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 #include <linux/random.h>
  87 #include <linux/uaccess.h>
  88 #include <asm/unaligned.h>
  89
  90 #include <linux/proc_fs.h>
  91 #include <linux/seq_file.h>
  92 #include <linux/export.h>
  93
  94 /* Set to 3 to get tracing... */
  95 #define ACONF_DEBUG 2
  96
  97 #if ACONF_DEBUG >= 3
  98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
  99 #else
 100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 101 #endif
 102
 103 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 104
 105 static inline u32 cstamp_delta(unsigned long cstamp)
 106 {
 107         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 108 }
 109
 110 #ifdef CONFIG_SYSCTL
 111 static void addrconf_sysctl_register(struct inet6_dev *idev);
 112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 113 #else
 114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 115 {
 116 }
 117
 118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 119 {
 120 }
 121 #endif
 122
 123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 125 static void ipv6_regen_rndid(unsigned long data);
 126
 127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 128 static int ipv6_count_addresses(struct inet6_dev *idev);
 129
 130 /*
 131  *      Configured unicast address hash table
 132  */
 133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 134 static DEFINE_SPINLOCK(addrconf_hash_lock);
 135
 136 static void addrconf_verify(unsigned long);
 137
 138 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 139 static DEFINE_SPINLOCK(addrconf_verify_lock);
 140
 141 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 142 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 143
 144 static void addrconf_type_change(struct net_device *dev,
 145                                  unsigned long event);
 146 static int addrconf_ifdown(struct net_device *dev, int how);
 147
 148 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 149                                                   int plen,
 150                                                   const struct net_device *dev,
 151                                                   u32 flags, u32 noflags);
 152
 153 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_timer(unsigned long data);
 155 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 156 static void addrconf_dad_run(struct inet6_dev *idev);
 157 static void addrconf_rs_timer(unsigned long data);
 158 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 160
 161 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 162                                 struct prefix_info *pinfo);
 163 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 164                                struct net_device *dev);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .mldv1_unsolicited_report_interval = 10 * HZ,
 175         .mldv2_unsolicited_report_interval = HZ,
 176         .dad_transmits          = 1,
 177         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 178         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 179         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199         .suppress_frag_ndisc    = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .force_mld_version      = 0,
 210         .mldv1_unsolicited_report_interval = 10 * HZ,
 211         .mldv2_unsolicited_report_interval = HZ,
 212         .dad_transmits          = 1,
 213         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 214         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 215         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 216         .use_tempaddr           = 0,
 217         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 218         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 219         .regen_max_retry        = REGEN_MAX_RETRY,
 220         .max_desync_factor      = MAX_DESYNC_FACTOR,
 221         .max_addresses          = IPV6_MAX_ADDRESSES,
 222         .accept_ra_defrtr       = 1,
 223         .accept_ra_pinfo        = 1,
 224 #ifdef CONFIG_IPV6_ROUTER_PREF
 225         .accept_ra_rtr_pref     = 1,
 226         .rtr_probe_interval     = 60 * HZ,
 227 #ifdef CONFIG_IPV6_ROUTE_INFO
 228         .accept_ra_rt_info_max_plen = 0,
 229 #endif
 230 #endif
 231         .proxy_ndp              = 0,
 232         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 233         .disable_ipv6           = 0,
 234         .accept_dad             = 1,
 235         .suppress_frag_ndisc    = 1,
 236 };
 237
 238 /* Check if a valid qdisc is available */
 239 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 240 {
 241         return !qdisc_tx_is_noop(dev);
 242 }
 243
 244 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 245 {
 246         if (del_timer(&idev->rs_timer))
 247                 __in6_dev_put(idev);
 248 }
 249
 250 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 251 {
 252         if (del_timer(&ifp->dad_timer))
 253                 __in6_ifa_put(ifp);
 254 }
 255
 256 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 257                                   unsigned long when)
 258 {
 259         if (!timer_pending(&idev->rs_timer))
 260                 in6_dev_hold(idev);
 261         mod_timer(&idev->rs_timer, jiffies + when);
 262 }
 263
 264 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 265                                    unsigned long when)
 266 {
 267         if (!timer_pending(&ifp->dad_timer))
 268                 in6_ifa_hold(ifp);
 269         mod_timer(&ifp->dad_timer, jiffies + when);
 270 }
 271
 272 static int snmp6_alloc_dev(struct inet6_dev *idev)
 273 {
 274         int i;
 275
 276         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 277                           sizeof(struct ipstats_mib),
 278                           __alignof__(struct ipstats_mib)) < 0)
 279                 goto err_ip;
 280
 281         for_each_possible_cpu(i) {
 282                 struct ipstats_mib *addrconf_stats;
 283                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i);
 284                 u64_stats_init(&addrconf_stats->syncp);
 285 #if SNMP_ARRAY_SZ == 2
 286                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i);
 287                 u64_stats_init(&addrconf_stats->syncp);
 288 #endif
 289         }
 290
 291
 292         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 293                                         GFP_KERNEL);
 294         if (!idev->stats.icmpv6dev)
 295                 goto err_icmp;
 296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 297                                            GFP_KERNEL);
 298         if (!idev->stats.icmpv6msgdev)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         kfree(idev->stats.icmpv6dev);
 305 err_icmp:
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 312 {
 313         struct inet6_dev *ndev;
 314
 315         ASSERT_RTNL();
 316
 317         if (dev->mtu < IPV6_MIN_MTU)
 318                 return NULL;
 319
 320         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 321
 322         if (ndev == NULL)
 323                 return NULL;
 324
 325         rwlock_init(&ndev->lock);
 326         ndev->dev = dev;
 327         INIT_LIST_HEAD(&ndev->addr_list);
 328         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 329                     (unsigned long)ndev);
 330         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 331         ndev->cnf.mtu6 = dev->mtu;
 332         ndev->cnf.sysctl = NULL;
 333         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 334         if (ndev->nd_parms == NULL) {
 335                 kfree(ndev);
 336                 return NULL;
 337         }
 338         if (ndev->cnf.forwarding)
 339                 dev_disable_lro(dev);
 340         /* We refer to the device */
 341         dev_hold(dev);
 342
 343         if (snmp6_alloc_dev(ndev) < 0) {
 344                 ADBG(KERN_WARNING
 345                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 346                         __func__, dev->name);
 347                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 348                 dev_put(dev);
 349                 kfree(ndev);
 350                 return NULL;
 351         }
 352
 353         if (snmp6_register_dev(ndev) < 0) {
 354                 ADBG(KERN_WARNING
 355                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 356                         __func__, dev->name);
 357                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 358                 ndev->dead = 1;
 359                 in6_dev_finish_destroy(ndev);
 360                 return NULL;
 361         }
 362
 363         /* One reference from device.  We must do this before
 364          * we invoke __ipv6_regen_rndid().
 365          */
 366         in6_dev_hold(ndev);
 367
 368         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 369                 ndev->cnf.accept_dad = -1;
 370
 371 #if IS_ENABLED(CONFIG_IPV6_SIT)
 372         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 373                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 374                 ndev->cnf.rtr_solicits = 0;
 375         }
 376 #endif
 377
 378         INIT_LIST_HEAD(&ndev->tempaddr_list);
 379         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 380         if ((dev->flags&IFF_LOOPBACK) ||
 381             dev->type == ARPHRD_TUNNEL ||
 382             dev->type == ARPHRD_TUNNEL6 ||
 383             dev->type == ARPHRD_SIT ||
 384             dev->type == ARPHRD_NONE) {
 385                 ndev->cnf.use_tempaddr = -1;
 386         } else {
 387                 in6_dev_hold(ndev);
 388                 ipv6_regen_rndid((unsigned long) ndev);
 389         }
 390
 391         ndev->token = in6addr_any;
 392
 393         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 394                 ndev->if_flags |= IF_READY;
 395
 396         ipv6_mc_init_dev(ndev);
 397         ndev->tstamp = jiffies;
 398         addrconf_sysctl_register(ndev);
 399         /* protected by rtnl_lock */
 400         rcu_assign_pointer(dev->ip6_ptr, ndev);
 401
 402         /* Join interface-local all-node multicast group */
 403         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 404
 405         /* Join all-node multicast group */
 406         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 407
 408         /* Join all-router multicast group if forwarding is set */
 409         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 410                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 411
 412         return ndev;
 413 }
 414
 415 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 416 {
 417         struct inet6_dev *idev;
 418
 419         ASSERT_RTNL();
 420
 421         idev = __in6_dev_get(dev);
 422         if (!idev) {
 423                 idev = ipv6_add_dev(dev);
 424                 if (!idev)
 425                         return NULL;
 426         }
 427
 428         if (dev->flags&IFF_UP)
 429                 ipv6_mc_up(idev);
 430         return idev;
 431 }
 432
 433 static int inet6_netconf_msgsize_devconf(int type)
 434 {
 435         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 436                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 437
 438         /* type -1 is used for ALL */
 439         if (type == -1 || type == NETCONFA_FORWARDING)
 440                 size += nla_total_size(4);
 441 #ifdef CONFIG_IPV6_MROUTE
 442         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 443                 size += nla_total_size(4);
 444 #endif
 445         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 446                 size += nla_total_size(4);
 447
 448         return size;
 449 }
 450
 451 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 452                                       struct ipv6_devconf *devconf, u32 portid,
 453                                       u32 seq, int event, unsigned int flags,
 454                                       int type)
 455 {
 456         struct nlmsghdr  *nlh;
 457         struct netconfmsg *ncm;
 458
 459         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 460                         flags);
 461         if (nlh == NULL)
 462                 return -EMSGSIZE;
 463
 464         ncm = nlmsg_data(nlh);
 465         ncm->ncm_family = AF_INET6;
 466
 467         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 468                 goto nla_put_failure;
 469
 470         /* type -1 is used for ALL */
 471         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 472             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 473                 goto nla_put_failure;
 474 #ifdef CONFIG_IPV6_MROUTE
 475         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 476             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 477                         devconf->mc_forwarding) < 0)
 478                 goto nla_put_failure;
 479 #endif
 480         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 481             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 482                 goto nla_put_failure;
 483
 484         return nlmsg_end(skb, nlh);
 485
 486 nla_put_failure:
 487         nlmsg_cancel(skb, nlh);
 488         return -EMSGSIZE;
 489 }
 490
 491 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 492                                   struct ipv6_devconf *devconf)
 493 {
 494         struct sk_buff *skb;
 495         int err = -ENOBUFS;
 496
 497         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 498         if (skb == NULL)
 499                 goto errout;
 500
 501         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 502                                          RTM_NEWNETCONF, 0, type);
 503         if (err < 0) {
 504                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 505                 WARN_ON(err == -EMSGSIZE);
 506                 kfree_skb(skb);
 507                 goto errout;
 508         }
 509         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 510         return;
 511 errout:
 512         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 513 }
 514
 515 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 516         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 517         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 518         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 519 };
 520
 521 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 522                                      struct nlmsghdr *nlh)
 523 {
 524         struct net *net = sock_net(in_skb->sk);
 525         struct nlattr *tb[NETCONFA_MAX+1];
 526         struct netconfmsg *ncm;
 527         struct sk_buff *skb;
 528         struct ipv6_devconf *devconf;
 529         struct inet6_dev *in6_dev;
 530         struct net_device *dev;
 531         int ifindex;
 532         int err;
 533
 534         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 535                           devconf_ipv6_policy);
 536         if (err < 0)
 537                 goto errout;
 538
 539         err = EINVAL;
 540         if (!tb[NETCONFA_IFINDEX])
 541                 goto errout;
 542
 543         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 544         switch (ifindex) {
 545         case NETCONFA_IFINDEX_ALL:
 546                 devconf = net->ipv6.devconf_all;
 547                 break;
 548         case NETCONFA_IFINDEX_DEFAULT:
 549                 devconf = net->ipv6.devconf_dflt;
 550                 break;
 551         default:
 552                 dev = __dev_get_by_index(net, ifindex);
 553                 if (dev == NULL)
 554                         goto errout;
 555                 in6_dev = __in6_dev_get(dev);
 556                 if (in6_dev == NULL)
 557                         goto errout;
 558                 devconf = &in6_dev->cnf;
 559                 break;
 560         }
 561
 562         err = -ENOBUFS;
 563         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 564         if (skb == NULL)
 565                 goto errout;
 566
 567         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 568                                          NETLINK_CB(in_skb).portid,
 569                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 570                                          -1);
 571         if (err < 0) {
 572                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 573                 WARN_ON(err == -EMSGSIZE);
 574                 kfree_skb(skb);
 575                 goto errout;
 576         }
 577         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 578 errout:
 579         return err;
 580 }
 581
 582 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 583                                       struct netlink_callback *cb)
 584 {
 585         struct net *net = sock_net(skb->sk);
 586         int h, s_h;
 587         int idx, s_idx;
 588         struct net_device *dev;
 589         struct inet6_dev *idev;
 590         struct hlist_head *head;
 591
 592         s_h = cb->args[0];
 593         s_idx = idx = cb->args[1];
 594
 595         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 596                 idx = 0;
 597                 head = &net->dev_index_head[h];
 598                 rcu_read_lock();
 599                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 600                           net->dev_base_seq;
 601                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 602                         if (idx < s_idx)
 603                                 goto cont;
 604                         idev = __in6_dev_get(dev);
 605                         if (!idev)
 606                                 goto cont;
 607
 608                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 609                                                        &idev->cnf,
 610                                                        NETLINK_CB(cb->skb).portid,
 611                                                        cb->nlh->nlmsg_seq,
 612                                                        RTM_NEWNETCONF,
 613                                                        NLM_F_MULTI,
 614                                                        -1) <= 0) {
 615                                 rcu_read_unlock();
 616                                 goto done;
 617                         }
 618                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 619 cont:
 620                         idx++;
 621                 }
 622                 rcu_read_unlock();
 623         }
 624         if (h == NETDEV_HASHENTRIES) {
 625                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 626                                                net->ipv6.devconf_all,
 627                                                NETLINK_CB(cb->skb).portid,
 628                                                cb->nlh->nlmsg_seq,
 629                                                RTM_NEWNETCONF, NLM_F_MULTI,
 630                                                -1) <= 0)
 631                         goto done;
 632                 else
 633                         h++;
 634         }
 635         if (h == NETDEV_HASHENTRIES + 1) {
 636                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 637                                                net->ipv6.devconf_dflt,
 638                                                NETLINK_CB(cb->skb).portid,
 639                                                cb->nlh->nlmsg_seq,
 640                                                RTM_NEWNETCONF, NLM_F_MULTI,
 641                                                -1) <= 0)
 642                         goto done;
 643                 else
 644                         h++;
 645         }
 646 done:
 647         cb->args[0] = h;
 648         cb->args[1] = idx;
 649
 650         return skb->len;
 651 }
 652
 653 #ifdef CONFIG_SYSCTL
 654 static void dev_forward_change(struct inet6_dev *idev)
 655 {
 656         struct net_device *dev;
 657         struct inet6_ifaddr *ifa;
 658
 659         if (!idev)
 660                 return;
 661         dev = idev->dev;
 662         if (idev->cnf.forwarding)
 663                 dev_disable_lro(dev);
 664         if (dev->flags & IFF_MULTICAST) {
 665                 if (idev->cnf.forwarding) {
 666                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 667                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 668                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 669                 } else {
 670                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 671                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 672                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 673                 }
 674         }
 675
 676         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 677                 if (ifa->flags&IFA_F_TENTATIVE)
 678                         continue;
 679                 if (idev->cnf.forwarding)
 680                         addrconf_join_anycast(ifa);
 681                 else
 682                         addrconf_leave_anycast(ifa);
 683         }
 684         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 685                                      dev->ifindex, &idev->cnf);
 686 }
 687
 688
 689 static void addrconf_forward_change(struct net *net, __s32 newf)
 690 {
 691         struct net_device *dev;
 692         struct inet6_dev *idev;
 693
 694         for_each_netdev(net, dev) {
 695                 idev = __in6_dev_get(dev);
 696                 if (idev) {
 697                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 698                         idev->cnf.forwarding = newf;
 699                         if (changed)
 700                                 dev_forward_change(idev);
 701                 }
 702         }
 703 }
 704
 705 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 706 {
 707         struct net *net;
 708         int old;
 709
 710         if (!rtnl_trylock())
 711                 return restart_syscall();
 712
 713         net = (struct net *)table->extra2;
 714         old = *p;
 715         *p = newf;
 716
 717         if (p == &net->ipv6.devconf_dflt->forwarding) {
 718                 if ((!newf) ^ (!old))
 719                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 720                                                      NETCONFA_IFINDEX_DEFAULT,
 721                                                      net->ipv6.devconf_dflt);
 722                 rtnl_unlock();
 723                 return 0;
 724         }
 725
 726         if (p == &net->ipv6.devconf_all->forwarding) {
 727                 net->ipv6.devconf_dflt->forwarding = newf;
 728                 addrconf_forward_change(net, newf);
 729                 if ((!newf) ^ (!old))
 730                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 731                                                      NETCONFA_IFINDEX_ALL,
 732                                                      net->ipv6.devconf_all);
 733         } else if ((!newf) ^ (!old))
 734                 dev_forward_change((struct inet6_dev *)table->extra1);
 735         rtnl_unlock();
 736
 737         if (newf)
 738                 rt6_purge_dflt_routers(net);
 739         return 1;
 740 }
 741 #endif
 742
 743 /* Nobody refers to this ifaddr, destroy it */
 744 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 745 {
 746         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 747
 748 #ifdef NET_REFCNT_DEBUG
 749         pr_debug("%s\n", __func__);
 750 #endif
 751
 752         in6_dev_put(ifp->idev);
 753
 754         if (del_timer(&ifp->dad_timer))
 755                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 756
 757         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 758                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 759                 return;
 760         }
 761         ip6_rt_put(ifp->rt);
 762
 763         kfree_rcu(ifp, rcu);
 764 }
 765
 766 static void
 767 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 768 {
 769         struct list_head *p;
 770         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 771
 772         /*
 773          * Each device address list is sorted in order of scope -
 774          * global before linklocal.
