880                         break;
2881
2882                 default:
2883                         addrconf_dev_config(dev);
2884                         break;
2885                 }
2886
2887                 if (idev) {
2888                         if (run_pending)
2889                                 addrconf_dad_run(idev);
2890
2891                         /*
2892                          * If the MTU changed during the interface down,
2893                          * when the interface up, the changed MTU must be
2894                          * reflected in the idev as well as routers.
2895                          */
2896                         if (idev->cnf.mtu6 != dev->mtu &&
2897                             dev->mtu >= IPV6_MIN_MTU) {
2898                                 rt6_mtu_change(dev, dev->mtu);
2899                                 idev->cnf.mtu6 = dev->mtu;
2900                         }
2901                         idev->tstamp = jiffies;
2902                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2903
2904                         /*
2905                          * If the changed mtu during down is lower than
2906                          * IPV6_MIN_MTU stop IPv6 on this interface.
2907                          */
2908                         if (dev->mtu < IPV6_MIN_MTU)
2909                                 addrconf_ifdown(dev, 1);
2910                 }
2911                 break;
2912
2913         case NETDEV_CHANGEMTU:
2914                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2915                         rt6_mtu_change(dev, dev->mtu);
2916                         idev->cnf.mtu6 = dev->mtu;
2917                         break;
2918                 }
2919
2920                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2921                         idev = ipv6_add_dev(dev);
2922                         if (idev)
2923                                 break;
2924                 }
2925
2926                 /*
2927                  * MTU falled under IPV6_MIN_MTU.
2928                  * Stop IPv6 on this interface.
2929                  */
2930
2931         case NETDEV_DOWN:
2932         case NETDEV_UNREGISTER:
2933                 /*
2934                  *      Remove all addresses from this interface.
2935                  */
2936                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2937                 break;
2938
2939         case NETDEV_CHANGENAME:
2940                 if (idev) {
2941                         snmp6_unregister_dev(idev);
2942                         addrconf_sysctl_unregister(idev);
2943                         addrconf_sysctl_register(idev);
2944                         err = snmp6_register_dev(idev);
2945                         if (err)
2946                                 return notifier_from_errno(err);
2947                 }
2948                 break;
2949
2950         case NETDEV_PRE_TYPE_CHANGE:
2951         case NETDEV_POST_TYPE_CHANGE:
2952                 addrconf_type_change(dev, event);
2953                 break;
2954         }
2955
2956         return NOTIFY_OK;
2957 }
2958
2959 /*
2960  *      addrconf module should be notified of a device going up
2961  */
2962 static struct notifier_block ipv6_dev_notf = {
2963         .notifier_call = addrconf_notify,
2964 };
2965
2966 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2967 {
2968         struct inet6_dev *idev;
2969         ASSERT_RTNL();
2970
2971         idev = __in6_dev_get(dev);
2972
2973         if (event == NETDEV_POST_TYPE_CHANGE)
2974                 ipv6_mc_remap(idev);
2975         else if (event == NETDEV_PRE_TYPE_CHANGE)
2976                 ipv6_mc_unmap(idev);
2977 }
2978
2979 static int addrconf_ifdown(struct net_device *dev, int how)
2980 {
2981         struct net *net = dev_net(dev);
2982         struct inet6_dev *idev;
2983         struct inet6_ifaddr *ifa;
2984         int state, i;
2985
2986         ASSERT_RTNL();
2987
2988         rt6_ifdown(net, dev);
2989         neigh_ifdown(&nd_tbl, dev);
2990
2991         idev = __in6_dev_get(dev);
2992         if (idev == NULL)
2993                 return -ENODEV;
2994
2995         /*
2996          * Step 1: remove reference to ipv6 device from parent device.
2997          *         Do not dev_put!
2998          */
2999         if (how) {
3000                 idev->dead = 1;
3001
3002                 /* protected by rtnl_lock */
3003                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3004
3005                 /* Step 1.5: remove snmp6 entry */
3006                 snmp6_unregister_dev(idev);
3007
3008         }
3009
3010         /* Step 2: clear hash table */
3011         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3012                 struct hlist_head *h = &inet6_addr_lst[i];
3013
3014                 spin_lock_bh(&addrconf_hash_lock);
3015         restart:
3016                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3017                         if (ifa->idev == idev) {
3018                                 hlist_del_init_rcu(&ifa->addr_lst);
3019                                 addrconf_del_dad_timer(ifa);
3020                                 goto restart;
3021                         }
3022                 }
3023                 spin_unlock_bh(&addrconf_hash_lock);
3024         }
3025
3026         write_lock_bh(&idev->lock);
3027
3028         addrconf_del_rs_timer(idev);
3029
3030         /* Step 2: clear flags for stateless addrconf */
3031         if (!how)
3032                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3033
3034 #ifdef CONFIG_IPV6_PRIVACY
3035         if (how && del_timer(&idev->regen_timer))
3036                 in6_dev_put(idev);
3037
3038         /* Step 3: clear tempaddr list */
3039         while (!list_empty(&idev->tempaddr_list)) {
3040                 ifa = list_first_entry(&idev->tempaddr_list,
3041                                        struct inet6_ifaddr, tmp_list);
3042                 list_del(&ifa->tmp_list);
3043                 write_unlock_bh(&idev->lock);
3044                 spin_lock_bh(&ifa->lock);
3045
3046                 if (ifa->ifpub) {
3047                         in6_ifa_put(ifa->ifpub);
3048                         ifa->ifpub = NULL;
3049                 }
3050                 spin_unlock_bh(&ifa->lock);
3051                 in6_ifa_put(ifa);
3052                 write_lock_bh(&idev->lock);
3053         }
3054 #endif
3055
3056         while (!list_empty(&idev->addr_list)) {
3057                 ifa = list_first_entry(&idev->addr_list,
3058                                        struct inet6_ifaddr, if_list);
3059                 addrconf_del_dad_timer(ifa);
3060
3061                 list_del(&ifa->if_list);
3062
3063                 write_unlock_bh(&idev->lock);
3064
3065                 spin_lock_bh(&ifa->state_lock);
3066                 state = ifa->state;
3067                 ifa->state = INET6_IFADDR_STATE_DEAD;
3068                 spin_unlock_bh(&ifa->state_lock);
3069
3070                 if (state != INET6_IFADDR_STATE_DEAD) {
3071                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3072                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3073                 }
3074                 in6_ifa_put(ifa);
3075
3076                 write_lock_bh(&idev->lock);
3077         }
3078
3079         write_unlock_bh(&idev->lock);
3080
3081         /* Step 5: Discard multicast list */
3082         if (how)
3083                 ipv6_mc_destroy_dev(idev);
3084         else
3085                 ipv6_mc_down(idev);
3086
3087         idev->tstamp = jiffies;
3088
3089         /* Last: Shot the device (if unregistered) */
3090         if (how) {
3091                 addrconf_sysctl_unregister(idev);
3092                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3093                 neigh_ifdown(&nd_tbl, dev);
3094                 in6_dev_put(idev);
3095         }
3096         return 0;
3097 }
3098
3099 static void addrconf_rs_timer(unsigned long data)
3100 {
3101         struct inet6_dev *idev = (struct inet6_dev *)data;
3102         struct in6_addr lladdr;
3103
3104         write_lock(&idev->lock);
3105         if (idev->dead || !(idev->if_flags & IF_READY))
3106                 goto out;
3107
3108         if (!ipv6_accept_ra(idev))
3109                 goto out;
3110
3111         /* Announcement received after solicitation was sent */
3112         if (idev->if_flags & IF_RA_RCVD)
3113                 goto out;
3114
3115         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3116                 if (!__ipv6_get_lladdr(idev, &lladdr, IFA_F_TENTATIVE))
3117                         ndisc_send_rs(idev->dev, &lladdr,
3118                                       &in6addr_linklocal_allrouters);
3119                 else
3120                         goto out;
3121
3122                 /* The wait after the last probe can be shorter */
3123                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3124                                              idev->cnf.rtr_solicits) ?
3125                                       idev->cnf.rtr_solicit_delay :
3126                                       idev->cnf.rtr_solicit_interval);
3127         } else {
3128                 /*
3129                  * Note: we do not support deprecated "all on-link"
3130                  * assumption any longer.
