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