 775          */
 776         list_for_each(p, &idev->addr_list) {
 777                 struct inet6_ifaddr *ifa
 778                         = list_entry(p, struct inet6_ifaddr, if_list);
 779                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 780                         break;
 781         }
 782
 783         list_add_tail(&ifp->if_list, p);
 784 }
 785
 786 static u32 inet6_addr_hash(const struct in6_addr *addr)
 787 {
 788         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 789 }
 790
 791 /* On success it returns ifp with increased reference count */
 792
 793 static struct inet6_ifaddr *
 794 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 795               const struct in6_addr *peer_addr, int pfxlen,
 796               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 797 {
 798         struct inet6_ifaddr *ifa = NULL;
 799         struct rt6_info *rt;
 800         unsigned int hash;
 801         int err = 0;
 802         int addr_type = ipv6_addr_type(addr);
 803
 804         if (addr_type == IPV6_ADDR_ANY ||
 805             addr_type & IPV6_ADDR_MULTICAST ||
 806             (!(idev->dev->flags & IFF_LOOPBACK) &&
 807              addr_type & IPV6_ADDR_LOOPBACK))
 808                 return ERR_PTR(-EADDRNOTAVAIL);
 809
 810         rcu_read_lock_bh();
 811         if (idev->dead) {
 812                 err = -ENODEV;                  /*XXX*/
 813                 goto out2;
 814         }
 815
 816         if (idev->cnf.disable_ipv6) {
 817                 err = -EACCES;
 818                 goto out2;
 819         }
 820
 821         spin_lock(&addrconf_hash_lock);
 822
 823         /* Ignore adding duplicate addresses on an interface */
 824         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 825                 ADBG("ipv6_add_addr: already assigned\n");
 826                 err = -EEXIST;
 827                 goto out;
 828         }
 829
 830         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 831
 832         if (ifa == NULL) {
 833                 ADBG("ipv6_add_addr: malloc failed\n");
 834                 err = -ENOBUFS;
 835                 goto out;
 836         }
 837
 838         rt = addrconf_dst_alloc(idev, addr, false);
 839         if (IS_ERR(rt)) {
 840                 err = PTR_ERR(rt);
 841                 goto out;
 842         }
 843
 844         neigh_parms_data_state_setall(idev->nd_parms);
 845
 846         ifa->addr = *addr;
 847         if (peer_addr)
 848                 ifa->peer_addr = *peer_addr;
 849
 850         spin_lock_init(&ifa->lock);
 851         spin_lock_init(&ifa->state_lock);
 852         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 853                     (unsigned long)ifa);
 854         INIT_HLIST_NODE(&ifa->addr_lst);
 855         ifa->scope = scope;
 856         ifa->prefix_len = pfxlen;
 857         ifa->flags = flags | IFA_F_TENTATIVE;
 858         ifa->valid_lft = valid_lft;
 859         ifa->prefered_lft = prefered_lft;
 860         ifa->cstamp = ifa->tstamp = jiffies;
 861         ifa->tokenized = false;
 862
 863         ifa->rt = rt;
 864
 865         ifa->idev = idev;
 866         in6_dev_hold(idev);
 867         /* For caller */
 868         in6_ifa_hold(ifa);
 869
 870         /* Add to big hash table */
 871         hash = inet6_addr_hash(addr);
 872
 873         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 874         spin_unlock(&addrconf_hash_lock);
 875
 876         write_lock(&idev->lock);
 877         /* Add to inet6_dev unicast addr list. */
 878         ipv6_link_dev_addr(idev, ifa);
 879
 880         if (ifa->flags&IFA_F_TEMPORARY) {
 881                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 882                 in6_ifa_hold(ifa);
 883         }
 884
 885         in6_ifa_hold(ifa);
 886         write_unlock(&idev->lock);
 887 out2:
 888         rcu_read_unlock_bh();
 889
 890         if (likely(err == 0))
 891                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 892         else {
 893                 kfree(ifa);
 894                 ifa = ERR_PTR(err);
 895         }
 896
 897         return ifa;
 898 out:
 899         spin_unlock(&addrconf_hash_lock);
 900         goto out2;
 901 }
 902
 903 enum cleanup_prefix_rt_t {
 904         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
 905         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
 906         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
 907 };
 908
 909 /*
 910  * Check, whether the prefix for ifp would still need a prefix route
 911  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
 912  * constants.
 913  *
 914  * 1) we don't purge prefix if address was not permanent.
 915  *    prefix is managed by its own lifetime.
 916  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
 917  * 3) if there are no addresses, delete prefix.
 918  * 4) if there are still other permanent address(es),
 919  *    corresponding prefix is still permanent.
 920  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
 921  *    don't purge the prefix, assume user space is managing it.
 922  * 6) otherwise, update prefix lifetime to the
 923  *    longest valid lifetime among the corresponding
 924  *    addresses on the device.
 925  *    Note: subsequent RA will update lifetime.
 926  **/
 927 static enum cleanup_prefix_rt_t
 928 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
 929 {
 930         struct inet6_ifaddr *ifa;
 931         struct inet6_dev *idev = ifp->idev;
 932         unsigned long lifetime;
 933         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
 934
 935         *expires = jiffies;
 936
 937         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 938                 if (ifa == ifp)
 939                         continue;
 940                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 941                                        ifp->prefix_len))
 942                         continue;
 943                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
 944                         return CLEANUP_PREFIX_RT_NOP;
 945
 946                 action = CLEANUP_PREFIX_RT_EXPIRE;
 947
 948                 spin_lock(&ifa->lock);
 949
 950                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 951                 /*
 952                  * Note: Because this address is
 953                  * not permanent, lifetime <
 954                  * LONG_MAX / HZ here.
 955                  */
 956                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
 957                         *expires = ifa->tstamp + lifetime * HZ;
 958                 spin_unlock(&ifa->lock);
 959         }
 960
 961         return action;
 962 }
 963
 964 static void
 965 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
 966 {
 967         struct rt6_info *rt;
 968
 969         rt = addrconf_get_prefix_route(&ifp->addr,
 970                                        ifp->prefix_len,
 971                                        ifp->idev->dev,
 972                                        0, RTF_GATEWAY | RTF_DEFAULT);
 973         if (rt) {
 974                 if (del_rt)
 975                         ip6_del_rt(rt);
 976                 else {
 977                         if (!(rt->rt6i_flags & RTF_EXPIRES))
 978                                 rt6_set_expires(rt, expires);
 979                         ip6_rt_put(rt);
 980                 }
 981         }
 982 }
 983
 984
 985 /* This function wants to get referenced ifp and releases it before return */
 986
 987 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 988 {
 989         int state;
 990         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
 991         unsigned long expires;
 992
 993         spin_lock_bh(&ifp->state_lock);
 994         state = ifp->state;
 995         ifp->state = INET6_IFADDR_STATE_DEAD;
 996         spin_unlock_bh(&ifp->state_lock);
 997
 998         if (state == INET6_IFADDR_STATE_DEAD)
 999                 goto out;
1000
1001         spin_lock_bh(&addrconf_hash_lock);
1002         hlist_del_init_rcu(&ifp->addr_lst);
1003         spin_unlock_bh(&addrconf_hash_lock);
1004
1005         write_lock_bh(&ifp->idev->lock);
1006
1007         if (ifp->flags&IFA_F_TEMPORARY) {
1008                 list_del(&ifp->tmp_list);
1009                 if (ifp->ifpub) {
1010                         in6_ifa_put(ifp->ifpub);
1011                         ifp->ifpub = NULL;
1012                 }
1013                 __in6_ifa_put(ifp);
1014         }
1015
1016         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1017                 action = check_cleanup_prefix_route(ifp, &expires);
1018
1019         list_del_init(&ifp->if_list);
1020         __in6_ifa_put(ifp);
1021
1022         write_unlock_bh(&ifp->idev->lock);
1023
1024         addrconf_del_dad_timer(ifp);
1025
1026         ipv6_ifa_notify(RTM_DELADDR, ifp);
1027
1028         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1029
1030         if (action != CLEANUP_PREFIX_RT_NOP) {
1031                 cleanup_prefix_route(ifp, expires,
1032                         action == CLEANUP_PREFIX_RT_DEL);
1033         }
1034
1035         /* clean up prefsrc entries */
1036         rt6_remove_prefsrc(ifp);
1037 out:
1038         in6_ifa_put(ifp);
1039 }
1040
1041 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1042 {
1043         struct inet6_dev *idev = ifp->idev;
1044         struct in6_addr addr, *tmpaddr;
1045         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1046         unsigned long regen_advance;
1047         int tmp_plen;
1048         int ret = 0;
1049         u32 addr_flags;
1050         unsigned long now = jiffies;
1051
1052         write_lock_bh(&idev->lock);
1053         if (ift) {
1054                 spin_lock_bh(&ift->lock);
1055                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1056                 spin_unlock_bh(&ift->lock);
1057                 tmpaddr = &addr;
1058         } else {
1059                 tmpaddr = NULL;
1060         }
1061 retry:
1062         in6_dev_hold(idev);
1063         if (idev->cnf.use_tempaddr <= 0) {
1064                 write_unlock_bh(&idev->lock);
1065                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1066                 in6_dev_put(idev);
1067                 ret = -1;
1068                 goto out;
1069         }
1070         spin_lock_bh(&ifp->lock);
1071         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1072                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1073                 spin_unlock_bh(&ifp->lock);
1074                 write_unlock_bh(&idev->lock);
1075                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1076                         __func__);
1077                 in6_dev_put(idev);
1078                 ret = -1;
1079                 goto out;
1080         }
1081         in6_ifa_hold(ifp);
1082         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1083         __ipv6_try_regen_rndid(idev, tmpaddr);
1084         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1085         age = (now - ifp->tstamp) / HZ;
1086         tmp_valid_lft = min_t(__u32,
1087                               ifp->valid_lft,
1088                               idev->cnf.temp_valid_lft + age);
1089         tmp_prefered_lft = min_t(__u32,
1090                                  ifp->prefered_lft,
1091                                  idev->cnf.temp_prefered_lft + age -
1092                                  idev->cnf.max_desync_factor);
1093         tmp_plen = ifp->prefix_len;
1094         tmp_tstamp = ifp->tstamp;
1095         spin_unlock_bh(&ifp->lock);
1096
1097         regen_advance = idev->cnf.regen_max_retry *
1098                         idev->cnf.dad_transmits *
1099                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1100         write_unlock_bh(&idev->lock);
1101
1102         /* A temporary address is created only if this calculated Preferred
1103          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1104          * an implementation must not create a temporary address with a zero
1105          * Preferred Lifetime.
1106          */
1107         if (tmp_prefered_lft <= regen_advance) {
1108                 in6_ifa_put(ifp);
1109                 in6_dev_put(idev);
1110                 ret = -1;
1111                 goto out;
1112         }
1113
1114         addr_flags = IFA_F_TEMPORARY;
1115         /* set in addrconf_prefix_rcv() */
1116         if (ifp->flags & IFA_F_OPTIMISTIC)
1117                 addr_flags |= IFA_F_OPTIMISTIC;
1118
1119         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1120                             ipv6_addr_scope(&addr), addr_flags,
1121                             tmp_valid_lft, tmp_prefered_lft);
1122         if (IS_ERR(ift)) {
1123                 in6_ifa_put(ifp);
1124                 in6_dev_put(idev);
1125                 pr_info("%s: retry temporary address regeneration\n", __func__);
1126                 tmpaddr = &addr;
1127                 write_lock_bh(&idev->lock);
1128                 goto retry;
1129         }
1130
1131         spin_lock_bh(&ift->lock);
1132         ift->ifpub = ifp;
1133         ift->cstamp = now;
1134         ift->tstamp = tmp_tstamp;
1135         spin_unlock_bh(&ift->lock);
1136
1137         addrconf_dad_start(ift);
1138         in6_ifa_put(ift);
1139         in6_dev_put(idev);
1140 out:
1141         return ret;
1142 }
1143
1144 /*
1145  *      Choose an appropriate source address (RFC3484)
1146  */
1147 enum {
1148         IPV6_SADDR_RULE_INIT = 0,
1149         IPV6_SADDR_RULE_LOCAL,
1150         IPV6_SADDR_RULE_SCOPE,
1151         IPV6_SADDR_RULE_PREFERRED,
1152 #ifdef CONFIG_IPV6_MIP6
1153         IPV6_SADDR_RULE_HOA,
1154 #endif
1155         IPV6_SADDR_RULE_OIF,
1156         IPV6_SADDR_RULE_LABEL,
1157         IPV6_SADDR_RULE_PRIVACY,
1158         IPV6_SADDR_RULE_ORCHID,
1159         IPV6_SADDR_RULE_PREFIX,
1160         IPV6_SADDR_RULE_MAX
1161 };
1162
1163 struct ipv6_saddr_score {
1164         int                     rule;
1165         int                     addr_type;
1166         struct inet6_ifaddr     *ifa;
1167         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1168         int                     scopedist;
1169         int                     matchlen;
1170 };
1171
1172 struct ipv6_saddr_dst {
1173         const struct in6_addr *addr;
1174         int ifindex;
1175         int scope;
1176         int label;
1177         unsigned int prefs;
1178 };
1179
1180 static inline int ipv6_saddr_preferred(int type)
1181 {
1182         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1183                 return 1;
1184         return 0;
1185 }
1186
1187 static int ipv6_get_saddr_eval(struct net *net,
1188                                struct ipv6_saddr_score *score,
1189                                struct ipv6_saddr_dst *dst,
1190                                int i)
1191 {
1192         int ret;
1193
1194         if (i <= score->rule) {
1195                 switch (i) {
1196                 case IPV6_SADDR_RULE_SCOPE:
1197                         ret = score->scopedist;
1198                         break;
1199                 case IPV6_SADDR_RULE_PREFIX:
1200                         ret = score->matchlen;
1201                         break;
1202                 default:
1203                         ret = !!test_bit(i, score->scorebits);
1204                 }
1205                 goto out;
1206         }
1207
1208         switch (i) {
1209         case IPV6_SADDR_RULE_INIT:
1210                 /* Rule 0: remember if hiscore is not ready yet */
1211                 ret = !!score->ifa;
1212                 break;
1213         case IPV6_SADDR_RULE_LOCAL:
1214                 /* Rule 1: Prefer same address */
1215                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1216                 break;
1217         case IPV6_SADDR_RULE_SCOPE:
1218                 /* Rule 2: Prefer appropriate scope
1219                  *
1220                  *      ret
1221                  *       ^
1222                  *    -1 |  d 15
1223                  *    ---+--+-+---> scope
1224                  *       |
1225                  *       |             d is scope of the destination.
1226                  *  B-d  |  \
1227                  *       |   \      <- smaller scope is better if
1228                  *  B-15 |    \        if scope is enough for destination.
1229                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1230                  * d-C-1 | /
1231                  *       |/         <- greater is better
1232                  *   -C  /             if scope is not enough for destination.
1233                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1234                  *
1235                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1236                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1237                  * Assume B = 0 and we get C > 29.
1238                  */
1239                 ret = __ipv6_addr_src_scope(score->addr_type);
1240                 if (ret >= dst->scope)
1241                         ret = -ret;
1242                 else
1243                         ret -= 128;     /* 30 is enough */
1244                 score->scopedist = ret;
1245                 break;
1246         case IPV6_SADDR_RULE_PREFERRED:
1247                 /* Rule 3: Avoid deprecated and optimistic addresses */
1248                 ret = ipv6_saddr_preferred(score->addr_type) ||
1249                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1250                 break;
1251 #ifdef CONFIG_IPV6_MIP6
1252         case IPV6_SADDR_RULE_HOA:
1253             {
1254                 /* Rule 4: Prefer home address */
1255                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1256                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1257                 break;
1258             }
1259 #endif
1260         case IPV6_SADDR_RULE_OIF:
1261                 /* Rule 5: Prefer outgoing interface */
1262                 ret = (!dst->ifindex ||
1263                        dst->ifindex == score->ifa->idev->dev->ifindex);
1264                 break;
1265         case IPV6_SADDR_RULE_LABEL:
1266                 /* Rule 6: Prefer matching label */
1267                 ret = ipv6_addr_label(net,
1268                                       &score->ifa->addr, score->addr_type,
1269                                       score->ifa->idev->dev->ifindex) == dst->label;
1270                 break;
1271         case IPV6_SADDR_RULE_PRIVACY:
1272             {
1273                 /* Rule 7: Prefer public address
1274                  * Note: prefer temporary address if use_tempaddr >= 2
1275                  */
1276                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1277                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1278                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1279                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1280                 break;
1281             }
1282         case IPV6_SADDR_RULE_ORCHID:
1283                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1284                  *          non-ORCHID vs non-ORCHID
1285                  */
1286                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1287                         ipv6_addr_orchid(dst->addr));
1288                 break;
1289         case IPV6_SADDR_RULE_PREFIX:
1290                 /* Rule 8: Use longest matching prefix */
1291                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1292                 if (ret > score->ifa->prefix_len)
1293                         ret = score->ifa->prefix_len;
1294                 score->matchlen = ret;
1295                 break;
1296         default:
1297                 ret = 0;
1298         }
1299
1300         if (ret)
1301                 __set_bit(i, score->scorebits);
1302         score->rule = i;
1303 out:
1304         return ret;
1305 }
1306
1307 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1308                        const struct in6_addr *daddr, unsigned int prefs,
1309                        struct in6_addr *saddr)
1310 {
1311         struct ipv6_saddr_score scores[2],
1312                                 *score = &scores[0], *hiscore = &scores[1];
1313         struct ipv6_saddr_dst dst;
1314         struct net_device *dev;
1315         int dst_type;
1316
1317         dst_type = __ipv6_addr_type(daddr);
1318         dst.addr = daddr;
1319         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1320         dst.scope = __ipv6_addr_src_scope(dst_type);
1321         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1322         dst.prefs = prefs;
1323
1324         hiscore->rule = -1;
1325         hiscore->ifa = NULL;
1326
1327         rcu_read_lock();
1328
1329         for_each_netdev_rcu(net, dev) {
1330                 struct inet6_dev *idev;
1331
1332                 /* Candidate Source Address (section 4)
1333                  *  - multicast and link-local destination address,
1334                  *    the set of candidate source address MUST only
1335                  *    include addresses assigned to interfaces
1336                  *    belonging to the same link as the outgoing
1337                  *    interface.
1338                  * (- For site-local destination addresses, the
1339                  *    set of candidate source addresses MUST only
1340                  *    include addresses assigned to interfaces
1341                  *    belonging to the same site as the outgoing
1342                  *    interface.)
1343                  */
1344                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1345                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1346                     dst.ifindex && dev->ifindex != dst.ifindex)
1347                         continue;
1348
1349                 idev = __in6_dev_get(dev);
1350                 if (!idev)
1351                         continue;
1352
1353                 read_lock_bh(&idev->lock);
1354                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1355                         int i;
1356
1357                         /*
1358                          * - Tentative Address (RFC2462 section 5.4)
1359                          *  - A tentative address is not considered
1360                          *    "assigned to an interface" in the traditional
1361                          *    sense, unless it is also flagged as optimistic.
1362                          * - Candidate Source Address (section 4)
1363                          *  - In any case, anycast addresses, multicast
1364                          *    addresses, and the unspecified address MUST
1365                          *    NOT be included in a candidate set.
1366                          */
1367                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1368                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1369                                 continue;
1370
1371                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1372
1373                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1374                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1375                                 LIMIT_NETDEBUG(KERN_DEBUG
1376                                                "ADDRCONF: unspecified / multicast address "
1377                                                "assigned as unicast address on %s",
1378                                                dev->name);
1379                                 continue;
1380                         }
1381
1382                         score->rule = -1;
1383                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1384
1385                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1386                                 int minihiscore, miniscore;
1387
1388                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1389                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1390
1391                                 if (minihiscore > miniscore) {
1392                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1393                                             score->scopedist > 0) {
1394                                                 /*
1395                                                  * special case:
1396                                                  * each remaining entry
1397                                                  * has too small (not enough)
1398                                                  * scope, because ifa entries
1399                                                  * are sorted by their scope
1400                                                  * values.