3131                  */
3132                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3133         }
3134
3135 out:
3136         write_unlock(&idev->lock);
3137         in6_dev_put(idev);
3138 }
3139
3140 /*
3141  *      Duplicate Address Detection
3142  */
3143 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3144 {
3145         unsigned long rand_num;
3146         struct inet6_dev *idev = ifp->idev;
3147
3148         if (ifp->flags & IFA_F_OPTIMISTIC)
3149                 rand_num = 0;
3150         else
3151                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3152
3153         ifp->dad_probes = idev->cnf.dad_transmits;
3154         addrconf_mod_dad_timer(ifp, rand_num);
3155 }
3156
3157 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3158 {
3159         struct inet6_dev *idev = ifp->idev;
3160         struct net_device *dev = idev->dev;
3161
3162         addrconf_join_solict(dev, &ifp->addr);
3163
3164         net_srandom(ifp->addr.s6_addr32[3]);
3165
3166         read_lock_bh(&idev->lock);
3167         spin_lock(&ifp->lock);
3168         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3169                 goto out;
3170
3171         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3172             idev->cnf.accept_dad < 1 ||
3173             !(ifp->flags&IFA_F_TENTATIVE) ||
3174             ifp->flags & IFA_F_NODAD) {
3175                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3176                 spin_unlock(&ifp->lock);
3177                 read_unlock_bh(&idev->lock);
3178
3179                 addrconf_dad_completed(ifp);
3180                 return;
3181         }
3182
3183         if (!(idev->if_flags & IF_READY)) {
3184                 spin_unlock(&ifp->lock);
3185                 read_unlock_bh(&idev->lock);
3186                 /*
3187                  * If the device is not ready:
3188                  * - keep it tentative if it is a permanent address.
3189                  * - otherwise, kill it.
3190                  */
3191                 in6_ifa_hold(ifp);
3192                 addrconf_dad_stop(ifp, 0);
3193                 return;
3194         }
3195
3196         /*
3197          * Optimistic nodes can start receiving
3198          * Frames right away
3199          */
3200         if (ifp->flags & IFA_F_OPTIMISTIC)
3201                 ip6_ins_rt(ifp->rt);
3202
3203         addrconf_dad_kick(ifp);
3204 out:
3205         spin_unlock(&ifp->lock);
3206         read_unlock_bh(&idev->lock);
3207 }
3208
3209 static void addrconf_dad_timer(unsigned long data)
3210 {
3211         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3212         struct inet6_dev *idev = ifp->idev;
3213         struct in6_addr mcaddr;
3214
3215         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3216                 goto out;
3217
3218         write_lock(&idev->lock);
3219         if (idev->dead || !(idev->if_flags & IF_READY)) {
3220                 write_unlock(&idev->lock);
3221                 goto out;
3222         }
3223
3224         spin_lock(&ifp->lock);
3225         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3226                 spin_unlock(&ifp->lock);
3227                 write_unlock(&idev->lock);
3228                 goto out;
3229         }
3230
3231         if (ifp->dad_probes == 0) {
3232                 /*
3233                  * DAD was successful
3234                  */
3235
3236                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3237                 spin_unlock(&ifp->lock);
3238                 write_unlock(&idev->lock);
3239
3240                 addrconf_dad_completed(ifp);
3241
3242                 goto out;
3243         }
3244
3245         ifp->dad_probes--;
3246         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3247         spin_unlock(&ifp->lock);
3248         write_unlock(&idev->lock);
3249
3250         /* send a neighbour solicitation for our addr */
3251         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3252         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3253 out:
3254         in6_ifa_put(ifp);
3255 }
3256
3257 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3258 {
3259         struct net_device *dev = ifp->idev->dev;
3260         struct in6_addr lladdr;
3261         bool send_rs, send_mld;
3262
3263         addrconf_del_dad_timer(ifp);
3264
3265         /*
3266          *      Configure the address for reception. Now it is valid.
3267          */
3268
3269         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3270
3271         /* If added prefix is link local and we are prepared to process
3272            router advertisements, start sending router solicitations.
3273          */
3274
3275         read_lock_bh(&ifp->idev->lock);
3276         spin_lock(&ifp->lock);
3277         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3278                    ifp->idev->valid_ll_addr_cnt == 1;
3279         send_rs = send_mld &&
3280                   ipv6_accept_ra(ifp->idev) &&
3281                   ifp->idev->cnf.rtr_solicits > 0 &&
3282                   (dev->flags&IFF_LOOPBACK) == 0;
3283         spin_unlock(&ifp->lock);
3284         read_unlock_bh(&ifp->idev->lock);
3285
3286         /* While dad is in progress mld report's source address is in6_addrany.
3287          * Resend with proper ll now.
3288          */
3289         if (send_mld)
3290                 ipv6_mc_dad_complete(ifp->idev);
3291
3292         if (send_rs) {
3293                 /*
3294                  *      If a host as already performed a random delay
3295                  *      [...] as part of DAD [...] there is no need
3296                  *      to delay again before sending the first RS
3297                  */
3298                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3299                         return;
3300                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3301
3302                 write_lock_bh(&ifp->idev->lock);
3303                 spin_lock(&ifp->lock);
3304                 ifp->idev->rs_probes = 1;
3305                 ifp->idev->if_flags |= IF_RS_SENT;
3306                 addrconf_mod_rs_timer(ifp->idev,
3307                                       ifp->idev->cnf.rtr_solicit_interval);
3308                 spin_unlock(&ifp->lock);
3309                 write_unlock_bh(&ifp->idev->lock);
3310         }
3311 }
3312
3313 static void addrconf_dad_run(struct inet6_dev *idev)
3314 {
3315         struct inet6_ifaddr *ifp;
3316
3317         read_lock_bh(&idev->lock);
3318         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3319                 spin_lock(&ifp->lock);
3320                 if (ifp->flags & IFA_F_TENTATIVE &&
3321                     ifp->state == INET6_IFADDR_STATE_DAD)
3322                         addrconf_dad_kick(ifp);
3323                 spin_unlock(&ifp->lock);
3324         }
3325         read_unlock_bh(&idev->lock);
3326 }
3327
3328 #ifdef CONFIG_PROC_FS
3329 struct if6_iter_state {
3330         struct seq_net_private p;
3331         int bucket;
3332         int offset;
3333 };
3334
3335 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3336 {
3337         struct inet6_ifaddr *ifa = NULL;
3338         struct if6_iter_state *state = seq->private;
3339         struct net *net = seq_file_net(seq);
3340         int p = 0;
3341
3342         /* initial bucket if pos is 0 */
3343         if (pos == 0) {
3344                 state->bucket = 0;
3345                 state->offset = 0;
3346         }
3347
3348         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3349                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3350                                          addr_lst) {
3351                         if (!net_eq(dev_net(ifa->idev->dev), net))
3352                                 continue;
3353                         /* sync with offset */
3354                         if (p < state->offset) {
3355                                 p++;
3356                                 continue;
3357                         }
3358                         state->offset++;
3359                         return ifa;
3360                 }
3361
3362                 /* prepare for next bucket */
3363                 state->offset = 0;
3364                 p = 0;
3365         }
3366         return NULL;
3367 }
3368
3369 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3370                                          struct inet6_ifaddr *ifa)
3371 {
3372         struct if6_iter_state *state = seq->private;
3373         struct net *net = seq_file_net(seq);
3374
3375         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3376                 if (!net_eq(dev_net(ifa->idev->dev), net))
3377                         continue;
3378                 state->offset++;
3379                 return ifa;
3380         }
3381
3382         while (++state->bucket < IN6_ADDR_HSIZE) {
3383                 state->offset = 0;
3384                 hlist_for_each_entry_rcu_bh(ifa,
3385                                      &inet6_addr_lst[state->bucket], addr_lst) {
3386                         if (!net_eq(dev_net(ifa->idev->dev), net))
3387                                 continue;
3388                         state->offset++;
3389                         return ifa;
3390                 }
3391         }
3392
3393         return NULL;
3394 }
3395
3396 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3397         __acquires(rcu_bh)
3398 {
3399         rcu_read_lock_bh();
3400         return if6_get_first(seq, *pos);
3401 }
3402
3403 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3404 {
3405         struct inet6_ifaddr *ifa;
3406
3407         ifa = if6_get_next(seq, v);
3408         ++*pos;
3409         return ifa;
3410 }
3411
3412 static void if6_seq_stop(struct seq_file *seq, void *v)
3413         __releases(rcu_bh)
3414 {
3415         rcu_read_unlock_bh();
3416 }
3417
3418 static int if6_seq_show(struct seq_file *seq, void *v)
3419 {
3420         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3421         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3422                    &ifp->addr,
3423                    ifp->idev->dev->ifindex,
3424                    ifp->prefix_len,
3425                    ifp->scope,
3426                    ifp->flags,
3427                    ifp->idev->dev->name);
3428         return 0;
3429 }
3430
3431 static const struct seq_operations if6_seq_ops = {
3432         .start  = if6_seq_start,
3433         .next   = if6_seq_next,
3434         .show   = if6_seq_show,
3435         .stop   = if6_seq_stop,
3436 };
3437
3438 static int if6_seq_open(struct inode *inode, struct file *file)
3439 {
3440         return seq_open_net(inode, file, &if6_seq_ops,
3441                             sizeof(struct if6_iter_state));
3442 }
3443
3444 static const struct file_operations if6_fops = {
3445         .owner          = THIS_MODULE,
3446         .open           = if6_seq_open,
3447         .read           = seq_read,
3448         .llseek         = seq_lseek,
3449         .release        = seq_release_net,
3450 };
3451
3452 static int __net_init if6_proc_net_init(struct net *net)
3453 {
3454         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3455                 return -ENOMEM;
3456         return 0;
3457 }
3458
3459 static void __net_exit if6_proc_net_exit(struct net *net)
3460 {
3461         remove_proc_entry("if_inet6", net->proc_net);
3462 }
3463
3464 static struct pernet_operations if6_proc_net_ops = {
3465        .init = if6_proc_net_init,
3466        .exit = if6_proc_net_exit,