1401                                                  */
1402                                                 goto try_nextdev;
1403                                         }
1404                                         break;
1405                                 } else if (minihiscore < miniscore) {
1406                                         if (hiscore->ifa)
1407                                                 in6_ifa_put(hiscore->ifa);
1408
1409                                         in6_ifa_hold(score->ifa);
1410
1411                                         swap(hiscore, score);
1412
1413                                         /* restore our iterator */
1414                                         score->ifa = hiscore->ifa;
1415
1416                                         break;
1417                                 }
1418                         }
1419                 }
1420 try_nextdev:
1421                 read_unlock_bh(&idev->lock);
1422         }
1423         rcu_read_unlock();
1424
1425         if (!hiscore->ifa)
1426                 return -EADDRNOTAVAIL;
1427
1428         *saddr = hiscore->ifa->addr;
1429         in6_ifa_put(hiscore->ifa);
1430         return 0;
1431 }
1432 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1433
1434 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1435                       u32 banned_flags)
1436 {
1437         struct inet6_ifaddr *ifp;
1438         int err = -EADDRNOTAVAIL;
1439
1440         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1441                 if (ifp->scope == IFA_LINK &&
1442                     !(ifp->flags & banned_flags)) {
1443                         *addr = ifp->addr;
1444                         err = 0;
1445                         break;
1446                 }
1447         }
1448         return err;
1449 }
1450
1451 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1452                     u32 banned_flags)
1453 {
1454         struct inet6_dev *idev;
1455         int err = -EADDRNOTAVAIL;
1456
1457         rcu_read_lock();
1458         idev = __in6_dev_get(dev);
1459         if (idev) {
1460                 read_lock_bh(&idev->lock);
1461                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1462                 read_unlock_bh(&idev->lock);
1463         }
1464         rcu_read_unlock();
1465         return err;
1466 }
1467
1468 static int ipv6_count_addresses(struct inet6_dev *idev)
1469 {
1470         int cnt = 0;
1471         struct inet6_ifaddr *ifp;
1472
1473         read_lock_bh(&idev->lock);
1474         list_for_each_entry(ifp, &idev->addr_list, if_list)
1475                 cnt++;
1476         read_unlock_bh(&idev->lock);
1477         return cnt;
1478 }
1479
1480 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1481                   const struct net_device *dev, int strict)
1482 {
1483         struct inet6_ifaddr *ifp;
1484         unsigned int hash = inet6_addr_hash(addr);
1485
1486         rcu_read_lock_bh();
1487         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1488                 if (!net_eq(dev_net(ifp->idev->dev), net))
1489                         continue;
1490                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1491                     !(ifp->flags&IFA_F_TENTATIVE) &&
1492                     (dev == NULL || ifp->idev->dev == dev ||
1493                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1494                         rcu_read_unlock_bh();
1495                         return 1;
1496                 }
1497         }
1498
1499         rcu_read_unlock_bh();
1500         return 0;
1501 }
1502 EXPORT_SYMBOL(ipv6_chk_addr);
1503
1504 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1505                                struct net_device *dev)
1506 {
1507         unsigned int hash = inet6_addr_hash(addr);
1508         struct inet6_ifaddr *ifp;
1509
1510         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1511                 if (!net_eq(dev_net(ifp->idev->dev), net))
1512                         continue;
1513                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1514                         if (dev == NULL || ifp->idev->dev == dev)
1515                                 return true;
1516                 }
1517         }
1518         return false;
1519 }
1520
1521 /* Compares an address/prefix_len with addresses on device @dev.
1522  * If one is found it returns true.
1523  */
1524 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1525         const unsigned int prefix_len, struct net_device *dev)
1526 {
1527         struct inet6_dev *idev;
1528         struct inet6_ifaddr *ifa;
1529         bool ret = false;
1530
1531         rcu_read_lock();
1532         idev = __in6_dev_get(dev);
1533         if (idev) {
1534                 read_lock_bh(&idev->lock);
1535                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1536                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1537                         if (ret)
1538                                 break;
1539                 }
1540                 read_unlock_bh(&idev->lock);
1541         }
1542         rcu_read_unlock();
1543
1544         return ret;
1545 }
1546 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1547
1548 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1549 {
1550         struct inet6_dev *idev;
1551         struct inet6_ifaddr *ifa;
1552         int     onlink;
1553
1554         onlink = 0;
1555         rcu_read_lock();
1556         idev = __in6_dev_get(dev);
1557         if (idev) {
1558                 read_lock_bh(&idev->lock);
1559                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1560                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1561                                                    ifa->prefix_len);
1562                         if (onlink)
1563                                 break;
1564                 }
1565                 read_unlock_bh(&idev->lock);
1566         }
1567         rcu_read_unlock();
1568         return onlink;
1569 }
1570 EXPORT_SYMBOL(ipv6_chk_prefix);
1571
1572 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1573                                      struct net_device *dev, int strict)
1574 {
1575         struct inet6_ifaddr *ifp, *result = NULL;
1576         unsigned int hash = inet6_addr_hash(addr);
1577
1578         rcu_read_lock_bh();
1579         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1580                 if (!net_eq(dev_net(ifp->idev->dev), net))
1581                         continue;
1582                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1583                         if (dev == NULL || ifp->idev->dev == dev ||
1584                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1585                                 result = ifp;
1586                                 in6_ifa_hold(ifp);
1587                                 break;
1588                         }
1589                 }
1590         }
1591         rcu_read_unlock_bh();
1592
1593         return result;
1594 }
1595
1596 /* Gets referenced address, destroys ifaddr */
1597
1598 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1599 {
1600         if (ifp->flags&IFA_F_PERMANENT) {
1601                 spin_lock_bh(&ifp->lock);
1602                 addrconf_del_dad_timer(ifp);
1603                 ifp->flags |= IFA_F_TENTATIVE;
1604                 if (dad_failed)
1605                         ifp->flags |= IFA_F_DADFAILED;
1606                 spin_unlock_bh(&ifp->lock);
1607                 if (dad_failed)
1608                         ipv6_ifa_notify(0, ifp);
1609                 in6_ifa_put(ifp);
1610         } else if (ifp->flags&IFA_F_TEMPORARY) {
1611                 struct inet6_ifaddr *ifpub;
1612                 spin_lock_bh(&ifp->lock);
1613                 ifpub = ifp->ifpub;
1614                 if (ifpub) {
1615                         in6_ifa_hold(ifpub);
1616                         spin_unlock_bh(&ifp->lock);
1617                         ipv6_create_tempaddr(ifpub, ifp);
1618                         in6_ifa_put(ifpub);
1619                 } else {
1620                         spin_unlock_bh(&ifp->lock);
1621                 }
1622                 ipv6_del_addr(ifp);
1623         } else
1624                 ipv6_del_addr(ifp);
1625 }
1626
1627 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1628 {
1629         int err = -ENOENT;
1630
1631         spin_lock(&ifp->state_lock);
1632         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1633                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1634                 err = 0;
1635         }
1636         spin_unlock(&ifp->state_lock);
1637
1638         return err;
1639 }
1640
1641 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1642 {
1643         struct inet6_dev *idev = ifp->idev;
1644
1645         if (addrconf_dad_end(ifp)) {
1646                 in6_ifa_put(ifp);
1647                 return;
1648         }
1649
1650         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1651                              ifp->idev->dev->name, &ifp->addr);
1652
1653         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1654                 struct in6_addr addr;
1655
1656                 addr.s6_addr32[0] = htonl(0xfe800000);
1657                 addr.s6_addr32[1] = 0;
1658
1659                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1660                     ipv6_addr_equal(&ifp->addr, &addr)) {
1661                         /* DAD failed for link-local based on MAC address */
1662                         idev->cnf.disable_ipv6 = 1;
1663
1664                         pr_info("%s: IPv6 being disabled!\n",
1665                                 ifp->idev->dev->name);
1666                 }
1667         }
1668
1669         addrconf_dad_stop(ifp, 1);
1670 }
1671
1672 /* Join to solicited addr multicast group. */
1673
1674 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1675 {
1676         struct in6_addr maddr;
1677
1678         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1679                 return;
1680
1681         addrconf_addr_solict_mult(addr, &maddr);
1682         ipv6_dev_mc_inc(dev, &maddr);
1683 }
1684
1685 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1686 {
1687         struct in6_addr maddr;
1688
1689         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1690                 return;
1691
1692         addrconf_addr_solict_mult(addr, &maddr);
1693         __ipv6_dev_mc_dec(idev, &maddr);
1694 }
1695
1696 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1697 {
1698         struct in6_addr addr;
1699         if (ifp->prefix_len >= 127) /* RFC 6164 */
1700                 return;
1701         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1702         if (ipv6_addr_any(&addr))
1703                 return;
1704         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1705 }
1706
1707 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1708 {
1709         struct in6_addr addr;
1710         if (ifp->prefix_len >= 127) /* RFC 6164 */
1711                 return;
1712         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1713         if (ipv6_addr_any(&addr))
1714                 return;
1715         __ipv6_dev_ac_dec(ifp->idev, &addr);
1716 }
1717
1718 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1719 {
1720         if (dev->addr_len != ETH_ALEN)
1721                 return -1;
1722         memcpy(eui, dev->dev_addr, 3);
1723         memcpy(eui + 5, dev->dev_addr + 3, 3);
1724
1725         /*
1726          * The zSeries OSA network cards can be shared among various
1727          * OS instances, but the OSA cards have only one MAC address.
1728          * This leads to duplicate address conflicts in conjunction
1729          * with IPv6 if more than one instance uses the same card.
1730          *
1731          * The driver for these cards can deliver a unique 16-bit
1732          * identifier for each instance sharing the same card.  It is
1733          * placed instead of 0xFFFE in the interface identifier.  The
1734          * "u" bit of the interface identifier is not inverted in this
1735          * case.  Hence the resulting interface identifier has local
1736          * scope according to RFC2373.
1737          */
1738         if (dev->dev_id) {
1739                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1740                 eui[4] = dev->dev_id & 0xFF;
1741         } else {
1742                 eui[3] = 0xFF;
1743                 eui[4] = 0xFE;
1744                 eui[0] ^= 2;
1745         }
1746         return 0;
1747 }
1748
1749 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1750 {
1751         if (dev->addr_len != IEEE802154_ADDR_LEN)
1752                 return -1;
1753         memcpy(eui, dev->dev_addr, 8);
1754         eui[0] ^= 2;
1755         return 0;
1756 }
1757
1758 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1759 {
1760         union fwnet_hwaddr *ha;
1761
1762         if (dev->addr_len != FWNET_ALEN)
1763                 return -1;
1764
1765         ha = (union fwnet_hwaddr *)dev->dev_addr;
1766
1767         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1768         eui[0] ^= 2;
1769         return 0;
1770 }
1771
1772 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1773 {
1774         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1775         if (dev->addr_len != ARCNET_ALEN)
1776                 return -1;
1777         memset(eui, 0, 7);
1778         eui[7] = *(u8 *)dev->dev_addr;
1779         return 0;
1780 }
1781
1782 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1783 {
1784         if (dev->addr_len != INFINIBAND_ALEN)
1785                 return -1;
1786         memcpy(eui, dev->dev_addr + 12, 8);
1787         eui[0] |= 2;
1788         return 0;
1789 }
1790
1791 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1792 {
1793         if (addr == 0)
1794                 return -1;
1795         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1796                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1797                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1798                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1799                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1800                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1801         eui[1] = 0;
1802         eui[2] = 0x5E;
1803         eui[3] = 0xFE;
1804         memcpy(eui + 4, &addr, 4);
1805         return 0;
1806 }
1807
1808 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1809 {
1810         if (dev->priv_flags & IFF_ISATAP)
1811                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1812         return -1;
1813 }
1814
1815 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1816 {
1817         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1818 }
1819
1820 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1821 {
1822         memcpy(eui, dev->perm_addr, 3);
1823         memcpy(eui + 5, dev->perm_addr + 3, 3);
1824         eui[3] = 0xFF;
1825         eui[4] = 0xFE;
1826         eui[0] ^= 2;
1827         return 0;
1828 }
1829
1830 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1831 {
1832         switch (dev->type) {
1833         case ARPHRD_ETHER:
1834         case ARPHRD_FDDI:
1835                 return addrconf_ifid_eui48(eui, dev);
1836         case ARPHRD_ARCNET:
1837                 return addrconf_ifid_arcnet(eui, dev);
1838         case ARPHRD_INFINIBAND:
1839                 return addrconf_ifid_infiniband(eui, dev);
1840         case ARPHRD_SIT:
1841                 return addrconf_ifid_sit(eui, dev);
1842         case ARPHRD_IPGRE:
1843                 return addrconf_ifid_gre(eui, dev);
1844         case ARPHRD_6LOWPAN:
1845         case ARPHRD_IEEE802154:
1846                 return addrconf_ifid_eui64(eui, dev);
1847         case ARPHRD_IEEE1394:
1848                 return addrconf_ifid_ieee1394(eui, dev);
1849         case ARPHRD_TUNNEL6:
1850                 return addrconf_ifid_ip6tnl(eui, dev);
1851         }
1852         return -1;
1853 }
1854
1855 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1856 {
1857         int err = -1;
1858         struct inet6_ifaddr *ifp;
1859
1860         read_lock_bh(&idev->lock);
1861         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1862                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1863                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1864                         err = 0;
1865                         break;
1866                 }
1867         }
1868         read_unlock_bh(&idev->lock);
1869         return err;
1870 }
1871
1872 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1873 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1874 {
1875 regen:
1876         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1877         idev->rndid[0] &= ~0x02;
1878
1879         /*
1880          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1881          * check if generated address is not inappropriate
1882          *
1883          *  - Reserved subnet anycast (RFC 2526)
1884          *      11111101 11....11 1xxxxxxx
1885          *  - ISATAP (RFC4214) 6.1
1886          *      00-00-5E-FE-xx-xx-xx-xx
1887          *  - value 0
1888          *  - XXX: already assigned to an address on the device
1889          */
1890         if (idev->rndid[0] == 0xfd &&
1891             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1892             (idev->rndid[7]&0x80))
1893                 goto regen;
1894         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1895                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1896                         goto regen;
1897                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1898                         goto regen;
1899         }
1900 }
1901
1902 static void ipv6_regen_rndid(unsigned long data)
1903 {
1904         struct inet6_dev *idev = (struct inet6_dev *) data;
1905         unsigned long expires;
1906
1907         rcu_read_lock_bh();
1908         write_lock_bh(&idev->lock);
1909
1910         if (idev->dead)
1911                 goto out;
1912
1913         __ipv6_regen_rndid(idev);
1914
1915         expires = jiffies +
1916                 idev->cnf.temp_prefered_lft * HZ -
1917                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1918                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1919                 idev->cnf.max_desync_factor * HZ;
1920         if (time_before(expires, jiffies)) {
1921                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1922                         __func__, idev->dev->name);
1923                 goto out;
1924         }
1925
1926         if (!mod_timer(&idev->regen_timer, expires))
1927                 in6_dev_hold(idev);
1928
1929 out:
1930         write_unlock_bh(&idev->lock);
1931         rcu_read_unlock_bh();
1932         in6_dev_put(idev);
1933 }
1934
1935 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1936 {
1937         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1938                 __ipv6_regen_rndid(idev);
1939 }
1940
1941 /*
1942  *      Add prefix route.
1943  */
1944
1945 static void
1946 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1947                       unsigned long expires, u32 flags)
1948 {
1949         struct fib6_config cfg = {
1950                 .fc_table = RT6_TABLE_PREFIX,
1951                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1952                 .fc_ifindex = dev->ifindex,
1953                 .fc_expires = expires,
1954                 .fc_dst_len = plen,
1955                 .fc_flags = RTF_UP | flags,
1956                 .fc_nlinfo.nl_net = dev_net(dev),
1957                 .fc_protocol = RTPROT_KERNEL,
1958         };
1959
1960         cfg.fc_dst = *pfx;
1961
1962         /* Prevent useless cloning on PtP SIT.
1963            This thing is done here expecting that the whole
1964            class of non-broadcast devices need not cloning.
1965          */
1966 #if IS_ENABLED(CONFIG_IPV6_SIT)
1967         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1968                 cfg.fc_flags |= RTF_NONEXTHOP;
1969 #endif
1970
1971         ip6_route_add(&cfg);
1972 }
1973
1974
1975 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1976                                                   int plen,
1977                                                   const struct net_device *dev,
1978                                                   u32 flags, u32 noflags)
1979 {
1980         struct fib6_node *fn;
1981         struct rt6_info *rt = NULL;
1982         struct fib6_table *table;
1983
1984         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1985         if (table == NULL)
1986                 return NULL;
1987
1988         read_lock_bh(&table->tb6_lock);
1989         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1990         if (!fn)
1991                 goto out;
1992         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1993                 if (rt->dst.dev->ifindex != dev->ifindex)
1994                         continue;
1995                 if ((rt->rt6i_flags & flags) != flags)
1996                         continue;
1997                 if ((rt->rt6i_flags & noflags) != 0)
1998                         continue;
1999                 dst_hold(&rt->dst);
2000                 break;
2001         }
2002 out:
2003         read_unlock_bh(&table->tb6_lock);
2004         return rt;
2005 }
2006
2007
2008 /* Create "default" multicast route to the interface */
2009
2010 static void addrconf_add_mroute(struct net_device *dev)
2011 {
2012         struct fib6_config cfg = {
2013                 .fc_table = RT6_TABLE_LOCAL,
2014                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2015                 .fc_ifindex = dev->ifindex,
2016                 .fc_dst_len = 8,
2017                 .fc_flags = RTF_UP,
2018                 .fc_nlinfo.nl_net = dev_net(dev),
2019         };
2020
2021         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2022
2023         ip6_route_add(&cfg);
2024 }
2025
2026 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2027 {
2028         struct inet6_dev *idev;
2029
2030         ASSERT_RTNL();
2031
2032         idev = ipv6_find_idev(dev);
2033         if (!idev)
2034                 return ERR_PTR(-ENOBUFS);
2035
2036         if (idev->cnf.disable_ipv6)
2037                 return ERR_PTR(-EACCES);
2038
2039         /* Add default multicast route */
2040         if (!(dev->flags & IFF_LOOPBACK))
2041                 addrconf_add_mroute(dev);
2042
2043         return idev;
2044 }
2045
2046 static void manage_tempaddrs(struct inet6_dev *idev,
2047                              struct inet6_ifaddr *ifp,
2048                              __u32 valid_lft, __u32 prefered_lft,
2049                              bool create, unsigned long now)
2050 {
2051         u32 flags;
2052         struct inet6_ifaddr *ift;
2053
2054         read_lock_bh(&idev->lock);
2055         /* update all temporary addresses in the list */
2056         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2057                 int age, max_valid, max_prefered;
2058
2059                 if (ifp != ift->ifpub)
2060                         continue;
2061
2062                 /* RFC 4941 section 3.3:
2063                  * If a received option will extend the lifetime of a public
2064                  * address, the lifetimes of temporary addresses should
2065                  * be extended, subject to the overall constraint that no
2066                  * temporary addresses should ever remain "valid" or "preferred"
2067                  * for a time longer than (TEMP_VALID_LIFETIME) or
2068                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2069                  */
2070                 age = (now - ift->cstamp) / HZ;
2071                 max_valid = idev->cnf.temp_valid_lft - age;
2072                 if (max_valid < 0)
2073                         max_valid = 0;
2074
2075                 max_prefered = idev->cnf.temp_prefered_lft -
2076                                idev->cnf.max_desync_factor - age;
2077                 if (max_prefered < 0)
2078                         max_prefered = 0;
2079
2080                 if (valid_lft > max_valid)
2081                         valid_lft = max_valid;
2082
2083                 if (prefered_lft > max_prefered)
2084                         prefered_lft = max_prefered;
2085
2086                 spin_lock(&ift->lock);
2087                 flags = ift->flags;
2088                 ift->valid_lft = valid_lft;
2089                 ift->prefered_lft = prefered_lft;
2090                 ift->tstamp = now;
2091                 if (prefered_lft > 0)
2092                         ift->flags &= ~IFA_F_DEPRECATED;
2093
2094                 spin_unlock(&ift->lock);
2095                 if (!(flags&IFA_F_TENTATIVE))
2096                         ipv6_ifa_notify(0, ift);
2097         }
2098
2099         if ((create || list_empty(&idev->tempaddr_list)) &&
2100             idev->cnf.use_tempaddr > 0) {
2101                 /* When a new public address is created as described
2102                  * in [ADDRCONF], also create a new temporary address.