3467 };
3468
3469 int __init if6_proc_init(void)
3470 {
3471         return register_pernet_subsys(&if6_proc_net_ops);
3472 }
3473
3474 void if6_proc_exit(void)
3475 {
3476         unregister_pernet_subsys(&if6_proc_net_ops);
3477 }
3478 #endif  /* CONFIG_PROC_FS */
3479
3480 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3481 /* Check if address is a home address configured on any interface. */
3482 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3483 {
3484         int ret = 0;
3485         struct inet6_ifaddr *ifp = NULL;
3486         unsigned int hash = inet6_addr_hash(addr);
3487
3488         rcu_read_lock_bh();
3489         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3490                 if (!net_eq(dev_net(ifp->idev->dev), net))
3491                         continue;
3492                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3493                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3494                         ret = 1;
3495                         break;
3496                 }
3497         }
3498         rcu_read_unlock_bh();
3499         return ret;
3500 }
3501 #endif
3502
3503 /*
3504  *      Periodic address status verification
3505  */
3506
3507 static void addrconf_verify(unsigned long foo)
3508 {
3509         unsigned long now, next, next_sec, next_sched;
3510         struct inet6_ifaddr *ifp;
3511         int i;
3512
3513         rcu_read_lock_bh();
3514         spin_lock(&addrconf_verify_lock);
3515         now = jiffies;
3516         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3517
3518         del_timer(&addr_chk_timer);
3519
3520         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3521 restart:
3522                 hlist_for_each_entry_rcu_bh(ifp,
3523                                          &inet6_addr_lst[i], addr_lst) {
3524                         unsigned long age;
3525
3526                         if (ifp->flags & IFA_F_PERMANENT)
3527                                 continue;
3528
3529                         spin_lock(&ifp->lock);
3530                         /* We try to batch several events at once. */
3531                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3532
3533                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3534                             age >= ifp->valid_lft) {
3535                                 spin_unlock(&ifp->lock);
3536                                 in6_ifa_hold(ifp);
3537                                 ipv6_del_addr(ifp);
3538                                 goto restart;
3539                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3540                                 spin_unlock(&ifp->lock);
3541                                 continue;
3542                         } else if (age >= ifp->prefered_lft) {
3543                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3544                                 int deprecate = 0;
3545
3546                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3547                                         deprecate = 1;
3548                                         ifp->flags |= IFA_F_DEPRECATED;
3549                                 }
3550
3551                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3552                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3553
3554                                 spin_unlock(&ifp->lock);
3555
3556                                 if (deprecate) {
3557                                         in6_ifa_hold(ifp);
3558
3559                                         ipv6_ifa_notify(0, ifp);
3560                                         in6_ifa_put(ifp);
3561                                         goto restart;
3562                                 }
3563 #ifdef CONFIG_IPV6_PRIVACY
3564                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3565                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3566                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3567                                         ifp->idev->cnf.dad_transmits *
3568                                         ifp->idev->nd_parms->retrans_time / HZ;
3569
3570                                 if (age >= ifp->prefered_lft - regen_advance) {
3571                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3572                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3573                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3574                                         if (!ifp->regen_count && ifpub) {
3575                                                 ifp->regen_count++;
3576                                                 in6_ifa_hold(ifp);
3577                                                 in6_ifa_hold(ifpub);
3578                                                 spin_unlock(&ifp->lock);
3579
3580                                                 spin_lock(&ifpub->lock);
3581                                                 ifpub->regen_count = 0;
3582                                                 spin_unlock(&ifpub->lock);
3583                                                 ipv6_create_tempaddr(ifpub, ifp);
3584                                                 in6_ifa_put(ifpub);
3585                                                 in6_ifa_put(ifp);
3586                                                 goto restart;
3587                                         }
3588                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3589                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3590                                 spin_unlock(&ifp->lock);
3591 #endif
3592                         } else {
3593                                 /* ifp->prefered_lft <= ifp->valid_lft */
3594                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3595                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3596                                 spin_unlock(&ifp->lock);
3597                         }
3598                 }
3599         }
3600
3601         next_sec = round_jiffies_up(next);
3602         next_sched = next;
3603
3604         /* If rounded timeout is accurate enough, accept it. */
3605         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3606                 next_sched = next_sec;
3607
3608         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3609         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3610                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3611
3612         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3613               now, next, next_sec, next_sched);
3614
3615         addr_chk_timer.expires = next_sched;
3616         add_timer(&addr_chk_timer);
3617         spin_unlock(&addrconf_verify_lock);
3618         rcu_read_unlock_bh();
3619 }
3620
3621 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3622                                      struct in6_addr **peer_pfx)
3623 {
3624         struct in6_addr *pfx = NULL;
3625
3626         *peer_pfx = NULL;
3627
3628         if (addr)
3629                 pfx = nla_data(addr);
3630
3631         if (local) {
3632                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3633                         *peer_pfx = pfx;
3634                 pfx = nla_data(local);
3635         }
3636
3637         return pfx;
3638 }
3639
3640 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3641         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3642         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3643         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3644 };
3645
3646 static int
3647 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3648 {
3649         struct net *net = sock_net(skb->sk);
3650         struct ifaddrmsg *ifm;
3651         struct nlattr *tb[IFA_MAX+1];
3652         struct in6_addr *pfx, *peer_pfx;
3653         int err;
3654
3655         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3656         if (err < 0)
3657                 return err;
3658
3659         ifm = nlmsg_data(nlh);
3660         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3661         if (pfx == NULL)
3662                 return -EINVAL;
3663
3664         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3665 }
3666
3667 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3668                              u32 prefered_lft, u32 valid_lft)
3669 {
3670         u32 flags;
3671         clock_t expires;
3672         unsigned long timeout;
3673
3674         if (!valid_lft || (prefered_lft > valid_lft))
3675                 return -EINVAL;
3676
3677         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3678         if (addrconf_finite_timeout(timeout)) {
3679                 expires = jiffies_to_clock_t(timeout * HZ);
3680                 valid_lft = timeout;
3681                 flags = RTF_EXPIRES;
3682         } else {
3683                 expires = 0;
3684                 flags = 0;
3685                 ifa_flags |= IFA_F_PERMANENT;
3686         }
3687
3688         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3689         if (addrconf_finite_timeout(timeout)) {
3690                 if (timeout == 0)
3691                         ifa_flags |= IFA_F_DEPRECATED;
3692                 prefered_lft = timeout;
3693         }
3694
3695         spin_lock_bh(&ifp->lock);
3696         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3697         ifp->tstamp = jiffies;
3698         ifp->valid_lft = valid_lft;
3699         ifp->prefered_lft = prefered_lft;
3700
3701         spin_unlock_bh(&ifp->lock);
3702         if (!(ifp->flags&IFA_F_TENTATIVE))
3703                 ipv6_ifa_notify(0, ifp);
3704
3705         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3706                               expires, flags);
3707         addrconf_verify(0);
3708
3709         return 0;
3710 }
3711
3712 static int
3713 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3714 {
3715         struct net *net = sock_net(skb->sk);
3716         struct ifaddrmsg *ifm;
3717         struct nlattr *tb[IFA_MAX+1];
3718         struct in6_addr *pfx, *peer_pfx;
3719         struct inet6_ifaddr *ifa;
3720         struct net_device *dev;
3721         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3722         u8 ifa_flags;
3723         int err;
3724
3725         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3726         if (err < 0)
3727                 return err;
3728
3729         ifm = nlmsg_data(nlh);
3730         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3731         if (pfx == NULL)
3732                 return -EINVAL;
3733
3734         if (tb[IFA_CACHEINFO]) {
3735                 struct ifa_cacheinfo *ci;
3736
3737                 ci = nla_data(tb[IFA_CACHEINFO]);
3738                 valid_lft = ci->ifa_valid;
3739                 preferred_lft = ci->ifa_prefered;
3740         } else {
3741                 preferred_lft = INFINITY_LIFE_TIME;
3742                 valid_lft = INFINITY_LIFE_TIME;
3743         }
3744
3745         dev =  __dev_get_by_index(net, ifm->ifa_index);
3746         if (dev == NULL)
3747                 return -ENODEV;
3748
3749         /* We ignore other flags so far. */
3750         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3751
3752         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3753         if (ifa == NULL) {
3754                 /*
3755                  * It would be best to check for !NLM_F_CREATE here but
3756                  * userspace alreay relies on not having to provide this.