2103                  * Also create a temporary address if it's enabled but
2104                  * no temporary address currently exists.
2105                  */
2106                 read_unlock_bh(&idev->lock);
2107                 ipv6_create_tempaddr(ifp, NULL);
2108         } else {
2109                 read_unlock_bh(&idev->lock);
2110         }
2111 }
2112
2113 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2114 {
2115         struct prefix_info *pinfo;
2116         __u32 valid_lft;
2117         __u32 prefered_lft;
2118         int addr_type;
2119         struct inet6_dev *in6_dev;
2120         struct net *net = dev_net(dev);
2121
2122         pinfo = (struct prefix_info *) opt;
2123
2124         if (len < sizeof(struct prefix_info)) {
2125                 ADBG("addrconf: prefix option too short\n");
2126                 return;
2127         }
2128
2129         /*
2130          *      Validation checks ([ADDRCONF], page 19)
2131          */
2132
2133         addr_type = ipv6_addr_type(&pinfo->prefix);
2134
2135         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2136                 return;
2137
2138         valid_lft = ntohl(pinfo->valid);
2139         prefered_lft = ntohl(pinfo->prefered);
2140
2141         if (prefered_lft > valid_lft) {
2142                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2143                 return;
2144         }
2145
2146         in6_dev = in6_dev_get(dev);
2147
2148         if (in6_dev == NULL) {
2149                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2150                                     dev->name);
2151                 return;
2152         }
2153
2154         /*
2155          *      Two things going on here:
2156          *      1) Add routes for on-link prefixes
2157          *      2) Configure prefixes with the auto flag set
2158          */
2159
2160         if (pinfo->onlink) {
2161                 struct rt6_info *rt;
2162                 unsigned long rt_expires;
2163
2164                 /* Avoid arithmetic overflow. Really, we could
2165                  * save rt_expires in seconds, likely valid_lft,
2166                  * but it would require division in fib gc, that it
2167                  * not good.
2168                  */
2169                 if (HZ > USER_HZ)
2170                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2171                 else
2172                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2173
2174                 if (addrconf_finite_timeout(rt_expires))
2175                         rt_expires *= HZ;
2176
2177                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2178                                                pinfo->prefix_len,
2179                                                dev,
2180                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2181                                                RTF_GATEWAY | RTF_DEFAULT);
2182
2183                 if (rt) {
2184                         /* Autoconf prefix route */
2185                         if (valid_lft == 0) {
2186                                 ip6_del_rt(rt);
2187                                 rt = NULL;
2188                         } else if (addrconf_finite_timeout(rt_expires)) {
2189                                 /* not infinity */
2190                                 rt6_set_expires(rt, jiffies + rt_expires);
2191                         } else {
2192                                 rt6_clean_expires(rt);
2193                         }
2194                 } else if (valid_lft) {
2195                         clock_t expires = 0;
2196                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2197                         if (addrconf_finite_timeout(rt_expires)) {
2198                                 /* not infinity */
2199                                 flags |= RTF_EXPIRES;
2200                                 expires = jiffies_to_clock_t(rt_expires);
2201                         }
2202                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2203                                               dev, expires, flags);
2204                 }
2205                 ip6_rt_put(rt);
2206         }
2207
2208         /* Try to figure out our local address for this prefix */
2209
2210         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2211                 struct inet6_ifaddr *ifp;
2212                 struct in6_addr addr;
2213                 int create = 0, update_lft = 0;
2214                 bool tokenized = false;
2215
2216                 if (pinfo->prefix_len == 64) {
2217                         memcpy(&addr, &pinfo->prefix, 8);
2218
2219                         if (!ipv6_addr_any(&in6_dev->token)) {
2220                                 read_lock_bh(&in6_dev->lock);
2221                                 memcpy(addr.s6_addr + 8,
2222                                        in6_dev->token.s6_addr + 8, 8);
2223                                 read_unlock_bh(&in6_dev->lock);
2224                                 tokenized = true;
2225                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2226                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2227                                 in6_dev_put(in6_dev);
2228                                 return;
2229                         }
2230                         goto ok;
2231                 }
2232                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2233                                     pinfo->prefix_len);
2234                 in6_dev_put(in6_dev);
2235                 return;
2236
2237 ok:
2238
2239                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2240
2241                 if (ifp == NULL && valid_lft) {
2242                         int max_addresses = in6_dev->cnf.max_addresses;
2243                         u32 addr_flags = 0;
2244
2245 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2246                         if (in6_dev->cnf.optimistic_dad &&
2247                             !net->ipv6.devconf_all->forwarding && sllao)
2248                                 addr_flags = IFA_F_OPTIMISTIC;
2249 #endif
2250
2251                         /* Do not allow to create too much of autoconfigured
2252                          * addresses; this would be too easy way to crash kernel.
2253                          */
2254                         if (!max_addresses ||
2255                             ipv6_count_addresses(in6_dev) < max_addresses)
2256                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2257                                                     pinfo->prefix_len,
2258                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2259                                                     addr_flags, valid_lft,
2260                                                     prefered_lft);
2261
2262                         if (IS_ERR_OR_NULL(ifp)) {
2263                                 in6_dev_put(in6_dev);
2264                                 return;
2265                         }
2266
2267                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2268                         update_lft = 0;
2269                         create = 1;
2270                         ifp->cstamp = jiffies;
2271                         ifp->tokenized = tokenized;
2272                         addrconf_dad_start(ifp);
2273                 }
2274
2275                 if (ifp) {
2276                         u32 flags;
2277                         unsigned long now;
2278                         u32 stored_lft;
2279
2280                         /* update lifetime (RFC2462 5.5.3 e) */
2281                         spin_lock(&ifp->lock);
2282                         now = jiffies;
2283                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2284                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2285                         else
2286                                 stored_lft = 0;
2287                         if (!update_lft && !create && stored_lft) {
2288                                 const u32 minimum_lft = min(
2289                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2290                                 valid_lft = max(valid_lft, minimum_lft);
2291
2292                                 /* RFC4862 Section 5.5.3e:
2293                                  * "Note that the preferred lifetime of the
2294                                  *  corresponding address is always reset to
2295                                  *  the Preferred Lifetime in the received
2296                                  *  Prefix Information option, regardless of
2297                                  *  whether the valid lifetime is also reset or
2298                                  *  ignored."
2299                                  *
2300                                  * So we should always update prefered_lft here.
2301                                  */
2302                                 update_lft = 1;
2303                         }
2304
2305                         if (update_lft) {
2306                                 ifp->valid_lft = valid_lft;
2307                                 ifp->prefered_lft = prefered_lft;
2308                                 ifp->tstamp = now;
2309                                 flags = ifp->flags;
2310                                 ifp->flags &= ~IFA_F_DEPRECATED;
2311                                 spin_unlock(&ifp->lock);
2312
2313                                 if (!(flags&IFA_F_TENTATIVE))
2314                                         ipv6_ifa_notify(0, ifp);
2315                         } else
2316                                 spin_unlock(&ifp->lock);
2317
2318                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2319                                          create, now);
2320
2321                         in6_ifa_put(ifp);
2322                         addrconf_verify(0);
2323                 }
2324         }
2325         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2326         in6_dev_put(in6_dev);
2327 }
2328
2329 /*
2330  *      Set destination address.
2331  *      Special case for SIT interfaces where we create a new "virtual"
2332  *      device.
2333  */
2334 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2335 {
2336         struct in6_ifreq ireq;
2337         struct net_device *dev;
2338         int err = -EINVAL;
2339
2340         rtnl_lock();
2341
2342         err = -EFAULT;
2343         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2344                 goto err_exit;
2345
2346         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2347
2348         err = -ENODEV;
2349         if (dev == NULL)
2350                 goto err_exit;
2351
2352 #if IS_ENABLED(CONFIG_IPV6_SIT)
2353         if (dev->type == ARPHRD_SIT) {
2354                 const struct net_device_ops *ops = dev->netdev_ops;
2355                 struct ifreq ifr;
2356                 struct ip_tunnel_parm p;
2357
2358                 err = -EADDRNOTAVAIL;
2359                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2360                         goto err_exit;
2361
2362                 memset(&p, 0, sizeof(p));
2363                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2364                 p.iph.saddr = 0;
2365                 p.iph.version = 4;
2366                 p.iph.ihl = 5;
2367                 p.iph.protocol = IPPROTO_IPV6;
2368                 p.iph.ttl = 64;
2369                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2370
2371                 if (ops->ndo_do_ioctl) {
2372                         mm_segment_t oldfs = get_fs();
2373
2374                         set_fs(KERNEL_DS);
2375                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2376                         set_fs(oldfs);
2377                 } else
2378                         err = -EOPNOTSUPP;
2379
2380                 if (err == 0) {
2381                         err = -ENOBUFS;
2382                         dev = __dev_get_by_name(net, p.name);
2383                         if (!dev)
2384                                 goto err_exit;
2385                         err = dev_open(dev);
2386                 }
2387         }
2388 #endif
2389
2390 err_exit:
2391         rtnl_unlock();
2392         return err;
2393 }
2394
2395 /*
2396  *      Manual configuration of address on an interface
2397  */
2398 static int inet6_addr_add(struct net *net, int ifindex,
2399                           const struct in6_addr *pfx,
2400                           const struct in6_addr *peer_pfx,
2401                           unsigned int plen, __u32 ifa_flags,
2402                           __u32 prefered_lft, __u32 valid_lft)
2403 {
2404         struct inet6_ifaddr *ifp;
2405         struct inet6_dev *idev;
2406         struct net_device *dev;
2407         int scope;
2408         u32 flags;
2409         clock_t expires;
2410         unsigned long timeout;
2411
2412         ASSERT_RTNL();
2413
2414         if (plen > 128)
2415                 return -EINVAL;
2416
2417         /* check the lifetime */
2418         if (!valid_lft || prefered_lft > valid_lft)
2419                 return -EINVAL;
2420
2421         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2422                 return -EINVAL;
2423
2424         dev = __dev_get_by_index(net, ifindex);
2425         if (!dev)
2426                 return -ENODEV;
2427
2428         idev = addrconf_add_dev(dev);
2429         if (IS_ERR(idev))
2430                 return PTR_ERR(idev);
2431
2432         scope = ipv6_addr_scope(pfx);
2433
2434         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2435         if (addrconf_finite_timeout(timeout)) {
2436                 expires = jiffies_to_clock_t(timeout * HZ);
2437                 valid_lft = timeout;
2438                 flags = RTF_EXPIRES;
2439         } else {
2440                 expires = 0;
2441                 flags = 0;
2442                 ifa_flags |= IFA_F_PERMANENT;
2443         }
2444
2445         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2446         if (addrconf_finite_timeout(timeout)) {
2447                 if (timeout == 0)
2448                         ifa_flags |= IFA_F_DEPRECATED;
2449                 prefered_lft = timeout;
2450         }
2451
2452         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2453                             valid_lft, prefered_lft);
2454
2455         if (!IS_ERR(ifp)) {
2456                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2457                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2458                                               expires, flags);
2459                 }
2460
2461                 /*
2462                  * Note that section 3.1 of RFC 4429 indicates
2463                  * that the Optimistic flag should not be set for
2464                  * manually configured addresses
2465                  */
2466                 addrconf_dad_start(ifp);
2467                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2468                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2469                                          true, jiffies);
2470                 in6_ifa_put(ifp);
2471                 addrconf_verify(0);
2472                 return 0;
2473         }
2474
2475         return PTR_ERR(ifp);
2476 }
2477
2478 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2479                           unsigned int plen)
2480 {
2481         struct inet6_ifaddr *ifp;
2482         struct inet6_dev *idev;
2483         struct net_device *dev;
2484
2485         if (plen > 128)
2486                 return -EINVAL;
2487
2488         dev = __dev_get_by_index(net, ifindex);
2489         if (!dev)
2490                 return -ENODEV;
2491
2492         if ((idev = __in6_dev_get(dev)) == NULL)
2493                 return -ENXIO;
2494
2495         read_lock_bh(&idev->lock);
2496         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2497                 if (ifp->prefix_len == plen &&
2498                     ipv6_addr_equal(pfx, &ifp->addr)) {
2499                         in6_ifa_hold(ifp);
2500                         read_unlock_bh(&idev->lock);
2501
2502                         ipv6_del_addr(ifp);
2503                         return 0;
2504                 }
2505         }
2506         read_unlock_bh(&idev->lock);
2507         return -EADDRNOTAVAIL;
2508 }
2509
2510
2511 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2512 {
2513         struct in6_ifreq ireq;
2514         int err;
2515
2516         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2517                 return -EPERM;
2518
2519         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2520                 return -EFAULT;
2521
2522         rtnl_lock();
2523         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2524                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2525                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2526         rtnl_unlock();
2527         return err;
2528 }
2529
2530 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2531 {
2532         struct in6_ifreq ireq;
2533         int err;
2534
2535         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2536                 return -EPERM;
2537
2538         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2539                 return -EFAULT;
2540
2541         rtnl_lock();
2542         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2543                              ireq.ifr6_prefixlen);
2544         rtnl_unlock();
2545         return err;
2546 }
2547
2548 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2549                      int plen, int scope)
2550 {
2551         struct inet6_ifaddr *ifp;
2552
2553         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2554                             scope, IFA_F_PERMANENT,
2555                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2556         if (!IS_ERR(ifp)) {
2557                 spin_lock_bh(&ifp->lock);
2558                 ifp->flags &= ~IFA_F_TENTATIVE;
2559                 spin_unlock_bh(&ifp->lock);
2560                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2561                 in6_ifa_put(ifp);
2562         }
2563 }
2564
2565 #if IS_ENABLED(CONFIG_IPV6_SIT)
2566 static void sit_add_v4_addrs(struct inet6_dev *idev)
2567 {
2568         struct in6_addr addr;
2569         struct net_device *dev;
2570         struct net *net = dev_net(idev->dev);
2571         int scope, plen;
2572         u32 pflags = 0;
2573
2574         ASSERT_RTNL();
2575
2576         memset(&addr, 0, sizeof(struct in6_addr));
2577         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2578
2579         if (idev->dev->flags&IFF_POINTOPOINT) {
2580                 addr.s6_addr32[0] = htonl(0xfe800000);
2581                 scope = IFA_LINK;
2582                 plen = 64;
2583         } else {
2584                 scope = IPV6_ADDR_COMPATv4;
2585                 plen = 96;
2586                 pflags |= RTF_NONEXTHOP;
2587         }
2588
2589         if (addr.s6_addr32[3]) {
2590                 add_addr(idev, &addr, plen, scope);
2591                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2592                 return;
2593         }
2594
2595         for_each_netdev(net, dev) {
2596                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2597                 if (in_dev && (dev->flags & IFF_UP)) {
2598                         struct in_ifaddr *ifa;
2599
2600                         int flag = scope;
2601
2602                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2603
2604                                 addr.s6_addr32[3] = ifa->ifa_local;
2605
2606                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2607                                         continue;
2608                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2609                                         if (idev->dev->flags&IFF_POINTOPOINT)
2610                                                 continue;
2611                                         flag |= IFA_HOST;
2612                                 }
2613
2614                                 add_addr(idev, &addr, plen, flag);
2615                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2616                                                       pflags);
2617                         }
2618                 }
2619         }
2620 }
2621 #endif
2622
2623 static void init_loopback(struct net_device *dev)
2624 {
2625         struct inet6_dev  *idev;
2626         struct net_device *sp_dev;
2627         struct inet6_ifaddr *sp_ifa;
2628         struct rt6_info *sp_rt;
2629
2630         /* ::1 */
2631
2632         ASSERT_RTNL();
2633
2634         if ((idev = ipv6_find_idev(dev)) == NULL) {
2635                 pr_debug("%s: add_dev failed\n", __func__);
2636                 return;
2637         }
2638
2639         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2640
2641         /* Add routes to other interface's IPv6 addresses */
2642         for_each_netdev(dev_net(dev), sp_dev) {
2643                 if (!strcmp(sp_dev->name, dev->name))
2644                         continue;
2645
2646                 idev = __in6_dev_get(sp_dev);
2647                 if (!idev)
2648                         continue;
2649
2650                 read_lock_bh(&idev->lock);
2651                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2652
2653                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2654                                 continue;
2655
2656                         if (sp_ifa->rt)
2657                                 continue;
2658
2659                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2660
2661                         /* Failure cases are ignored */
2662                         if (!IS_ERR(sp_rt)) {
2663                                 sp_ifa->rt = sp_rt;
2664                                 ip6_ins_rt(sp_rt);
2665                         }
2666                 }
2667                 read_unlock_bh(&idev->lock);
2668         }
2669 }
2670
2671 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2672 {
2673         struct inet6_ifaddr *ifp;
2674         u32 addr_flags = IFA_F_PERMANENT;
2675
2676 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2677         if (idev->cnf.optimistic_dad &&
2678             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2679                 addr_flags |= IFA_F_OPTIMISTIC;
2680 #endif
2681
2682
2683         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2684                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2685         if (!IS_ERR(ifp)) {
2686                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2687                 addrconf_dad_start(ifp);
2688                 in6_ifa_put(ifp);
2689         }
2690 }
2691
2692 static void addrconf_dev_config(struct net_device *dev)
2693 {
2694         struct in6_addr addr;
2695         struct inet6_dev *idev;
2696
2697         ASSERT_RTNL();
2698
2699         if ((dev->type != ARPHRD_ETHER) &&
2700             (dev->type != ARPHRD_FDDI) &&
2701             (dev->type != ARPHRD_ARCNET) &&
2702             (dev->type != ARPHRD_INFINIBAND) &&
2703             (dev->type != ARPHRD_IEEE802154) &&
2704             (dev->type != ARPHRD_IEEE1394) &&
2705             (dev->type != ARPHRD_TUNNEL6) &&
2706             (dev->type != ARPHRD_6LOWPAN)) {
2707                 /* Alas, we support only Ethernet autoconfiguration. */
2708                 return;
2709         }
2710
2711         idev = addrconf_add_dev(dev);
2712         if (IS_ERR(idev))
2713                 return;
2714
2715         memset(&addr, 0, sizeof(struct in6_addr));
2716         addr.s6_addr32[0] = htonl(0xFE800000);
2717
2718         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2719                 addrconf_add_linklocal(idev, &addr);
2720 }
2721
2722 #if IS_ENABLED(CONFIG_IPV6_SIT)
2723 static void addrconf_sit_config(struct net_device *dev)
2724 {
2725         struct inet6_dev *idev;
2726
2727         ASSERT_RTNL();
2728
2729         /*
2730          * Configure the tunnel with one of our IPv4
2731          * addresses... we should configure all of
2732          * our v4 addrs in the tunnel
2733          */
2734
2735         if ((idev = ipv6_find_idev(dev)) == NULL) {
2736                 pr_debug("%s: add_dev failed\n", __func__);
2737                 return;
2738         }
2739
2740         if (dev->priv_flags & IFF_ISATAP) {
2741                 struct in6_addr addr;
2742
2743                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2744                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2745                         addrconf_add_linklocal(idev, &addr);
2746                 return;
2747         }
2748
2749         sit_add_v4_addrs(idev);
2750
2751         if (dev->flags&IFF_POINTOPOINT)
2752                 addrconf_add_mroute(dev);
2753 }
2754 #endif
2755
2756 #if IS_ENABLED(CONFIG_NET_IPGRE)
2757 static void addrconf_gre_config(struct net_device *dev)
2758 {
2759         struct inet6_dev *idev;
2760         struct in6_addr addr;
2761
2762         ASSERT_RTNL();
2763
2764         if ((idev = ipv6_find_idev(dev)) == NULL) {
2765                 pr_debug("%s: add_dev failed\n", __func__);
2766                 return;
2767         }
2768
2769         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2770         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2771                 addrconf_add_linklocal(idev, &addr);
2772 }
2773 #endif
2774
2775 static inline int
2776 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2777 {
2778         struct in6_addr lladdr;
2779
2780         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2781                 addrconf_add_linklocal(idev, &lladdr);
2782                 return 0;
2783         }
2784         return -1;
2785 }
2786
2787 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2788                            void *ptr)
2789 {
2790         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2791         struct inet6_dev *idev = __in6_dev_get(dev);
2792         int run_pending = 0;
2793         int err;
2794
2795         switch (event) {
2796         case NETDEV_REGISTER:
2797                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2798                         idev = ipv6_add_dev(dev);
2799                         if (!idev)
2800                                 return notifier_from_errno(-ENOMEM);
2801                 }
2802                 break;
2803
2804         case NETDEV_UP:
2805         case NETDEV_CHANGE:
2806                 if (dev->flags & IFF_SLAVE)
2807                         break;
2808
2809                 if (event == NETDEV_UP) {
2810                         if (!addrconf_qdisc_ok(dev)) {
2811                                 /* device is not ready yet. */
2812                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2813                                         dev->name);
2814                                 break;
2815                         }
2816
2817                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2818                                 idev = ipv6_add_dev(dev);
2819
2820                         if (idev) {
2821                                 idev->if_flags |= IF_READY;
2822                                 run_pending = 1;
2823                         }
2824                 } else {
2825                         if (!addrconf_qdisc_ok(dev)) {
2826                                 /* device is still not ready. */
2827                                 break;
2828                         }
2829
2830                         if (idev) {
2831                                 if (idev->if_flags & IF_READY)
2832                                         /* device is already configured. */
2833                                         break;
2834                                 idev->if_flags |= IF_READY;
2835                         }
2836
2837                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2838                                 dev->name);
2839
2840                         run_pending = 1;
2841                 }
2842
2843                 switch (dev->type) {
2844 #if IS_ENABLED(CONFIG_IPV6_SIT)
2845                 case ARPHRD_SIT:
2846                         addrconf_sit_config(dev);
2847                         break;
2848 #endif
2849 #if IS_ENABLED(CONFIG_NET_IPGRE)
2850                 case ARPHRD_IPGRE:
2851                         addrconf_gre_config(dev);
2852                         break;
2853 #endif
2854                 case ARPHRD_LOOPBACK:
2855                         init_loopback(dev);
2856                         break;
2857
2858                 default:
2859                         addrconf_dev_config(dev);
2860                         break;
2861                 }
2862
2863                 if (idev) {
2864                         if (run_pending)
2865                                 addrconf_dad_run(idev);
2866
2867                         /*
2868                          * If the MTU changed during the interface down,
2869                          * when the interface up, the changed MTU must be
2870                          * reflected in the idev as well as routers.
2871                          */
2872                         if (idev->cnf.mtu6 != dev->mtu &&
2873                             dev->mtu >= IPV6_MIN_MTU) {
2874                                 rt6_mtu_change(dev, dev->mtu);
2875                                 idev->cnf.mtu6 = dev->mtu;
2876                         }
2877                         idev->tstamp = jiffies;
2878                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2879
2880                         /*
2881                          * If the changed mtu during down is lower than
2882                          * IPV6_MIN_MTU stop IPv6 on this interface.
2883                          */
2884                         if (dev->mtu < IPV6_MIN_MTU)
2885                                 addrconf_ifdown(dev, 1);
2886                 }
2887                 break;
2888
2889         case NETDEV_CHANGEMTU:
2890                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2891                         rt6_mtu_change(dev, dev->mtu);
2892                         idev->cnf.mtu6 = dev->mtu;
2893                         break;
2894                 }
2895
2896                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2897                         idev = ipv6_add_dev(dev);
2898                         if (idev)
2899                                 break;
2900                 }
2901
2902                 /*
2903                  * if MTU under IPV6_MIN_MTU.
2904                  * Stop IPv6 on this interface.
2905                  */
2906
2907         case NETDEV_DOWN:
2908         case NETDEV_UNREGISTER:
2909                 /*
2910                  *      Remove all addresses from this interface.
2911                  */
2912                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2913                 break;
2914
2915         case NETDEV_CHANGENAME:
2916                 if (idev) {
2917                         snmp6_unregister_dev(idev);
2918                         addrconf_sysctl_unregister(idev);
2919                         addrconf_sysctl_register(idev);
2920                         err = snmp6_register_dev(idev);
2921                         if (err)
2922                                 return notifier_from_errno(err);
2923                 }
2924                 break;
2925
2926         case NETDEV_PRE_TYPE_CHANGE:
2927         case NETDEV_POST_TYPE_CHANGE:
2928                 addrconf_type_change(dev, event);
2929                 break;
2930         }
2931
2932         return NOTIFY_OK;
2933 }
2934
2935 /*
2936  *      addrconf module should be notified of a device going up
2937  */
2938 static struct notifier_block ipv6_dev_notf = {
2939         .notifier_call = addrconf_notify,
2940 };
2941
2942 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2943 {
2944         struct inet6_dev *idev;
2945         ASSERT_RTNL();
2946
2947         idev = __in6_dev_get(dev);
2948
2949         if (event == NETDEV_POST_TYPE_CHANGE)
2950                 ipv6_mc_remap(idev);
2951         else if (event == NETDEV_PRE_TYPE_CHANGE)
2952                 ipv6_mc_unmap(idev);
2953 }
2954
2955 static int addrconf_ifdown(struct net_device *dev, int how)
2956 {
2957         struct net *net = dev_net(dev);
2958         struct inet6_dev *idev;
2959         struct inet6_ifaddr *ifa;
2960         int state, i;
2961
2962         ASSERT_RTNL();
2963
2964         rt6_ifdown(net, dev);
2965         neigh_ifdown(&nd_tbl, dev);
2966
2967         idev = __in6_dev_get(dev);
2968         if (idev == NULL)
2969                 return -ENODEV;
2970
2971         /*
2972          * Step 1: remove reference to ipv6 device from parent device.
2973          *         Do not dev_put!
2974          */
2975         if (how) {
2976                 idev->dead = 1;
2977
2978                 /* protected by rtnl_lock */
2979                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2980
2981                 /* Step 1.5: remove snmp6 entry */
2982                 snmp6_unregister_dev(idev);
2983
2984         }
2985
2986         /* Step 2: clear hash table */
2987         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2988                 struct hlist_head *h = &inet6_addr_lst[i];
2989
2990                 spin_lock_bh(&addrconf_hash_lock);
2991         restart:
2992                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2993                         if (ifa->idev == idev) {
2994                                 hlist_del_init_rcu(&ifa->addr_lst);
2995                                 addrconf_del_dad_timer(ifa);
2996                                 goto restart;
2997                         }
2998                 }
2999                 spin_unlock_bh(&addrconf_hash_lock);
3000         }
3001
3002         write_lock_bh(&idev->lock);
3003
3004         addrconf_del_rs_timer(idev);
3005
3006         /* Step 2: clear flags for stateless addrconf */
3007         if (!how)
3008                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3009
3010         if (how && del_timer(&idev->regen_timer))
3011                 in6_dev_put(idev);
3012
3013         /* Step 3: clear tempaddr list */
3014         while (!list_empty(&idev->tempaddr_list)) {
3015                 ifa = list_first_entry(&idev->tempaddr_list,
3016                                        struct inet6_ifaddr, tmp_list);
3017                 list_del(&ifa->tmp_list);
3018                 write_unlock_bh(&idev->lock);
3019                 spin_lock_bh(&ifa->lock);
3020
3021                 if (ifa->ifpub) {
3022                         in6_ifa_put(ifa->ifpub);
3023                         ifa->ifpub = NULL;
3024                 }
3025                 spin_unlock_bh(&ifa->lock);
3026                 in6_ifa_put(ifa);
3027                 write_lock_bh(&idev->lock);
3028         }
3029
3030         while (!list_empty(&idev->addr_list)) {
3031                 ifa = list_first_entry(&idev->addr_list,
3032                                        struct inet6_ifaddr, if_list);
3033                 addrconf_del_dad_timer(ifa);
3034
3035                 list_del(&ifa->if_list);
3036
3037                 write_unlock_bh(&idev->lock);
3038
3039                 spin_lock_bh(&ifa->state_lock);
3040                 state = ifa->state;
3041                 ifa->state = INET6_IFADDR_STATE_DEAD;
3042                 spin_unlock_bh(&ifa->state_lock);
3043
3044                 if (state != INET6_IFADDR_STATE_DEAD) {
3045                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3046                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3047                 }
3048                 in6_ifa_put(ifa);
3049
3050                 write_lock_bh(&idev->lock);
3051         }
3052
3053         write_unlock_bh(&idev->lock);
3054
3055         /* Step 5: Discard multicast list */
3056         if (how)
3057                 ipv6_mc_destroy_dev(idev);
3058         else
3059                 ipv6_mc_down(idev);
3060
3061         idev->tstamp = jiffies;
3062
3063         /* Last: Shot the device (if unregistered) */
3064         if (how) {
3065                 addrconf_sysctl_unregister(idev);
3066                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3067                 neigh_ifdown(&nd_tbl, dev);
3068                 in6_dev_put(idev);
3069         }
3070         return 0;
3071 }
3072
3073 static void addrconf_rs_timer(unsigned long data)
3074 {
3075         struct inet6_dev *idev = (struct inet6_dev *)data;
3076         struct net_device *dev = idev->dev;
3077         struct in6_addr lladdr;
3078
3079         write_lock(&idev->lock);
3080         if (idev->dead || !(idev->if_flags & IF_READY))
3081                 goto out;
3082
3083         if (!ipv6_accept_ra(idev))
3084                 goto out;
3085
3086         /* Announcement received after solicitation was sent */
3087         if (idev->if_flags & IF_RA_RCVD)
3088                 goto out;
3089
3090         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3091                 write_unlock(&idev->lock);
3092                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3093                         ndisc_send_rs(dev, &lladdr,
3094                                       &in6addr_linklocal_allrouters);
3095                 else
3096                         goto put;
3097
3098                 write_lock(&idev->lock);
3099                 /* The wait after the last probe can be shorter */
3100                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3101                                              idev->cnf.rtr_solicits) ?
3102                                       idev->cnf.rtr_solicit_delay :
3103                                       idev->cnf.rtr_solicit_interval);
3104         } else {
3105                 /*
3106                  * Note: we do not support deprecated "all on-link"
3107                  * assumption any longer.
3108                  */
3109                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3110         }
3111
3112 out:
3113         write_unlock(&idev->lock);
3114 put:
3115         in6_dev_put(idev);
3116 }
3117
3118 /*
3119  *      Duplicate Address Detection
3120  */
3121 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3122 {
3123         unsigned long rand_num;
3124         struct inet6_dev *idev = ifp->idev;
3125
3126         if (ifp->flags & IFA_F_OPTIMISTIC)
3127                 rand_num = 0;
3128         else
3129                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3130
3131         ifp->dad_probes = idev->cnf.dad_transmits;
3132         addrconf_mod_dad_timer(ifp, rand_num);
3133 }
3134
3135 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3136 {
3137         struct inet6_dev *idev = ifp->idev;
3138         struct net_device *dev = idev->dev;
3139
3140         addrconf_join_solict(dev, &ifp->addr);
3141
3142         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3143
3144         read_lock_bh(&idev->lock);
3145         spin_lock(&ifp->lock);
3146         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3147                 goto out;
3148
3149         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3150             idev->cnf.accept_dad < 1 ||
3151             !(ifp->flags&IFA_F_TENTATIVE) ||
3152             ifp->flags & IFA_F_NODAD) {
3153                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3154                 spin_unlock(&ifp->lock);
3155                 read_unlock_bh(&idev->lock);
3156
3157                 addrconf_dad_completed(ifp);
3158                 return;
3159         }
3160
3161         if (!(idev->if_flags & IF_READY)) {
3162                 spin_unlock(&ifp->lock);
3163                 read_unlock_bh(&idev->lock);
3164                 /*
3165                  * If the device is not ready:
3166                  * - keep it tentative if it is a permanent address.
3167                  * - otherwise, kill it.
3168                  */
3169                 in6_ifa_hold(ifp);
3170                 addrconf_dad_stop(ifp, 0);
3171                 return;
3172         }
3173
3174         /*
3175          * Optimistic nodes can start receiving
3176          * Frames right away
3177          */
3178         if (ifp->flags & IFA_F_OPTIMISTIC)
3179                 ip6_ins_rt(ifp->rt);
3180
3181         addrconf_dad_kick(ifp);
3182 out:
3183         spin_unlock(&ifp->lock);
3184         read_unlock_bh(&idev->lock);
3185 }
3186
3187 static void addrconf_dad_timer(unsigned long data)
3188 {
3189         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3190         struct inet6_dev *idev = ifp->idev;
3191         struct in6_addr mcaddr;
3192
3193         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3194                 goto out;
3195
3196         write_lock(&idev->lock);
3197         if (idev->dead || !(idev->if_flags & IF_READY)) {
3198                 write_unlock(&idev->lock);
3199                 goto out;
3200         }
3201
3202         spin_lock(&ifp->lock);
3203         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3204                 spin_unlock(&ifp->lock);
3205                 write_unlock(&idev->lock);
3206                 goto out;
3207         }
3208
3209         if (ifp->dad_probes == 0) {
3210                 /*
3211                  * DAD was successful
3212                  */
3213
3214                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3215                 spin_unlock(&ifp->lock);
3216                 write_unlock(&idev->lock);
3217
3218                 addrconf_dad_completed(ifp);
3219
3220                 goto out;
3221         }
3222
3223         ifp->dad_probes--;
3224         addrconf_mod_dad_timer(ifp,
3225                                NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3226         spin_unlock(&ifp->lock);
3227         write_unlock(&idev->lock);
3228
3229         /* send a neighbour solicitation for our addr */
3230         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3231         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3232 out:
3233         in6_ifa_put(ifp);
3234 }
3235
3236 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3237 {
3238         struct net_device *dev = ifp->idev->dev;
3239         struct in6_addr lladdr;
3240         bool send_rs, send_mld;
3241
3242         addrconf_del_dad_timer(ifp);
3243
3244         /*
3245          *      Configure the address for reception. Now it is valid.
3246          */
3247
3248         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3249
3250         /* If added prefix is link local and we are prepared to process
3251            router advertisements, start sending router solicitations.
3252          */
3253
3254         read_lock_bh(&ifp->idev->lock);
3255         spin_lock(&ifp->lock);
3256         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3257                    ifp->idev->valid_ll_addr_cnt == 1;
3258         send_rs = send_mld &&
3259                   ipv6_accept_ra(ifp->idev) &&
3260                   ifp->idev->cnf.rtr_solicits > 0 &&
3261                   (dev->flags&IFF_LOOPBACK) == 0;
3262         spin_unlock(&ifp->lock);
3263         read_unlock_bh(&ifp->idev->lock);
3264
3265         /* While dad is in progress mld report's source address is in6_addrany.
3266          * Resend with proper ll now.