3757                  */
3758                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3759                                       ifm->ifa_prefixlen, ifa_flags,
3760                                       preferred_lft, valid_lft);
3761         }
3762
3763         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3764             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3765                 err = -EEXIST;
3766         else
3767                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3768
3769         in6_ifa_put(ifa);
3770
3771         return err;
3772 }
3773
3774 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3775                           u8 scope, int ifindex)
3776 {
3777         struct ifaddrmsg *ifm;
3778
3779         ifm = nlmsg_data(nlh);
3780         ifm->ifa_family = AF_INET6;
3781         ifm->ifa_prefixlen = prefixlen;
3782         ifm->ifa_flags = flags;
3783         ifm->ifa_scope = scope;
3784         ifm->ifa_index = ifindex;
3785 }
3786
3787 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3788                          unsigned long tstamp, u32 preferred, u32 valid)
3789 {
3790         struct ifa_cacheinfo ci;
3791
3792         ci.cstamp = cstamp_delta(cstamp);
3793         ci.tstamp = cstamp_delta(tstamp);
3794         ci.ifa_prefered = preferred;
3795         ci.ifa_valid = valid;
3796
3797         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3798 }
3799
3800 static inline int rt_scope(int ifa_scope)
3801 {
3802         if (ifa_scope & IFA_HOST)
3803                 return RT_SCOPE_HOST;
3804         else if (ifa_scope & IFA_LINK)
3805                 return RT_SCOPE_LINK;
3806         else if (ifa_scope & IFA_SITE)
3807                 return RT_SCOPE_SITE;
3808         else
3809                 return RT_SCOPE_UNIVERSE;
3810 }
3811
3812 static inline int inet6_ifaddr_msgsize(void)
3813 {
3814         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3815                + nla_total_size(16) /* IFA_LOCAL */
3816                + nla_total_size(16) /* IFA_ADDRESS */
3817                + nla_total_size(sizeof(struct ifa_cacheinfo));
3818 }
3819
3820 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3821                              u32 portid, u32 seq, int event, unsigned int flags)
3822 {
3823         struct nlmsghdr  *nlh;
3824         u32 preferred, valid;
3825
3826         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3827         if (nlh == NULL)
3828                 return -EMSGSIZE;
3829
3830         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3831                       ifa->idev->dev->ifindex);
3832
3833         if (!(ifa->flags&IFA_F_PERMANENT)) {
3834                 preferred = ifa->prefered_lft;
3835                 valid = ifa->valid_lft;
3836                 if (preferred != INFINITY_LIFE_TIME) {
3837                         long tval = (jiffies - ifa->tstamp)/HZ;
3838                         if (preferred > tval)
3839                                 preferred -= tval;
3840                         else
3841                                 preferred = 0;
3842                         if (valid != INFINITY_LIFE_TIME) {
3843                                 if (valid > tval)
3844                                         valid -= tval;
3845                                 else
3846                                         valid = 0;
3847                         }
3848                 }
3849         } else {
3850                 preferred = INFINITY_LIFE_TIME;
3851                 valid = INFINITY_LIFE_TIME;
3852         }
3853
3854         if (!ipv6_addr_any(&ifa->peer_addr)) {
3855                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3856                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3857                         goto error;
3858         } else
3859                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3860                         goto error;
3861
3862         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3863                 goto error;
3864
3865         return nlmsg_end(skb, nlh);
3866
3867 error:
3868         nlmsg_cancel(skb, nlh);
3869         return -EMSGSIZE;
3870 }
3871
3872 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3873                                 u32 portid, u32 seq, int event, u16 flags)
3874 {
3875         struct nlmsghdr  *nlh;
3876         u8 scope = RT_SCOPE_UNIVERSE;
3877         int ifindex = ifmca->idev->dev->ifindex;
3878
3879         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3880                 scope = RT_SCOPE_SITE;
3881
3882         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3883         if (nlh == NULL)
3884                 return -EMSGSIZE;
3885
3886         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3887         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3888             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3889                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3890                 nlmsg_cancel(skb, nlh);
3891                 return -EMSGSIZE;
3892         }
3893
3894         return nlmsg_end(skb, nlh);
3895 }
3896
3897 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3898                                 u32 portid, u32 seq, int event, unsigned int flags)
3899 {
3900         struct nlmsghdr  *nlh;
3901         u8 scope = RT_SCOPE_UNIVERSE;
3902         int ifindex = ifaca->aca_idev->dev->ifindex;
3903
3904         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3905                 scope = RT_SCOPE_SITE;
3906
3907         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3908         if (nlh == NULL)
3909                 return -EMSGSIZE;
3910
3911         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3912         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3913             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3914                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3915                 nlmsg_cancel(skb, nlh);
3916                 return -EMSGSIZE;
3917         }
3918
3919         return nlmsg_end(skb, nlh);
3920 }
3921
3922 enum addr_type_t {
3923         UNICAST_ADDR,
3924         MULTICAST_ADDR,
3925         ANYCAST_ADDR,
3926 };
3927
3928 /* called with rcu_read_lock() */
3929 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3930                           struct netlink_callback *cb, enum addr_type_t type,
3931                           int s_ip_idx, int *p_ip_idx)
3932 {
3933         struct ifmcaddr6 *ifmca;
3934         struct ifacaddr6 *ifaca;
3935         int err = 1;
3936         int ip_idx = *p_ip_idx;
3937
3938         read_lock_bh(&idev->lock);
3939         switch (type) {
3940         case UNICAST_ADDR: {
3941                 struct inet6_ifaddr *ifa;
3942
3943                 /* unicast address incl. temp addr */
3944                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3945                         if (++ip_idx < s_ip_idx)
3946                                 continue;
3947                         err = inet6_fill_ifaddr(skb, ifa,
3948                                                 NETLINK_CB(cb->skb).portid,
3949                                                 cb->nlh->nlmsg_seq,
3950                                                 RTM_NEWADDR,
3951                                                 NLM_F_MULTI);
3952                         if (err <= 0)
3953                                 break;
3954                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3955                 }
3956                 break;
3957         }
3958         case MULTICAST_ADDR:
3959                 /* multicast address */
3960                 for (ifmca = idev->mc_list; ifmca;
3961                      ifmca = ifmca->next, ip_idx++) {
3962                         if (ip_idx < s_ip_idx)
3963                                 continue;
3964                         err = inet6_fill_ifmcaddr(skb, ifmca,
3965                                                   NETLINK_CB(cb->skb).portid,
3966                                                   cb->nlh->nlmsg_seq,
3967                                                   RTM_GETMULTICAST,
3968                                                   NLM_F_MULTI);
3969                         if (err <= 0)
3970                                 break;
3971                 }
3972                 break;
3973         case ANYCAST_ADDR:
3974                 /* anycast address */
3975                 for (ifaca = idev->ac_list; ifaca;
3976                      ifaca = ifaca->aca_next, ip_idx++) {
3977                         if (ip_idx < s_ip_idx)
3978                                 continue;
3979                         err = inet6_fill_ifacaddr(skb, ifaca,
3980                                                   NETLINK_CB(cb->skb).portid,
3981                                                   cb->nlh->nlmsg_seq,
3982                                                   RTM_GETANYCAST,
3983                                                   NLM_F_MULTI);
3984                         if (err <= 0)
3985                                 break;
3986                 }
3987                 break;
3988         default:
3989                 break;
3990         }
3991         read_unlock_bh(&idev->lock);
3992         *p_ip_idx = ip_idx;
3993         return err;
3994 }
3995
3996 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3997                            enum addr_type_t type)
3998 {
3999         struct net *net = sock_net(skb->sk);
4000         int h, s_h;
4001         int idx, ip_idx;
4002         int s_idx, s_ip_idx;
4003         struct net_device *dev;
4004         struct inet6_dev *idev;
4005         struct hlist_head *head;
4006
4007         s_h = cb->args[0];
4008         s_idx = idx = cb->args[1];
4009         s_ip_idx = ip_idx = cb->args[2];
4010
4011         rcu_read_lock();
4012         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4013         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4014                 idx = 0;
4015                 head = &net->dev_index_head[h];
4016                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4017                         if (idx < s_idx)
4018                                 goto cont;
4019                         if (h > s_h || idx > s_idx)
4020                                 s_ip_idx = 0;
4021                         ip_idx = 0;
4022                         idev = __in6_dev_get(dev);
4023                         if (!idev)
4024                                 goto cont;
4025
4026                         if (in6_dump_addrs(idev, skb, cb, type,
4027                                            s_ip_idx, &ip_idx) <= 0)
4028                                 goto done;
4029 cont:
4030                         idx++;
4031                 }
4032         }
4033 done:
4034         rcu_read_unlock();
4035         cb->args[0] = h;
4036         cb->args[1] = idx;
4037         cb->args[2] = ip_idx;
4038
4039         return skb->len;
4040 }
4041
4042 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4043 {
4044         enum addr_type_t type = UNICAST_ADDR;
4045
4046         return inet6_dump_addr(skb, cb, type);
4047 }
4048
4049 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4050 {
4051         enum addr_type_t type = MULTICAST_ADDR;
4052
4053         return inet6_dump_addr(skb, cb, type);
4054 }
4055
4056
4057 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4058 {
4059         enum addr_type_t type = ANYCAST_ADDR;
4060
4061         return inet6_dump_addr(skb, cb, type);
4062 }
4063
4064 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4065 {
4066         struct net *net = sock_net(in_skb->sk);
4067         struct ifaddrmsg *ifm;
4068         struct nlattr *tb[IFA_MAX+1];
4069         struct in6_addr *addr = NULL, *peer;
4070         struct net_device *dev = NULL;
4071         struct inet6_ifaddr *ifa;
4072         struct sk_buff *skb;
4073         int err;
4074
4075         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4076         if (err < 0)
4077                 goto errout;
4078
4079         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4080         if (addr == NULL) {
4081                 err = -EINVAL;
4082                 goto errout;
4083         }
4084
4085         ifm = nlmsg_data(nlh);
4086         if (ifm->ifa_index)
4087                 dev = __dev_get_by_index(net, ifm->ifa_index);
4088
4089         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4090         if (!ifa) {
4091                 err = -EADDRNOTAVAIL;
4092                 goto errout;
4093         }
4094
4095         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4096         if (!skb) {
4097                 err = -ENOBUFS;
4098                 goto errout_ifa;
4099         }
4100
4101         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4102                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4103         if (err < 0) {
4104                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4105                 WARN_ON(err == -EMSGSIZE);
4106                 kfree_skb(skb);
4107                 goto errout_ifa;
4108         }
4109         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4110 errout_ifa:
4111         in6_ifa_put(ifa);
4112 errout:
4113         return err;
4114 }
4115
4116 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4117 {
4118         struct sk_buff *skb;
4119         struct net *net = dev_net(ifa->idev->dev);
4120         int err = -ENOBUFS;
4121
4122         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4123         if (skb == NULL)
4124                 goto errout;
4125
4126         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4127         if (err < 0) {
4128                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4129                 WARN_ON(err == -EMSGSIZE);
4130                 kfree_skb(skb);
4131                 goto errout;
4132         }
4133         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4134         return;
4135 errout:
4136         if (err < 0)
4137                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4138 }
4139
4140 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4141                                 __s32 *array, int bytes)
4142 {
4143         BUG_ON(bytes < (DEVCONF_MAX * 4));
4144
4145         memset(array, 0, bytes);
4146         array[DEVCONF_FORWARDING] = cnf->forwarding;
4147         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4148         array[DEVCONF_MTU6] = cnf->mtu6;
4149         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4150         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4151         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4152         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4153         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4154         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4155                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4156         array[DEVCONF_RTR_SOLICIT_DELAY] =
4157                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4158         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4159         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4160                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4161         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4162                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4163 #ifdef CONFIG_IPV6_PRIVACY
4164         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4165         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4166         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4167         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4168         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4169 #endif
4170         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4171         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4172         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4173 #ifdef CONFIG_IPV6_ROUTER_PREF
4174         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4175         array[DEVCONF_RTR_PROBE_INTERVAL] =
4176                 jiffies_to_msecs(cnf->rtr_probe_interval);
4177 #ifdef CONFIG_IPV6_ROUTE_INFO
4178         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4179 #endif
4180 #endif
4181         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4182         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4183 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4184         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4185 #endif
4186 #ifdef CONFIG_IPV6_MROUTE
4187         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4188 #endif
4189         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4190         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4191         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4192         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4193         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4194 }
4195
4196 static inline size_t inet6_ifla6_size(void)
4197 {
4198         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4199              + nla_total_size(sizeof(struct ifla_cacheinfo))
4200              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4201              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4202              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4203              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4204 }
4205
4206 static inline size_t inet6_if_nlmsg_size(void)
4207 {
4208         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4209                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4210                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4211                + nla_total_size(4) /* IFLA_MTU */
4212                + nla_total_size(4) /* IFLA_LINK */
4213                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4214 }
4215
4216 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4217                                       int items, int bytes)
4218 {
4219         int i;
4220         int pad = bytes - sizeof(u64) * items;
4221         BUG_ON(pad < 0);
4222
4223         /* Use put_unaligned() because stats may not be aligned for u64. */
4224         put_unaligned(items, &stats[0]);
4225         for (i = 1; i < items; i++)
4226                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4227
4228         memset(&stats[items], 0, pad);
4229 }
4230
4231 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4232                                       int items, int bytes, size_t syncpoff)
4233 {
4234         int i;
4235         int pad = bytes - sizeof(u64) * items;
4236         BUG_ON(pad < 0);
4237
4238         /* Use put_unaligned() because stats may not be aligned for u64. */
4239         put_unaligned(items, &stats[0]);
4240         for (i = 1; i < items; i++)
4241                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4242
4243         memset(&stats[items], 0, pad);
4244 }
4245
4246 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4247                              int bytes)
4248 {
4249         switch (attrtype) {
4250         case IFLA_INET6_STATS:
4251                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4252                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4253                 break;
4254         case IFLA_INET6_ICMP6STATS:
4255                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4256                 break;
4257         }
4258 }
4259
4260 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4261 {
4262         struct nlattr *nla;
4263         struct ifla_cacheinfo ci;
4264
4265         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4266                 goto nla_put_failure;
4267         ci.max_reasm_len = IPV6_MAXPLEN;
4268         ci.tstamp = cstamp_delta(idev->tstamp);
4269         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4270         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4271         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4272                 goto nla_put_failure;
4273         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4274         if (nla == NULL)
4275                 goto nla_put_failure;
4276         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4277
4278         /* XXX - MC not implemented */
4279
4280         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4281         if (nla == NULL)
4282                 goto nla_put_failure;
4283         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4284
4285         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4286         if (nla == NULL)
4287                 goto nla_put_failure;
4288         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4289
4290         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4291         if (nla == NULL)
4292                 goto nla_put_failure;
4293         read_lock_bh(&idev->lock);
4294         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4295         read_unlock_bh(&idev->lock);
4296
4297         return 0;
4298
4299 nla_put_failure:
4300         return -EMSGSIZE;
4301 }
4302
4303 static size_t inet6_get_link_af_size(const struct net_device *dev)
4304 {
4305         if (!__in6_dev_get(dev))
4306                 return 0;
4307
4308         return inet6_ifla6_size();
4309 }
4310
4311 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4312 {
4313         struct inet6_dev *idev = __in6_dev_get(dev);
4314
4315         if (!idev)
4316                 return -ENODATA;
4317
4318         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4319                 return -EMSGSIZE;
4320
4321         return 0;
4322 }
4323
4324 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4325 {
4326         struct inet6_ifaddr *ifp;
4327         struct net_device *dev = idev->dev;
4328         bool update_rs = false;
4329         struct in6_addr ll_addr;
4330
4331         if (token == NULL)
4332                 return -EINVAL;
4333         if (ipv6_addr_any(token))
4334                 return -EINVAL;
4335         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4336                 return -EINVAL;
4337         if (!ipv6_accept_ra(idev))
4338                 return -EINVAL;
4339         if (idev->cnf.rtr_solicits <= 0)
4340                 return -EINVAL;
4341
4342         write_lock_bh(&idev->lock);
4343
4344         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4345         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4346
4347         write_unlock_bh(&idev->lock);
4348
4349         if (!idev->dead && (idev->if_flags & IF_READY) &&
4350             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4351                              IFA_F_OPTIMISTIC)) {
4352
4353                 /* If we're not ready, then normal ifup will take care
4354                  * of this. Otherwise, we need to request our rs here.