3267          */
3268         if (send_mld)
3269                 ipv6_mc_dad_complete(ifp->idev);
3270
3271         if (send_rs) {
3272                 /*
3273                  *      If a host as already performed a random delay
3274                  *      [...] as part of DAD [...] there is no need
3275                  *      to delay again before sending the first RS
3276                  */
3277                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3278                         return;
3279                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3280
3281                 write_lock_bh(&ifp->idev->lock);
3282                 spin_lock(&ifp->lock);
3283                 ifp->idev->rs_probes = 1;
3284                 ifp->idev->if_flags |= IF_RS_SENT;
3285                 addrconf_mod_rs_timer(ifp->idev,
3286                                       ifp->idev->cnf.rtr_solicit_interval);
3287                 spin_unlock(&ifp->lock);
3288                 write_unlock_bh(&ifp->idev->lock);
3289         }
3290 }
3291
3292 static void addrconf_dad_run(struct inet6_dev *idev)
3293 {
3294         struct inet6_ifaddr *ifp;
3295
3296         read_lock_bh(&idev->lock);
3297         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3298                 spin_lock(&ifp->lock);
3299                 if (ifp->flags & IFA_F_TENTATIVE &&
3300                     ifp->state == INET6_IFADDR_STATE_DAD)
3301                         addrconf_dad_kick(ifp);
3302                 spin_unlock(&ifp->lock);
3303         }
3304         read_unlock_bh(&idev->lock);
3305 }
3306
3307 #ifdef CONFIG_PROC_FS
3308 struct if6_iter_state {
3309         struct seq_net_private p;
3310         int bucket;
3311         int offset;
3312 };
3313
3314 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3315 {
3316         struct inet6_ifaddr *ifa = NULL;
3317         struct if6_iter_state *state = seq->private;
3318         struct net *net = seq_file_net(seq);
3319         int p = 0;
3320
3321         /* initial bucket if pos is 0 */
3322         if (pos == 0) {
3323                 state->bucket = 0;
3324                 state->offset = 0;
3325         }
3326
3327         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3328                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3329                                          addr_lst) {
3330                         if (!net_eq(dev_net(ifa->idev->dev), net))
3331                                 continue;
3332                         /* sync with offset */
3333                         if (p < state->offset) {
3334                                 p++;
3335                                 continue;
3336                         }
3337                         state->offset++;
3338                         return ifa;
3339                 }
3340
3341                 /* prepare for next bucket */
3342                 state->offset = 0;
3343                 p = 0;
3344         }
3345         return NULL;
3346 }
3347
3348 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3349                                          struct inet6_ifaddr *ifa)
3350 {
3351         struct if6_iter_state *state = seq->private;
3352         struct net *net = seq_file_net(seq);
3353
3354         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3355                 if (!net_eq(dev_net(ifa->idev->dev), net))
3356                         continue;
3357                 state->offset++;
3358                 return ifa;
3359         }
3360
3361         while (++state->bucket < IN6_ADDR_HSIZE) {
3362                 state->offset = 0;
3363                 hlist_for_each_entry_rcu_bh(ifa,
3364                                      &inet6_addr_lst[state->bucket], addr_lst) {
3365                         if (!net_eq(dev_net(ifa->idev->dev), net))
3366                                 continue;
3367                         state->offset++;
3368                         return ifa;
3369                 }
3370         }
3371
3372         return NULL;
3373 }
3374
3375 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3376         __acquires(rcu_bh)
3377 {
3378         rcu_read_lock_bh();
3379         return if6_get_first(seq, *pos);
3380 }
3381
3382 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3383 {
3384         struct inet6_ifaddr *ifa;
3385
3386         ifa = if6_get_next(seq, v);
3387         ++*pos;
3388         return ifa;
3389 }
3390
3391 static void if6_seq_stop(struct seq_file *seq, void *v)
3392         __releases(rcu_bh)
3393 {
3394         rcu_read_unlock_bh();
3395 }
3396
3397 static int if6_seq_show(struct seq_file *seq, void *v)
3398 {
3399         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3400         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3401                    &ifp->addr,
3402                    ifp->idev->dev->ifindex,
3403                    ifp->prefix_len,
3404                    ifp->scope,
3405                    (u8) ifp->flags,
3406                    ifp->idev->dev->name);
3407         return 0;
3408 }
3409
3410 static const struct seq_operations if6_seq_ops = {
3411         .start  = if6_seq_start,
3412         .next   = if6_seq_next,
3413         .show   = if6_seq_show,
3414         .stop   = if6_seq_stop,
3415 };
3416
3417 static int if6_seq_open(struct inode *inode, struct file *file)
3418 {
3419         return seq_open_net(inode, file, &if6_seq_ops,
3420                             sizeof(struct if6_iter_state));
3421 }
3422
3423 static const struct file_operations if6_fops = {
3424         .owner          = THIS_MODULE,
3425         .open           = if6_seq_open,
3426         .read           = seq_read,
3427         .llseek         = seq_lseek,
3428         .release        = seq_release_net,
3429 };
3430
3431 static int __net_init if6_proc_net_init(struct net *net)
3432 {
3433         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3434                 return -ENOMEM;
3435         return 0;
3436 }
3437
3438 static void __net_exit if6_proc_net_exit(struct net *net)
3439 {
3440         remove_proc_entry("if_inet6", net->proc_net);
3441 }
3442
3443 static struct pernet_operations if6_proc_net_ops = {
3444        .init = if6_proc_net_init,
3445        .exit = if6_proc_net_exit,
3446 };
3447
3448 int __init if6_proc_init(void)
3449 {
3450         return register_pernet_subsys(&if6_proc_net_ops);
3451 }
3452
3453 void if6_proc_exit(void)
3454 {
3455         unregister_pernet_subsys(&if6_proc_net_ops);
3456 }
3457 #endif  /* CONFIG_PROC_FS */
3458
3459 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3460 /* Check if address is a home address configured on any interface. */
3461 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3462 {
3463         int ret = 0;
3464         struct inet6_ifaddr *ifp = NULL;
3465         unsigned int hash = inet6_addr_hash(addr);
3466
3467         rcu_read_lock_bh();
3468         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3469                 if (!net_eq(dev_net(ifp->idev->dev), net))
3470                         continue;
3471                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3472                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3473                         ret = 1;
3474                         break;
3475                 }
3476         }
3477         rcu_read_unlock_bh();
3478         return ret;
3479 }
3480 #endif
3481
3482 /*
3483  *      Periodic address status verification
3484  */
3485
3486 static void addrconf_verify(unsigned long foo)
3487 {
3488         unsigned long now, next, next_sec, next_sched;
3489         struct inet6_ifaddr *ifp;
3490         int i;
3491
3492         rcu_read_lock_bh();
3493         spin_lock(&addrconf_verify_lock);
3494         now = jiffies;
3495         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3496
3497         del_timer(&addr_chk_timer);
3498
3499         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3500 restart:
3501                 hlist_for_each_entry_rcu_bh(ifp,
3502                                          &inet6_addr_lst[i], addr_lst) {
3503                         unsigned long age;
3504
3505                         /* When setting preferred_lft to a value not zero or
3506                          * infinity, while valid_lft is infinity
3507                          * IFA_F_PERMANENT has a non-infinity life time.
3508                          */
3509                         if ((ifp->flags & IFA_F_PERMANENT) &&
3510                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3511                                 continue;
3512
3513                         spin_lock(&ifp->lock);
3514                         /* We try to batch several events at once. */
3515                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3516
3517                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3518                             age >= ifp->valid_lft) {
3519                                 spin_unlock(&ifp->lock);
3520                                 in6_ifa_hold(ifp);
3521                                 ipv6_del_addr(ifp);
3522                                 goto restart;
3523                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3524                                 spin_unlock(&ifp->lock);
3525                                 continue;
3526                         } else if (age >= ifp->prefered_lft) {
3527                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3528                                 int deprecate = 0;
3529
3530                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3531                                         deprecate = 1;
3532                                         ifp->flags |= IFA_F_DEPRECATED;
3533                                 }
3534
3535                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3536                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3537                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3538
3539                                 spin_unlock(&ifp->lock);
3540
3541                                 if (deprecate) {
3542                                         in6_ifa_hold(ifp);
3543
3544                                         ipv6_ifa_notify(0, ifp);
3545                                         in6_ifa_put(ifp);
3546                                         goto restart;
3547                                 }
3548                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3549                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3550                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3551                                         ifp->idev->cnf.dad_transmits *
3552                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3553
3554                                 if (age >= ifp->prefered_lft - regen_advance) {
3555                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3556                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3557                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3558                                         if (!ifp->regen_count && ifpub) {
3559                                                 ifp->regen_count++;
3560                                                 in6_ifa_hold(ifp);
3561                                                 in6_ifa_hold(ifpub);
3562                                                 spin_unlock(&ifp->lock);
3563
3564                                                 spin_lock(&ifpub->lock);
3565                                                 ifpub->regen_count = 0;
3566                                                 spin_unlock(&ifpub->lock);
3567                                                 ipv6_create_tempaddr(ifpub, ifp);
3568                                                 in6_ifa_put(ifpub);
3569                                                 in6_ifa_put(ifp);
3570                                                 goto restart;
3571                                         }
3572                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3573                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3574                                 spin_unlock(&ifp->lock);
3575                         } else {
3576                                 /* ifp->prefered_lft <= ifp->valid_lft */
3577                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3578                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3579                                 spin_unlock(&ifp->lock);
3580                         }
3581                 }
3582         }
3583
3584         next_sec = round_jiffies_up(next);
3585         next_sched = next;
3586
3587         /* If rounded timeout is accurate enough, accept it. */
3588         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3589                 next_sched = next_sec;
3590
3591         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3592         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3593                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3594
3595         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3596               now, next, next_sec, next_sched);
3597
3598         addr_chk_timer.expires = next_sched;
3599         add_timer(&addr_chk_timer);
3600         spin_unlock(&addrconf_verify_lock);
3601         rcu_read_unlock_bh();
3602 }
3603
3604 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3605                                      struct in6_addr **peer_pfx)
3606 {
3607         struct in6_addr *pfx = NULL;
3608
3609         *peer_pfx = NULL;
3610
3611         if (addr)
3612                 pfx = nla_data(addr);
3613
3614         if (local) {
3615                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3616                         *peer_pfx = pfx;
3617                 pfx = nla_data(local);
3618         }
3619
3620         return pfx;
3621 }
3622
3623 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3624         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3625         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3626         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3627         [IFA_FLAGS]             = { .len = sizeof(u32) },
3628 };
3629
3630 static int
3631 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3632 {
3633         struct net *net = sock_net(skb->sk);
3634         struct ifaddrmsg *ifm;
3635         struct nlattr *tb[IFA_MAX+1];
3636         struct in6_addr *pfx, *peer_pfx;
3637         int err;
3638
3639         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3640         if (err < 0)
3641                 return err;
3642
3643         ifm = nlmsg_data(nlh);
3644         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3645         if (pfx == NULL)
3646                 return -EINVAL;
3647
3648         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3649 }
3650
3651 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3652                              u32 prefered_lft, u32 valid_lft)
3653 {
3654         u32 flags;
3655         clock_t expires;
3656         unsigned long timeout;
3657         bool was_managetempaddr;
3658         bool had_prefixroute;
3659
3660         if (!valid_lft || (prefered_lft > valid_lft))
3661                 return -EINVAL;
3662
3663         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3664             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3665                 return -EINVAL;
3666
3667         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3668         if (addrconf_finite_timeout(timeout)) {
3669                 expires = jiffies_to_clock_t(timeout * HZ);
3670                 valid_lft = timeout;
3671                 flags = RTF_EXPIRES;
3672         } else {
3673                 expires = 0;
3674                 flags = 0;
3675                 ifa_flags |= IFA_F_PERMANENT;
3676         }
3677
3678         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3679         if (addrconf_finite_timeout(timeout)) {
3680                 if (timeout == 0)
3681                         ifa_flags |= IFA_F_DEPRECATED;
3682                 prefered_lft = timeout;
3683         }
3684
3685         spin_lock_bh(&ifp->lock);
3686         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3687         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3688                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
3689         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3690                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3691                         IFA_F_NOPREFIXROUTE);
3692         ifp->flags |= ifa_flags;
3693         ifp->tstamp = jiffies;
3694         ifp->valid_lft = valid_lft;
3695         ifp->prefered_lft = prefered_lft;
3696
3697         spin_unlock_bh(&ifp->lock);
3698         if (!(ifp->flags&IFA_F_TENTATIVE))
3699                 ipv6_ifa_notify(0, ifp);
3700
3701         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3702                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3703                                       expires, flags);
3704         } else if (had_prefixroute) {
3705                 enum cleanup_prefix_rt_t action;
3706                 unsigned long rt_expires;
3707
3708                 write_lock_bh(&ifp->idev->lock);
3709                 action = check_cleanup_prefix_route(ifp, &rt_expires);
3710                 write_unlock_bh(&ifp->idev->lock);
3711
3712                 if (action != CLEANUP_PREFIX_RT_NOP) {
3713                         cleanup_prefix_route(ifp, rt_expires,
3714                                 action == CLEANUP_PREFIX_RT_DEL);
3715                 }
3716         }
3717
3718         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3719                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3720                         valid_lft = prefered_lft = 0;
3721                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3722                                  !was_managetempaddr, jiffies);
3723         }
3724
3725         addrconf_verify(0);
3726
3727         return 0;
3728 }
3729
3730 static int
3731 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3732 {
3733         struct net *net = sock_net(skb->sk);
3734         struct ifaddrmsg *ifm;
3735         struct nlattr *tb[IFA_MAX+1];
3736         struct in6_addr *pfx, *peer_pfx;
3737         struct inet6_ifaddr *ifa;
3738         struct net_device *dev;
3739         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3740         u32 ifa_flags;
3741         int err;
3742
3743         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3744         if (err < 0)
3745                 return err;
3746
3747         ifm = nlmsg_data(nlh);
3748         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3749         if (pfx == NULL)
3750                 return -EINVAL;
3751
3752         if (tb[IFA_CACHEINFO]) {
3753                 struct ifa_cacheinfo *ci;
3754
3755                 ci = nla_data(tb[IFA_CACHEINFO]);
3756                 valid_lft = ci->ifa_valid;
3757                 preferred_lft = ci->ifa_prefered;
3758         } else {
3759                 preferred_lft = INFINITY_LIFE_TIME;
3760                 valid_lft = INFINITY_LIFE_TIME;
3761         }
3762
3763         dev =  __dev_get_by_index(net, ifm->ifa_index);
3764         if (dev == NULL)
3765                 return -ENODEV;
3766
3767         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3768
3769         /* We ignore other flags so far. */
3770         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3771                      IFA_F_NOPREFIXROUTE;
3772
3773         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3774         if (ifa == NULL) {
3775                 /*
3776                  * It would be best to check for !NLM_F_CREATE here but
3777                  * userspace already relies on not having to provide this.
3778                  */
3779                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3780                                       ifm->ifa_prefixlen, ifa_flags,
3781                                       preferred_lft, valid_lft);
3782         }
3783
3784         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3785             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3786                 err = -EEXIST;
3787         else
3788                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3789
3790         in6_ifa_put(ifa);
3791
3792         return err;
3793 }
3794
3795 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3796                           u8 scope, int ifindex)
3797 {
3798         struct ifaddrmsg *ifm;
3799
3800         ifm = nlmsg_data(nlh);
3801         ifm->ifa_family = AF_INET6;
3802         ifm->ifa_prefixlen = prefixlen;
3803         ifm->ifa_flags = flags;
3804         ifm->ifa_scope = scope;
3805         ifm->ifa_index = ifindex;
3806 }
3807
3808 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3809                          unsigned long tstamp, u32 preferred, u32 valid)
3810 {
3811         struct ifa_cacheinfo ci;
3812
3813         ci.cstamp = cstamp_delta(cstamp);
3814         ci.tstamp = cstamp_delta(tstamp);
3815         ci.ifa_prefered = preferred;
3816         ci.ifa_valid = valid;
3817
3818         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3819 }
3820
3821 static inline int rt_scope(int ifa_scope)
3822 {
3823         if (ifa_scope & IFA_HOST)
3824                 return RT_SCOPE_HOST;
3825         else if (ifa_scope & IFA_LINK)
3826                 return RT_SCOPE_LINK;
3827         else if (ifa_scope & IFA_SITE)
3828                 return RT_SCOPE_SITE;
3829         else
3830                 return RT_SCOPE_UNIVERSE;
3831 }
3832
3833 static inline int inet6_ifaddr_msgsize(void)
3834 {
3835         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3836                + nla_total_size(16) /* IFA_LOCAL */
3837                + nla_total_size(16) /* IFA_ADDRESS */
3838                + nla_total_size(sizeof(struct ifa_cacheinfo))
3839                + nla_total_size(4)  /* IFA_FLAGS */;
3840 }
3841
3842 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3843                              u32 portid, u32 seq, int event, unsigned int flags)
3844 {
3845         struct nlmsghdr  *nlh;
3846         u32 preferred, valid;
3847
3848         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3849         if (nlh == NULL)
3850                 return -EMSGSIZE;
3851
3852         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3853                       ifa->idev->dev->ifindex);
3854
3855         if (!((ifa->flags&IFA_F_PERMANENT) &&
3856               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3857                 preferred = ifa->prefered_lft;
3858                 valid = ifa->valid_lft;
3859                 if (preferred != INFINITY_LIFE_TIME) {
3860                         long tval = (jiffies - ifa->tstamp)/HZ;
3861                         if (preferred > tval)
3862                                 preferred -= tval;
3863                         else
3864                                 preferred = 0;
3865                         if (valid != INFINITY_LIFE_TIME) {
3866                                 if (valid > tval)
3867                                         valid -= tval;
3868                                 else
3869                                         valid = 0;
3870                         }
3871                 }
3872         } else {
3873                 preferred = INFINITY_LIFE_TIME;
3874                 valid = INFINITY_LIFE_TIME;
3875         }
3876
3877         if (!ipv6_addr_any(&ifa->peer_addr)) {
3878                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3879                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3880                         goto error;
3881         } else
3882                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3883                         goto error;
3884
3885         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3886                 goto error;
3887
3888         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
3889                 goto error;
3890
3891         return nlmsg_end(skb, nlh);
3892
3893 error:
3894         nlmsg_cancel(skb, nlh);
3895         return -EMSGSIZE;
3896 }
3897
3898 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3899                                 u32 portid, u32 seq, int event, u16 flags)
3900 {
3901         struct nlmsghdr  *nlh;
3902         u8 scope = RT_SCOPE_UNIVERSE;
3903         int ifindex = ifmca->idev->dev->ifindex;
3904
3905         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3906                 scope = RT_SCOPE_SITE;
3907
3908         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3909         if (nlh == NULL)
3910                 return -EMSGSIZE;
3911
3912         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3913         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3914             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3915                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3916                 nlmsg_cancel(skb, nlh);
3917                 return -EMSGSIZE;
3918         }
3919
3920         return nlmsg_end(skb, nlh);
3921 }
3922
3923 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3924                                 u32 portid, u32 seq, int event, unsigned int flags)
3925 {
3926         struct nlmsghdr  *nlh;
3927         u8 scope = RT_SCOPE_UNIVERSE;
3928         int ifindex = ifaca->aca_idev->dev->ifindex;
3929
3930         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3931                 scope = RT_SCOPE_SITE;
3932
3933         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3934         if (nlh == NULL)
3935                 return -EMSGSIZE;
3936
3937         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3938         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3939             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3940                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3941                 nlmsg_cancel(skb, nlh);
3942                 return -EMSGSIZE;
3943         }
3944
3945         return nlmsg_end(skb, nlh);
3946 }
3947
3948 enum addr_type_t {
3949         UNICAST_ADDR,
3950         MULTICAST_ADDR,
3951         ANYCAST_ADDR,
3952 };
3953
3954 /* called with rcu_read_lock() */
3955 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3956                           struct netlink_callback *cb, enum addr_type_t type,
3957                           int s_ip_idx, int *p_ip_idx)
3958 {
3959         struct ifmcaddr6 *ifmca;
3960         struct ifacaddr6 *ifaca;
3961         int err = 1;
3962         int ip_idx = *p_ip_idx;
3963
3964         read_lock_bh(&idev->lock);
3965         switch (type) {
3966         case UNICAST_ADDR: {
3967                 struct inet6_ifaddr *ifa;
3968
3969                 /* unicast address incl. temp addr */
3970                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3971                         if (++ip_idx < s_ip_idx)
3972                                 continue;
3973                         err = inet6_fill_ifaddr(skb, ifa,
3974                                                 NETLINK_CB(cb->skb).portid,