4355                  */
4356                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4357                 update_rs = true;
4358         }
4359
4360         write_lock_bh(&idev->lock);
4361
4362         if (update_rs) {
4363                 idev->if_flags |= IF_RS_SENT;
4364                 idev->rs_probes = 1;
4365                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4366         }
4367
4368         /* Well, that's kinda nasty ... */
4369         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4370                 spin_lock(&ifp->lock);
4371                 if (ifp->tokenized) {
4372                         ifp->valid_lft = 0;
4373                         ifp->prefered_lft = 0;
4374                 }
4375                 spin_unlock(&ifp->lock);
4376         }
4377
4378         write_unlock_bh(&idev->lock);
4379         addrconf_verify(0);
4380         return 0;
4381 }
4382
4383 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4384 {
4385         int err = -EINVAL;
4386         struct inet6_dev *idev = __in6_dev_get(dev);
4387         struct nlattr *tb[IFLA_INET6_MAX + 1];
4388
4389         if (!idev)
4390                 return -EAFNOSUPPORT;
4391
4392         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4393                 BUG();
4394
4395         if (tb[IFLA_INET6_TOKEN])
4396                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4397
4398         return err;
4399 }
4400
4401 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4402                              u32 portid, u32 seq, int event, unsigned int flags)
4403 {
4404         struct net_device *dev = idev->dev;
4405         struct ifinfomsg *hdr;
4406         struct nlmsghdr *nlh;
4407         void *protoinfo;
4408
4409         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4410         if (nlh == NULL)
4411                 return -EMSGSIZE;
4412
4413         hdr = nlmsg_data(nlh);
4414         hdr->ifi_family = AF_INET6;
4415         hdr->__ifi_pad = 0;
4416         hdr->ifi_type = dev->type;
4417         hdr->ifi_index = dev->ifindex;
4418         hdr->ifi_flags = dev_get_flags(dev);
4419         hdr->ifi_change = 0;
4420
4421         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4422             (dev->addr_len &&
4423              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4424             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4425             (dev->ifindex != dev->iflink &&
4426              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4427                 goto nla_put_failure;
4428         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4429         if (protoinfo == NULL)
4430                 goto nla_put_failure;
4431
4432         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4433                 goto nla_put_failure;
4434
4435         nla_nest_end(skb, protoinfo);
4436         return nlmsg_end(skb, nlh);
4437
4438 nla_put_failure:
4439         nlmsg_cancel(skb, nlh);
4440         return -EMSGSIZE;
4441 }
4442
4443 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4444 {
4445         struct net *net = sock_net(skb->sk);
4446         int h, s_h;
4447         int idx = 0, s_idx;
4448         struct net_device *dev;
4449         struct inet6_dev *idev;
4450         struct hlist_head *head;
4451
4452         s_h = cb->args[0];
4453         s_idx = cb->args[1];
4454
4455         rcu_read_lock();
4456         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4457                 idx = 0;
4458                 head = &net->dev_index_head[h];
4459                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4460                         if (idx < s_idx)
4461                                 goto cont;
4462                         idev = __in6_dev_get(dev);
4463                         if (!idev)
4464                                 goto cont;
4465                         if (inet6_fill_ifinfo(skb, idev,
4466                                               NETLINK_CB(cb->skb).portid,
4467                                               cb->nlh->nlmsg_seq,
4468                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4469                                 goto out;
4470 cont:
4471                         idx++;
4472                 }
4473         }
4474 out:
4475         rcu_read_unlock();
4476         cb->args[1] = idx;
4477         cb->args[0] = h;
4478
4479         return skb->len;
4480 }
4481
4482 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4483 {
4484         struct sk_buff *skb;
4485         struct net *net = dev_net(idev->dev);
4486         int err = -ENOBUFS;
4487
4488         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4489         if (skb == NULL)
4490                 goto errout;
4491
4492         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4493         if (err < 0) {
4494                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4495                 WARN_ON(err == -EMSGSIZE);
4496                 kfree_skb(skb);
4497                 goto errout;
4498         }
4499         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4500         return;
4501 errout:
4502         if (err < 0)
4503                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4504 }
4505
4506 static inline size_t inet6_prefix_nlmsg_size(void)
4507 {
4508         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4509                + nla_total_size(sizeof(struct in6_addr))
4510                + nla_total_size(sizeof(struct prefix_cacheinfo));
4511 }
4512
4513 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4514                              struct prefix_info *pinfo, u32 portid, u32 seq,
4515                              int event, unsigned int flags)
4516 {
4517         struct prefixmsg *pmsg;
4518         struct nlmsghdr *nlh;
4519         struct prefix_cacheinfo ci;
4520
4521         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4522         if (nlh == NULL)
4523                 return -EMSGSIZE;
4524
4525         pmsg = nlmsg_data(nlh);
4526         pmsg->prefix_family = AF_INET6;
4527         pmsg->prefix_pad1 = 0;
4528         pmsg->prefix_pad2 = 0;
4529         pmsg->prefix_ifindex = idev->dev->ifindex;
4530         pmsg->prefix_len = pinfo->prefix_len;
4531         pmsg->prefix_type = pinfo->type;
4532         pmsg->prefix_pad3 = 0;
4533         pmsg->prefix_flags = 0;
4534         if (pinfo->onlink)
4535                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4536         if (pinfo->autoconf)
4537                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4538
4539         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4540                 goto nla_put_failure;
4541         ci.preferred_time = ntohl(pinfo->prefered);
4542         ci.valid_time = ntohl(pinfo->valid);
4543         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4544                 goto nla_put_failure;
4545         return nlmsg_end(skb, nlh);
4546
4547 nla_put_failure:
4548         nlmsg_cancel(skb, nlh);
4549         return -EMSGSIZE;
4550 }
4551
4552 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4553                          struct prefix_info *pinfo)
4554 {
4555         struct sk_buff *skb;
4556         struct net *net = dev_net(idev->dev);
4557         int err = -ENOBUFS;
4558
4559         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4560         if (skb == NULL)
4561                 goto errout;
4562
4563         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4564         if (err < 0) {
4565                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4566                 WARN_ON(err == -EMSGSIZE);
4567                 kfree_skb(skb);
4568                 goto errout;
4569         }
4570         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4571         return;
4572 errout:
4573         if (err < 0)
4574                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4575 }
4576
4577 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4578 {
4579         write_lock_bh(&ifp->idev->lock);
4580         spin_lock(&ifp->lock);
4581         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4582                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4583             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4584                 ifp->idev->valid_ll_addr_cnt += count;
4585         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4586         spin_unlock(&ifp->lock);
4587         write_unlock_bh(&ifp->idev->lock);
4588 }
4589
4590 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4591 {
4592         struct net *net = dev_net(ifp->idev->dev);
4593
4594         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4595
4596         switch (event) {
4597         case RTM_NEWADDR:
4598                 update_valid_ll_addr_cnt(ifp, 1);
4599
4600                 /*
4601                  * If the address was optimistic
4602                  * we inserted the route at the start of
4603                  * our DAD process, so we don't need
4604                  * to do it again
4605                  */
4606                 if (!(ifp->rt->rt6i_node))
4607                         ip6_ins_rt(ifp->rt);
4608                 if (ifp->idev->cnf.forwarding)
4609                         addrconf_join_anycast(ifp);
4610                 if (!ipv6_addr_any(&ifp->peer_addr))
4611                         addrconf_prefix_route(&ifp->peer_addr, 128,
4612                                               ifp->idev->dev, 0, 0);
4613                 break;
4614         case RTM_DELADDR:
4615                 update_valid_ll_addr_cnt(ifp, -1);
4616
4617                 if (ifp->idev->cnf.forwarding)
4618                         addrconf_leave_anycast(ifp);
4619                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4620                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4621                         struct rt6_info *rt;
4622                         struct net_device *dev = ifp->idev->dev;
4623
4624                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4625                                         dev->ifindex, 1);
4626                         if (rt) {
4627                                 dst_hold(&rt->dst);
4628                                 if (ip6_del_rt(rt))
4629                                         dst_free(&rt->dst);
4630                         }
4631                 }
4632                 dst_hold(&ifp->rt->dst);
4633
4634                 if (ip6_del_rt(ifp->rt))
4635                         dst_free(&ifp->rt->dst);
4636                 break;
4637         }
4638         atomic_inc(&net->ipv6.dev_addr_genid);
4639         rt_genid_bump_ipv6(net);
4640 }
4641
4642 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4643 {
4644         rcu_read_lock_bh();
4645         if (likely(ifp->idev->dead == 0))
4646                 __ipv6_ifa_notify(event, ifp);
4647         rcu_read_unlock_bh();
4648 }
4649
4650 #ifdef CONFIG_SYSCTL
4651
4652 static
4653 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4654                            void __user *buffer, size_t *lenp, loff_t *ppos)
4655 {
4656         int *valp = ctl->data;
4657         int val = *valp;
4658         loff_t pos = *ppos;
4659         struct ctl_table lctl;
4660         int ret;
4661
4662         /*
4663          * ctl->data points to idev->cnf.forwarding, we should
4664          * not modify it until we get the rtnl lock.
4665          */
4666         lctl = *ctl;
4667         lctl.data = &val;
4668
4669         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4670
4671         if (write)
4672                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4673         if (ret)
4674                 *ppos = pos;
4675         return ret;
4676 }
4677
4678 static void dev_disable_change(struct inet6_dev *idev)
4679 {
4680         struct netdev_notifier_info info;
4681
4682         if (!idev || !idev->dev)
4683                 return;
4684
4685         netdev_notifier_info_init(&info, idev->dev);
4686         if (idev->cnf.disable_ipv6)
4687                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4688         else
4689                 addrconf_notify(NULL, NETDEV_UP, &info);
4690 }
4691
4692 static void addrconf_disable_change(struct net *net, __s32 newf)
4693 {
4694         struct net_device *dev;
4695         struct inet6_dev *idev;
4696
4697         rcu_read_lock();
4698         for_each_netdev_rcu(net, dev) {
4699                 idev = __in6_dev_get(dev);
4700                 if (idev) {
4701                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4702                         idev->cnf.disable_ipv6 = newf;
4703                         if (changed)
4704                                 dev_disable_change(idev);
4705                 }
4706         }
4707         rcu_read_unlock();
4708 }
4709
4710 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4711 {
4712         struct net *net;
4713         int old;
4714
4715         if (!rtnl_trylock())
4716                 return restart_syscall();
4717
4718         net = (struct net *)table->extra2;
4719         old = *p;
4720         *p = newf;
4721
4722         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4723                 rtnl_unlock();
4724                 return 0;
4725         }
4726
4727         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4728                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4729                 addrconf_disable_change(net, newf);
4730         } else if ((!newf) ^ (!old))
4731                 dev_disable_change((struct inet6_dev *)table->extra1);
4732
4733         rtnl_unlock();
4734         return 0;
4735 }
4736
4737 static
4738 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4739                             void __user *buffer, size_t *lenp, loff_t *ppos)
4740 {
4741         int *valp = ctl->data;
4742         int val = *valp;
4743         loff_t pos = *ppos;
4744         struct ctl_table lctl;
4745         int ret;
4746
4747         /*
4748          * ctl->data points to idev->cnf.disable_ipv6, we should
4749          * not modify it until we get the rtnl lock.
4750          */
4751         lctl = *ctl;
4752         lctl.data = &val;
4753
4754         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4755
4756         if (write)
4757                 ret = addrconf_disable_ipv6(ctl, valp, val);
4758         if (ret)
4759                 *ppos = pos;
4760         return ret;
4761 }
4762
4763 static struct addrconf_sysctl_table
4764 {
4765         struct ctl_table_header *sysctl_header;
4766         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4767 } addrconf_sysctl __read_mostly = {
4768         .sysctl_header = NULL,
4769         .addrconf_vars = {
4770                 {
4771                         .procname       = "forwarding",
4772                         .data           = &ipv6_devconf.forwarding,
4773                         .maxlen         = sizeof(int),
4774                         .mode           = 0644,
4775                         .proc_handler   = addrconf_sysctl_forward,
4776                 },
4777                 {
4778                         .procname       = "hop_limit",
4779                         .data           = &ipv6_devconf.hop_limit,
4780                         .maxlen         = sizeof(int),
4781                         .mode           = 0644,
4782                         .proc_handler   = proc_dointvec,
4783                 },
4784                 {
4785                         .procname       = "mtu",
4786                         .data           = &ipv6_devconf.mtu6,
4787                         .maxlen         = sizeof(int),
4788                         .mode           = 0644,
4789                         .proc_handler   = proc_dointvec,
4790                 },
4791                 {
4792                         .procname       = "accept_ra",
4793                         .data           = &ipv6_devconf.accept_ra,
4794                         .maxlen         = sizeof(int),
4795                         .mode           = 0644,
4796                         .proc_handler   = proc_dointvec,
4797                 },
4798                 {
4799                         .procname       = "accept_redirects",
4800                         .data           = &ipv6_devconf.accept_redirects,
4801                         .maxlen         = sizeof(int),
4802                         .mode           = 0644,
4803                         .proc_handler   = proc_dointvec,
4804                 },
4805                 {
4806                         .procname       = "autoconf",
4807                         .data           = &ipv6_devconf.autoconf,
4808                         .maxlen         = sizeof(int),
4809                         .mode           = 0644,
4810                         .proc_handler   = proc_dointvec,
4811                 },
4812                 {
4813                         .procname       = "dad_transmits",
4814                         .data           = &ipv6_devconf.dad_transmits,
4815                         .maxlen         = sizeof(int),
4816                         .mode           = 0644,
4817                         .proc_handler   = proc_dointvec,
4818                 },
4819                 {
4820                         .procname       = "router_solicitations",
4821                         .data           = &ipv6_devconf.rtr_solicits,
4822                         .maxlen         = sizeof(int),
4823                         .mode           = 0644,
4824                         .proc_handler   = proc_dointvec,
4825                 },
4826                 {
4827                         .procname       = "router_solicitation_interval",
4828                         .data           = &ipv6_devconf.rtr_solicit_interval,
4829                         .maxlen         = sizeof(int),
4830                         .mode           = 0644,
4831                         .proc_handler   = proc_dointvec_jiffies,
4832                 },
4833                 {
4834                         .procname       = "router_solicitation_delay",
4835                         .data           = &ipv6_devconf.rtr_solicit_delay,
4836                         .maxlen         = sizeof(int),
4837                         .mode           = 0644,
4838                         .proc_handler   = proc_dointvec_jiffies,
4839                 },
4840                 {
4841                         .procname       = "force_mld_version",
4842                         .data           = &ipv6_devconf.force_mld_version,
4843                         .maxlen         = sizeof(int),
4844                         .mode           = 0644,
4845                         .proc_handler   = proc_dointvec,
4846                 },
4847                 {
4848                         .procname       = "mldv1_unsolicited_report_interval",
4849                         .data           =
4850                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4851                         .maxlen         = sizeof(int),
4852                         .mode           = 0644,
4853                         .proc_handler   = proc_dointvec_ms_jiffies,
4854                 },
4855                 {
4856                         .procname       = "mldv2_unsolicited_report_interval",
4857                         .data           =
4858                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4859                         .maxlen         = sizeof(int),
4860                         .mode           = 0644,
4861                         .proc_handler   = proc_dointvec_ms_jiffies,
4862                 },
4863 #ifdef CONFIG_IPV6_PRIVACY
4864                 {
4865                         .procname       = "use_tempaddr",
4866                         .data           = &ipv6_devconf.use_tempaddr,
4867                         .maxlen         = sizeof(int),
4868                         .mode           = 0644,
4869                         .proc_handler   = proc_dointvec,
4870                 },
4871                 {
4872                         .procname       = "temp_valid_lft",
4873                         .data           = &ipv6_devconf.temp_valid_lft,
4874                         .maxlen         = sizeof(int),
4875                         .mode           = 0644,
4876                         .proc_handler   = proc_dointvec,
4877                 },
4878                 {
4879                         .procname       = "temp_prefered_lft",
4880                         .data           = &ipv6_devconf.temp_prefered_lft,
4881                         .maxlen         = sizeof(int),
4882                         .mode           = 0644,
4883                         .proc_handler   = proc_dointvec,
4884                 },
4885                 {
4886                         .procname       = "regen_max_retry",
4887                         .data           = &ipv6_devconf.regen_max_retry,
4888                         .maxlen         = sizeof(int),
4889                         .mode           = 0644,
4890                         .proc_handler   = proc_dointvec,
4891                 },
4892                 {
4893                         .procname       = "max_desync_factor",
4894                         .data           = &ipv6_devconf.max_desync_factor,
4895                         .maxlen         = sizeof(int),
4896                         .mode           = 0644,
4897                         .proc_handler   = proc_dointvec,
4898                 },
4899 #endif
4900                 {
4901                         .procname       = "max_addresses",
4902                         .data           = &ipv6_devconf.max_addresses,
4903                         .maxlen         = sizeof(int),
4904                         .mode           = 0644,
4905                         .proc_handler   = proc_dointvec,
4906                 },
4907                 {
4908                         .procname       = "accept_ra_defrtr",
4909                         .data           = &ipv6_devconf.accept_ra_defrtr,
4910                         .maxlen         = sizeof(int),
4911                         .mode           = 0644,
4912                         .proc_handler   = proc_dointvec,
4913                 },
4914                 {
4915                         .procname       = "accept_ra_pinfo",
4916                         .data           = &ipv6_devconf.accept_ra_pinfo,
4917                         .maxlen         = sizeof(int),
4918                         .mode           = 0644,
4919                         .proc_handler   = proc_dointvec,
4920                 },
4921 #ifdef CONFIG_IPV6_ROUTER_PREF
4922                 {
4923                         .procname       = "accept_ra_rtr_pref",
4924                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4925                         .maxlen         = sizeof(int),
4926                         .mode           = 0644,
4927                         .proc_handler   = proc_dointvec,
4928                 },
4929                 {
4930                         .procname       = "router_probe_interval",
4931                         .data           = &ipv6_devconf.rtr_probe_interval,
4932                         .maxlen         = sizeof(int),
4933                         .mode           = 0644,
4934                         .proc_handler   = proc_dointvec_jiffies,
4935                 },
4936 #ifdef CONFIG_IPV6_ROUTE_INFO
4937                 {
4938                         .procname       = "accept_ra_rt_info_max_plen",
4939                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4940                         .maxlen         = sizeof(int),
4941                         .mode           = 0644,
4942                         .proc_handler   = proc_dointvec,
4943                 },
4944 #endif
4945 #endif
4946                 {
4947                         .procname       = "proxy_ndp",
4948                         .data           = &ipv6_devconf.proxy_ndp,
4949                         .maxlen         = sizeof(int),
4950                         .mode           = 0644,
4951                         .proc_handler   = proc_dointvec,
4952                 },
4953                 {
4954                         .procname       = "accept_source_route",
4955                         .data           = &ipv6_devconf.accept_source_route,
4956                         .maxlen         = sizeof(int),
4957                         .mode           = 0644,
4958                         .proc_handler   = proc_dointvec,
4959                 },
4960 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4961                 {
4962                         .procname       = "optimistic_dad",
4963                         .data           = &ipv6_devconf.optimistic_dad,
4964                         .maxlen         = sizeof(int),
4965                         .mode           = 0644,
4966                         .proc_handler   = proc_dointvec,
4967
4968                 },