3975                                                 cb->nlh->nlmsg_seq,
3976                                                 RTM_NEWADDR,
3977                                                 NLM_F_MULTI);
3978                         if (err <= 0)
3979                                 break;
3980                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3981                 }
3982                 break;
3983         }
3984         case MULTICAST_ADDR:
3985                 /* multicast address */
3986                 for (ifmca = idev->mc_list; ifmca;
3987                      ifmca = ifmca->next, ip_idx++) {
3988                         if (ip_idx < s_ip_idx)
3989                                 continue;
3990                         err = inet6_fill_ifmcaddr(skb, ifmca,
3991                                                   NETLINK_CB(cb->skb).portid,
3992                                                   cb->nlh->nlmsg_seq,
3993                                                   RTM_GETMULTICAST,
3994                                                   NLM_F_MULTI);
3995                         if (err <= 0)
3996                                 break;
3997                 }
3998                 break;
3999         case ANYCAST_ADDR:
4000                 /* anycast address */
4001                 for (ifaca = idev->ac_list; ifaca;
4002                      ifaca = ifaca->aca_next, ip_idx++) {
4003                         if (ip_idx < s_ip_idx)
4004                                 continue;
4005                         err = inet6_fill_ifacaddr(skb, ifaca,
4006                                                   NETLINK_CB(cb->skb).portid,
4007                                                   cb->nlh->nlmsg_seq,
4008                                                   RTM_GETANYCAST,
4009                                                   NLM_F_MULTI);
4010                         if (err <= 0)
4011                                 break;
4012                 }
4013                 break;
4014         default:
4015                 break;
4016         }
4017         read_unlock_bh(&idev->lock);
4018         *p_ip_idx = ip_idx;
4019         return err;
4020 }
4021
4022 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4023                            enum addr_type_t type)
4024 {
4025         struct net *net = sock_net(skb->sk);
4026         int h, s_h;
4027         int idx, ip_idx;
4028         int s_idx, s_ip_idx;
4029         struct net_device *dev;
4030         struct inet6_dev *idev;
4031         struct hlist_head *head;
4032
4033         s_h = cb->args[0];
4034         s_idx = idx = cb->args[1];
4035         s_ip_idx = ip_idx = cb->args[2];
4036
4037         rcu_read_lock();
4038         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4039         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4040                 idx = 0;
4041                 head = &net->dev_index_head[h];
4042                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4043                         if (idx < s_idx)
4044                                 goto cont;
4045                         if (h > s_h || idx > s_idx)
4046                                 s_ip_idx = 0;
4047                         ip_idx = 0;
4048                         idev = __in6_dev_get(dev);
4049                         if (!idev)
4050                                 goto cont;
4051
4052                         if (in6_dump_addrs(idev, skb, cb, type,
4053                                            s_ip_idx, &ip_idx) <= 0)
4054                                 goto done;
4055 cont:
4056                         idx++;
4057                 }
4058         }
4059 done:
4060         rcu_read_unlock();
4061         cb->args[0] = h;
4062         cb->args[1] = idx;
4063         cb->args[2] = ip_idx;
4064
4065         return skb->len;
4066 }
4067
4068 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4069 {
4070         enum addr_type_t type = UNICAST_ADDR;
4071
4072         return inet6_dump_addr(skb, cb, type);
4073 }
4074
4075 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4076 {
4077         enum addr_type_t type = MULTICAST_ADDR;
4078
4079         return inet6_dump_addr(skb, cb, type);
4080 }
4081
4082
4083 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4084 {
4085         enum addr_type_t type = ANYCAST_ADDR;
4086
4087         return inet6_dump_addr(skb, cb, type);
4088 }
4089
4090 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4091 {
4092         struct net *net = sock_net(in_skb->sk);
4093         struct ifaddrmsg *ifm;
4094         struct nlattr *tb[IFA_MAX+1];
4095         struct in6_addr *addr = NULL, *peer;
4096         struct net_device *dev = NULL;
4097         struct inet6_ifaddr *ifa;
4098         struct sk_buff *skb;
4099         int err;
4100
4101         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4102         if (err < 0)
4103                 goto errout;
4104
4105         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4106         if (addr == NULL) {
4107                 err = -EINVAL;
4108                 goto errout;
4109         }
4110
4111         ifm = nlmsg_data(nlh);
4112         if (ifm->ifa_index)
4113                 dev = __dev_get_by_index(net, ifm->ifa_index);
4114
4115         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4116         if (!ifa) {
4117                 err = -EADDRNOTAVAIL;
4118                 goto errout;
4119         }
4120
4121         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4122         if (!skb) {
4123                 err = -ENOBUFS;
4124                 goto errout_ifa;
4125         }
4126
4127         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4128                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4129         if (err < 0) {
4130                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4131                 WARN_ON(err == -EMSGSIZE);
4132                 kfree_skb(skb);
4133                 goto errout_ifa;
4134         }
4135         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4136 errout_ifa:
4137         in6_ifa_put(ifa);
4138 errout:
4139         return err;
4140 }
4141
4142 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4143 {
4144         struct sk_buff *skb;
4145         struct net *net = dev_net(ifa->idev->dev);
4146         int err = -ENOBUFS;
4147
4148         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4149         if (skb == NULL)
4150                 goto errout;
4151
4152         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4153         if (err < 0) {
4154                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4155                 WARN_ON(err == -EMSGSIZE);
4156                 kfree_skb(skb);
4157                 goto errout;
4158         }
4159         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4160         return;
4161 errout:
4162         if (err < 0)
4163                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4164 }
4165
4166 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4167                                 __s32 *array, int bytes)
4168 {
4169         BUG_ON(bytes < (DEVCONF_MAX * 4));
4170
4171         memset(array, 0, bytes);
4172         array[DEVCONF_FORWARDING] = cnf->forwarding;
4173         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4174         array[DEVCONF_MTU6] = cnf->mtu6;
4175         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4176         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4177         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4178         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4179         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4180         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4181                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4182         array[DEVCONF_RTR_SOLICIT_DELAY] =
4183                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4184         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4185         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4186                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4187         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4188                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4189         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4190         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4191         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4192         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4193         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4194         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4195         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4196         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4197 #ifdef CONFIG_IPV6_ROUTER_PREF
4198         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4199         array[DEVCONF_RTR_PROBE_INTERVAL] =
4200                 jiffies_to_msecs(cnf->rtr_probe_interval);
4201 #ifdef CONFIG_IPV6_ROUTE_INFO
4202         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4203 #endif
4204 #endif
4205         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4206         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4207 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4208         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4209 #endif
4210 #ifdef CONFIG_IPV6_MROUTE
4211         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4212 #endif
4213         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4214         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4215         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4216         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4217         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4218 }
4219
4220 static inline size_t inet6_ifla6_size(void)
4221 {
4222         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4223              + nla_total_size(sizeof(struct ifla_cacheinfo))
4224              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4225              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4226              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4227              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4228 }
4229
4230 static inline size_t inet6_if_nlmsg_size(void)
4231 {
4232         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4233                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4234                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4235                + nla_total_size(4) /* IFLA_MTU */
4236                + nla_total_size(4) /* IFLA_LINK */
4237                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4238 }
4239
4240 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4241                                       int items, int bytes)
4242 {
4243         int i;
4244         int pad = bytes - sizeof(u64) * items;
4245         BUG_ON(pad < 0);
4246
4247         /* Use put_unaligned() because stats may not be aligned for u64. */
4248         put_unaligned(items, &stats[0]);
4249         for (i = 1; i < items; i++)
4250                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4251
4252         memset(&stats[items], 0, pad);
4253 }
4254
4255 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4256                                       int items, int bytes, size_t syncpoff)
4257 {
4258         int i;
4259         int pad = bytes - sizeof(u64) * items;
4260         BUG_ON(pad < 0);
4261
4262         /* Use put_unaligned() because stats may not be aligned for u64. */
4263         put_unaligned(items, &stats[0]);
4264         for (i = 1; i < items; i++)
4265                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4266
4267         memset(&stats[items], 0, pad);
4268 }
4269
4270 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4271                              int bytes)
4272 {
4273         switch (attrtype) {
4274         case IFLA_INET6_STATS:
4275                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4276                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4277                 break;
4278         case IFLA_INET6_ICMP6STATS:
4279                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4280                 break;
4281         }
4282 }
4283
4284 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4285 {
4286         struct nlattr *nla;
4287         struct ifla_cacheinfo ci;
4288
4289         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4290                 goto nla_put_failure;
4291         ci.max_reasm_len = IPV6_MAXPLEN;
4292         ci.tstamp = cstamp_delta(idev->tstamp);
4293         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4294         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4295         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4296                 goto nla_put_failure;
4297         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4298         if (nla == NULL)
4299                 goto nla_put_failure;
4300         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4301
4302         /* XXX - MC not implemented */
4303
4304         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4305         if (nla == NULL)
4306                 goto nla_put_failure;
4307         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4308
4309         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4310         if (nla == NULL)
4311                 goto nla_put_failure;
4312         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4313
4314         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4315         if (nla == NULL)
4316                 goto nla_put_failure;
4317         read_lock_bh(&idev->lock);
4318         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4319         read_unlock_bh(&idev->lock);
4320
4321         return 0;
4322
4323 nla_put_failure:
4324         return -EMSGSIZE;
4325 }
4326
4327 static size_t inet6_get_link_af_size(const struct net_device *dev)
4328 {
4329         if (!__in6_dev_get(dev))
4330                 return 0;
4331
4332         return inet6_ifla6_size();
4333 }
4334
4335 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4336 {
4337         struct inet6_dev *idev = __in6_dev_get(dev);
4338
4339         if (!idev)
4340                 return -ENODATA;
4341
4342         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4343                 return -EMSGSIZE;
4344
4345         return 0;
4346 }
4347
4348 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4349 {
4350         struct inet6_ifaddr *ifp;
4351         struct net_device *dev = idev->dev;
4352         bool update_rs = false;
4353         struct in6_addr ll_addr;
4354
4355         if (token == NULL)
4356                 return -EINVAL;
4357         if (ipv6_addr_any(token))
4358                 return -EINVAL;
4359         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4360                 return -EINVAL;
4361         if (!ipv6_accept_ra(idev))
4362                 return -EINVAL;
4363         if (idev->cnf.rtr_solicits <= 0)
4364                 return -EINVAL;
4365
4366         write_lock_bh(&idev->lock);
4367
4368         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4369         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4370
4371         write_unlock_bh(&idev->lock);
4372
4373         if (!idev->dead && (idev->if_flags & IF_READY) &&
4374             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4375                              IFA_F_OPTIMISTIC)) {
4376
4377                 /* If we're not ready, then normal ifup will take care
4378                  * of this. Otherwise, we need to request our rs here.
4379                  */
4380                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4381                 update_rs = true;
4382         }
4383
4384         write_lock_bh(&idev->lock);
4385
4386         if (update_rs) {
4387                 idev->if_flags |= IF_RS_SENT;
4388                 idev->rs_probes = 1;
4389                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4390         }
4391
4392         /* Well, that's kinda nasty ... */
4393         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4394                 spin_lock(&ifp->lock);
4395                 if (ifp->tokenized) {
4396                         ifp->valid_lft = 0;
4397                         ifp->prefered_lft = 0;
4398                 }
4399                 spin_unlock(&ifp->lock);
4400         }
4401
4402         write_unlock_bh(&idev->lock);
4403         addrconf_verify(0);
4404         return 0;
4405 }
4406
4407 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4408 {
4409         int err = -EINVAL;
4410         struct inet6_dev *idev = __in6_dev_get(dev);
4411         struct nlattr *tb[IFLA_INET6_MAX + 1];
4412
4413         if (!idev)
4414                 return -EAFNOSUPPORT;
4415
4416         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4417                 BUG();
4418
4419         if (tb[IFLA_INET6_TOKEN])
4420                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4421
4422         return err;
4423 }
4424
4425 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4426                              u32 portid, u32 seq, int event, unsigned int flags)
4427 {
4428         struct net_device *dev = idev->dev;
4429         struct ifinfomsg *hdr;
4430         struct nlmsghdr *nlh;
4431         void *protoinfo;
4432
4433         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4434         if (nlh == NULL)
4435                 return -EMSGSIZE;
4436
4437         hdr = nlmsg_data(nlh);
4438         hdr->ifi_family = AF_INET6;
4439         hdr->__ifi_pad = 0;
4440         hdr->ifi_type = dev->type;
4441         hdr->ifi_index = dev->ifindex;
4442         hdr->ifi_flags = dev_get_flags(dev);
4443         hdr->ifi_change = 0;
4444
4445         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4446             (dev->addr_len &&
4447              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4448             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4449             (dev->ifindex != dev->iflink &&
4450              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4451                 goto nla_put_failure;
4452         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4453         if (protoinfo == NULL)
4454                 goto nla_put_failure;
4455
4456         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4457                 goto nla_put_failure;
4458
4459         nla_nest_end(skb, protoinfo);
4460         return nlmsg_end(skb, nlh);
4461
4462 nla_put_failure:
4463         nlmsg_cancel(skb, nlh);
4464         return -EMSGSIZE;
4465 }
4466
4467 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4468 {
4469         struct net *net = sock_net(skb->sk);
4470         int h, s_h;
4471         int idx = 0, s_idx;
4472         struct net_device *dev;
4473         struct inet6_dev *idev;
4474         struct hlist_head *head;
4475
4476         s_h = cb->args[0];
4477         s_idx = cb->args[1];
4478
4479         rcu_read_lock();
4480         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4481                 idx = 0;
4482                 head = &net->dev_index_head[h];
4483                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4484                         if (idx < s_idx)
4485                                 goto cont;
4486                         idev = __in6_dev_get(dev);
4487                         if (!idev)
4488                                 goto cont;
4489                         if (inet6_fill_ifinfo(skb, idev,
4490                                               NETLINK_CB(cb->skb).portid,
4491                                               cb->nlh->nlmsg_seq,
4492                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4493                                 goto out;
4494 cont:
4495                         idx++;
4496                 }
4497         }
4498 out:
4499         rcu_read_unlock();
4500         cb->args[1] = idx;
4501         cb->args[0] = h;
4502
4503         return skb->len;
4504 }
4505
4506 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4507 {
4508         struct sk_buff *skb;
4509         struct net *net = dev_net(idev->dev);
4510         int err = -ENOBUFS;
4511
4512         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4513         if (skb == NULL)
4514                 goto errout;
4515
4516         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4517         if (err < 0) {
4518                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4519                 WARN_ON(err == -EMSGSIZE);
4520                 kfree_skb(skb);
4521                 goto errout;
4522         }
4523         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4524         return;
4525 errout:
4526         if (err < 0)
4527                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4528 }
4529
4530 static inline size_t inet6_prefix_nlmsg_size(void)
4531 {
4532         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4533                + nla_total_size(sizeof(struct in6_addr))
4534                + nla_total_size(sizeof(struct prefix_cacheinfo));
4535 }
4536
4537 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4538                              struct prefix_info *pinfo, u32 portid, u32 seq,
4539                              int event, unsigned int flags)
4540 {
4541         struct prefixmsg *pmsg;
4542         struct nlmsghdr *nlh;
4543         struct prefix_cacheinfo ci;
4544
4545         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4546         if (nlh == NULL)
4547                 return -EMSGSIZE;
4548
4549         pmsg = nlmsg_data(nlh);
4550         pmsg->prefix_family = AF_INET6;
4551         pmsg->prefix_pad1 = 0;
4552         pmsg->prefix_pad2 = 0;
4553         pmsg->prefix_ifindex = idev->dev->ifindex;
4554         pmsg->prefix_len = pinfo->prefix_len;
4555         pmsg->prefix_type = pinfo->type;
4556         pmsg->prefix_pad3 = 0;
4557         pmsg->prefix_flags = 0;
4558         if (pinfo->onlink)
4559                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4560         if (pinfo->autoconf)
4561                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4562
4563         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4564                 goto nla_put_failure;
4565         ci.preferred_time = ntohl(pinfo->prefered);
4566         ci.valid_time = ntohl(pinfo->valid);
4567         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4568                 goto nla_put_failure;
4569         return nlmsg_end(skb, nlh);
4570
4571 nla_put_failure:
4572         nlmsg_cancel(skb, nlh);
4573         return -EMSGSIZE;
4574 }
4575
4576 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4577                          struct prefix_info *pinfo)
4578 {
4579         struct sk_buff *skb;
4580         struct net *net = dev_net(idev->dev);
4581         int err = -ENOBUFS;
4582
4583         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4584         if (skb == NULL)
4585                 goto errout;
4586
4587         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4588         if (err < 0) {
4589                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4590                 WARN_ON(err == -EMSGSIZE);
4591                 kfree_skb(skb);
4592                 goto errout;
4593         }
4594         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4595         return;
4596 errout:
4597         if (err < 0)
4598                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4599 }
4600
4601 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4602 {
4603         write_lock_bh(&ifp->idev->lock);
4604         spin_lock(&ifp->lock);
4605         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4606                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4607             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4608                 ifp->idev->valid_ll_addr_cnt += count;
4609         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4610         spin_unlock(&ifp->lock);
4611         write_unlock_bh(&ifp->idev->lock);
4612 }
4613
4614 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4615 {
4616         struct net *net = dev_net(ifp->idev->dev);
4617
4618         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4619
4620         switch (event) {
4621         case RTM_NEWADDR:
4622                 update_valid_ll_addr_cnt(ifp, 1);
4623
4624                 /*
4625                  * If the address was optimistic
4626                  * we inserted the route at the start of
4627                  * our DAD process, so we don't need
4628                  * to do it again
4629                  */
4630                 if (!(ifp->rt->rt6i_node))
4631                         ip6_ins_rt(ifp->rt);
4632                 if (ifp->idev->cnf.forwarding)
4633                         addrconf_join_anycast(ifp);
4634                 if (!ipv6_addr_any(&ifp->peer_addr))
4635                         addrconf_prefix_route(&ifp->peer_addr, 128,
4636                                               ifp->idev->dev, 0, 0);
4637                 break;
4638         case RTM_DELADDR:
4639                 update_valid_ll_addr_cnt(ifp, -1);
4640
4641                 if (ifp->idev->cnf.forwarding)
4642                         addrconf_leave_anycast(ifp);
4643                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4644                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4645                         struct rt6_info *rt;
4646                         struct net_device *dev = ifp->idev->dev;
4647
4648                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4649                                         dev->ifindex, 1);
4650                         if (rt) {
4651                                 dst_hold(&rt->dst);
4652                                 if (ip6_del_rt(rt))
4653                                         dst_free(&rt->dst);
4654                         }
4655                 }
4656                 dst_hold(&ifp->rt->dst);
4657
4658                 if (ip6_del_rt(ifp->rt))
4659                         dst_free(&ifp->rt->dst);
4660                 break;
4661         }
4662         atomic_inc(&net->ipv6.dev_addr_genid);
4663         rt_genid_bump_ipv6(net);
4664 }
4665
4666 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4667 {
4668         rcu_read_lock_bh();
4669         if (likely(ifp->idev->dead == 0))
4670                 __ipv6_ifa_notify(event, ifp);
4671         rcu_read_unlock_bh();
4672 }
4673
4674 #ifdef CONFIG_SYSCTL
4675
4676 static
4677 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4678                            void __user *buffer, size_t *lenp, loff_t *ppos)
4679 {
4680         int *valp = ctl->data;
4681         int val = *valp;
4682         loff_t pos = *ppos;
4683         struct ctl_table lctl;
4684         int ret;
4685
4686         /*
4687          * ctl->data points to idev->cnf.forwarding, we should
4688          * not modify it until we get the rtnl lock.
4689          */
4690         lctl = *ctl;
4691         lctl.data = &val;
4692
4693         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4694
4695         if (write)
4696                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4697         if (ret)
4698                 *ppos = pos;
4699         return ret;
4700 }
4701
4702 static void dev_disable_change(struct inet6_dev *idev)
4703 {
4704         struct netdev_notifier_info info;
4705
4706         if (!idev || !idev->dev)
4707                 return;
4708
4709         netdev_notifier_info_init(&info, idev->dev);
4710         if (idev->cnf.disable_ipv6)
4711                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4712         else
4713                 addrconf_notify(NULL, NETDEV_UP, &info);
4714 }
4715
4716 static void addrconf_disable_change(struct net *net, __s32 newf)
4717 {
4718         struct net_device *dev;
4719         struct inet6_dev *idev;
4720
4721         rcu_read_lock();
4722         for_each_netdev_rcu(net, dev) {
4723                 idev = __in6_dev_get(dev);
4724                 if (idev) {
4725                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4726                         idev->cnf.disable_ipv6 = newf;
4727                         if (changed)
4728                                 dev_disable_change(idev);
4729                 }
4730         }
4731         rcu_read_unlock();
4732 }
4733
4734 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4735 {
4736         struct net *net;
4737         int old;
4738
4739         if (!rtnl_trylock())
4740                 return restart_syscall();
4741
4742         net = (struct net *)table->extra2;
4743         old = *p;
4744         *p = newf;
4745
4746         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4747                 rtnl_unlock();
4748                 return 0;
4749         }
4750
4751         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4752                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4753                 addrconf_disable_change(net, newf);
4754         } else if ((!newf) ^ (!old))
4755                 dev_disable_change((struct inet6_dev *)table->extra1);
4756
4757         rtnl_unlock();
4758         return 0;
4759 }
4760
4761 static
4762 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4763                             void __user *buffer, size_t *lenp, loff_t *ppos)
4764 {
4765         int *valp = ctl->data;
4766         int val = *valp;
4767         loff_t pos = *ppos;
4768         struct ctl_table lctl;
4769         int ret;
4770
4771         /*
4772          * ctl->data points to idev->cnf.disable_ipv6, we should
4773          * not modify it until we get the rtnl lock.
4774          */
4775         lctl = *ctl;
4776         lctl.data = &val;
4777
4778         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4779
4780         if (write)
4781                 ret = addrconf_disable_ipv6(ctl, valp, val);
4782         if (ret)
4783                 *ppos = pos;
4784         return ret;
4785 }
4786
4787 static
4788 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4789                               void __user *buffer, size_t *lenp, loff_t *ppos)
4790 {
4791         int *valp = ctl->data;
4792         int ret;
4793         int old, new;
4794
4795         old = *valp;
4796         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4797         new = *valp;
4798
4799         if (write && old != new) {
4800                 struct net *net = ctl->extra2;
4801
4802                 if (!rtnl_trylock())
4803                         return restart_syscall();
4804
4805                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4806                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4807                                                      NETCONFA_IFINDEX_DEFAULT,
4808                                                      net->ipv6.devconf_dflt);
4809                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4810                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4811                                                      NETCONFA_IFINDEX_ALL,
4812                                                      net->ipv6.devconf_all);
4813                 else {
4814                         struct inet6_dev *idev = ctl->extra1;
4815
4816                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4817                                                      idev->dev->ifindex,
4818                                                      &idev->cnf);
4819                 }
4820                 rtnl_unlock();
4821         }
4822
4823         return ret;
4824 }
4825
4826
4827 static struct addrconf_sysctl_table
4828 {
4829         struct ctl_table_header *sysctl_header;
4830         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4831 } addrconf_sysctl __read_mostly = {
4832         .sysctl_header = NULL,
4833         .addrconf_vars = {
4834                 {
4835                         .procname       = "forwarding",
4836                         .data           = &ipv6_devconf.forwarding,
4837                         .maxlen         = sizeof(int),
4838                         .mode           = 0644,
4839                         .proc_handler   = addrconf_sysctl_forward,
4840                 },
4841                 {
4842                         .procname       = "hop_limit",
4843                         .data           = &ipv6_devconf.hop_limit,
4844                         .maxlen         = sizeof(int),
4845                         .mode           = 0644,
4846                         .proc_handler   = proc_dointvec,
4847                 },
4848                 {
4849                         .procname       = "mtu",
4850                         .data           = &ipv6_devconf.mtu6,
4851                         .maxlen         = sizeof(int),
4852                         .mode           = 0644,
4853                         .proc_handler   = proc_dointvec,
4854                 },
4855                 {
4856                         .procname       = "accept_ra",
4857                         .data           = &ipv6_devconf.accept_ra,
4858                         .maxlen         = sizeof(int),
4859                         .mode           = 0644,
4860                         .proc_handler   = proc_dointvec,
4861                 },
4862                 {
4863                         .procname       = "accept_redirects",
4864                         .data           = &ipv6_devconf.accept_redirects,
4865                         .maxlen         = sizeof(int),
4866                         .mode           = 0644,
4867                         .proc_handler   = proc_dointvec,
4868                 },
4869                 {
4870                         .procname       = "autoconf",
4871                         .data           = &ipv6_devconf.autoconf,
4872                         .maxlen         = sizeof(int),
4873                         .mode           = 0644,
4874                         .proc_handler   = proc_dointvec,
4875                 },
4876                 {
4877                         .procname       = "dad_transmits",
4878                         .data           = &ipv6_devconf.dad_transmits,
4879                         .maxlen         = sizeof(int),
4880                         .mode           = 0644,
4881                         .proc_handler   = proc_dointvec,
4882                 },
4883                 {
4884                         .procname       = "router_solicitations",
4885                         .data           = &ipv6_devconf.rtr_solicits,
4886                         .maxlen         = sizeof(int),
4887                         .mode           = 0644,
4888                         .proc_handler   = proc_dointvec,
4889                 },
4890                 {
4891                         .procname       = "router_solicitation_interval",
4892                         .data           = &ipv6_devconf.rtr_solicit_interval,
4893                         .maxlen         = sizeof(int),
4894                         .mode           = 0644,
4895                         .proc_handler   = proc_dointvec_jiffies,
4896                 },
4897                 {
4898                         .procname       = "router_solicitation_delay",
4899                         .data           = &ipv6_devconf.rtr_solicit_delay,
4900                         .maxlen         = sizeof(int),
4901                         .mode           = 0644,
4902                         .proc_handler   = proc_dointvec_jiffies,
4903                 },
4904                 {
4905                         .procname       = "force_mld_version",
4906                         .data           = &ipv6_devconf.force_mld_version,
4907                         .maxlen         = sizeof(int),
4908                         .mode           = 0644,
4909                         .proc_handler   = proc_dointvec,
4910                 },
4911                 {
4912                         .procname       = "mldv1_unsolicited_report_interval",
4913                         .data           =
4914                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4915                         .maxlen         = sizeof(int),
4916                         .mode           = 0644,
4917                         .proc_handler   = proc_dointvec_ms_jiffies,
4918                 },
4919                 {
4920                         .procname       = "mldv2_unsolicited_report_interval",
4921                         .data           =
4922                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4923                         .maxlen         = sizeof(int),
4924                         .mode           = 0644,
4925                         .proc_handler   = proc_dointvec_ms_jiffies,
4926                 },
4927                 {
4928                         .procname       = "use_tempaddr",
4929                         .data           = &ipv6_devconf.use_tempaddr,
4930                         .maxlen         = sizeof(int),
4931                         .mode           = 0644,
4932                         .proc_handler   = proc_dointvec,
4933                 },
4934                 {
4935                         .procname       = "temp_valid_lft",
4936                         .data           = &ipv6_devconf.temp_valid_lft,
4937                         .maxlen         = sizeof(int),
4938                         .mode           = 0644,
4939                         .proc_handler   = proc_dointvec,
4940                 },
4941                 {
4942                         .procname       = "temp_prefered_lft",
4943                         .data           = &ipv6_devconf.temp_prefered_lft,
4944                         .maxlen         = sizeof(int),
4945                         .mode           = 0644,
4946                         .proc_handler   = proc_dointvec,
4947                 },
4948                 {
4949                         .procname       = "regen_max_retry",
4950                         .data           = &ipv6_devconf.regen_max_retry,
4951                         .maxlen         = sizeof(int),
4952                         .mode           = 0644,
4953                         .proc_handler   = proc_dointvec,
4954                 },
4955                 {
4956                         .procname       = "max_desync_factor",
4957                         .data           = &ipv6_devconf.max_desync_factor,
4958                         .maxlen         = sizeof(int),
4959                         .mode           = 0644,
4960                         .proc_handler   = proc_dointvec,
4961                 },
4962                 {
4963                         .procname       = "max_addresses",
4964                         .data           = &ipv6_devconf.max_addresses,
4965                         .maxlen         = sizeof(int),
4966                         .mode           = 0644,
4967                         .proc_handler   = proc_dointvec,
4968                 },
4969                 {
4970                         .procname       = "accept_ra_defrtr",
4971                         .data           = &ipv6_devconf.accept_ra_defrtr,
4972                         .maxlen         = sizeof(int),
4973                         .mode           = 0644,
4974                         .proc_handler   = proc_dointvec,
4975                 },
4976                 {
4977                         .procname       = "accept_ra_pinfo",
4978                         .data           = &ipv6_devconf.accept_ra_pinfo,
4979                         .maxlen         = sizeof(int),
4980                         .mode           = 0644,
4981                         .proc_handler   = proc_dointvec,
4982                 },
4983 #ifdef CONFIG_IPV6_ROUTER_PREF
4984                 {
4985                         .procname       = "accept_ra_rtr_pref",
4986                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4987                         .maxlen         = sizeof(int),
4988                         .mode           = 0644,
4989                         .proc_handler   = proc_dointvec,
4990                 },
4991                 {
4992                         .procname       = "router_probe_interval",
4993                         .data           = &ipv6_devconf.rtr_probe_interval,
4994                         .maxlen         = sizeof(int),
4995                         .mode           = 0644,
4996                         .proc_handler   = proc_dointvec_jiffies,
4997                 },
4998 #ifdef CONFIG_IPV6_ROUTE_INFO
4999                 {
5000                         .procname       = "accept_ra_rt_info_max_plen",
5001                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5002                         .maxlen         = sizeof(int),
5003                         .mode           = 0644,
5004                         .proc_handler   = proc_dointvec,
5005                 },
5006 #endif
5007 #endif
5008                 {
5009                         .procname       = "proxy_ndp",
5010                         .data           = &ipv6_devconf.proxy_ndp,
5011                         .maxlen         = sizeof(int),
5012                         .mode           = 0644,
5013                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5014                 },
5015                 {
5016                         .procname       = "accept_source_route",
5017                         .data           = &ipv6_devconf.accept_source_route,
5018                         .maxlen         = sizeof(int),
5019                         .mode           = 0644,
5020                         .proc_handler   = proc_dointvec,
5021                 },
5022 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5023                 {
5024                         .procname       = "optimistic_dad",
5025                         .data           = &ipv6_devconf.optimistic_dad,
5026                         .maxlen         = sizeof(int),
5027                         .mode           = 0644,
5028                         .proc_handler   = proc_dointvec,
5029
5030                 },
5031 #endif
5032 #ifdef CONFIG_IPV6_MROUTE
5033                 {
5034                         .procname       = "mc_forwarding",
5035                         .data           = &ipv6_devconf.mc_forwarding,
5036                         .maxlen         = sizeof(int),
5037                         .mode           = 0444,
5038                         .proc_handler   = proc_dointvec,
5039                 },
5040 #endif
5041                 {
5042                         .procname       = "disable_ipv6",
5043                         .data           = &ipv6_devconf.disable_ipv6,
5044                         .maxlen         = sizeof(int),
5045                         .mode           = 0644,
5046                         .proc_handler   = addrconf_sysctl_disable,
5047                 },
5048                 {
5049                         .procname       = "accept_dad",
5050                         .data           = &ipv6_devconf.accept_dad,
5051                         .maxlen         = sizeof(int),
5052                         .mode           = 0644,
5053                         .proc_handler   = proc_dointvec,
5054                 },
5055                 {
5056                         .procname       = "force_tllao",
5057                         .data           = &ipv6_devconf.force_tllao,
5058                         .maxlen         = sizeof(int),
5059                         .mode           = 0644,
5060                         .proc_handler   = proc_dointvec
5061                 },
5062                 {
5063                         .procname       = "ndisc_notify",
5064                         .data           = &ipv6_devconf.ndisc_notify,
5065                         .maxlen         = sizeof(int),
5066                         .mode           = 0644,
5067                         .proc_handler   = proc_dointvec
5068                 },
5069                 {
5070                         .procname       = "suppress_frag_ndisc",
5071                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5072                         .maxlen         = sizeof(int),
5073                         .mode           = 0644,
5074                         .proc_handler   = proc_dointvec
5075                 },
5076                 {
5077                         /* sentinel */
5078                 }
5079         },
5080 };
5081
5082 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5083                 struct inet6_dev *idev, struct ipv6_devconf *p)
5084 {
5085         int i;
5086         struct addrconf_sysctl_table *t;
5087         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5088
5089         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5090         if (t == NULL)
5091                 goto out;
5092
5093         for (i = 0; t->addrconf_vars[i].data; i++) {
5094                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5095                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5096                 t->addrconf_vars[i].extra2 = net;
5097         }
5098
5099         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5100
5101         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5102         if (t->sysctl_header == NULL)
5103                 goto free;
5104
5105         p->sysctl = t;
5106         return 0;
5107
5108 free:
5109         kfree(t);
5110 out:
5111         return -ENOBUFS;
5112 }
5113
5114 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5115 {
5116         struct addrconf_sysctl_table *t;
5117
5118         if (p->sysctl == NULL)
5119                 return;
5120
5121         t = p->sysctl;
5122         p->sysctl = NULL;
5123         unregister_net_sysctl_table(t->sysctl_header);
5124         kfree(t);
5125 }
5126
5127 static void addrconf_sysctl_register(struct inet6_dev *idev)
5128 {
5129         neigh_sysctl_register(idev->dev, idev->nd_parms,
5130                               &ndisc_ifinfo_sysctl_change);
5131         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5132                                         idev, &idev->cnf);
5133 }
5134
5135 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5136 {
5137         __addrconf_sysctl_unregister(&idev->cnf);
5138         neigh_sysctl_unregister(idev->nd_parms);
5139 }
5140
5141
5142 #endif
5143
5144 static int __net_init addrconf_init_net(struct net *net)
5145 {
5146         int err = -ENOMEM;
5147         struct ipv6_devconf *all, *dflt;
5148
5149         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5150         if (all == NULL)
5151                 goto err_alloc_all;
5152
5153         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5154         if (dflt == NULL)
5155                 goto err_alloc_dflt;
5156
5157         /* these will be inherited by all namespaces */
5158         dflt->autoconf = ipv6_defaults.autoconf;
5159         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5160
5161         net->ipv6.devconf_all = all;
5162         net->ipv6.devconf_dflt = dflt;
5163
5164 #ifdef CONFIG_SYSCTL
5165         err = __addrconf_sysctl_register(net, "all", NULL, all);
5166         if (err < 0)
5167                 goto err_reg_all;
5168
5169         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5170         if (err < 0)
5171                 goto err_reg_dflt;
5172 #endif
5173         return 0;
5174
5175 #ifdef CONFIG_SYSCTL
5176 err_reg_dflt:
5177         __addrconf_sysctl_unregister(all);
5178 err_reg_all:
5179         kfree(dflt);
5180 #endif
5181 err_alloc_dflt:
5182         kfree(all);
5183 err_alloc_all:
5184         return err;
5185 }
5186
5187 static void __net_exit addrconf_exit_net(struct net *net)
5188 {
5189 #ifdef CONFIG_SYSCTL
5190         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5191         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5192 #endif
5193         if (!net_eq(net, &init_net)) {
5194                 kfree(net->ipv6.devconf_dflt);
5195                 kfree(net->ipv6.devconf_all);
5196         }
5197 }
5198
5199 static struct pernet_operations addrconf_ops = {
5200         .init = addrconf_init_net,
5201         .exit = addrconf_exit_net,
5202 };
5203
5204 static struct rtnl_af_ops inet6_ops = {
5205         .family           = AF_INET6,
5206         .fill_link_af     = inet6_fill_link_af,
5207         .get_link_af_size = inet6_get_link_af_size,
5208         .set_link_af      = inet6_set_link_af,
5209 };
5210
5211 /*
5212  *      Init / cleanup code
5213  */
5214
5215 int __init addrconf_init(void)
5216 {
5217         int i, err;
5218
5219         err = ipv6_addr_label_init();
5220         if (err < 0) {
5221                 pr_crit("%s: cannot initialize default policy table: %d\n",
5222                         __func__, err);
5223                 goto out;
5224         }
5225
5226         err = register_pernet_subsys(&addrconf_ops);
5227         if (err < 0)
5228                 goto out_addrlabel;
5229
5230         /* The addrconf netdev notifier requires that loopback_dev
5231          * has it's ipv6 private information allocated and setup
5232          * before it can bring up and give link-local addresses
5233          * to other devices which are up.
5234          *
5235          * Unfortunately, loopback_dev is not necessarily the first
5236          * entry in the global dev_base list of net devices.  In fact,
5237          * it is likely to be the very last entry on that list.
5238          * So this causes the notifier registry below to try and
5239          * give link-local addresses to all devices besides loopback_dev
5240          * first, then loopback_dev, which cases all the non-loopback_dev
5241          * devices to fail to get a link-local address.
5242          *
5243          * So, as a temporary fix, allocate the ipv6 structure for
5244          * loopback_dev first by hand.
5245          * Longer term, all of the dependencies ipv6 has upon the loopback
5246          * device and it being up should be removed.
5247          */
5248         rtnl_lock();
5249         if (!ipv6_add_dev(init_net.loopback_dev))
5250                 err = -ENOMEM;
5251         rtnl_unlock();
5252         if (err)
5253                 goto errlo;
5254
5255         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5256                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5257
5258         register_netdevice_notifier(&ipv6_dev_notf);
5259
5260         addrconf_verify(0);
5261
5262         rtnl_af_register(&inet6_ops);
5263
5264         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5265                               NULL);
5266         if (err < 0)
5267                 goto errout;
5268
5269         /* Only the first call to __rtnl_register can fail */
5270         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5271         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5272         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5273                         inet6_dump_ifaddr, NULL);
5274         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5275                         inet6_dump_ifmcaddr, NULL);
5276         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5277                         inet6_dump_ifacaddr, NULL);
5278         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5279                         inet6_netconf_dump_devconf, NULL);
5280
5281         ipv6_addr_label_rtnl_register();
5282
5283         return 0;
5284 errout:
5285         rtnl_af_unregister(&inet6_ops);
5286         unregister_netdevice_notifier(&ipv6_dev_notf);
5287 errlo:
5288         unregister_pernet_subsys(&addrconf_ops);
5289 out_addrlabel:
5290         ipv6_addr_label_cleanup();
5291 out:
5292         return err;
5293 }
5294
5295 void addrconf_cleanup(void)
5296 {
5297         struct net_device *dev;
5298         int i;
5299
5300         unregister_netdevice_notifier(&ipv6_dev_notf);
5301         unregister_pernet_subsys(&addrconf_ops);
5302         ipv6_addr_label_cleanup();
5303
5304         rtnl_lock();
5305
5306         __rtnl_af_unregister(&inet6_ops);
5307
5308         /* clean dev list */
5309         for_each_netdev(&init_net, dev) {
5310                 if (__in6_dev_get(dev) == NULL)
5311                         continue;
5312                 addrconf_ifdown(dev, 1);
5313         }
5314         addrconf_ifdown(init_net.loopback_dev, 2);
5315
5316         /*
5317          *      Check hash table.
5318          */
5319         spin_lock_bh(&addrconf_hash_lock);
5320         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5321                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5322         spin_unlock_bh(&addrconf_hash_lock);
5323
5324         del_timer(&addr_chk_timer);
5325         rtnl_unlock();
5326 }