4969 #endif
4970 #ifdef CONFIG_IPV6_MROUTE
4971                 {
4972                         .procname       = "mc_forwarding",
4973                         .data           = &ipv6_devconf.mc_forwarding,
4974                         .maxlen         = sizeof(int),
4975                         .mode           = 0444,
4976                         .proc_handler   = proc_dointvec,
4977                 },
4978 #endif
4979                 {
4980                         .procname       = "disable_ipv6",
4981                         .data           = &ipv6_devconf.disable_ipv6,
4982                         .maxlen         = sizeof(int),
4983                         .mode           = 0644,
4984                         .proc_handler   = addrconf_sysctl_disable,
4985                 },
4986                 {
4987                         .procname       = "accept_dad",
4988                         .data           = &ipv6_devconf.accept_dad,
4989                         .maxlen         = sizeof(int),
4990                         .mode           = 0644,
4991                         .proc_handler   = proc_dointvec,
4992                 },
4993                 {
4994                         .procname       = "force_tllao",
4995                         .data           = &ipv6_devconf.force_tllao,
4996                         .maxlen         = sizeof(int),
4997                         .mode           = 0644,
4998                         .proc_handler   = proc_dointvec
4999                 },
5000                 {
5001                         .procname       = "ndisc_notify",
5002                         .data           = &ipv6_devconf.ndisc_notify,
5003                         .maxlen         = sizeof(int),
5004                         .mode           = 0644,
5005                         .proc_handler   = proc_dointvec
5006                 },
5007                 {
5008                         .procname       = "suppress_frag_ndisc",
5009                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5010                         .maxlen         = sizeof(int),
5011                         .mode           = 0644,
5012                         .proc_handler   = proc_dointvec
5013                 },
5014                 {
5015                         /* sentinel */
5016                 }
5017         },
5018 };
5019
5020 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5021                 struct inet6_dev *idev, struct ipv6_devconf *p)
5022 {
5023         int i;
5024         struct addrconf_sysctl_table *t;
5025         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5026
5027         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5028         if (t == NULL)
5029                 goto out;
5030
5031         for (i = 0; t->addrconf_vars[i].data; i++) {
5032                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5033                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5034                 t->addrconf_vars[i].extra2 = net;
5035         }
5036
5037         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5038
5039         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5040         if (t->sysctl_header == NULL)
5041                 goto free;
5042
5043         p->sysctl = t;
5044         return 0;
5045
5046 free:
5047         kfree(t);
5048 out:
5049         return -ENOBUFS;
5050 }
5051
5052 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5053 {
5054         struct addrconf_sysctl_table *t;
5055
5056         if (p->sysctl == NULL)
5057                 return;
5058
5059         t = p->sysctl;
5060         p->sysctl = NULL;
5061         unregister_net_sysctl_table(t->sysctl_header);
5062         kfree(t);
5063 }
5064
5065 static void addrconf_sysctl_register(struct inet6_dev *idev)
5066 {
5067         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5068                               &ndisc_ifinfo_sysctl_change);
5069         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5070                                         idev, &idev->cnf);
5071 }
5072
5073 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5074 {
5075         __addrconf_sysctl_unregister(&idev->cnf);
5076         neigh_sysctl_unregister(idev->nd_parms);
5077 }
5078
5079
5080 #endif
5081
5082 static int __net_init addrconf_init_net(struct net *net)
5083 {
5084         int err = -ENOMEM;
5085         struct ipv6_devconf *all, *dflt;
5086
5087         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5088         if (all == NULL)
5089                 goto err_alloc_all;
5090
5091         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5092         if (dflt == NULL)
5093                 goto err_alloc_dflt;
5094
5095         /* these will be inherited by all namespaces */
5096         dflt->autoconf = ipv6_defaults.autoconf;
5097         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5098
5099         net->ipv6.devconf_all = all;
5100         net->ipv6.devconf_dflt = dflt;
5101
5102 #ifdef CONFIG_SYSCTL
5103         err = __addrconf_sysctl_register(net, "all", NULL, all);
5104         if (err < 0)
5105                 goto err_reg_all;
5106
5107         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5108         if (err < 0)
5109                 goto err_reg_dflt;
5110 #endif
5111         return 0;
5112
5113 #ifdef CONFIG_SYSCTL
5114 err_reg_dflt:
5115         __addrconf_sysctl_unregister(all);
5116 err_reg_all:
5117         kfree(dflt);
5118 #endif
5119 err_alloc_dflt:
5120         kfree(all);
5121 err_alloc_all:
5122         return err;
5123 }
5124
5125 static void __net_exit addrconf_exit_net(struct net *net)
5126 {
5127 #ifdef CONFIG_SYSCTL
5128         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5129         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5130 #endif
5131         if (!net_eq(net, &init_net)) {
5132                 kfree(net->ipv6.devconf_dflt);
5133                 kfree(net->ipv6.devconf_all);
5134         }
5135 }
5136
5137 static struct pernet_operations addrconf_ops = {
5138         .init = addrconf_init_net,
5139         .exit = addrconf_exit_net,
5140 };
5141
5142 static struct rtnl_af_ops inet6_ops = {
5143         .family           = AF_INET6,
5144         .fill_link_af     = inet6_fill_link_af,
5145         .get_link_af_size = inet6_get_link_af_size,
5146         .set_link_af      = inet6_set_link_af,
5147 };
5148
5149 /*
5150  *      Init / cleanup code
5151  */
5152
5153 int __init addrconf_init(void)
5154 {
5155         int i, err;
5156
5157         err = ipv6_addr_label_init();
5158         if (err < 0) {
5159                 pr_crit("%s: cannot initialize default policy table: %d\n",
5160                         __func__, err);
5161                 goto out;
5162         }
5163
5164         err = register_pernet_subsys(&addrconf_ops);
5165         if (err < 0)
5166                 goto out_addrlabel;
5167
5168         /* The addrconf netdev notifier requires that loopback_dev
5169          * has it's ipv6 private information allocated and setup
5170          * before it can bring up and give link-local addresses
5171          * to other devices which are up.
5172          *
5173          * Unfortunately, loopback_dev is not necessarily the first
5174          * entry in the global dev_base list of net devices.  In fact,
5175          * it is likely to be the very last entry on that list.
5176          * So this causes the notifier registry below to try and
5177          * give link-local addresses to all devices besides loopback_dev
5178          * first, then loopback_dev, which cases all the non-loopback_dev
5179          * devices to fail to get a link-local address.
5180          *
5181          * So, as a temporary fix, allocate the ipv6 structure for
5182          * loopback_dev first by hand.
5183          * Longer term, all of the dependencies ipv6 has upon the loopback
5184          * device and it being up should be removed.
5185          */
5186         rtnl_lock();
5187         if (!ipv6_add_dev(init_net.loopback_dev))
5188                 err = -ENOMEM;
5189         rtnl_unlock();
5190         if (err)
5191                 goto errlo;
5192
5193         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5194                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5195
5196         register_netdevice_notifier(&ipv6_dev_notf);
5197
5198         addrconf_verify(0);
5199
5200         err = rtnl_af_register(&inet6_ops);
5201         if (err < 0)
5202                 goto errout_af;
5203
5204         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5205                               NULL);
5206         if (err < 0)
5207                 goto errout;
5208
5209         /* Only the first call to __rtnl_register can fail */
5210         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5211         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5212         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5213                         inet6_dump_ifaddr, NULL);
5214         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5215                         inet6_dump_ifmcaddr, NULL);
5216         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5217                         inet6_dump_ifacaddr, NULL);
5218         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5219                         inet6_netconf_dump_devconf, NULL);
5220
5221         ipv6_addr_label_rtnl_register();
5222
5223         return 0;
5224 errout:
5225         rtnl_af_unregister(&inet6_ops);
5226 errout_af:
5227         unregister_netdevice_notifier(&ipv6_dev_notf);
5228 errlo:
5229         unregister_pernet_subsys(&addrconf_ops);
5230 out_addrlabel:
5231         ipv6_addr_label_cleanup();
5232 out:
5233         return err;
5234 }
5235
5236 void addrconf_cleanup(void)
5237 {
5238         struct net_device *dev;
5239         int i;
5240
5241         unregister_netdevice_notifier(&ipv6_dev_notf);
5242         unregister_pernet_subsys(&addrconf_ops);
5243         ipv6_addr_label_cleanup();
5244
5245         rtnl_lock();
5246
5247         __rtnl_af_unregister(&inet6_ops);
5248
5249         /* clean dev list */
5250         for_each_netdev(&init_net, dev) {
5251                 if (__in6_dev_get(dev) == NULL)
5252                         continue;
5253                 addrconf_ifdown(dev, 1);
5254         }
5255         addrconf_ifdown(init_net.loopback_dev, 2);
5256
5257         /*
5258          *      Check hash table.
5259          */
5260         spin_lock_bh(&addrconf_hash_lock);
5261         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5262                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5263         spin_unlock_bh(&addrconf_hash_lock);
5264
5265         del_timer(&addr_chk_timer);
5266         rtnl_unlock();
5267 }